Use of Wicking Geotextile To Dehydrate Road Embankments

Scholars' Mine
Doctoral Dissertations Student Theses and Dissertations
Spring 2019
Use of wicking geotextile to dehydrate road embankments

Chuang Lin
Follow this and additional works at: https://scholarsmine.mst.edu/doctoral_dissertations
Part of the Civil Engineering Commons, and the Geotechnical Engineering Commons
Department: Civil, Architectural and Environmental Engineering
Recommended Citation
Lin, Chuang, "Use of wicking geotextile to dehydrate road embankments" (2019). Doctoral Dissertations.
3102.
https://scholarsmine.mst.edu/doctoral_dissertations/3102
This thesis is brought to you by Scholars' Mine, a service of the Missouri S&T Library and Learning Resources. This
work is protected by U. S. Copyright Law. Unauthorized use including reproduction for redistribution requires the
permission of the copyright holder. For more information, please contact [email protected].
USE OF WICKING GEOTEXTILE TO DEHYDRATE ROAD EMBANKMENTS
by
CHUANG LIN
A DISSERTATION
Presented to the Faculty of the Graduate School of the
MISSOURI UNIVERSITY OF SCIENCE AND TECHNOLOGY
In Partial Fulfillment of the Requirements for the Degree
DOCTOR OF PHILOSOPHY
in
CIVIL ENGINEERING
2019
Approved by:
Xiong Zhang, Advisor

Guney Olgun
Jenny Liu
Jie Han
Wen Deng
© 2019
CHUANG LIN
All Rights Reserved

iii
PUBLICATION DISSERTATION OPTION
This dissertation consists of the following eight articles, formatted in the style
used by the Missouri University of Science and Technology:
Paper I (pages 8-57) is under review for publication in the journal of Geotextiles
and Geomembranes.
Paper II (pages 58-107) has been published in the Journal of Materials in Civil
Engineering.
Paper III (pages 108-165) is under review for publication in the journal of
Geotextiles and Geomembranes.
Paper IV (pages 166-222) is under review for publication in the Journal of
Transportation Engineering.
Paper V (pages 223-271) is under review for publication in the International
Journal of Geomechanics.
Paper VI (pages 272-311) has been published in the Journal of Materials in Civil
Engineering.
Paper VII (pages 312-357) has been published in the Journal of Performance of
Constructed Facilities.
Paper VIII (pages 358-400) has been published in the Journal of Cleaner
Production.
iv
ABSTRACT
Excess water within a pavement structure is the major cause of pavement
deteriorations. A roadway is often constructed with soils compacted at the optimum water
contents to achieve the best performance. After construction, the soil water content
variation is controlled by the ambient environment. The base course aggregate is very
sensitive to water content variations. Unfortunately, conventional drainage system can
only drain gravity water, but not capillary water, and it cannot work under unsaturated
conditions. Consequently, no matter how well the road is constructed, the water content is
expected to increase with time and the excess water will inevitably accelerate pavement
deteriorations under repetitive traffic load. This study aims at solving the excess water
induced problems using a new wicking geotextile. Firstly, a series of laboratory test
results were performed to characterize the mechanical and hydraulic properties of the
wicking geotextile, the soil, and their interactions. After that, the laboratory test results
were used as numerical simulation inputs to evaluate the performance of the wicking
geotextile under different working and climatic conditions. The benefits of the wicking
geotextile were quantified and incorporated into the existing pavement design methods.
Then, the field performance of the wicking geotextile was continuously monitored and
potential issues that might influence the long-term performance of the wicking geotextile
are evaluated. Eventually, a new-bio-wicking system is proposed to further improve the
drainage efficiency of the wicking geotextile. The results from laboratory tests, numerical
simulations, and field observations validated the efficiency of the wicking geotextile to
dehydrate road embankments.

v
ACKNOWLEDGMENTS
I would like to express my sincere gratitude to my advisor, Dr. Xiong Zhang, for
his professional guidance, support, and encouragement. It is my pleasure and privilege to
be one of his students and his guidance becomes my life-long precious heritage. Words
cannot express how grateful I am to have such an amazing advisor.
I would like to thank all the committee members, Dr. Jenny Liu, Dr. Jie Han, Dr.
Guney Olgun, and Dr Wen Deng, for their insightful suggestions during my Ph.D. study.
Part of my study was sponsored by TenCate Geosynthetics. I would like to express my
thanks to Mr. John Lostumbo, Mr. Brett Odgers for their support and guidance. Thanks
also extended to Dr. George Koerner and Dr. Robert Koerner for awarding me the GSI
Fellowship for the 2017-18 academic year.
Thanks further extended to the geotechnical research group at Missouri S&T,
Beshoy Riad, Javad Galinmoghadam, Xiaolong Xia, Chao Zeng, and Elieh Mohtashami,
for their insightful advice and endless support. Special thanks go to Yipeng Guo, a
visiting scholar from Central South University, for the collaboration work on the wicking
geotextile. In addition, I also received great help from the stuff working in the civil,
architectural, and environmental engineering department, Mr. Brian Swift, Gary Abbott,
John Bullock, and Greg Leckrone.
I would like to wholeheartedly thank my wife, Meng Zhang, and our parents.
Their love and unconditional support encourage me to overcome any difficulties in my
life. I feel honored to be their loved one. This dissertation cannot be accomplished
without their love.

vi
TABLE OF CONTENTS
Page
PUBLICATION DISSERTATION OPTION ................................................................... iii
ABSTRACT ....................................................................................................................... iv
ACKNOWLEDGMENTS ...................................................................................................v
LIST OF ILLUSTRATIONS .............................................................................................xv
LIST OF TABLES .............................................................................................................xx
SECTION
1. INTRODUCTION ...................................................................................................... 1
1.1. PROBLEM STATEMENT ......................................................................... 1
1.1.1. Adverse Effect of Excess Water in Road Embankments. .............. 1
1.1.2. Limitations of Conventional Subsurface Drainage Systems. ........ 2
1.2. A NEW DRAINAGE SYSTEM WITH WICKING GEOTEXTILE ......... 4
1.3. RESEARCH OBJECTIVES AND METHODOLOGIES ........................... 6
1.4. DISSERTATION OUTLINE ...................................................................... 7
PAPER
I. CHARACTERIZATIONS OF WATER RETENTION CURVES FOR SEVERAL

TYPES OF WOVEN GEOTEXTILES ...................................................................... 8
ABSTRACT ................................................................................................................... 8
1. INTRODUCTION ...................................................................................................... 9
2. RESEARCH OBJECTIVES AND METHODOLOGIES ........................................ 13
3. TESTING MATERIALS ......................................................................................... 15
4.TESTING PROCEDURES ....................................................................................... 18

vii
4.1. CAPILLARY RISE TEST ........................................................................ 19
4.2. PRESSURE PLATE TEST ....................................................................... 21
4.3. SALT CONCENTRATION TEST ........................................................... 24
5. TEST RESULTS AND DISCUSSIONS .................................................................. 25
5.1. CAPILLARY RISE TEST ........................................................................ 25
5.2. PRESSURE PLATE TEST ....................................................................... 29
5.3. SALT CONCENTRATION TEST ........................................................... 31
6. COMPARISONS OF TESTING TECHNIQUES .................................................... 33
7. WATER RETENTION CURVES (WRCs) ............................................................. 36
8. CONCLUSIONS ...................................................................................................... 39
REFERENCES ............................................................................................................. 55
II. COMPREHENSIVE MATERIAL CHARACTERIZATIONS OF PAVEMENT

STRUCTURE INSTALLED WITH WICKING FABRICS ................................... 58
ABSTRACT ................................................................................................................. 58
1. INTRODUCTION .................................................................................................... 59
2. A NEW GEOTEXTILE WITH WICKING FIBERS............................................... 62
3. RESEARCH OBJECTIVES AND METHODOLOGIES ........................................ 65
4. MATERIAL CHARACTERIZATIONS .................................................................. 66
4.1. SOIL PROPERTIES.................................................................................. 66
4.1.1. Physical Properties. ...................................................................... 66
4.1.2. Resilient Modulus. ....................................................................... 66
4.1.3. Soil Water Characteristic Curve (SWCC). .................................. 70
4.1.4. Saturated Hydraulic Conductivity. .............................................. 73
4.2. WICKING FABRIC PROPERTIES ......................................................... 73

viii
4.2.1. Initial Tangent Modulus............................................................... 73
4.2.2. Geotextile Water Characteristic Curve (GWCC). ....................... 74
4.2.3. Saturated Hydraulic Conductivity. .............................................. 77
4.3. SOIL-GEOTEXTILE INTERFACE FRICTIONAL ANGLE .................. 78
5. PERFORMANCE OF THE SOIL-GEOTEXTILE SYSTEM ................................. 79
5.1. WORKING MECHANISM AND FUNCTIONAL RANGE ................... 79
5.2. THEORETICAL WATER REMOVAL CAPABILITY ........................... 83
5.3. NUMERICAL VALIDATION ................................................................. 85
6. CONCLUSIONS ...................................................................................................... 88
REFERENCES ........................................................................................................... 102
III. LABORATORY EVALUATION OF DIFFERENT GEOSYNTHETICS FOR

WATER DRAINAGE .......................................................................................... 108
ABSTRACT ............................................................................................................... 108
1. INTRODUCTION .................................................................................................. 109
2. TEST MATERIALS AND TEST PROCEDURES ............................................... 113
2.1. TESTING MATERIALS ........................................................................ 113
2.2. EXPERIMENT SETUP .......................................................................... 115
2.3. INSTRUMENTATION ........................................................................... 116
2.4. EXPERIMENTAL DESIGN ................................................................... 117
2.5. TESTING PROCEDURES ..................................................................... 118
3. RESULTS AND DISCUSSIONS .......................................................................... 120
3.1. GENERAL VISUAL OBSERVATIONS FOR WATER DRAINAGE . 120
3.2. VARIATIONS OF VOLUMETRIC WATER CONTENT .................... 121

ix
3.3. VOLUMETRIC WATER CONTENT CONTOURS FOR

DIFFERENT TESTS............................................................................... 123
3.4. CUMULATIVE AMOUNT OF WATER ............................................... 128
3.5. GRAVIMETRIC WATER CONTENT AFTER 10 DAYS .................... 130
4. WATER RETENTION CURVES FOR DIFFERENT MATERIALS .................. 133
5. WORKING MECHANISMS FOR DIFFERENT MATERIALS .......................... 136
6. CONCLUSIONS .................................................................................................... 140
ACKNOWLEDGEMENTS ....................................................................................... 141
REFERENCES ........................................................................................................... 162
IV. NUMERICAL SIMULATION OF SEASONAL VARIATIONS OF BASE

COURSE RESILIENT MODULUS IN PAVEMENT STRUCTURES .............. 166
ABSTRACT ............................................................................................................... 166
1. INTRODUCTION .................................................................................................. 167
2. FINITE ELEMENT MODEL ................................................................................ 171
2.1. GOVERNING EQUATIONS ................................................................. 171
2.2. CLIMATIC EFFECT .............................................................................. 172
2.3. MODEL VALIDATION ......................................................................... 179
3. MODEL CONFIGURATIONS AND MATERIAL INPUTS ............................... 181
4. SIMULATION RESULTS AND DISCUSSIONS ................................................ 184
4.1. CASE I: NEW PAVEMENT WITHOUT CRACKING ......................... 185
4.2. CASE II: MODERATELY DETERIORATED PAVEMENT WITH

RAINFALL EVENTS ............................................................................. 188
4.3. CASE III: PAVEMENT WITH CRACKS AND REAL TIME

CLIMATIC DATA.................................................................................. 194
5. CONCLUSIONS .................................................................................................... 198

x
REFERENCES ........................................................................................................... 219
V. WORKING MECHANISM OF A NEW WICKING GEOTEXTILE IN

ROADWAY APPLICATIONS: A NUMERICAL STUDY ................................ 223
ABSTRACT ............................................................................................................... 223
1. INTRODUCTION .................................................................................................. 225
2. A NEW WOVEN GEOTEXTILE WITH LATERAL DRAINAGE

CAPABILITY ........................................................................................................ 227
3. RESEARCH OBJECTIVES AND METHODOLOGIES ...................................... 231
4. NUMERICAL MODEL AND MODEL CALIBRATION .................................... 232
4.1. NUMERICAL MODEL .......................................................................... 232
4.2. MATERIAL PROPERTIES .................................................................... 233
4.3. MODEL CALIBRATION ....................................................................... 236
5. THE DRAINAGE PERFORMANCE OF THE WICKING GEOTEXTILE

UNDER DIFFERENT WORKING CONDITIONS .............................................. 238
5.1. RISING OF GROUNDWATER TABLE ............................................... 239
5.2. PRECIPITATION INFILTRATION ...................................................... 241
5.2.1. Light Rainfall Event................................................................... 242
5.2.2. Heavy Rainfall Event. ................................................................ 245
5.3. DRAINAGE PERFORMANCE OF A ROADWAY INSTALLED

WITH THE WICKING GEOTEXTILE. ................................................. 247
6. FIELD VALIDATIONS......................................................................................... 250
7. CONCLUSIONS .................................................................................................... 252
REFERENCES ........................................................................................................... 269
VI. LABORATORY DRAINAGE PERFORMANCE OF A NEW GEOTEXTILE

WITH WICKING FABRIC ................................................................................. 272
xi
ABSTRACT ............................................................................................................... 272
1. INTRODUCTION .................................................................................................. 273
2. A NEW GEOTEXTILE WITH WICKING FABRIC ............................................ 276
2.1. PROPOSED DRAINAGE DESIGN WITH WICKING FABRIC ......... 276
2.2. LAB VALIDATION ............................................................................... 277
2.3. FIELD APPLICATIONS ........................................................................ 278
3. RESEARCH OBJECTIVES................................................................................... 279
4. TESTING FLUME CONSTRUCTION ................................................................. 280
5. TEST RESULTS AND DISCUSSIONS ................................................................ 282
5.1. TESTING FLUME FOR SAND ............................................................. 283
5.1.1. Wicking Fabric Overall Performance in Sand. .......................... 283
5.1.2. Wicking (Drying) Test. .............................................................. 286
5.1.3. Wetting Test. .............................................................................. 287
5.1.4. Influence of Wicking Fabric Splice Area. ................................. 289
5.2. TESTING FLUME FOR E-1 AGGREGATE ......................................... 290
5.2.1. Wicking Fabric Overall Performance in E-1 Aggregate. .......... 290
5.2.2. Wicking (Drying) Process. ........................................................ 291
5.2.3. Wetting Process. ........................................................................ 293
5.3. CLOGGING EFFECT ............................................................................. 293
6. CONCLUSIONS .................................................................................................... 295
REFERENCES ........................................................................................................... 309
VII. LONG-TERM PERFORMANCE OF WICKING FABRIC IN ALASKAN

PAVEMENTS .................................................................................................... 312
ABSTRACT ............................................................................................................... 312

xii
1. INTRODUCTION .................................................................................................. 313
2. TEST SECTION CONSTRUCTION AND INSTRUMENTATION .................... 318
3. RESULTS AND ANALYSIS ................................................................................ 319
3.1. GENERAL CLIMATIC CONDITIONS................................................. 319
3.2. SOIL TEMPERATURE CHANGES ...................................................... 320
3.3. SOIL MOISTURE CHANGES ............................................................... 323
3.4. PERFORMANCE OF WICKING FABRIC AT DIFFERENT

CLIMATIC CONDITIONS .................................................................... 326
3.4.1. During Rainfall Events. ............................................................. 326
3.4.2. During Freezing Process. ........................................................... 328
3.4.3. During Thawing Process. ........................................................... 329
3.5. CLOGGING EFFECT ............................................................................. 330
3.6. PERMANENT DEFORMATION AND MECHANICAL FAILURE ... 332
3.7. AGING .................................................................................................... 333
4. DISCUSSIONS ...................................................................................................... 334
4.1. FACTORS INFLUENCING THE OCCURRENCE OF FROST

BOILS/SOFT SPOTS ............................................................................. 334
4.2. WICKING FABRIC LONG-TERM PERFORMANCE......................... 336
5. CONCLUSIONS .................................................................................................... 337
REFERENCES ........................................................................................................... 356
VIII. A BIO-WICKING SYSTEM TO DEHYDRATE ROAD EMBANKMENT ... 358
ABSTRACT ............................................................................................................... 358
1. INTRODUCTION .................................................................................................. 359

xiii
2. TEST MATERIALS AND TEST METHODOLOGIES ....................................... 364
2.1. TEST MATERIALS................................................................................ 365
2.1.1. E-1 Aggregate Properties. .......................................................... 365
2.1.2. Vegetation Properties. ................................................................ 365
2.2. TEST DESIGNS AND CONSTRUCTION PROCESSES ..................... 366
2.2.1. Elemental Level Test. ................................................................ 366
2.2.2. Full-Scale Test. .......................................................................... 368
2.3. TEST SCHEDULE.................................................................................. 371
3. TEST RESULTS AND DISCUSSIONS ................................................................ 372
3.1. CLIMATIC DATA.................................................................................. 373
3.2. ELEMENTAL LEVEL TEST ................................................................. 373
3.2.1. Test Boxes Net Water Gain. ...................................................... 373
3.2.2. Moisture Contours in Testing Boxes. ........................................ 375
3.3. FULL-SCALE TEST .............................................................................. 376
3.3.1. Testing Flumes Net Water Gain. ............................................... 376
3.3.2. Effect of Rainfall Event. ............................................................ 378
3.4. BIO-WICKING SYSTEM WORKING MECHANISM ........................ 379
3.4.1. Macroscopic Analysis. ............................................................... 380
3.4.2. Microscopic Analysis. ............................................................... 380
4. CONCLUSIONS .................................................................................................... 382
REFERENCES ........................................................................................................... 398

xiv
SECTION
2. CONCLUSIONS AND RECOMMENDATIONS................................................. 401
2.1. CONCLUSIONS ..................................................................................... 401
2.1.1. Laboratory Characterizations. .................................................... 401
2.1.2. Numerical Simulations. ............................................................. 402
2.1.3. Engineering Applications. ......................................................... 403
2.2. RECOMMENDATIONS ........................................................................ 405
BIBLIOGRAPHY ............................................................................................................407
VITA ................................................................................................................................409
xv
LIST OF ILLUSTRATIONS
SECTION Page
Figure 1.1. Wicking geotextile at different magnification ratios ........................................ 5
Figure 1.2. Proposed subsurface drainage system with wicking geotextile ....................... 5
PAPER I
Figure 1. Testing materials ............................................................................................... 42
Figure 2. Capillary rise test ............................................................................................... 44
Figure 3. Pressure plate test .............................................................................................. 45
Figure 4. Salt concentration test........................................................................................ 47
Figure 5. Capillary rise test results ................................................................................... 48
Figure 6. Pressure plate test results ................................................................................... 50
Figure 7. Salt concentration test results ............................................................................ 52
Figure 8. Summary of test results for non-wicking geotextiles ........................................ 52
Figure 9. Summary of test results for wicking geotextiles ............................................... 53
Figure 10. Water retention curves (WRCs) of woven geotextiles .................................... 54
PAPER II
Figure 1. A pavement structure with the conventional drainage system .......................... 92
Figure 2. A pavement structure with the wicking fabric .................................................. 92
Figure 3. Resilient modulus test results ............................................................................ 93
Figure 4. Determination of SWCC ................................................................................... 94
Figure 5. GWCC determination ........................................................................................ 95
Figure 6. Wicking fabric constant head test...................................................................... 97

xvi
Figure 7. Large-scale direct shear test .............................................................................. 98
Figure 8. Performance of the soil-geotextile system ........................................................ 99
Figure 9. Numerical model configurations ..................................................................... 100
Figure 10. Simulation results .......................................................................................... 100
PAPER III
Figure 1. Influence of soil water content on resilient modulus and permanent

deformation ...................................................................................................... 144
Figure 2. Drainage materials used in the laboratory tests ............................................... 145
Figure 3. Test apparatus .................................................................................................. 147
Figure 4. Wetting fronts after 10 days of drainage ......................................................... 148
Figure 5. Monitored volumetric water content variations with time under different
testing conditions ............................................................................................ 150
Figure 6. Volumetric water content contours under different testing conditions ........... 152
Figure 7. Cumulative amount of water under different testing conditions ..................... 156
Figure 8. Gravimetric water content contours after the test............................................ 157
Figure 9. Relative gravimetric water content difference contours .................................. 158
Figure 10. Water Retention Curves (WRCs) of different testing materials .................... 159
Figure 11. Comparisons of suction distributions with elevations in T1-T4 after the
test ................................................................................................................. 160
Figure 12. Working mechanisms for different drainage materials ................................. 161
PAPER IV
Figure 1. Meteorological data ......................................................................................... 201
Figure 2. Water balance analysis for soil-climatic interactions ...................................... 203
Figure 3. FEM model calibration .................................................................................... 204
Figure 4. Model configurations....................................................................................... 205

xvii
Figure 5. Stress distributions at different locations ........................................................ 206
Figure 6. Simulation results for Case I (a new pavement without cracks) ..................... 206
Figure 7. Suction, water content, and resilient modulus distributions at tire

centerlines ....................................................................................................... 208
Figure 8. Simulation results for Case II (moderately deteriorated pavement with fully
developed cracks)............................................................................................. 210
Figure 9. Comparisons of total amount of water infiltrated into the pavement

structure........................................................................................................... 212
Figure 10. Resilient modulus contours during rainfall events ........................................ 214
Figure 11. Simulation results for Case III (moderately deteriorated pavement with
real-time meteorological data) ....................................................................... 216
Figure 12. Water content variations with time at selected vegetation areas ................... 218
PAPER V
Figure 1. Capillary break effect ...................................................................................... 256
Figure 2. Macro- and microstructures of the wicking geotextile ................................. 257
Figure 3. A wicking geotextile with lateral drainage ability .......................................... 258
Figure 4. Successful field applications ........................................................................... 259
Figure 5. Hydraulic Properties ........................................................................................ 260
Figure 6. Model calibration............................................................................................. 261
Figure 7. Simulation results of rising groundwater table ................................................ 262
Figure 8. Comparisons of simulation results during a light rainfall event (wicking vs.
non-wicking geotextiles)................................................................................. 263
Figure 9. Simulation results of suction distributions during a heavy rainfall event ....... 265
Figure 10. Schematic plot of a road embankment installed with the wicking geotextile
(not to scale) ................................................................................................... 266
Figure 11. Simulation results of a road embankment installed with the wicking
geotextile ....................................................................................................... 267
xviii
Figure 12. Field evidence of the effectiveness of the wicking geotextile ....................... 268
PAPER VI
Figure 1. New geotextile structure and conceptual application ...................................... 298
Figure 2. Wetting front movement tests ......................................................................... 298
Figure 3. Wicking fabric field application at Beaver Slide on Dalton Highway, AK .... 299
Figure 4. Testing flume design and construction: ........................................................... 300
Figure 5. Total amount of water (within the monitored area) in the testing flume for
sand ................................................................................................................. 302
Figure 6. Moisture content contours of the testing flume for sand ................................. 303
Figure 7. Effect of wicking fabric splice area ................................................................. 304
Figure 8. Total amount of water (within the monitored area) for the testing flume for
E-1 aggregate .................................................................................................. 306
Figure 9. Moisture content contours of the testing flume for E-1 aggregate .................. 307
Figure 10. SEM images of wicking fabric samples ........................................................ 308
PAPER VII
Figure 1. Profile of the test section (Zhang et al., 2014) ................................................ 342
Figure 2. Hourly climatic data at beaver slide test section ............................................. 342
Figure 3. Soil temperature changes ................................................................................. 343
Figure 4. Soil moisture changes ...................................................................................... 345
Figure 5. Soil moisture contours during rainfall events .................................................. 348
Figure 6. Moisture contours before freezing................................................................... 350
Figure 7. Moisture contours during thawing process...................................................... 351
Figure 8. SEM images of clogging effect ....................................................................... 353
Figure 9. SEM images of mechanical failure.................................................................. 354

xix
Figure 10. Aging effect ................................................................................................... 355
PAPER VIII
Figure 1. Comparisons of the original and the bio-wicking system designs .................. 385
Figure 2. Gradation curve for E-1 aggregate .................................................................. 386
Figure 3. Schematic plot of elemental level test designs (not to scale) .......................... 387
Figure 4. Elemental level testing box construction process ............................................ 388
Figure 5. Full scale test design (not to scale) .................................................................. 389
Figure 6. Vegetation area construction process .............................................................. 390
Figure 7. Testing flumes construction process ............................................................... 391
Figure 8. Climatic data during testing period ................................................................. 392
Figure 9. Elemental level test results .............................................................................. 393
Figure 10. Moisture contours for small box test ............................................................. 394
Figure 11. Comparisons of net water gain for testing flumes ......................................... 395
Figure 12. Moisture contours for full-scale test .............................................................. 396
Figure 13. Grass root and wicking fabric interactions .................................................... 397
Figure 14. SEM images of grass and wicking fabric ...................................................... 397
xx
LIST OF TABLES
PAPER I Page
Table 1. Material properties of T1 (conventional geotextile) and T2 (new geotextile) .... 41
Table 2. Salt concentrations and the corresponding suctions ........................................... 41
Table 3. Regression parameters ........................................................................................ 41
PAPER II
Table 1. Specifications for the new geotextile with wicking fibers (TenCate 2015) ....... 91
Table 2. Summary of laboratory tests ............................................................................... 91
Table 3. Material properties for numerical simulation models ......................................... 92
PAPER III
Table 1. Wicking geotextile specifications ..................................................................... 142
Table 2. WRC parameters for different drainage materials ............................................ 143
PAPER IV
Table 1. Material inputs .................................................................................................. 200
Table 2. Effective soil resilient modulus ........................................................................ 200
PAPER V
Table 1. Soil and geotextile properties ........................................................................... 255
PAPER VI
Table 1. Geotextile specifications ................................................................................... 297
PAPER VII
Table 1. Geotextile specification .................................................................................... 340
Table 2. Rainfall events summary .................................................................................. 341
Table 3. SEM analyses summary .................................................................................... 341

1
SECTION
1. INTRODUCTION
1.1. PROBLEM STATEMENT
1.1.1. Adverse Effect of Excess Water in Road Embankments. The subsurface
water in a road embankment exists in four forms: water vapor, bounded water, capillary
water, and “free” (or gravitational) water (Kochina and Ya 1952; Aravin and Numerov
1953; and Muskat 1946). The water vapor in most cases stores inside soil pores where
above the saturation zone. In the existing subsurface drainage designs, water vapor
transmission is negligible. For bounded water, it is relatively hard to move from the soil
particles and can be considered as part of the soil particles. This part of the water phase in
soil also cannot move under gravity force and therefore is not considered in the designs.
Capillary water also exists in the soil pores where above the saturation zone. However,
different from water vapor, it can flow under the action of surface tension. The height of
capillary rise is a function of the soil particle distribution, which relates to the pore size
distribution and density of the soils (Lane and Washburn 1946; and Barber and Sawyer
1952). Since capillary water cannot be drained out by gravity, the most common way to
control capillary water is to lower the water table or use capillary barrier, which blocks the
upward capillary flow. Last but the most common type of water, namely free water, is the
water in liquid form that flows under the force of gravity and obeys Darcy’s law. Control
free water becomes the major concern in the existing subsurface drainage designs. The
subsurface water comes from a variety of sources and mainly falls into two categories:
2
groundwater and infiltration (Brown et al. 2001). Groundwater refers to the water exists in
the saturation zone below the water table. The major source of groundwater is
precipitation. Infiltration water is defined as the water seeps into the pavement structure
through pavement surface, shoulders, or median. Precipitation is also the major source for
infiltration water. For bituminous pavements, the primary infiltration water source is
longitudinal joints at shoulders and construction joints between strips of paving. As for
concrete slabs, infiltration water takes place through cracks, joints, and shoulders
(Cedergren 1973 and 1974).
Excessive water in a pavement structure is recognized as one of the major adverse
factors that influence its overall performance (Cedergren 1994; Christopher and
McGuffey 1997; and Henry and Holtz 2001). Specific problems associated with water are
causing soil expansion and collapsing, soil stiffness reduction, pumping, stripping asphalt
pavement and crack generation (Han and Zhang 2014). Water related problems are
responsible for decreased pavement life, increased costs for maintenance, and increased
pavement roughness. The water induced roadway deteriorations occur to some extent
throughout regions and climates of the United States. A recent NCHRP (National
Cooperative Highway Research Program) study estimated that the excess water reduces
the life expectancy by more than half (Christopher and McGuffey 1997). In addition,
government transportation engineers in cold regions have credited a minimum of half of
road maintenance expenditures to the effect of freezing and thawing.
1.1.2. Limitations of Conventional Subsurface Drainage Systems. Regarding a
road embankment, the soil is compacted at the optimum water content to achieve the
maximum dry density and the best performance. After construction, the soil moisture
3
content is con nature has the ability to thoroughly air dry the soil in the embankment. The
suction in the air is often higher than 14 MPa since the relative humidity is normally lower
than 90%. The surface layer is air-dried quickly under such high suction level and
becomes impermeable as the degree of saturation decreased. The air-dried surface layer
serves as a plastic membrane on top of the road embankment and impedes the water
exchange between the ambient environment and the soils inside. On the other hand, the
soil water content will increase with time due to precipitation infiltration, capillary action,
rising of water table, and infiltration from cracks at the slope.
One conventional treatment includes using better materials that was not water
sensitive and contains limited fines. Unfortunately, the good quality material may not be
readily available on site. Moreover, the price of good quality material is more expensive
than marginal one, which makes it economically infeasible considering a road that
extends hundreds of miles long. Another treatment uses woven geotextiles to
mechanically reinforce the road embankment. The soil-geotextile composite mainly
works under unsaturated conditions. Due to positive pore water pressure built up in the
overlying soil, the benefits of the mechanical reinforcement cannot be fully taken under
unsaturated conditions. In addition, the reinforcing range is limited considering the
relatively thin layer of the geotextile to the overall thickness of the base course. In
summary, none of the conventional treatments can effectively resolve the problem caused
by excess water.
In order to obtain sufficient pavement drainage, the major design considerations
involve: preventing the amount of water entering the pavement structure and quickly
removing water that enters the pavement system, using materials that are insensitive to
4
the effect of moisture, and incorporating design methods to minimize water damage
(ARA 2004; FHWA 1980; and AASHTO 1993). A subsurface drainage system is often
consisted of a layer of coarse-grained soil or geosynthetics. Conventional drainage
systems rely on gravity to drain water out of soils, which cannot drain the capillary water.
In addition, the drainage systems cannot work under unsaturated conditions.
Consequently, no matter how well the road is constructed, the excess water will soften
the soil with time and will accelerate the road deterioration under repetitive traffic load.
1.2. A NEW DRAINAGE SYSTEM WITH WICKING GEOTEXTILE
A new type of woven geotextile with wicking fibers has been recently developed
which has the potential to drain both capillary water and “free” water in road
embankments. Figure 1.1 shows the images of the wicking geotextile at different
magnification magnitudes. The wicking geotextile is made of a special type of
hydrophilic nylon fibers with multichannel cross-sections. Figure 1.1a shows the top
view of the wicking fabric. Figures 1.1b-c show the scanning electron microscopy (SEM)
photo of multiple fibers and a single fiber at different magnification ratios, respectively.
The multichannel cross-section has a high shape factors and great number of channels per
fiber (specific surface area = 3650 cm2/g), which give wicking geotextile great potential
for maximizing capillary action and water transport in an unsaturated environment. Most
importantly, it can keep saturated under low RH (high suction) environment.
By installing a layer of wicking geotextile horizontally in the pavement, as shown
in Figure 1.2, the water in the pavement structure can be absorbed from the soils,
transported along the wicking geotextile to the shoulder, and vaporized to the
5
surrounding atmosphere which has much higher suction. It is likely to generate a
relatively dry zone in the pavement structure which not only can help improve the
performance pavement structure, but also help prevent the frost heave and subsequent
thaw-weakening in cold regions.
Figure 1.1. Wicking geotextile at different magnification ratios
Water Infiltration via Cracks

Suction > 14,000kPa
(RH < 90%)
Relatively Dry Zone
Air-Dried Soil Wicking Fabric
(Impermeable)
Capillary Water
Figure 1.2. Proposed subsurface drainage system with wicking geotextile

6
1.3. RESEARCH OBJECTIVES AND METHODOLOGIES
The primary objective of this research is to evaluate the efficiency of the wicking
geotextile to dehydrate road embankments. There are some fundamental questions need
to be answered:(1) what is the working mechanism of the wicking geotextile to dehydrate
road embankments, (2) how to quantify the benefits of the drainage system with the
wicking geotextile, (3) how to incorporate the benefits into the existing pavement design
methods, and (4) what is the field performance of the wicking geotextile and what are the
potential issues that will influence the long-term performance of the wicking geotextile.
To answer those questions, the scope of this research has been divided into three
phases, including material characterizations, numerical simulations, and field
applications. For material characterizations, the mechanical and hydraulic properties of
the wicking geotextile, the soil, and the interactions of the soil-geotextile system have
been systematically studied. The laboratory testing techniques were proposed to
determine the saturated and unsaturated hydraulic properties of the wicking geotextiles.
Comparisons of the water retention capabilities of different geosynthetics were compared
to demonstrate the unique hydrological characteristics of the wicking geotextile in terms
of the abilities to hold and transport water under unsaturated conditions. In addition, the
drainage performance of different geosynthetics was also tested in the laboratory.
The test results from the material characterizations were used as inputs for
numerical simulations. A coupled hydro-mechanical model with considerations of
climatic effect was proposed and calibrated. Then, elemental level model was established
to demonstrate the working mechanism of the wicking geotextile. After that, a full-scale
model was established to predict the performance of a road embankment installed with
7
the wicking geotextile. Last but not the least, the simulation results were used as critical
material inputs for the AASHTO (American Association of State Highway and
Transportation Officials) 1993 and MEPDG (Mechanistic-Empirical Pavement Design
Guide) to predict the pavement performance. Detailed discussions were presented
regarding the incorporation of the benefits of the wicking geotextile into the existing
pavement design approaches.
Meanwhile, a field test section was selected at Beaver Slide on the Dalton
Highway, Alaska. Two layers of the wicking geotextiles were installed to mitigate frost
heave and the subsequent thaw weakening issues. The soil temperature and moisture was
closely monitored since 2010 and the field performance of the wicking geotextile was
illustrated. Samples were collected from the field to perform the SEM analyses and the
potential issues that may influence the drainage performance of the wicking geotextile
were assessed. In addition, a bio-wicking system was proposed to further improve the
drainage performance of the wicking geotextile and the drainage performance of the bio-
wicking system was evaluated.
1.4. DISSERTATION OUTLINE
This dissertation contains three sections. Section 1 includes a brief introduction of
the problem, research significance, primary objective, and the scope of work. Paper
section includes nine journal papers that present the material characterizations, numerical
simulations, and field applications of the wicking geotextile. Section 3 summarizes the
work has been accomplished, major findings and conclusions, and also a proposal for
future research.
8
PAPER
I. CHARACTERIZATIONS OF WATER RETENTION CURVES FOR

SEVERAL TYPES OF WOVEN GEOTEXTILES
Chuang Lin1, Yipeng Guo2, Xiong Zhang3, and Javad Galinmoghadam4
1
Graduate Research Assistant, S.M. ASCE, Department of Civil, Architectural, and
Environmental Engineering, Missouri University of Science and Technology, Rolla, MO
65409-0030, Email: [email protected]
2a
Graduate Research Assistant, S.M. ASCE, Department of Civil Engineering, Central
South University, Changsha, Hunan, China, 410075, Email: [email protected]
2b
Visting Scholar, Department of Civil, Architectural, and Environmental Engineering,
Missouri University of Science and Technology, Rolla, MO 65409-0030, Email:
[email protected]
3
(Corresponding Author) Associate Professor, Ph.D., P.E., M. ASCE, Department of
Civil, Architectural, and Environmental Engineering, Missouri University of Science and
Technology, Rolla, MO 65409-0030, Email: [email protected]
4
Graduate Research Assistant, Department of Civil, Architectural, and Environmental
Engineering, Missouri University of Science and Technology, Rolla, MO 65409-0030,
Email: [email protected]
ABSTRACT
Woven geotextiles have been widely used in soil infrastructures for reinforcement
purposes. However, the benefits of the reinforcing effect could be devastated due to the
capillary break effect under unsaturated conditions. Woven geotextiles impeded water
from percolating to the underlying soils and would cause excess water accumulated near
the geotextile installation location. The excess water could not be drained out by
9
conventional subsurface drainage systems. Therefore, no matter how well the road was
designed and constructed, the softened soil would reduce the stiffness of soil-geotextile
composite and the excess water would accelerate the deterioration of pavement systems
under repetitive traffic load.
Recently, a new type of woven geotextile with wicking fibers was developed
which could drain both gravitational and capillary water. The water retention curve
(WRC) was an important unsaturated hydraulic characteristic to depict the water
retention ability of a geotextile under unsaturated conditions. However, the existing
testing techniques were only suitable nonwoven geotextiles whose pore size distributions
were relatively uniform. This paper aimed at reexamining the existing testing techniques
and proposing the proper testing techniques for characterizing the WRCs of woven
geotextiles whose pore size distributions were anisotropic. Then, the WRCs of different
types of woven geotextiles were determined using the proposed methods. Finally, the
working mechanism of the wicking geotextile was demonstrated and the effect of
wicking fiber on the water retention capability of the wicking geotextile was evaluated.
KEY WORDS: woven geotextiles, testing techniques, water retention curve,
wicking fabric, and unsaturated condition
1. INTRODUCTION
Geotextiles have been widely used in soil infrastructures for separation,
stabilization, reinforcement, filtration, and drainage purposes (Bouazza et al. 2006a).
They are especially attractive in solving problems related to roads, embankment, landfill
10
covers and liners, and retaining walls (Iryo and Rowe 2003). Woven geotextiles are
manufactured using traditional weaving methods and are extensively used for
reinforcement purposes in roadway applications (Koerner 2012). Many studies have
proven the effectiveness of woven geotextiles in increasing the initial stiffness, improving
the cyclic fatigue behavior, and decreasing the rutting depth (Giroud and Noiray 1981;
Douglas 1993). However, the benefits of the reinforcing effect for woven geotextiles
could be devastated due to capillary break effect. Capillary break effect was observed
when two geomaterials with different pore sizes were in contact under unsaturated
conditions (Zornberg et al. 2010). Woven geotextiles impeded water from percolating to
the underlying soil and excess water accumulation was reported to occur near the
geotextile installation locations by numerous researchers (Clough and French 1982;
Giroud et al. 2000; McCartney et al. 2005; Stormont and Anderson 1999; Stormont and
Morris 2000). The excess water could not be drained out by conventional subsurface
drainage systems. Conventional drainage systems could only drain gravitational water
under saturated or nearly saturated conditions, but not capillary water under unsaturated
conditions. Therefore, the soil water content would inevitably increase with time. In
addition, the overall stiffness of a roadway relied on the stiffness of the soil-geotextile
composite, rather than a thin layer of geotextile. The reinforcing range of the geotextile
was limited due to such a thin layer (1-3 mm) when compared with the overall thickness
of a base course (300-500 mm). Consequently, no matter how well the road was designed
and constructed, the softened soil would reduce the stiffness of soil-geotextile composite
and the excess water would accelerate pavement deteriorations under repetitive traffic
load.
11
Recently, a new type of woven geotextile with wicking fibers (hereafter named as
“wicking geotextile”) was developed which was able to overcome the limitation of
woven geotextiles, which is the lack of lateral drainage ability. The wicking geotextile
was a dual functional woven geotextile for reinforcement and drainage purposes. The
unique characteristic of the wicking geotextile was that it could work as a capillary
barrier in the cross-plane direction meanwhile had lateral drainage capability in the in-
plane direction under unsaturated conditions. A detailed description of this type of woven
geotextile was presented in the Testing Material section. A series of laboratory tests have
proven its efficiency in reducing soil water content (Zhang and Belmont 2009). In
addition, the wicking geotextile was also successfully used in mitigating frost heave and
thaw weakening issues in Alaska (Currey 2016). It was reported that the saving for the
initial construction was $2.5 million and another $3.5 million savings were for the
maintenance cost.
Because the wicking geotextile was placed above the groundwater and often
worked under unsaturated conditions, it was necessary to characterize the hydraulic
properties of the wicking geotextile under unsaturated conditions. Key to assessing the
water flow through unsaturated geotextiles included two important hydraulic properties:
the water retention curve (WRC) and the hydraulic conductivity function (or K-function)
(Zornberg et al. 2010). The WRC depicted the ability of the geomaterial to hold water
under unsaturated conditions while the K-function represented its ability to transport
water under unsaturated conditions. The unsaturated K-function also relied on the
continuity of the flow path within the voids under unsaturated conditions. In other words,
the K-function was also controlled by the WRC and could be derived based on the
12
saturated hydraulic conductivity and the WRC (Mualem 1986; van Genuchten 1980;
Fredlund et al. 1994; Fredlund and Rahardjo 1993; Klute 1965).
Unfortunately, the existing laboratory testing techniques were designed for soils
or nonwoven geotextile that shared similar pore size distributions. Stormont and Morris
(2000) indicated that the nonwoven geotextiles exhibit properties and behaviors similar to
coarse, uniform soils. Therefore, the testing techniques to determine the WRC of soils
(ASTM D6836-16) have been adapted to obtain the WRCs of nonwoven geotextiles
experimentally (Knight and Kotha 2001; Nahlawi et al 2007; and Bathurst et al. 2007).
However, due to the existence of different weaving patterns for woven geotextiles, the
pore sizes were expected to be anisotropic and the hydraulic conductivities of a woven
geotextile were expected to the different in different directions. It required further
examination to determine if the existing testing techniques were suitable for woven
geotextiles. Different testing techniques might result in different test results for the same
type of geotextile. It was of great importance to evaluate the suitability of the existing
testing techniques to characterize the WRCs of woven geotextiles. To the authors’ best
knowledge, very limited researchers have studied the anisotropic hydraulic characteristics
of woven geotextiles, let alone the wicking geotextile. In addition, due to the existence of
the wicking fiber yarns in the wicking geotextile, the ability of the wicking geotextile to
hold and transport water under unsaturated conditions was expected to be dependent on
the number of wicking fibers within the wicking geotextile. To better understand the
influence of the wicking fibers on the hydrological characteristics of the wicking
geotextile, it was necessary to compare the WRC of the wicking those of other non-
wicking geotextiles. Finally, it was also important to establish the relationship between
13
the number of wicking fiber yarns and the regression parameters for the WRC of the
wicking geotextile. A proper estimation of the water retention ability of the wicking
geotextile in the field could save the cost and time performing laboratory tests.
2. RESEARCH OBJECTIVES AND METHODOLOGIES
The objectives of this paper are: 1) to reexamine the suitability of the existing
testing techniques for the determination of WRCs for woven geotextiles; 2) to propose
the proper testing techniques and to determine the WRCs of several types of woven
geotextiles using the proposed techniques; 3) to compare the ability of different woven
geotextiles to hold water under unsaturated conditions and to explain the working
mechanism of the wicking geotextile; 4) and to quantify the influence of the number of
wicking fiber yarns on the water retention ability of the wicking geotextile.
To achieve the objectives of this paper, the testing materials were first introduced.
In total, five types of materials were used in the laboratory tests, including two types of
conventional non-wicking geotextiles (Mirafi® HP570 as T1 and Mirafi®RS580i as T2),
the wicking geotextile (Mirafi® H2Ri as T3), a deep grooved 4DGTM wicking fiber yarn
as T4, and an artificially modified H2Ri as T5). T1 was a conventional woven geotextile
for soil reinforcement purposes and was selected to be a representative of non-wicking
woven geotextiles. T2 was also a conventional non-wicking woven geotextile but it
shared the same weaving pattern with T3, which was the wicking geotextile. T3 was a
dual functional woven geotextile for both reinforcement and drainage purposes. The
wicking fibers gave the geotextile the maximum potential for capillary action and water
14
transporting ability under unsaturated conditions. Comparisons between T2 and T3 would
demonstrate the unique hydrological characteristics of the wicking geotextile against the
non-wicking geotextile. T4 was only the wicking fiber yarn that was used in T3 for
drainage purposes. In this paper, T4 was selected to quantify the water storage ability of
the fibers and the influence of the wicking fibers on the water storage ability of the
wicking geotextile. As for T5, an additional wicking fiber yarn was artificially knitted to
T3 to increase the water storage ability of the wicking geotextile. T5 was used to evaluate
the influence of the number of wicking fiber yarns on the water storage ability of the
woven geotextiles.
A series of laboratory tests, including capillary rise, pressure plate, and salt
concentration tests, were performed to reexamine the suitability of the conventional
testing techniques for the determination of the WRCs of woven geotextiles. Geotextile
samples were tested in different directions to account for the effect of anisotropy on the
water retention ability and the test results were compared. Results from different testing
techniques were discussed and the proper testing technique was proposed. After that, the
proposed testing techniques were used to determine the WRCs of T1-T4 and the water
retention abilities of different types of woven geotextiles were discussed in detail.
Finally, the effect of the number of wicking fiber yarns on the water retention ability of
the wicking geotextile was evaluated and the methodology was proposed to predict the
water retention ability of a particular woven geotextile based upon the number of wicking
fiber yarns.
15
3. TESTING MATERIALS
The testing materials included four types of woven geotextiles and one type of
wicking fiber yarn. The specifications of the three woven geotextiles (T1-T3) are listed in
Table 1. The Apparent Opening Size (AOS) was determined by dry sieving uniform-
sized glass beads of a known standard sieve size through the geotextile until the weight of
beads passing through the geotextile was 5% or less (ASTM D4751 2016). It reflected
the approximate largest opening dimension of the geotextile (in the cross-plane
direction). Comparisons among T1-T3 indicated that T1 had the smallest opening size
among all three geotextiles. The AOS values for T2 and T3 were the same (0.425 mm)
because of the same weaving pattern. The permittivity of a geotextile was defined as the
volumetric flow rate of water per unit cross-section area per unit head under laminar flow
conditions in the cross-plane direction through a geotextile (ASTM D4491 2017). Since
T2 was designed for filtration, it was reasonable for T2 to have the largest value of
permittivity among all three geotextiles. In addition, even though T1 had a larger pore
size (AOS value of 0.60 mm), the permittivity and flow rate values for T1 were
comparable with T3 (AOS value of 0.425 mm).
Figure 1 shows the weaving patterns for the woven geotextiles and the
microstructures of the wicking fiber. Figure 1a shows the actual and schematic plots of
the weaving pattern for T1. As shown in the top two images, the woven geotextile was
composed of weft yarns with multifilament fibers (with a large cross-sectional area) in
the cross-machine direction (CD) and warp yarns with monofilament fibers in the
machine direction (MD). The perpendicular direction (PD) was perpendicular to the CD-
16
MD plane and was also known as the cross-plane direction. The bottom two images in
Figure 1a show the actual and schematic plot of the cross-section for T1. The warp yarns
were threaded into the weaving loom and the reed shed the warp yarn up, allowing a
shuttle to insert the weft yarn. The reeds then shed back upward and the process
continued as a cycle. The pores in the CD were relatively more continuous and connected
than those in the other two directions (MD and PD). Note that the weaving pattern caused
different pore-size distributions that would have a significant influence on the selection of
the proper testing techniques to determine the WRC of a woven geotextile. Comparisons
of the test results using different testing techniques were presented in the test results
section.
Figure 1b shows the weaving pattern for T2, which was composed of a double
layer geotextile formed from a single weave. The top two images in Figure 1b show that
the black multifilament in the CD and the black monofilament in the MD were woven
together as in conventional woven geotextiles. However, the monofilament (in orange
color) was made of high-tenacity polypropylene, which could provide superior
reinforcement strength. The bottom two images in Figure 1b show the cross-section of
T2. A weft yarn set was selected and marked in the image to better demonstrate the
weaving pattern. Each set was comprised of a first weft yarn (multifilament in black
color), a second weft yarn (monofilament in orange color), and a warp yarn
(monofilament in black color) woven in the CD. In addition, a warp yarn interwove the
first weft yarns and the second weft yarns. The two first weft yarns were stacked together
and three-second weft yarns were distributed at the top, middle, and bottom of the woven
geotextile.
17
Figure 1c shows the weaving pattern for T3, which was the wicking geotextile.
The weaving pattern for T3 was the same as for T2. However, the second weft yarn
(monofilament in orange color) in T2 was replaced by the multifilament wicking fiber
yarns in T3. The bottom two images in Figure 1c show the cross-section of the wicking
geotextile. Similar to T2, the first weft yarn and wrap yarn were all made of high-tenacity
filaments for reinforcement purposes. However, different from T2, the second weft yarn
in T3 was made of multifilament deep grooved (4DG) nylon wicking fibers. Detailed
demonstration regarding the microstructure of the wicking fiber can be found in Figure
1d and will be discussed in the following paragraph. By installing the wicking fibers in
the CD, it was expected that the ability of the geotextile to hold and transport water in the
CD would be higher than that in the other two directions.
To further demonstrate the unique microstructure of the wicking fibers, Figure 1d
shows the macro- and micro-structure of the wicking fiber. The top two images in Figure
1d show the SEM (Scanning Electron Microscopy) image and the schematic plot of a
single wicking fiber. The nylon wicking fiber was hydrophilic and hygroscopic and was
specially designed with multichannel shape cross-section. The multichannel shape cross-
section could maximize the surface tension and allow higher capillary force. In addition,
the SEM image of the wicking fiber indicated that the average diameter of each filament
was 30-50 microns, and the average opening was 5-12 microns. The wicking fiber was
able to maintain saturation under unsaturated conditions and allowed water to be
transported along the long axis of the filament. The bottom two images in Figure 1d show
the configurations and schematic plots of a single wicking fabric yarn. There were 144
filaments within a single wicking fiber yarn and the filaments were twisted twice per
18
linear inch. The twisting design of the wicking fabric yarn ensured a stronger wettability
for the polymer. The voids were structured in the shape of multiple continuous channels
down the long axis of the filament. The greater efficiency in filling the voids, the more
effective wicking would be along the length of the monofilament.
In addition, to further investigate and quantify the effect of wicking fiber yarns on
the water retention ability of the wicking geotextile, an additional wicking fiber yarn was
artificially knitted into a set of the original wicking geotextile. This type of artificially
modified wicking geotextile was denoted as T5. Figure 1e shows the weaving pattern of
T5. An additional wicking fiber yarn was added to each weft yarn set. Besides the
additional wicking fiber yarns, T3 and T5 shared the same weaving pattern.
4. TESTING PROCEDURES
This paper focused on the determination of the WRCs of woven geotextiles. The
principles of water flow through unsaturated geomaterials were more complex than the
water flow through saturated geomaterials, partially due to the non-constant capability of
the geomaterials to hold water (Zornberg et al. 2010). The WRC of a geomaterial was
used to describe its ability to hold water under unsaturated conditions and was sensitive
to pore size distribution, chemical components of the material, density, and pore structure
(Bouazza et al. 2006b). The WRC showed a significant difference in the drying and
wetting paths (also referred to as hysteresis). Geotextile wetting curves obtained via
different methods (either hanging column test or capillary rise test) showed varying
amount of hysteresis (Krisdani et al. 2006). To reduce the uncertainty and the complexity
19
in determining the WRC of woven geotextiles, only the WRCs in the drying path was
determined. During the sample preparation process, the geotextile samples were
submerged in water before the test. Meanwhile, an 80.0 kPa vacuum pressure was applied
for 24 hours to accelerate the extraction of air bubbles entrapped within the samples.
Several testing techniques have been developed to determine the WRCs of soils
and have recently been modified for the determination of the WRCs of nonwoven
geotextiles. In this research, three conventional testing techniques were reexamined,
including the capillary rise test (suction ≤ 10 kPa), the pressure plate test (10 kPa <
suction ≤ 1500 kPa), and the salt concentration test (suction > 1500 kPa). Detailed
information regarding the test apparatus and test procedures will be discussed as follows.
4.1. CAPILLARY RISE TEST
Capillary rise test was essentially a physical testing technique that expelled the
excess water of an initially saturated geotextile sample by imposing the suction boundary.
Figure 2 shows the schematic plot and laboratory setup of the testing apparatus. Because
the thickness of the geotextile was very small compared with the other two dimensions,
the samples were only tested in the MD and CD. Figure 2b shows the test setup when the
drainage direction was in parallel to the CD (hereafter denoted as “CD test”). In
comparison, if the drainage direction was in parallel to the MD, the test was denoted as
the “MD test”.
The geotextile samples were cut into 0.1 m-wide by 1.0 m-long strips and
saturated before testing. The entire sample was covered with plastic wrap to minimize
water loss caused by the evaporation process. One end of the sample was submerged into
20
a water reservoir and the water surface was considered as the datum plane. Meanwhile,
the other end of the sample was suspended vertically above the reservoir and a ruler was
also hanging in parallel beside the sample so that the elevation of the sample could be
determined. Under the influence of gravity, the water within the saturated sample would
flow downward along the CD. The water within the testing specimen would flow
downward under the influence of gravity. The pore water pressure below the datum plane
was positive and above the datum plane was negative. Under steady-state, the pore water
pressure shall be linearly distributed in the vertical direction, as expressed in Equation 1.
 = w gh (1)
where,  = matric suction, kPa;  w = density of water, kg/m3; g = gravitational
acceleration, m/s2; and h = specimen elevation above the water table, m.
Krisdani et al. (2006) performed the capillary rise test using the same testing
techniques with different testing durations (1, 2, 3, and 5 days) and the test results were
consistent. This fact indicated that testing durations were not a major factor that would
influence the test results. In this study, the test was still conducted for 7 days to ensure
that the testing system reached steady-state at the end of the test. After the test, the
geotextile sample was cut into small pieces and the corresponding elevations at the
centroid of each piece were recorded. Then, the strips were put into a desiccator to be air
dried and the corresponding water contents were determined. According to Equation 1,
the suction values for each piece of the geotextile samples could be calculated and the
WRC of the geotextile at low suction range could be determined.
For each type of geotextile (T1-T3), the capillary rise test was performed twice,
including the CD test and the MD test, respectively. However, the wicking fiber yarns
21
(T4) were twisted together and put into a straw to prevent water loss from the evaporation
process. Then, the wicking fabric yarns were tested using the same technique described
above. Because the wicking fiber yarns did not have any weaving pattern, the capillary
rise test for T4 was only tested along the direction of the long axis.
4.2. PRESSURE PLATE TEST
Figure 3 shows the test apparatus for the pressure plate test. Pressure plate test
was used to determine the WRCs in the medium suction range (10-1500 kPa). The test
samples were placed on top of the ceramic plate. The ceramic disc only conducts water as
long as the air pressure applied to the specimen was lower than the air-entry value (AEV)
of the ceramic disc. The air pressure forced the pore water to flow through the water
conductive ceramic disc under the pressure difference within and outside the testing
apparatus. The excess water was collected by a 50 ml beaker and the mass of beaker was
regularly measured till a constant value was obtained, indicating that an equilibrium
condition was achieved. At the equilibrium condition, the air pressure corresponds to the
matric suction value and the water content of the specimen was determined after the test.
This approach was repeated successively for higher suction levels and the WRCs for the
medium suction range (< 1500 kPa) could be determined. Knight and Kotha (2001)
modified the technique used for soils to measure the WRC for nonwoven geotextiles and
Nahlawi et al. (2007) and Bathurst et al. (2009) further improved the testing technique
using a modified capillary pressure cell. However, all the described techniques required
significant modifications of the testing device and can only be used to determine the
WRCs of nonwoven geotextiles. Due to the anisotropic pore size distributions within the
22
woven geotextile, the conventional testing techniques for soils or nonwoven geotextiles
may not be suitable for woven geotextile, as discussed in the following paragraph.
Figure 3a shows the test setup for the PD test. The saturated samples were cut into
30 mm by 80 mm pieces and were placed on top of the saturated ceramic plate with the
water draining in the PD direction. The water within the geotextile samples was expected
to flow in the PD (or cross-plane) direction from the geotextile to the ceramic disc under
the imposed air pressure. A good contact between the geotextile samples and the ceramic
plate was necessary so that the channels for water flow were connected and continuous.
However, the contact between the geotextile sample and the ceramic plate was point to
point and the continuity of the flow channels relied on the testing technique (using CD,
MD, or PD tests). For example, in the PD test, the long-axis of the wicking fiber
perpendicular to the direction of water and the deep grooves within a fiber were relatively
isolated from each other. Even if the water flow channel at the bottom of the wicking
fiber (refer to Figure 3a) was in direct contact with the porous ceramic plate and the water
within the bottom channel could flow out of the wicking fiber, most of the water was
retained within the deep grooves that were not in contact with the ceramic plate and could
not be drained out no matter how long the PD test lasted. Therefore, the PD test might
cause excess water retained in the wicking geotextile and was not the appropriate testing
technique to determine the WRC for the wicking geotextile. Azevedo and Zornberg
(2013) reported that the woven geotextile was not in good contact with the porous
stone/ceramic disc due to the uneven weaving texture and the wicking geotextile barely
absorbed water upon wetting.

23
For this paper, the authors would like to propose a simplified testing technique to
determine the WRC of the wicking geotextile without modifications of the testing
apparatus. Figure 3b shows the modified testing technique. To overcome the
disadvantages of the testing technique in Figure 3a, the saturated geotextile samples were
sandwiched by two pieces of acrylic boards and clamps were used to erect and fasten the
samples in the vertical direction. Then, a thin layer of soil slurry (kaolinite: water = 1:2
by weight) was coated at the bottom of the sample to ensure a good contact area between
the sample and the ceramic disc. Note that the soil slurry should not be too thick to
impede water flow nor too thin to flow away. The soil slurry layer ensured a continuous
liquid water flow path and the excess pore water could be freely forced out of the sample
when applying the regulated air pressure. The excess water was expelled out of the
system to a 50 ml beaker and the mass of the water was continuously measured until a
constant value was obtained. At the equilibrium condition, the lower part of the specimen
was contaminated and was cut off. The upper part of the specimen was used to determine
the water content.
To account for the anisotropic characteristics of the wicking geotextile in the in-
plane direction, both the MD and CD tests were performed using the proposed testing
technique, as shown in Figure 3b. Similar to the PD test, the MD test was not a proper
testing technique for the wicking geotextile. For the MD test, even though the contact
problem can be resolved by coating the soil slurry, the long-axis of the wicking fiber was
still perpendicular to the direction of water flow and the water was trapped within the
isolated deep grooves under unsaturated conditions. In contrast, during the CD test, the
direction of water flow was in parallel with the long-axis of the wicking fiber and the soil
24
slurry ensured a good contact between the geotextile sample and the ceramic plate. In this
way, the excess water was able to continuously flow from the deep grooves to the
ceramic plate via the soil slurry, and eventually expelled out of the system under the
applied air pressure. Given the same applied suction level, the water contents obtained
from the MD test were expected to be higher than those from the CD test. Detailed
discussion will be presented in the test results section.
As for other non-wicking woven geotextiles, T1 and T2, the geotextile samples
were tested using the same testing techniques (PD, MD, and CD tests) and the test results
will be presented in the discussion section. In addition, for T4, which was the wicking
fibers, the wicking fiber yarns were directly sandwiched by two transparent acrylic
boards. Then, the bottom of the wicking fiber yarns was submerged into the soil slurry.
T4 was only tested in the long-axis direction.
4.3. SALT CONCENTRATION TEST
Salt concentration test was a thermodynamic based technique to determine the
WRC at high suction level (> 1500 kPa). Different from the capillary rise and pressure
plate tests, the water flow directions did not influence the final water contents since the
evaporation process did not have a directional preference. Therefore, for the salt
concentration test, the anisotropic hydraulic properties of woven geotextiles could not be
reflected. Figure 4 shows the test apparatus for the salt concentration test. A controlled
relative humidity environment was used to establish a constant total suction. The pore
water within the sample was allowed to evaporate and reach in equilibrium with the
surrounding vapor pressure. The relative humidity was controlled by the salt
25
concentration in the MgCl2 solute. The relationships between the concentration and the
corresponding suction are presented in Table 2, according to the Lord Kelvin equation
(Fredlund and Rahardjo 1993). In total six glass jars were filled with solute with different
salt concentrations and the corresponding suction values ranged from 1303 kPa to 14554
kPa. The saturated samples were placed in tinfoil sample holder with punched holes at
the bottom to shorten the time required to reach equilibrium. After the samples were put
in the jar, electrical tape was used to seal the glass jar. It took about 7 days to reach
equilibrium and the water contents of the samples were determined after the test. Given
suction values and the corresponding water contents, the WRC at high suction range
could be determined.
5. TEST RESULTS AND DISCUSSIONS
5.1. CAPILLARY RISE TEST
Figure 5 shows the capillary rise test results for different types of geotextiles. In
general two drainage mechanisms will influence the water content distribution at the end
of the test: including gravitational drainage and capillary action. Firstly, the degree of
saturation was 100% at the beginning of the test and water trapped in the test sample was
firstly drained downward under the influence of gravity. Secondly, as the degree of
saturation started to decrease, capillary action dominated the water flow and water would
flow upward along the continuous pores along the geotextiles. Eventually, the water flow
ceased and the steady-state was obtained when downward gravitational force equaled to
26
the upward capillary force. Both the MD and CD testing techniques were used in the
capillary test for T1 to T3 while T4 was only tested in the long-axis direction.
Figure 5a shows the capillary rise test results for T1. In the x-axis, the higher the
suction value, the drier the geotextile sample. In general, the water content decreased
with increasing suction. At suction of 0.1 kPa where the sample was nearly saturated, the
water contents determined by both MD and CD tests were consistent and the saturation
water content was 48.0%. As the suction increased, the water contents significantly
decreased with increasing suction. For example, the water content decreased to 0.0% as
the suction value reached 3.0 kPa. This fact indicated that T1 could not hold water under
unsaturated conditions. In addition, it is important to point out that the test results
obtained from the CD test gave higher water content values that those from the MD test,
even though the testing material was the same. For a typical geomaterial, its water storage
ability (or the ability to hold water) under the same suction level shall be the same
regardless of the difference in testing techniques. Then the question becomes which
testing technique was proper for woven geotextiles? The answer lies in the weaving
pattern of woven geotextiles and the continuity of the pores within the woven geotextile.
For example, Figure 1a shows the weaving pattern and the cross section for T1. The
pores were continuous and connected in the CD while relatively isolated and occluded in
the MD. The schematic plot of the cross section also indicated that there were more
connected deep grooves in the CD. In comparison, due to the isolation caused by the
multifilament fibers with a large cross-sectional area, it would be very difficult for water
to flow along the MD. In addition, when the test samples were hanging vertically (as
shown in Figure 2), it was easier for the geotextile samples to absorb water due to
27
capillary action. This fact explained the test results in Figure 5a that the water contents
were higher in the CD test within the suction range of 0.5-3.0 kPa. Therefore, the authors
recommended that the capillary test should be performed in such a way that the direction
of the continuous and connected pores was parallel to the direction of water flow. As the
suction exceeded 3.0 kPa, the water contents within the geotextile samples were close to
0.0%, indicating that very limited water was retained in the geotextile samples. Similar
test results were observed in Figure 5b which shows the capillary rise test results for T2.
As indicated in Figure 1b, the pores within T2 were continuous and connected in the CD
than in the MD, resulting in relatively high water content values in the CD test. The
saturated water content for T2 was 52% and the water content significantly decreased
with increasing suction. As suction reached 3.0 kPa, the water content was close to 0.0%.
In comparison, Figure 5c shows the capillary rise test results for T3, which was
the wicking geotextile (refer to Figure 1c). It is worthwhile to point out that the weaving
patterns for T2 and T3 were the same and the difference in the measured WRCs reflected
the influence of the wicking fibers on the water storage ability of the woven geotextile.
Firstly, the test results for T3 were consistent with that for T2 in the MD test. The
saturated water content was 54.1% and the water content significantly decreased with
increasing suction. As the suction was higher than 3.0 kPa, the water content was 0.0% in
the MD test. This results indicated that the existence of the wicking fibers had limited
influence of the water storage ability in the MD. In comparison, the water contents for T3
were much higher than those for T2 in the CD test. For example, the water content for T3
was 9.4% when the suction value was 6.4 kPa. In contrast, the water content for T2 was
only 0.2% at the same suction level. The significant difference in water content was the
28
cause by the existence of the wicking fiber in the CD. Figure 1d shows the schematic plot
of the wicking fibers. The multichannel cross-section had a high shape factor and great
numbers of channel per fiber, which gave the wicking fibers great potential for
maximizing capillary action and water storage ability under unsaturated conditions. The
test results indicated that T3 had much stronger ability to hold water under unsaturated
conditions when compared with T1 and T2, which are conventional non-wicking
geotextiles.
To further demonstrate the water storage ability of the wicking fibers, Figure 5d
compares the CD test results of T2 (non-wicking geotextile), T3 (wicking geotextile), and
T4 (wicking fiber yarn). Firstly, the saturated water content for T4 was 81.0%, which was
the highest among the three testing materials. This was caused by the hydrophilic and
hydroscopic characteristics of the wicking fiber. The wicking fibers could maximize the
capillary force and absorb a large amount of water under unsaturated conditions.
Secondly, as the degree of saturation decreased, the advantage of wicking fiber to hold
water became more obvious. For example, when the suction value was lower than 1.0
kPa, both T2 and T3 were nearly saturated and the corresponding water contents varied
from 50.0% to 54.0%. As the degree of saturation continued to decrease and the
corresponding suction exceeded 1.0 kPa, the water content for T3 became higher than
that for T2, given the same suction value. Thirdly, comparisons of the test results clearly
demonstrated that the wicking fibers improved the ability of the woven geotextile to hold
water under unsaturated conditions. For example, the water content for T2 was 1.3%
when the suction was 3.0 kPa. While the water contents for T3 and T4 were much higher
than T2 and the corresponding water contents were 17.8% and 55.8%, respectively.
29
In summary, to determine the WRC of the wicking (a woven) geotextile, the
direction of the wicking fiber (the most continuous and connected pores) shall be in
parallel with the direction of water flow. More importantly, test results indicated that the
wicking fibers indeed increased the ability of a woven geotextile to hold water under
unsaturated conditions.
5.2. PRESSURE PLATE TEST
Figure 6 shows the results of the pressure plate test by using different testing
techniques (MD, CD, and PD tests). Figure 6a shows the test results for T1. In general,
the water content did not vary significantly with increasing suction values. Regardless of
the testing techniques, the water contents obtained in the MD, CD, and PD tests were all
lower than the saturated values, but to different extents. For example, the average water
content was 30.0% in the MD test while this value was 7.5% and 0.0% in the PD and CD
tests, respectively. This fact indicated that the degree of continuity of the pores in
different directions was different. The water contents obtained from the CD and MD tests
were consistent in the pressure plate and the capillary rise tests. As discussed in the
capillary rise test results, the pores within T1 was continuous and connected in the CD,
but relatively occluded and isolated in the MD. Water could easily flow along the
continuous pores in the CD to the ceramic disc, and eventually being expelled out of the
system under the imposed air pressure. In addition, Figure 5a proved that T1 could not
hold water under unsaturated conditions because the water content reduced to 0.0% when
the suction exceeded 3.0 kPa. This is the reason why the water content remained to be
0.0% in the CD test during the pressure plate test. Moreover, the results of the pressure
30
plate test also reflected the relative continuity of the pores in each direction. The more
continuous of the pores, the lower the water content would be expected. According to this
criteria, the continuity of the pores for T1, from high to low, should be CD, PD, and MD,
respectively. Figure 6b shows the results of the pressure plate test for T2. The average
water content in the MD test was still the highest (22.5%) among all three directions,
indicating that the relative continuity of the pores was still the lowest in the MD. In
addition, the average water contents in the CD test remained to be close to 0.0%, which
was consistent with the results obtained from the capillary rise test. Moreover, the
average water content in the MD test for T2 (22.5%) was lower than that for T1 (30.0%),
indicating that the continuity of the pores in T2 was higher than those in T1.
Figure 6c shows the result of the pressure plate test for T3, which is the wicking
geotextile. It is important to emphasize again that T2 and T3 shared the same weaving
pattern but T3 had the wicking fiber yarns in the CD. Firstly, the average water content in
the MD test was 29.0%, which was 6.5% higher than that for T2. The wicking fibers in
T3 could hold a measurable amount of water under unsaturated conditions. During the
MD test, the wicking fiber yarns were perpendicular to the drainage direction (PD) and
the excess water within the wicking fibers was very difficult to be drained out, resulting
in the highest water contents among all three testing techniques. Compared with the
limited contact area between the geotextile sample and the ceramic plate in the MD test,
the contact area for the PD test was much larger. When the saturated sample was placed
on the ceramic plate, part of the water could be drained out under such a high degree of
saturation. However, as the degree of saturation continued to decrease and the water
within the large pores among the weaving yarns was replaced with air, the excess water
31
could only be retained in the small pores in the deep grooves which were not in direct
contact with the ceramic disc. This is the reason why the water contents obtained from
the PD test were lower than those obtained from the MD test. In contrast, the water
contents obtained from the CD test gradually decreased with increasing suctions,
indicating that the water flow path between the geotextile sample and the ceramic plate
was continuous and the excess water held in the wicking fibers could be gradually
drained out. It is important to point out that the ability of T3 to hold water under
unsaturated conditions was a unique feature compared with other non-wicking
geotextiles. This feature ensured the wicking geotextile to be able to hold and transport
water under unsaturated conditions.
Figure 6d showed the test results for T4, which was the wicking fiber yarn.
Because the wicking fiber yarns did not have a weaving pattern and water would flow
along the long-axis, the pressure plate test was only performed in the wicking direction.
In general, the water contents gradually decreased with increasing suction values. The
water content was 10.2% when suction was 700 kPa. In retrospect of the WRC of non-
wicking geotextile (Figure 5a-b), the water content was 0.0% when suction was greater
than 3.0 kPa. This significant difference in water contents under unsaturated conditions
reflected a much stronger ability of the wicking fibers to hold and transport water under
unsaturated conditions.
5.3. SALT CONCENTRATION TEST
Since the salt concentration test was a thermodynamic testing technique, there
was no directional preference for water to be vaporized. Note that the thermodynamic
32
techniques have not been used for geosynthetics because the water content of geotextiles
at high suction values is so low that its measurements have not been needed for practical
applications (Zornberg et al. 2010). However, one of the study purpose of this research
was to determine the WRC of the wicking geotextile, it was required to cover a much
wider suction range (from completely saturated to complete dry). In addition, due to the
existence of the wicking fibers, it was also necessary to evaluate the effect of the wicking
fiber on the water storage ability of the wicking geotextile at high suction level.
Therefore, the salt concentration test results were equally important in understanding the
working mechanism of the wicking geotextile.
Figure 7 shows the results of the salt concentration tests. The test results for T1
and T2 also indicated that the water contents for non-wicking geotextiles remained to be
0.0%, which proved Zornberg’s (Zornberg et al. 2010) assertion it was not necessary to
determine the WRC of non-wicking geotextile at high suction level. In fact, the water
contents for T1 and T2 already reduced to 0.0% at a suction value of 3.0 kPa, according
to the results of the capillary rise test (Figure 5a-b). In contrast, there was still a
measurable amount of water retained in the deep grooves for T3 and T5 due to the
existence of wicking fibers. For example, the water contents for T2 gradually decreased
from 1.2% at 1303 kPa to 0.4% at 39,776 kPa. As for T4, which was the wicking fabric
yarn, the water contents gradually decreased from 9.5% at 1303 kPa to 5.4% at 39776
kPa. The test results indicated that the WRC of the wicking geotextile at the high suction
range was not negligible and was important in better understanding its hydraulic behavior
under unsaturated conditions.

33
6. COMPARISONS OF TESTING TECHNIQUES
An accurate mathematical representation of the WRCs of woven geotextiles relies
on reliable and appropriate laboratory testing techniques. To select the most appropriate
testing technique, the results from the capillary rise, pressure plate, and salt concentration
tests were presented together. Since the non-wicking and wicking geotextiles have shown
a significant different water storage abilities under unsaturated conditions, the test results
for the two non-wicking geotextiles (T1 and T2) were compared together while the test
results contained the wicking fibers (T3 and T4) were discussed together. It is important
to point out that the ability of a geomaterial to hold water under unsaturated conditions is
a unique hydraulic characteristic and can be depicted via the WRC. Despite the testing
techniques, the WRC shall be a smooth, continuous, and unique curve that covers the
entire suction range, from 0 kPa (fully saturated) to 105 kPa (fully dried).
Figure 8a shows the test results that covered the full suction range (0.1 kPa to 105
kPa) for T1. The test results obtained with different testing techniques were plotted in the
same Figure. For the capillary rise test, the test results obtained from the CD test showed
higher water contents compared with those obtained from the MD test. As discussed in
the capillary rise test results, this deviation was caused by the difference in pore-size
distribution and the continuity of the pores in different directions. Since the CD was
parallel to the drainage direction and water could freely to flow in or flow out of the test
sample, the test results obtained from the CD were considered more representative. In
addition, the test results from the pressure plate test also validated the authors’ argument.
At low suction range (< 10 kPa), the water contents remained to be 0.0% as the suction
34
value was greater than 3.0 kPa in both the CD and MD tests. However, at the medium
suction range (10kPa-1500 kPa), only the test results obtained from the CD test were
consistent with the capillary rise test results. As for the test results obtained from the MD
and PD tests, depending on the continuity of the pores in different directions, only part of
the water could be drained out and the test results could not be considered representative.
As for the WRC at high suction range (> 1500 kPa), there was no directional preference
for the thermodynamic testing technique so that the test results were consistent. In
summary, the test results obtained from the CD test were more representative and the CD
test was recommended to be the proper testing technique to determine the WRCs of for
non-wicking geotextiles. A similar WRC for T2 could be obtained, as shown in Figure
8b. Since both T1 and T2 were conventional non-wicking geotextiles, the WRCs for T1
and T2 share the same characteristic, with significant reduction in water content within a
narrow suction range and remaining at 0.0% water content at higher suction level. For
T2, the test results obtained from the CD test were also considered to be representative.
The results of pressure plate test in the CD were consistent with those obtained from the
capillary rise and salt concentration tests and remained to be 0.0%.
In comparison, Figure 9 shows the WRCs for T3 (wicking geotextile) and T4
(wicking fiber yarn) for the full suction range. Note that due to the same weaving pattern
for T2 and T3, the continuity and the connectivity of the pores were expected to be the
same. Figure 9a shows that only the test results obtained from the CD test were
continuous and representative. At the low suction range (0-10 kPa), the water content for
T3 decreased significantly with increasing suction values, but not as fast as in T1 and T2.
For example, the saturated water contents for T1, T2, and T3 were 48.0%, 52.0%, and
35
54.1%, respectively. However, at the suction value of 3.0 kPa, the water content
decreased to 0.0% for T1 and T2 but was still 17.5% for T3. The additional water was
retained within the deep grooves of the wicking fibers and the water storage ability for T3
was higher than T1 and T2. As suction continued increasing to the medium range (10-
1500 kPa), the water contents for T3 continued decreasing but did not reduced to 0.0%.
This fact indicated that the wicking fibers still had the ability to hold a measurable
amount of water under such medium suction level. When the suction increased to the
high level (>1500 kPa), the wicking fibers became further desaturated and limited
amount of water could be retained in the deep grooves.
Figure 9b demonstrates the WRC of T4 for the full suction range. Because T4
consisted of pure wicking fiber yarns, the saturated water content was the highest among
all the tested materials and was 81.0%. The results obtained from the capillary rise,
pressure plate, and salt concentration tests were consistent. It is worthwhile to point out
that the water storage ability of T4 was very high even under high suction range. For
example, the water content was 5.4% at the suction of 4×104 kPa. Such a high water
storage ability was beneficial for water to flow under unsaturated conditions. A higher
water content indicated a higher degree of saturation and a higher hydraulic conductivity.
This unique hydraulic characteristic allows the wicking fibers to hold and transport water

36
7. WATER RETENTION CURVES (WRCs)
Based on the discussions in the previous section, the results obtained in the CD
test (capillary rise and pressure plate tests) combined with the results from the salt
concentration test have been selected to represent the WRCs of each geotextile. Since the
three-parameter equations provided great flexibility for the best-fitting analysis, the
authors used Fredlund and Xing’s (1994) equations to perform the regression analysis:
ws
w( ) = C ( ) (2)
{ln[e + ( / a)n ]}m
ln(1 + /  r )
C ( ) = 1 − (3)
ln[1 + 106 /  r ]
where,  = suction, kPa;  r = suction corresponding to residual water content, kPa;
w( ) = corresponding water content, %; ws = saturation water content, %; C ( ) =
correction factor; a = fitting parameter related to AEV; n = fitting parameter related to
the rate of desaturation; and m = fitting parameter related to the curvature near residual
conditions.
The WRCs for T1-T4 were presented in Figure 10. Fredlund et al. (2000)
proposed two types of models, namely unimodal and bimodal, to depict the WRC of soil
based upon grain-size distribution data. For conventional non-wicking woven geotextiles
such as T1 and T2, the pore sizes within the geotextiles were relatively uniform and the
WRCs of non-wicking woven geotextiles were comparable to those of uniform sand with
narrow pore-size distribution. As shown in Figure 10, the WRCs of T1 and T2 remained
saturated up to the air entry value (AEV), where the largest pores started draining
37
(Brooks and Corey 1964). When passing the AEV, the slopes were steep and the water
content rapidly decreased with increasing suctions. Once the second bending point (given
by the residual suction) was reached, large increments in suction would have limited
effect on water content variations. Due to the relatively uniform pore size distributions in
T1 and T2, a unimodal was sufficient to define the WRCs of non-wicking woven
geotextiles.
As for the WRC of T3 which is the wicking geotextile, two types of pores with
obviously different pore sizes could be observed, as shown in Figure 10. On one hand,
the first type of pores referred to the relatively large ones among the weaving yarns and
can be reflected by the inter-yarn AEV. The larger the pore sizes were, the smaller the
AEV shall be. On the other hand, the second type of pores referred to the deep grooves
within the wicking fiber. The average opening of the groove was 5-12 microns,
corresponding to a larger inner-yarn AEV of 254 kPa. In other words, the WRC of the
wicking geotextile was similar to that of a gap-graded soil with only two types of particle
sizes. The WRC of gap-graded soil was often depicted by the bimodal curves (Durner
1994). Two different AEVs and two distinct residual points can be defined for the
bimodal WRC. As for the wicking fibers, there was only one type of pores within the
wicking fiber and a unimodal was used to determine the WRC for T4.
Because the WRCs of the three tested materials (T2-T4) showed a high
dependence on the number of the wicking fiber yarns, it was of great interest to establish
the relationship between the numbers of wicking fiber yarns and the water storage ability
of the geotextile. Since T2 shared the same weaving pattern with T3 but without any
wicking fiber yarns, the number of wicking fiber yarns for T2 and T3 were 0 and 3,
38
respectively. In addition, a modified wicking geotextile, T5, was artificially fabricated in
the lab by knitting an additional wicking fiber yarn to one set of the weft yarns and the
number of fiber yarns for T5 was denoted as 4.
The regression parameters for T1-T5 were presented in Table 3. To quantify the
effect of numbers of wicking fiber yarns on the water storage ability of the wicking
geotextile, simple linear regression equations were established between the number of
yarns and the regression parameters (a, m, and n). Note that the water storage ability was
controlled by the wicking fibers under relatively high suction level (> 50 kPa). The
second set of parameters (suction > 50 kPa) for T3 and T5 were used for regression
analysis. The regression equations are presented in Equation 4 and the coefficient of
determinations was acceptable (R2 = 0.98~0.99). The regression equations were useful in
quantifying the water storage ability of the wicking geotextile. For example, based on
engineering necessities for drainage conditions, the ability of the geotextile to hold water
could be adjusted by adding or subtracting certain numbers of wicking fiber yarns to meet
the criteria. The hydraulic properties (WRC and K-function) of the material could be
reasonably predicted by using Equation 4. One can save a considerable amount of time
and money since it was not necessary to physically perform any laboratory tests.
a = 77.015x + 5.2308 (4a)
m = −2938x + 1.9823 (4b)
n = −0.3273x + 2.5204 (4c)
where, a , n and m = fitting parameters; and x = number of wicking fiber yarns.

39
8. CONCLUSIONS
This paper reexamined the existing testing techniques for the suitability to
determine the WRCs of woven geotextiles. The authors also proposed the proper testing
techniques for determining the WRC of a new wicking geotextile. The WRCs of different
types of woven geotextiles, both wicking and non-wicking, were determined using the
proposed testing technique. The advantages of the wicking geotextile to hold and
transport water under unsaturated conditions were discussed in detail. The effect of the
number of wicking fiber yarns on the water storage ability of the wicking geotextile has
been quantified. The major conclusions are summarized as follows:
1. The conventional testing techniques of capillary rise and pressure plate tests for soil
or nonwoven geotextiles requires modifications in determining the WRCs of woven
geotextiles. Due to the anisotropic pore-size distributions within a woven
geotextile, the CD test shall be used in which the water flow direction was in
parallel with the direction of the pores that had the highest continuity and
connectivity. In addition, no modification was required for the salt concentration
test because the test relied on the controlled relative humidity within the chamber
and had no directional preference.
2. Based upon the proposed testing techniques, five WRCs of the woven geotextiles,
including non-wicking geotextile, wicking geotextile, and wicking fiber, have been
determined. According to the test results, the non-wicking geotextiles did not have
the ability to hold water under unsaturated conditions. In contrast, due to the
existence of the wicking fibers, the wicking geotextile had very strong water storage
40
ability which ensures the wicking geotextile to be able to hold and transport water
3. Due to the relative uniform pore-size distributions within non-wicking geotextiles,
a unimodal was sufficient to depict the WRC. However, there were two types of
pores with obviously different pore sizes within the wicking geotextile, including
the large pores among the weaving yarns and the small pores within the openings
of the deep grooves of the wicking fiber. Therefore, two AEVs, namely inter-yarn
AEV (1.1 kPa) and inner-yarn AEV (254 kPa), were used to depict the WRC of the
wicking geotextile.
4. According to the regression results, the numbers of yarns have a significant
influence on the water storage ability of the wicking geotextile. The higher number
of yarns in the geotextile, the higher the ability to hold and transport water under
unsaturated conditions. The relationships between the regression parameters and
the numbers of wicking fiber yarns were established. The relationship allowed it
possible to modify and adjust the hydraulic properties of the new woven geotextile
according to engineering necessities.

41
Table 1. Material properties of T1 (conventional geotextile) and T2 (new geotextile)

Materials
Properties Unit T1 T2 T3
(HP570) (RS580i) (H2Ri)
40
Apparent Opening Size (AOS) U.S. Sieve (mm) 30 (0.60) 40 (0.425)
(0.425)
Permittivity sec-1 0.5 1.0 0.4
Flow Rate l/min/m2 1222 3055 1222
Table 2. Salt concentrations and the corresponding suctions
Mg Cl2 (g/L) Suction (kPa)

19.050 1303
38.100 2739
47.626 3523
66.676 5244
95.251 8249
142.877 14554
Table 3. Regression parameters
T3 T4
Parameters T1 T2 Suction Suction Suction Suction T5
≤ 50 kPa > 50 kPa ≤ 50 kPa > 50 kPa
ws 41.37 50.58 52.8 5 54.51 8 81
ψr 2.2 3 7 600 8 800 23
a 0.8 0.8 0.83 254 0.89 300 1.76
n 4.47 2.5 2.19 1.62 0.65 1.15 1.69
m 1.25 2 0.99 1.03 3.68 0.86 0.76
42
Perpendicular Direction(PD)
Cross-Machine Direction Machine Direction

(CD) (MD)
(a)
Perpendicular Direction (PD)

(CD) (MD)
(b)
Figure 1. Testing materials: (a) T1 (HP570), (b) T2 (RS580i), (c) T3 (H2Ri), (d) T4
(4DG wicking fiber), and (e) T5 (modified wicking geotextile)
43
Perpendicular Direction (PD)

(CD) (MD)
(c)
(d)
Figure 1. Test materials: (a) T1 (HP570), (b) T2 (RS580i), (c) T3 (H2Ri), (d) T4 (4DG
wicking fiber), and (e) T5 (modified wicking geotextile) (cont.)
44
(e)
Figure 1. Test materials: (a) T1 (HP570), (b) T2 (RS580i), (c) T3 (H2Ri), (d) T4 (4DG
wicking fiber), and (e) T5 (modified wicking geotextile) (cont.)
(a) (b)
Figure 2. Capillary rise test: (a) schematic plot, and (b) test setup (CD test)
45
(a)
Figure 3. Pressure plate test: (a) test setup (PD test), and (b) test setup (CD test and MD
test)
46
(b)
Figure 3. Pressure plate test: (a) test setup (PD test), and (b) test setup (CD test and MD
test) (cont.)
47
(a)
(b)
Figure 4. Salt concentration test: (a) schematic plot, and (b) test setup
48
(a) T1
(b) T2
Figure 5. Capillary rise test results: (a) T1, (b) T2, (c) T3, and (d) comparisons among
T2, T3, and T4
49
(c) T3
(d) T4
Figure 5. Capillary rise test results: (a) T1, (b) T2, (c) T3, and (d) comparisons among
T2, T3, and T4 (cont.)
50
(a) T1
(b) T2
Figure 6. Pressure plate test results: (a) T1, (b) T2, (c) T3, and (d) T5
51
(c) T3
(d) T5
Figure 6. Pressure plate test results: (a) T1, (b) T2, (c) T3, and (d) T5 (cont.)
52
Figure 7. Salt concentration test results
(a) T1
Figure 8. Summary of test results for non-wicking geotextiles: (a) T1, and (b) T2
53
(b) T2
Figure 8. Summary of test results for non-wicking geotextiles: (a) T1, and (b) T2 (cont.)
(a) T3
Figure 9. Summary of test results for wicking geotextiles: (a) T3, and (b) T4
54
(b) T4
Figure 9. Summary of test results for wicking geotextiles: (a) T3, and (b) T4 (cont.)
Figure 10. Water retention curves (WRCs) of woven geotextiles

55
REFERENCES
ASTM D4491-17 (2017). “Standard Test Methods for Water Permeability of Geotextiles
by Permittivity.” ASTM International, West Conshohocken, PA, 2017.
ASTM D6836-16 (2016). “Standard Test Methods for Determination of the Soil Water
Characteristic Curve for Desorption Using Hanging Column, Pressure Extractor,
Chilled Mirror Hygrometer, or Centrifuge.” ASTM International, West
Conshohocken, PA. https://doi.org/10.1520/D6836-16.
ASTM. D4751-16 (2016). “Standard Test Method for Determining Apparent Opening
Size of A Geotextile”. ASTM International, West Conshohocken, PA.
Azevedo, M., and Zornberg, J. (2013). “Capillary Barrier Dissipation by New Wicking
Geotextile.” Proceedings, Panamerican Conference on Unsaturated Soils, 20-22.
Bathurst, R. J., Ho, A. F., and Siemens, G. (2007). “A Column Apparatus for
Investigation of 1-D Unsaturated-Saturated Response of Snad-Geotextile
Systems.” ASTM Geotechnical Testing Journal, 30(6), 433-441.
Bouazza, A., Zornberg, J. G., McCartney, J. S., and Nahlawi, H. (2006a). “Significance
of Unsaturated Behaviour of Geotextiles in Earthen Structures." Australian
Geomechanics, 41(3), 133-142.
Bouazza, A., Freund, M., and Nahlawi, H. (2006b). “Water Retention of Nonwoven
Polyester Geotextiles." Polymer Testing, 25(8), 1038-1043.
Brooks, R. H. and Corey, A. T. (1964). “Hydraulic Properties of Porous Media.”

Colorado State University, Hydrology Paper, 3(27).
Clough, I., and French, W. (1982). “Laboratory and Field Work Relating to the Use of
Geotextiles in Arid Regions.” Proc., Proceedings of the 2nd International
Conference on Geotextiles, Las Vegas, NV, 1-6.
Currey, J. (2016). “H2Ri Wicking Fabric Experimental Feature Final Report Daltong
Highway MP 197-209 Rehabilitation.” Alaska Department of Transportation and
Public Facilities, Project No. IM-DP-065-4(8)/61214, Juneau, AK, US.
Douglas, R. A. (1993). “Stiffness of Geosynthetics-Built Unpaved Road Structures:

Experimental Programme, Analysis and Results.” Proceedings of the
Geosynthetics 1993, Vancouver, Canada, 65-79.
56
Durner, W. (1994). “Hydraulic Conductivity Estimation for Soils with Heterogeneous

Pore Structyure.” Water Resources Research, 30(2), 211-223.
Fredlund, D. G., and Xing, A. (1994). “Equations for the Soil-Water Characteristic
Curve.” Canadian Geotechnical Journal, 31(4), 521-532.
Fredlund, D. G. and Rahardjo, H. (1993). “Soil Mechanics for Unsaturated Soils.” John
Wiley & Sons, New York, NY, US.
Fredlund, D. G., Xing, A., and Huang, S. Y. (1994). “Predicting the Permeability
Function for Unsaturated Soils Using the Soil-Water Characteristic Curve.”
Canadian Geotechnical Journal, 31(4), 533-546.
Fredlund, D. M., Fredlund, D. G., and Wilson, G. W. (2000). “An Equation to Represent
Grain-Size Distribution.” Canadian Geotechnical Journal, 37(4), 817-827.
Giroud, J. P. and Noiray, L. (1981). “Geotextile-Reinforced Unpaved Roads.” Journal of

Geotechnical Engineering Division, American Society of Civil Engineers,
Volume 107, No. GT9, 1233-1254.
Giroud, J., Zornberg, J., and Zhao, A. (2000). “Hydraulic Design of Geosynthetic and
Granular Liquid Collection Layers.” Geosynthetics International, 7(4-6), 285-380.
Iryo, T. and Rowe, K. R. (2003). “On the Hydraulic Behavior of Unsaturated Nonwoven
Geotexiles.” Geotextiles and Geomembranes, 21(6), 381-404.
Klute, A. (1965). “Laboratory Measurement of Hydraulic Conductivity of Unsaturated

Soil.” In Methods of Soil Analysis, Monogram 9, Part 1, American Society of
Agronomy, Madison, WI, 253-261.
Knight, M., and Kotha, S. (2001). “Measurement of geotextile-water characteristic curves

using a controlled outflow capillary pressure cell.” Geosynthetics International,
8(3), 271-282.
Koerner, R. M. (2012). “Designing with Geosynthetics.” Xlibris Corporation,

Bloomington, IN, US.
Krisdani, H., Rahardjo, H., and Leong, E. C. (2006). “Experimental Study of 1-D
Capillary Barrier Model Using Geosynthetic Material as the Coarse-Grained
Layer." Unsaturated Soils 2006, 1683-1694.
McCartney, J. S., Kuhn, J. A., and Zornberg, J. G. (2005). “Geosynthetic Drainage

Layers in Contact with Unsaturated Soils.” Proceedings of the 16th International
Conference on Soil Mechanics and Geotechncial Engineering (ISSMGE), Osaka,
Japan, 2301-2305.
57
Mualem, Y. (1986). “Hydraulic Conductivity of Unsaturated Soils: Prediction and

Formulas.” In Methods of Soil Analysis, Part 1: Physical and Mineralogical
Methods, Second Edition, Agronomy, Madison, WI, 799-823.
Nahlawi, H., Bouazza, A., and Kodikara, J. (2007). “Characterisation of Geotextiles

Water Retention Using a Modified Capillary Pressure Cell.” Geotextiles and
Geomembranes, 25(3), 186-193.
Stormont, J. C., and Anderson, C. E. (1999). “Capillary Barrier Effect from Underlying
Coarser Soil Layer.” Journal of Geotechnical and Geoenvironmental Engineering,
125(8), 641-648.
Stormont, J. C., and Morris, C. E. (1998). “Method to Estimate Water Storage Capacity
of Capillary Barriers.” Journal of Geotechnical and Geoenvironmental
Engineering, 124(4), 297-302.
Stormont, J. C., and Morris, C. E. (2000). “Characterization of Unsaturated Nonwoven

Geotextiles.” Advances in Unsaturated Geotechnics, 153-164.
Van Genuchten, M. T. (1980). “A Closed-Form Equation for Predicting the Hydraulic

Conductivity of Unsaturated Soils.” Journal of Soil Science Society of America,
44, 892-898.
Zhang, X., and Belmont, N. (2009). “Use of Mirafi Nylon Wicking Fabric to Help
Prevent Frost Heaving in Alaska Pavement: 1st, 2nd, 3rd, 4th and 5th Progress
Reports.” Progress Reports to TENCATE GEOSYNTHETICS (North America),
Institute of Northern Engineering (INE)/Alaska University Transportation Center
(AUTC), University of Alaska Fairbanks.
Zornberg, J., Bouazza, A., and McCartney, J. (2010). “Geosynthetic Capillary Barriers:
Current State of Knowledge.” Geosynthetics International, 17(5), 273-300.
58
II. COMPREHENSIVE MATERIAL CHARACTERIZATIONS OF PAVEMENT

STRUCTURE INSTALLED WITH WICKING FABRICS
Chuang Lin1, S.M. ASCE; Xiong Zhang2, M. ASCE; and Jie Han3, F. ASCE
1
Graduate Research Assistant, Dept. of Civil, Architectural, and Environmental
Engineering, Missouri Univ. of Science and Technology, Rolla, MO 65409-0030. Email:
[email protected]
2
(Corresponding Author) Associate Professor, Ph.D., P.E., Dept. of Civil,
Architectural, and Environmental Engineering, Missouri Univ. of Science and
Technology, Rolla, MO 65409- 0030. Email: [email protected]
3
Professor, Ph.D., P.E., Dept. of Civil, Environmental, and Architectural
Engineer- ing, Univ. of Kansas, 1530 W. 15th St., Lawrence, KS 66045-7609. Email:
[email protected]
ABSTRACT
When a road is constructed, soils within the embankment are often compacted at
the optimum water content to achieve the best performance. The post-construction water
content tends to increase with time due to capillary action, precipitation infiltration, and
water condensation. Since soils are sensitive to water content variations, a slight water
content increase will cause a significant reduction in soil moduli and a dramatic increase
in permanent deformation under cyclic loading. Materials with large pores, such as
granular materials and nonwoven geotextiles, are commonly used for the drainage of
gravitational water under a saturated condition, but not for the drainage of capillary water
under an unsaturated condition. Therefore, the excessive water in the soil accelerates the
deterioration of the road with time.

59
A new geotextile with wicking fibers, which can drain both gravitational and
capillary water under saturated and unsaturated conditions, has recently been developed.
Several field applications have proven its effectiveness in dehydrating the road
embankment. This paper aimed at quantifying the benefits of the new geotextile in terms
of water removal. Firstly, comprehensive laboratory tests were performed to characterize
the properties of the soil, the new geotextile, and the soil-geotextile interactions.
Secondly, the working mechanism and the functional range of the new geotextile were
determined and the drainage ability of the soil-geotextile system was quantified via
numerical simulations. The wicking fabric functions as an effective drainage material to
hold and transport capillary water in the in-plane direction. The soil-geotextile system is
able to reduce the water content of the base course by 2.2% from the optimum value, and
the corresponding resilient modulus can be increased by 2-3 times.
KEYWORDS: Subsurface drainage; Pavement; Geotextile; Unsaturated soil;
Suction; Resilient modulus; Reinforcement.
1. INTRODUCTION
Excess water within a pavement structure is one of the major causes of pavement
deterioration. Cedergren (1987) indicated that if the pavement structure is saturated for
only 10% of the time in its service life, the serviceability index will reduce to half of that
of a fully drained pavement structure. Water-induced pavement distresses include but not
limited to rutting, pothole, pumping, edge dropoff in expansive subgrades, and frost
heave in cold regions (Yilmaz and Sargin 2012). The modulus of a base course is
60
sensitive to small water content variations. Li et al. (2011) reported that the resilient
modulus of the D-1 aggregate (a typical granular base course used in Alaska with 0%-6%
fine content that is considered a high-quality geomaterial) decreased to half when the soil
water content increased from 3.5% to 6.0%. Severe rutting issues of granular aggregates
due to soil wetting were reported in a number of published papers (Arnold 2004; Corey
1994; Lekarp et al. 2000; Shoop and Henry 1991; Uzan 2004). In cold regions, the rise
and accumulation of water combined with freezing and thawing cycles will cause frost
heave, another devastating engineering problem (Henry 1998; Konrad and Morgenstern
1980).
The soil post-construction water content is expected to gradually increase with
time and the excess water will accelerate the rate of pavement deteriorations. When
pavement is constructed, soils are often compacted at the optimum water content to
achieve the maximum dry density and the best performance, as shown in Figure 1. After
construction, the moisture migration within a pavement structure is dominated by
ambient atmosphere variations. The relative humidity (RH) in the air is usually lower
than 90% during the daylight time and the corresponding suction can be as high as 14
MPa (T = 20ºC and RH = 90%) (Fredlund and Rahardjo 1993). The surface soil is
exposed to the surrounding atmosphere and can be quickly air dried under such a high
suction level. As soil becomes unsaturated, the liquid water in the soil pores becomes
discontinuous and its permeability significantly decreases. The soil permeability
decreases exponentially as the soil desaturates (Fredlund et al. 2012). Air can easily go
into soil pores and block the liquid water flow. The air-dried surface soil serves like a
layer of impermeable membrane, impeding the water exchange between the soil and the
61
surrounding atmosphere. In the meantime, the water content of the soil within the road
embankment will gradually increase due to capillary action if the groundwater table is
relatively shallow and the base course contains fine content over 4% (Siswosoebrotho
2005). In addition, cracks will develop at the road surface and the infiltration rate will be
increased. The excess water will be accumulated within the base course, resulting in a
significant reduction in the soil moduli and a dramatic increase in permanent deformation
under cyclic loading. The problem can become even worse with time as the asphalt
concrete layer continues to age and deteriorate. Cracks will become wider and more
abundant, resulting in more water infiltrating into the pavement structure and accelerating
the deterioration process (Hagen and Cochran 1996). Therefore, no matter how well the
road is constructed, excess water will accelerate the deterioration rate under the influence
of cyclic traffic loads.
The existing subsurface drainage designs (AASHTO 1993; AASHTO 2008;
Witczak et al. 2004) are based on water flow in saturated soils and do not consider
unsaturated conditions. However, soils within a pavement structure are under unsaturated
conditions during most of the time in its service life. The drainage capacity of the
pavement structure will be inevitably overestimated because the soil permeability
exponentially decreases as the soil desaturates, and its ability to transport water will be
extensively decreased under unsaturated conditions. A conventional drainage system
(e.g., a granular layer or geocomposite) can only drain gravitational water, not capillary
water. Capillary barriers are often used to prevent capillary water from rising from the
groundwater table to the base course (Bouazza et al. 2006). A capillary barrier consists of
a coarse-grained soil layer (e.g., sand or gravel) or porous geosynthetics (e.g., a

62
nonwoven geotextile), which have larger pores than the surrounding soils. Its hydraulic
conductivity also drops significantly with the decreasing degree of saturation. However,
capillary barriers cannot drain out capillary water, so excess water accumulates in the
overlying soil (Khire et al. 2000). In addition, a capillary barrier may lose its function and
water will gradually build up near the barrier as the surrounding soil approaches
saturation (Giroud et al. 2000; McCartney et al. 2005; Zornberg et al. 2010). Therefore,
the existing subsurface drainage systems have their limitations, and the conventional
treatments cannot work effectively in to drain capillary water.
2. A NEW GEOTEXTILE WITH WICKING FIBERS
Woven geotextiles have been widely used for reinforcement purposes in
pavement structures due to their high tensile strengths. However, the overall performance
of a pavement structure is not solely determined by the strength of the geotextile, but also
relies on the interactions between the soil and the geotextile. Since the thickness of the
geotextile is generally very thin compared with the entire thickness of the base course,
the reinforcement effect of the geotextile is limited. Moreover, as discussed in the
previous section, another important factor that affects the performance of the pavement
structure is the soil water content and shall be considered in pavement design procedures.
The overall performance of the soil-geotextile system can be very poor as the soil
approaches saturation, no matter how strong the geotextile is (Christopher, et al. 1998). In
addition, even though non-wicking geotextiles have been used for drainage purposes,
they are not effective in reducing capillary water and in fact may sometimes inadvertently
63
reduce the performance and longevity of the pavement structures (Christopher and
McGuffey 1997; Heckel 1997). Therefore, it is necessary to come up with an alternative
that can effectively reduce the excess water (especially capillary water) and improve the
long-term performance of the pavement structure.
A new geotextile with wicking fibers (hereafter referred to as the “wicking
fabric”) has been recently developed and has the ability to drain both gravitational and
capillary water under saturated and unsaturated conditions, as shown in Figure 2a. This
wicking fabric contains both high modulus polypropylene yarns (black color) for
reinforcement purposes and wicking fiber yarns (white color) for drainage purposes (as
shown in the first image). The multi-channels are clearly observed through an SEM
(scanning electron microscopy) image of the wicking fibers (second image). The
diameter of a single wicking fiber ranges from 30 μm to 50 μm and the spacing between
deep grooves varies from 5 μm to 12 μm. The specially designed multi-channel cross-
sectional shape ensures the wicking fiber will have a high shape factor and a great
number of channels per fiber (corresponding to a specific surface area of 3650 cm2/g).
Table 1 summarizes the specifications of the wicking fabric. The wide-width tensile
strengths of the wicking fabric are 15.8 kN/m in the cross-machine direction (CD), which
is the weaving direction of the wicking fiber, and 7.0 kN/m in the machine direction
(MD). The permittivity of the wicking fabric is 0.4 s-1, corresponding to an equivalent
flow rate of 1222 l/min/m2. A salient feature of the wicking fabric is that it can maintain
saturation and have a relatively high hydraulic conductivity to transport water under
unsaturated conditions, as schematically shown in the third image. In a laboratory
environment in which the relative humidity was approximately 50%, the wetting front
64
moved 1.86 m horizontally with zero hydraulic gradient in 983 minutes. With this
feature, the wicking fabric can connect the soil in a pavement structure with the ambient
atmosphere under both saturated and unsaturated conditions.
Figure 2b shows the conceptual subsurface drainage design with the wicking
fabric. A layer of the wicking fabric is installed below or within the base course. The
edges of the wicking fabric are exposed to the air for evaporation purposes. Both
gravitational and capillary water can be absorbed by the wicking fabric from surrounding
soils, transported along the wicking fibers to the facing of the road slopes, and are
eventually evaporated to the ambient atmosphere. Unlike conventional drainage systems
that will result in excess water accumulation in the overlying soils, the wicking fabric
serves as a “pipe” to connect the soils inside a pavement structure with the outside
atmosphere. Compared with the amount of water needed to saturate the earth’s
atmosphere, the amount of water in a pavement structure is very limited. If properly
designed, the atmosphere can work as a “natural pump” that works 24 hours a day and
365 days a year to dehydrate the road embankment.
The effectiveness of the drainage system that incorporates the wicking fabric has
been validated through several successful field applications. For example, Zhang and
Presler (2012) used two layers of wicking fabrics to prevent frost heave and the
subsequent thaw weakening issue for Alaskan pavements. Field observation indicated
that the wicking fabric still functions well five years after the completion of the project
(Lin et al. 2017). Currey (2016) also reported successful applications of the wicking
fabric to deal with soft subgrades, and the initial construction was saved $ 2.5 million.
65
Delgado (2015) also used the wicking fabric to deal with differential settlement induced
pavement deteriorations.
Although preliminary tests show that the wicking fabric is very effective in
dehydrating road embankments, the benefits of the wicking fabric need to be quantified
and incorporated in the existing pavement design to promote more applications. The
objectives of this paper are to: 1) comprehensively characterize the properties of the soil,
the wicking fabric, and their interactions under saturated and unsaturated conditions, 2)
theoretically explain the working mechanism of the wicking fabric and its functional
range, 3) quantify the ability of the soil-wicking fabric system in terms of water removal,
and 4) compare the performance differences of pavement structures with and without the
wicking fabric.
Table 2 summarizes the laboratory tests required to characterize the mechanical
and hydraulic properties of the soil, the wicking fabric, and the interactions between the
soil and fabric. The modulus of elasticity for the soil was determined by the resilient
modulus test. The initial tangent modulus of the wicking fabric was determined by a
wide-width tensile test (the test data was obtained from the specifications). Important
hydraulic properties of the geomaterials include their water characteristic curves and the
hydraulic conductivity functions (also named as K-functions). The water characteristic
curve depicts the ability of the geomaterial to hold water under unsaturated conditions.
The pressure plate and salt concentration tests were used to determine the Soil Water
66
Characteristic Curve (SWCC) of the soil. As for the wicking fabric, an additional
capillary rise test was used to determine the Geotextile Water Characteristic Curve
(GWCC). The saturated hydraulic conductivities of the geomaterials were determined via
the constant head test. The K-functions were predicted based on the theoretical
derivations that were based on the measured SWCC/GWCC and the saturated hydraulic
conductivities. In addition, the interface friction angle between the soil and the wicking
fabric was determined via the large-scale direct shear test.
4. MATERIAL CHARACTERIZATIONS
4.1. SOIL PROPERTIES
4.1.1. Physical Properties. The selected soil was Aggregate Base Class 3 (AB3),
a typical base course used in Kansas. The sieve analysis (ASTM D6913/D6913M-17) and
the modified Proctor compaction (ASTM D1557-12) tests were conducted. The coefficient
of uniformity (Cu) and the coefficient of gradation (Cc) of the AB3 were 54.7 and 2.9,
respectively. There were approximately 10% of fines in AB3, and the liquid limit and
plasticity index of the fines were 20% and 7, respectively. According to the Unified Soil
Classification System (USCS), the soil should be classified as well-graded gravel with
silty clay and sand (GW-GC). The optimum water content and maximum dry density of
the AB3 were 8.5% and 2.2×103 kg/m3, respectively.
4.1.2. Resilient Modulus. Resilient modulus has been widely recognized as the
primary mechanical property required in both the MEPDG (Mechanistic-Empirical
Pavement Design Guide) and the AASHTO (American Association of State Highway and
67
Transportation Officials) pavement designs (Gu et al. 2014). It is commonly used to
evaluate the dynamic behavior of base and subgrade materials under a traffic load. Factors
influencing the base course resilient behavior include stress state, moisture condition, and
matric suction (Yang et al. 2008). The stress state dependency of soil resilient behavior
has been well studied. However, it is also important to take water content variations into
consideration during the design process because the soil resilient modulus is sensitive to
water content variations (Yang et al. 2005). Although numerous studies have investigated
the influence of water content variations on the resilient behavior of soils, most of them
focused on a very limited water content range (Jin et al. 1994; Li and Qubain 2003; and
Khoury and Zaman 2004). Unlike previous studies, this research investigated the resilient
modulus variations for the entire water content range, from fully saturated (water content
= 12.5%) to completely dry (water content = 0%) at a 95% maximum density.
The resilient modulus test was performed according to AASHTO T309-99 (2003).
The structures of a compacted soil depend on the compaction effort and the water
content. The soil’s potential to absorb and retain water primarily depends on its initial
matric suction or the initial water content (Tripathy et al. 2005). To achieve a relatively
similar soil structure for each sample, all soil samples were first compacted at the
optimum water content (8.5%). For the target water content lower than 8.5%, the samples
were exposed to the air for approximately 15-20 minutes every day and then covered with
a plastic wrap for the rest of the day to ensure a uniform water content distribution.
Meanwhile, for the target water content higher than 8.5%, water was sprayed on the soil
surface and then the sample was also covered with a plastic wrap to achieve a uniform
68
water distribution. For each water content level, at least three samples were prepared and
tested.
Figure 3 shows the resilient modulus test results. In general, the resilient modulus
significantly decreased with increasing water contents. Since soils in the field were
compacted at the optimum water content (8.5%), the soil resilient modulus value at this
level was considered the designed value and would be used as a reference for comparison
purposes. The confining pressure of 34.5 kPa is taken as an example to demonstrate the
effect of water content variations on the soil resilient behavior (as shown in the red box in
Figure 3a). The resilient modulus was 95 MPa at the water content of 8.9%. As discussed
previously, the soil post-compaction water content tended to increase due to precipitation
infiltration, capillary action, and rising of the groundwater table. Suppose the soil water
content increased to 10.8% (1.9% higher than the reference value), then the resilient
modulus, would decrease to 56 MPa (i.e., approximately 50% reduction). In other words,
if the post-compaction water content could be maintained at its optimum value, the
modulus would be doubled compared with that at a water content of 10.8%. Moreover, if
the post-compaction water content could be further reduced to 6.5% (2% lower than the
optimum value), the soil resilient modulus would increase to 363 MPa, which is over 3
times higher than the modulus at the optimum water content. In other words, only ± 2 %
water content variations could lead to a difference of six times in the resilient modulus.
Moreover, the resilient modulus values also increased with increasing confining
pressures. For instance, the resilient modulus of the base course gradually increased from
333 MPa to 633 MPa (given the same water content of 6.9%) while the confining
69
pressure increased from 20.7 kPa to 137.9 kPa. However, the effect of confining pressure
variations on the resilient behavior was not as significant as that of the water content.
The permanent deformations also show high dependencies on water content
variations, as shown in Figure 3b. The average permanent deformation significantly
increased with increasing water content. Similarly, take the permanent deformation at a
water content of 8.8% as a reference for discussions. The permanent deformation
increased by 3 times (from 2.5 mm to 8.2 mm) as the water content increased from 8.8%
to 10.8%. In contrast, the permanent deformation decreased from 2.5 mm to 1.4 mm as
the water content decreased from 8.8% to 6.8%. In summary, both the resilient modulus
and the permanent deformation were sensitive to water content variations.
As discussed above, the resilient modulus is controlled by two factors: the stress
condition and the soil water content. The stress condition represents the stress state of the
soil, including bulk and shear stresses. Meanwhile, the soil water content is closely
related to the matrix suction of the soil (based on the soil water characteristic curve
(SWCC)), which is another independent stress state variable that describes the shear
strength of an unsaturated soil. The influence of soil suction on shear strength is
significant due to the additional confinement provided by the soil suction. However, as
the soil approaching saturation, water within the soil pores serves as a lubricant that
significantly reduces the strength of the soil. This is why the soil sample could not
complete the loading sequence when the water content was higher than 10.8%. Therefore,
the regression equation for the resilient modulus should be based upon two stress state
variables, including stress (bulk and shear stresses) and water content (closely related to
soil suction).
70
Various models have been proposed to predict the soil resilient behavior and the
most commonly used is the universal model proposed by Witczak (2003). In this study,
based on test results, the Witczak model was modified to include the influences of both
the water content and the stress state, as expressed in Equation 1. The R-squared value
was 0.76, which was acceptable considering such a wide range of water content
variations. In addition, the regression equation was only suitable to represent soils with
water content smaller than 10.8%, otherwise the predicted resilient modulus would be a
negative value. This regression curve was consistent with the observation during the test
that the soil samples with water contents higher than 10.8% failed during the loading
sequences, and the test was terminated as the vertical strain exceeded 5% (according to
AASHTO T307-99 (2003)).
( -0.9174+0.2058wc ) (0.854-0.5493wc )
M R = ( 30.448 - 2.854  wc ) Pa ( / Pa ) × ( OCT / Pa +1) (1)
where M R = resilient modulus, MPa; Pa = atmospheric pressure (i.e., 101 kPa);  =
bulk stress, 1 +  2 +  3 , kPa; wc = water content, %; and  OCT = octahedral shear
stress, τOCT = 13 (1 −  2 ) + ( 2 −  3 ) + (1 −  3 )

2 2 2
, kPa.
4.1.3. Soil Water Characteristic Curve (SWCC). The SWCC depicts the ability
of soil to hold water under unsaturated conditions. The SWCC is the base of engineering
practice for unsaturated soils since it correlates to the unsaturated hydraulic conductivity,
the soil shear strength, and the volume change of an unsaturated soil (Alim and Nishigaki
2009). The drying SWCC is easier to measure in the laboratory and has become the
primary curve used in estimating unsaturated soil property functions (Fredlund et al.
2002). Pressure plate (for suction ≤ 1500 kPa) and salt concentration (for suction > 1500
71
kPa) tests were performed to determine the SWCC and the test equipment is shown in
Figures 4.
Most SWCC tests using the standard pressure plate test procedures (ASTM
D6836-16) were performed on relatively small samples (approximately 50-70 mm
diameter and 10 mm tall) and larger particles were removed when preparing the sample
(Bareither and Benson 2013). Bouwer and Rice (1984) proposed a water content
correction procedure for a bulk soil containing large particles (largest particle size of
approximately 18 mm) and have been used in practice periodically. Bareither and Benson
(2013) further simplified the Bouwer-Rice methodology by only determining the SWCC
for the finer soil fraction (particle sizes smaller than the opening of a No. 4 U.S. sieve)
and correcting the SWCC based upon the saturated water content of the bulk soil
containing large particles. The largest particle size for the AB3 was 19 mm and the
correction procedure proposed by Bareither and Benson (2013) was used to correct the
results of the pressure plate test in this paper.
The pressure plate test was performed according to ASTM D6836-16 (2016).
Each soil sample was first compacted at the optimum water content with five layers. The
two adjacent layers were separated with a thin metal plate to generate a flat and smooth
contact surface (first image in Figure 4a). Then, the compacted sample was put into a
plastic mold with predrilled holes (second image in Figure 4a) and submerged into water
for saturation. After that, the saturated sample was placed on a ceramic plate in the
pressure plate extractor (third image in Figure 4a). Air pressure was applied to the system
and only water could be drained out through the ceramic plate. The test lasted for 7 days
until an equilibrium state was achieved. At the equilibrium condition, the soil water
72
content was measured and the applied pressure was recorded. This approach was
successively repeated for higher pressures and the SWCC at the low suction range
(<1500 kPa) was determined.
The salt concentration test is a thermodynamic-based test and is often used to
determine the SWCC at the high suction range (>1500 kPa). The test apparatus of the salt
concentration test is shown in Figure 4b. Salt solutions with different concentrations were
used to generate controlled relative humidity environments, or constant suctions, in the
desiccators. Water within the soil sample would evaporate in the sealed desiccator until
the equilibrium state was achieved. The relationships between suctions and the
corresponding salt concentrations were provided by Goldberg and Nuttall (1978). In total,
six salt solutions (MgCl2) were prepared at molar concentrations of 0.2M, 0.4 M, 0.5M,
0.7 M, 1.0 M, and 1.5 M, respectively. The corresponding suctions were 1303 kPa, 2739
kPa, 3523 kPa, 5244 kPa, 8249 kPa, and 14554 kPa, respectively. The large particles
were not removed since they would not impede the water vapor exchange between the
soil sample and the surrounding environment. The soil sample was first compacted at the
optimum water content. Then, the compacted sample was scattered into small pieces to
maximize the contact surface and reduce the time to reach equilibrium. The test lasted for
7 days, and the water content of the samples was determined after the test. Given the soil
water contents and the corresponding suction values based on the salt concentrations, the
SWCC at the high suction range (>1500 kPa) was determined.
Based on the results of the pressure plate and the salt concentration tests, the
SWCC could be regressed according to Fredlund and Xing’s (1994) equation, as
expressed in Equation 2.
73
 
0.412
wc = 12.50  C ( ) 1 / ln 2.718 + ( 2.35 ) 
1.241
(2)
 
where wc = soil water content, %; C( ) = 1 − ln(1 + /  r ) / ln(1 + 10 /  r ) ,  =

6
suction, kPa.; and  r = residual suction, kPa.
4.1.4. Saturated Hydraulic Conductivity. The saturated hydraulic conductivity
represents the ability of soil to transport water under a saturated condition. It can be
determined by the constant head test according to ASTM D5856-15 (2015). The soil
sample was first compacted at the optimum water content of 8.5%. After compaction, the
soil and the compaction mold were both submerged into the water tank for 24 hours till
the sample was fully saturated. Three samples were prepared and tested in the same way
described above. The average value of the saturated hydraulic conductivity was 5.2 × 10-6
m/s.
4.2. WICKING FABRIC PROPERTIES
4.2.1. Initial Tangent Modulus. The initial slope of the stress and strain curve for
many woven geotextiles, including the wicking fabric, which is a typical woven
multifilament geotextile that falls into this category, are quite linear, and the initial tangent
modulus can be considered to be a reasonably accurate modulus (Koerner 2012). The
initial tangent modulus of the wicking fabric was calculated based on the wide-width
tensile strength test and the test results were provided in the geotextile specifications (see
Table 1). The wide-width tensile strengths of the wicking fabric at 2% strain were 15.8
kN/m (CD) and 7.0 kN/m (MD), respectively. To obtain the true stress unit, these values
74
were divided by the thickness of the wicking fabric (1.6 mm), and the initial tangent
modulus values were 493 MPa (CD) and 218 MPa (MD), respectively.
4.2.2. Geotextile Water Characteristic Curve (GWCC). Similar to the SWCC,
the GWCC depicts the ability of the wicking fabric to hold water under unsaturated
conditions. The GWCC of the wicking fabric is expected to be directional with higher
values in the wicking direction and lower values in the cross-plane direction. The water
characteristic curve of a geomaterial is closely related to its pore size distribution
(Fredlund et al. 2002). A woven geotextile is expected to have a uniform pore size, or
AOS (Apparent Opening Size) is defined as the opening of a U.S. standard sieve number
that has openings closest in size to the openings in the geotextile. Two woven geotextiles
with the same AOS values are expected to have the same opening sizes and therefore have
similar water characteristic curves in the cross-plane direction. The drying GWCC of
another type of non-wicking geotextile, Mirafi HP570, was determined using the same
testing techniques. The non-wicking geotextile is developed by the same manufacturer,
and has the same weaving pattern and a very close AOS value (0.6 mm) to the wicking
fabric (0.425 mm). Therefore, the drying GWCC of the non-wicking geotextile was used
to represent the ability of the wicking geotextile to hold water in the cross-plane direction,
as expressed in Equation 3.
 
1.45
wc = 28.0  C ( ) 1/ ln 2.718 + ( 1.10 ) 
4.47
(3)
 
where wc = water content, %;  = suction, kPa.;
C( ) = 1 − ln(1 + /  r ) / ln(1 + 106 /  r ) , and  r = residual suction, kPa.

75
The capillary rise, pressure plate, and salt concentration tests were performed to
determine the drying GWCC of the wicking fabric in the in-plane (wicking) direction.
The testing technique for the pressure plate test was modified to be suitable for the
wicking fabric, which has specially designed wicking fibers. Detailed information
regarding the test procedures, modifications of the testing equipment, and comparisons of
test results between the modified and the traditional testing techniques is presented in
another paper (Lin et al., “Comparisons of Water Retention Curve for a New Wicking
Geotextile against Conventional Types”, working paper, Missouri S&T, Rolla, Missouri).
For the completeness of this paper, the test procedures are briefly introduced and the test
results are presented.
For the capillary rise test, a 0.1 m (width) × 1 m (length) saturated wicking fabric
strip was pinned to a wooden board, as shown in Figure 5a. The strip, together with the
wooden board, was raised vertically with the bottom end submerged into the water
reservoir. The strip was covered with plastic wrap to minimize the evaporation effect.
The test lasted for 7 days until an equilibrium condition was achieved. Then, the wicking
fabric strip was cut into 0.1 m-long pieces and the water content of each piece was
determined. The corresponding suction value,  , could be determined according to
Equation 4. Given the water contents and the corresponding suction values, the drying
GWCC (suction < 10 kPa) was determined.
 = w h (4)
where  = suction, kPa;  w = unit weight of water, KN/m3; and h = distance from the
centroid of the wicking fabric piece to the water surface, m.

76
Figure 5b shows the apparatus of the pressure plate test. Due to the special woven
texture of the wicking fabric, the fibers could not be in good contact with the ceramic
plate. To ensure a continuous water flow between the fibers and the ceramic plate, the
bottom of the saturated wicking fabric sample was immersed into a soil slurry mixed with
kaolinite and water (1:2 by weight). The soil slurry should not be too thick to impede
water flow, nor too thin to flow away. The wicking fabric samples were sandwiched by
two pieces of clear acrylic boards and fastened with two clamps. The test procedures of
the wicking fabric were the same as for soils. The air pressure was applied to the system
till an equilibrium condition was achieved. After the test was completed, the lower part of
the sample, which was contaminated by the soil slurry, was cut off and the water content
of the remaining part was determined. Given the suction values (equal to the values of
applied air pressures) and the corresponding water content, the drying GWCC at the low
suction range (< 1500 kPa) was determined.
The salt concentration test apparatus for the wicking fabric is shown in Figure 5c.
The saturated wicking fabric samples were cut into 5 cm by 5 cm square pieces and
saturated before the test. Then, the test was followed by the same procedures as for the
soil. The test lasted for 7 days till an equilibrium condition was achieved. Then, the water
contents of the samples were determined and the corresponding suction values were
recorded, and the GWCC at the high suction range (> 1500 kPa) was determined.
Based on the test results, the drying GWCC of the wicking fabric in the in-plane
direction was a segmented function with two sets of parameters and the regression
equations are expressed in Equation 5.

77
 
0.780
wc = 33.0  C ( ) 1/ ln 2.718 + ( 1.33 ) 
2.947
(ψ ≤ 50 kPa) (5a)
 
 
1.174
wc = 5.19  C ( ) 1/ ln 2.718 + ( 254.06 ) 
1.732
(ψ > 50 kPa) (5b)
 
where wc = water content, %; C( ) = 1 − ln(1 + /  r ) / ln(1 + 10 /  r ) ,  = suction,

6
kPa, and  r = residual suction, kPa.
4.2.3. Saturated Hydraulic Conductivity. The hydraulic conductivities of the
wicking fabric are directional due to the existence of the wicking fibers and the in-plane
hydraulic conductivity is expected to be higher than the cross-plane value. The
specifications of the wicking fabric (shown in Table 1) only give the permittivity value
(0.4 s-1) under a saturated condition, and the corresponding saturated cross-plane hydraulic
conductivity is 6.4  10-4 m/s. However, the saturated in-plane hydraulic conductivity or
transmissivity of the wicking fabric is not provided in the specifications. The wicking
fibers are a major component for the purpose of drainage, and determination of the
transmissivity in the in-plane (wicking) direction is necessary to depict the ability of the
wicking fabric to transport water.
The apparatus for the constant head test was modified and used to determine the
saturated in-plane hydraulic conductivity of the wicking fabric, as shown in Figure 6. The
water heads within the reservoir and the 50 ml beaker were maintained at a constant so
that the head difference, h , was also constant throughout the test. The water path length,
L , was determined by measuring the length between the water surfaces in the water
reservoir and in the 50 ml beaker. Therefore, the corresponding hydraulic gradient, i ,
could be calculated ( i = h / L ). The discharge volume of the water was collected in the
78
1000 ml beaker and the corresponding testing time was recorded. To minimize the
variation caused by water evaporation, the wicking fabric sample was covered with a
plastic sheet. The constant head test was replicated three times, and the average in-plane
saturated hydraulic conductivity of the wicking fabric was 6.20  10-4 m/s.
4.3. SOIL-GEOTEXTILE INTERFACE FRICTIONAL ANGLE
The interface frictional angle between the wicking fabric and the AB3 was
determined by the large-scale direct shear test and the test apparatus is shown in Figure
7a. The shear box was composed of two parts (upper and lower shear box) with
dimensions of 300 mm (length) × 300 (width) mm × 75 mm (height) each. The test was
performed according to ASTM D5321 (2002). The AB3 was first mixed with water and
hand compacted in the lower shear box. Then, a layer of the wicking fabric was placed on
top of the compacted AB3 with the machine direction of the wicking fibers parallel to the
shear direction. After that, the upper shear box was also backfilled with AB3 and hand
compacted. The AB3 was carefully compacted to obtain a consistent density (95% of the
maximum density) for each test. The normal load was first applied to the target value and
then the shear force was developed by hydraulic cylinders at a rate of 6 mm/min. Vertical
and horizontal displacements were measured by analogue indicators. The water contents
of the AB3 were determined after the test.
The test results are shown in Figure 7b. The AB3 was a granular soil with limited
fines (approximately 10%) and the soil cohesion, c , was assumed to be zero. The
interface frictional angle,  , between the AB3 and the wicking fabric first increased to a
peak value of 47.5° (at the water content of 2%), and then reduced to 39.4° as the water
79
content reached 10.5%. At the 2% water content, the soil was unsaturated and soil voids
were partially occupied with water. The soil suction developed and resulted in an
increment in the effective stress in the soil. The increased effective stress served as an
additional confinement that contributed to the highest interface frictional angle. However,
the interface frictional angle decreased after the water content exceeded 2%. As the water
content increased, the corresponding suction decreased and water detained in soil voids
acted as a lubricant that reduced the interface frictional angle. As the water content
continued to increase, the suction would significantly decrease and the confining effect
diminished. At this moment, the lubricating effect dominated the interfacial shear
behavior and the frictional angle further decreased. However, considering such a wide
range of water content variations, the interface frictional angle was not very sensitive to
water content variations.
5. PERFORMANCE OF THE SOIL-GEOTEXTILE SYSTEM
5.1. WORKING MECHANISM AND FUNCTIONAL RANGE
The performance of the soil-geotextile system mainly relies on its ability to hold
and transport water. Figure 8a plots the drying SWCC and GWCCs (in-plane and cross-
plane) according to Equations 2, 3, and 6. Two terminologies are often used to identify
the main features of the water characteristic curve, namely the air-entry value (AEV) and
the residual water content (Fredlund et al. 2012). The AEV is the matric suction where air
starts to displace water in the largest pores. The AEV was closely related to the largest
pore size within a geomaterial. The residual water content is defined as the water content
80
where a larger suction change is required to remove additional water from the
geomaterial. Based on the drying SWCC, the AEV and residual water content are 2.35
kPa and 5%, respectively. During the initial drying process of the AB3, the soil suction
builds up, but water does not drain out from the soil until the AEV is reached. After the
soil suction exceeds the AEV (2.35 kPa), air enters the soil pores and the soil water
content decreases. The soil water content continues to decrease until the residual water
content reaches 5%. The water becomes occluded within the soil pores if the soil water
content is lower than 5%. A large suction change is required to further reduce the soil
water content because there are no available pathways for water flow. In summary, the
AB3 shows low capability to hold water under unsaturated conditions.
The cross-plane and in-plane drying GWCCs of the wicking fabric are also shown
in Figure 8a. Because there are no wicking fibers in the cross-plane direction, the drying
GWCC of the wicking fabric in this direction is similar to that of other non-wicking
geotextiles. The cross-plane GWCC shows a highly nonlinear response, with a significant
decrement in water content within a comparatively narrow range of suction (1-3 kPa).
The shape of the curvature indicates that air will displace water after the suction exceeds
the AEV (1.1 kPa), and the ability of the wicking fabric to hold water in the cross-plane
direction will diminish after the suction value exceeds 3 kPa (residual suction value).
Actually, the above characteristic of the wicking fabric is also observed in other non-
wicking geotextiles which are often used as capillary barriers.
In comparison, the in-plane drying GWCC of the wicking fabric shows a much
stronger capability to hold water under unsaturated conditions, as shown in Figure 8a.
Two sets of regression parameters are used in describing the characteristics of the drying
81
GWCC. There are two AEVs for the wicking fabric, namely inter-yarn AEV and the
inner-yarn AEV. The inter-yarn AEV is 1.1 kPa and is mainly controlled by the large
pores among the weaving yarns. The AOS (apparent opening size) of the wicking fabric
is 0.425 mm (refer to Table 1), which is equivalent to the opening size of a U.S. No. 40
sieve or the particle size of a medium sand (0.2 mm -0.63 mm). The relatively large pore
sizes among the weaving yarns indicate that air can easily enter into the pores (not the
pores in the wicking fibers) as suction exceeds 1.1 kPa. However, due to the existence of
the wicking fibers, the deep grooves, also known as the multi-channels, remain saturated
and can continue to transport water. This explains why the slope of drying GWCC is
much flatter in the in-plane direction than in the cross-plane direction.
The other AEV of the wicking fabric, namely inner-yarn AEV, is 254 kPa and is
mainly controlled by the spacing of the deep grooves within the wicking fibers. The deep
grooves serve as inner-yarn drainage pathways that remain saturated until the inner-yarn
AEV is exceeded. As the suction further exceeds 254 kPa, air enters into the inner-
drainage pathways and the water within the deep grooves becomes disconnected. After
this point, the inner-drainage pathways will be desaturated and the ability of the wicking
fibers to transport water will significantly decrease. Therefore, the functional range of the
wicking fabric shall be between 0 kPa and 254 kPa. Within this range, the wicking fabric
is expected to work effectively to remove water out of pavement structures.
The K-functions of the AB3 and the wicking fabric could be derived from the
combination of the saturated hydraulic conductivity and the SWCC/GWCC based on the
method proposed by Kunze et al. (1968). The K-functions are expressed in Equations 6-8
(Equation 6 for the AB3, Equation the 7 for wicking fabric in the cross-plane direction,
82
and Equation 8 for the wicking fabric in the in-plane direction) and plotted in Figure 8b.
In general, the K-functions of the geomaterials remains at the saturated values as the
suction values are lower than the AEV. After the suction exceeds the AEV, the hydraulic
conductivities all exponentially decrease with increasing suctions. In addition, the
wicking fabric is able to work as a capillary barrier in the cross-plane direction while
working as a drainage geomaterial with a high efficiency in the in-plane direction. First
of all, similar to other capillary barriers, the wicking fabric has relatively larger pore sizes
compared with the AB3, and kCP decreases faster than k AB 3 . As long as the suction is
greater than 3 kPa ( kCP < k AB 3 ), the wicking fabric will impede the water flow in the
cross-plane direction and the water (especially capillary water) cannot flow through the
wicking fabric to the overlying soil layer. However, when the soil and the wicking fabric
are nearly saturated (suction ≤ 3 kPa) and the large pores of the wicking fabric are filled
with water, the kCP will become greater than k AB 3 and the barrier effect of the wicking
fabric vanishes. This critical suction value is often named “breakthrough” suction
(Stormont and Anderson 1999). In comparison, due to the existence of the wicking fibers
in the in-plane direction, the ability of the wicking fabric to hold water is improved
(Figure 8a) and the corresponding k IP is always greater than k AB 3 (Figure 8b). This fact
indicates that the wicking fabric is able to drain the water under both saturated and
unsaturated conditions in the in-plane direction. This unique feature allows the wicking
fabric to work effectively to wick water from the surrounding AB3 and continuously
transport along the wicking fabric to the face of the road embankment.
( −0.0727log
10 ( )
4
+ 0.6053log10 ( )3 −1.7158log10 ( ) 2 −0.6322log10 ( ) −5.2956 )
k AB3 = 10 (6)
83
( −0.2801log 10 ( )
3
+1.8869log10 ( ) 2 −6.5629log10 ( ) −3.0524 )
kCP = 10 (7)
( −0.3925log
10 ( )4 + 2.1792log10 ( )3 −3.2393log10 ( )2 − 2.0705log10 ( ) −3.1491 )
kIP = 10 (if   254 kPa)
(8a)
( 0.1810log10 ( )
2
−3.9730log10 ( ) −1.1708 )
kIP = 10 (if   254 kPa) (8b)
where k AB 3 , k IP , and kCP = the hydraulic conductivities of the AB3, the wicking fabric in
the in-plane and cross-plane directions, m/s; and  = suction.
5.2. THEORETICAL WATER REMOVAL CAPABILITY
The theoretical water removal capability of the soil-geotextile system can be
explained as follows. Suppose the AB3 is used as the base course in the field and is
compacted at the optimum water content of 8.5% (represented by the red dashed line in
Figure 8a). Based on the SWCC, the corresponding soil suction is 15 kPa. The post
construction of the AB3 is expected to be maintained at the optimum value and the
corresponding resilient modulus, which is the designed modulus, is 98 MPa according to
Equation 1 (the confining pressure is 34.5 kPa).
After construction, the suction gradient between the AB3 and the wicking fabric
is the major driving force under unsaturated conditions, and the AB3 tends to reach
equilibrium with the wicking fabric. At the equilibrium state, the suction values for both
the AB3 and the wicking fabric are expected to be 254 kPa, which is the maximum
suction value of the wicking fabric to effectively drain water out of pavement structures.
Given that the soil suction is 254 kPa, the corresponding soil water content can be
84
determined via the SWCC, and the value is 6.5% (represented by the blue line in Figure
8a). Compared with the designed water content (8.5%), the wicking fabric is theoretically
able to reduce the water content of the AB3 by 2% from the optimum value (from 8.5%
to 6.5%). The corresponding resilient modulus will be increased to 363 MPa (given the
same stress condition), which is over three times higher than the designed value.
Meanwhile, the permanent deformation also decreases from 2.5 mm (at the designed
water content of 8.5%) to 1.4 mm (at the water content of 6.5%).
It is important to point out that the reinforcing effects of the soil-wicking fabric
system involve two parts: mechanical reinforcement and hydraulic reinforcement.
Mechanical reinforcement is mainly provided by the high modulus polyethylene yarns.
Hydraulic reinforcement is the result of the wicking effect of the hydrophilic and
hygroscopic wicking fibers. The above discussions regarding the reinforcement effect of
the soil-geotextile system only consider the effect of hydraulic reinforcement, not
mechanical reinforcement. If mechanical reinforcement is also considered, the resilient
modulus of the AB3 is expected to be even higher. Moreover, for a pavement structure
without the wicking fabric, the post-construction water content of the AB3 will increase
with time because conventional drainage systems are not able to deal with capillary
water. The actual resilient modulus of the AB3 in the field is expected to be lower than
the designed value. Therefore, the resilient modulus difference for pavement structures
with and without the wicking fabric will be larger and the benefits of the wicking fabric
in improving the resilient modulus of the base course will be more obvious.
85
5.3. NUMERICAL VALIDATION
The objectives of numerical simulations are to evaluate the pavement structure
performance with and without the wicking fabric and to quantify the benefits of the
wicking fabric in terms of water removal. Figure 9a shows the configurations of the
numerical model. Due to the symmetrical geometry, half of a two-lane road was
simulated. The width of the roadway was 5.5 m and the road embankment had a slope of
1:3 (Vertical:Horizontal). The pavement structure is composed of a 0.1 m-thick asphalt
concrete (AC), a 0.5 m-thick base course, and a 1.0 m-thick subgrade. To simplify the
numerical model, both the base course and the subgrade were assumed to be AB3. The
pavement structures were assumed to be new or reconstructed for all models, implying
that precipitation infiltration was not considered in the model.
The material properties used in the numerical simulation are presented in Table 3.
Because the numerical models aimed to simulate the moisture migration within the base
course, the AC layer was assumed to be elastic. The resilient modulus of the AB3
depended on the stress state and the soil water content, as indicated in Equation 1. The
hydraulic properties of the AB3 are the SWCC and the K-function, representing the
ability to hold and transport water. For the wicking fabric, the modulus values were
constants and the Poisson’s ratio was assumed to be close to zero (0.001) due to the
limited deformation in the vertical direction. The hydraulic properties of the wicking
fabric also included two aspects, including the GWCCs (in-plane and cross-plane) and
the corresponding K-functions. Because the interface frictional angle was not sensitive to
water content variations, the frictional angle was assumed to be 40.3° (corresponding to a
water content of 8.5%).

86
A typical 18-kip, dual-tire single axial load (equivalent to a tire pressure of 620
kPa) was applied on top of the AC layer. At the slope of the road, a constant suction of
250 kPa was applied. The groundwater table was assumed to be at the bottom of the
subgrade layer (1.6 m from the road surface). The AB3 was assumed to be compacted at
the optimum water content (8.5%) and the corresponding initial suction condition was set
to be at 15 kPa. A 3D finite element analysis, using Abaqus Version 6.14-2 (Hibbitt et al.
2014), was established and the mesh configurations of the numerical model are given in
Figure 9b. The AC layer and the base course material were modeled using 8-node
trilinear displacement and temperature elements (C3D8T). The geotextile reinforcement
was modeled using 8-node shell, biquadratic displacement, and bilinear temperature in
the shell surface elements (S8RT). Due to the similarities between the coupled thermal
stress problem and the coupled hydro mechanical problem, Zhang (2005) proposed and
validated a new coupled hydro-mechanical model for both saturated and unsaturated
soils, and the model was used in this simulation. A coupled temperature-displacement
analysis was conducted till a steady state was obtained.
The suction, water content, and resilient modulus distributions beneath the center
of the wheel tire were exported and plotted in Figure 10. The bottom of the subgrade
layer was selected as the datum plane. Solid lines represent the distributions for the
pavement with the wicking fabric, while dashed lines represent the distributions without
the wicking fabric. The red-dashed line indicates the wicking fabric installation elevation.
As shown in Figure 10a, for the pavement structure without the wicking fabric, the
suction was linearly distributed with the highest suction (16.6 kPa) at 1.5 m and lowest
suction (0 kPa) at 0 m. In comparison, the suction distribution for the pavement structure
87
with the wicking fabric was divided into two sections. The suction values above the
wicking fabric ranged from 127 kPa to 133 kPa. Higher suction values corresponded to
lower water contents and higher resilient modulus values. Moreover, the wicking effect
was observed up to a depth of 0.4 m below the wicking fabric. For soils with elevations
lower than 0.6 m, the suction differences were less than 1.0 kPa for pavement structures,
both with and without the wicking fabric.
The water content distributions are shown in Figure 10b. For the pavement
structure without the wicking fabric, the water content distribution was highly nonlinear,
with the lowest value of 8.5% at 1.5 m and the highest value of 12.5% at 0.0 m. In
addition, the average water content above the wicking fabric (1.0 m-1.5 m) was 8.7% and
was slightly higher than the optimum value due to the capillary action. In contrast, for the
pavement structure with the wicking fabric, the average water content was 6.5%, which
was 2.2% lower than the pavement without the wicking fabric. This result further
validated the theoretical water removal ability of the soil-geotextile system discussed in
the previous section. In addition, for soils beneath the wicking fabric, the water content
differences varied from 0.9% (at 1 m) to 0.1% (0.6 m) for pavements with and without
the wicking fabric.
The resilient modulus distributions are shown in Figure 10c. For soil above the
wicking fabric (1.0 m-1.5 m), the average resilient modulus value increased from 249.8
MPa (without the wicking fabric) to 556.6 MPa (with the wicking fabric) and the
correspondent improvement factor of 2.2. As for soils beneath the wicking fabric, the
resilient modulus increased from 207 MPa to 659 MPa at 1 m and the improvement
factor was 3.2. This improvement factor was the results of the combination of both
88
mechanical and hydraulic reinforcements. As the wicking effect diminished at 0.6 m, the
improvement factor reduced to 1.4. At this elevation and below, the mechanical
reinforcement was the major factor that contributed to the resilient modulus
improvement.
In summary, the numerical simulation results further validated the ability of the
soil-geotextile system in terms of water removal and showed that the soil post-
construction water content can be reduced by 2% from the optimum value and that the
corresponding soil resilient modulus was increased by 2.2 to 3.2 times.
6. CONCLUSIONS
This paper is aimed at evaluating the improved performance of a pavement
structure installed with wicking fabrics. A series of laboratory tests were performed to
characterize the saturated and unsaturated properties of the AB3 base course, the wicking
fabric, and the interactions between the bae course and the fabric. Based on the test
results, the working mechanism of the soil-geotextile system was demonstrated and the
ability of the soil-geotextile system in terms of water removal was discussed. Numerical
models of pavement structures with and without the wicking fabric were established to
quantify the benefits of the soil-geotextile system. The major conclusions are summarized
as follows:
1. The AB3 was sensitive to water content variations. A ± 2% water content variation
from optimum value would result in a difference of 6 times in the resilient modulus
and the permanent deformation.

89
2. A series of lab tests were performed to characterize the soil and geotextile
properties under both saturated and unsaturated conditions. The soil/geotextile
abilities to hold and transport water were represented by the SWCC/GWCC and the
K-functions. The soil and geotextile unsaturated properties were highly nonlinear
and their abilities to hold and transport water decreased exponentially with
increasing suctions. The interface frictional angle between the soil and the wicking
fabric was not sensitive to water content variations.
3. Based on lab test results, the soil-geotextile system’s working mechanism and its
functional range were theoretically explained. The wicking fabric can serve as a
capillary barrier (similar to other non-wicking nonwoven geotextiles) to impede
capillary water rising upward to the base course, while working as an effective
drainage material in the in-plane direction to wick water out of pavement structures.
The deep grooves within the wicking fibers have a high specific area and a high
ability to hold and transport water under unsaturated conditions. This unique feature
allows the wicking fabric to be effective until the suction value reaches the inner-
yarn AEV of 254 kPa.
4. Theoretically, the soil-geotextile system has the ability to reduce the water content
of the AB3 by 2% from the optimum value. The corresponding resilient modulus
is expected to increase by three times.
5. Numerical models were used to evaluate the performance of pavement structures
with and without the wicking fabric. The simulation results validated the
effectiveness of the wicking fabric in terms of water removal. The average water
content above the wicking fabric was 2.2% lower than that for the pavement
90
structure without. In addition, the wicking effect could be extended to a depth of
0.4 m beneath the wicking. For soils above the wicking fabric, the resilient modulus
increased by 2.2 times. For soil beneath the wicking fabric, the resilient modulus
increased by 3.2 times (right beneath the wicking fabric), but reduced to 1.4 at 0.4
m.
In summary, the wicking fabric functions as a non-wicking capillary barrier in the
cross-plane direction and serves as an effective drainage material to hold and transport
water in the in-plane direction. The soil-geotextile system is able to reduce the water
content of the base course and improve its resilient behavior.

91
Table 1. Specifications for the new geotextile with wicking fibers (TenCate 2015)
Mechanical Properties Test Method Unit Average Roll Value
Tensile Modulus at 2%
ASTM D4595-17 kN/m 15.8
Strain (CD)
Tensile Modulus at 2%
ASTM D4595-17 kN/m 7
Strain (MD)
Permittivity ASTM D4491-17 Sec-1 0.4
Flow Rate ASTM D4491-17 l/min/m2 1222
Pore Size (050) ASTM D6767-16 microns 85
Pore Size (095) ASTM D6767-16 microns 195
Apparent Opening Size
ASTM D4751-16 mm 0.425
(AOS)
Wet Front Movement 15.2
ASTM C1559-15 cm
(24 minutes) (Vertical Direction)
(983 minutes) Zero ASTM C1559-15 cm (Horizontal
Gradient Direction)
Table 2. Summary of laboratory tests
Properties Soil Wicking Fabric Interaction

Wide-Width Tensile Large-Scale
Mechanical Resilient Modulus Test Strength Test Direct Shear
(from Specifications) Test
Constant Head Test Constant Head Test
Capillary
Pressure Rise Test
Plate Test
Hydraulic -
SWCC GWCC Pressure
Plate Test
Salt
Concentration Salt
Test Concentration
Test
92
Table 3. Material properties for numerical simulation models

Modulus (E) Poisson’s
Material SWCC/GWCC K-function
(MPa) ratio (ν)
AC 1000 0.3 N/A 0 (impermeable)
AB3 Equation 1 0.35 Equation 2 Equation 6
Equation Equation Equation
Wicking 493 218 Equation 7
0.001 3 5 8
Fabric (CD1) (MD2) (CP3)
(CP3) (IP4) (IP4)
Note:
1
CD: cross machine direction; 2MD: machine direction; 3CP: cross-plane; and 4IP: in-
plane
Groundwater Table
Figure 1. A pavement structure with the conventional drainage system
(a)
Figure 2. A pavement structure with the wicking fabric: (a) wicking fabric, and (b)
subsurface drainage design with the wicking fabric
93
(b)
Figure 2. A pavement structure with the wicking fabric: (a) wicking fabric, and (b)
subsurface drainage design with the wicking fabric (cont.)
(a)
Figure 3. Resilient modulus test results: (a) resilient modulus variations with water
contents, and (b) permanent deformation variations with water contents
94
(b)
Figure 3. Resilient modulus test results: (a) resilient modulus variations with water
contents, and (b) permanent deformation variations with water contents (cont.)
(a)
Figure 4. Determination of SWCC: (a) pressure plate test, and (b) salt concentration test
95
(b)
Figure 4. Determination of SWCC: (a) pressure plate test, and (b) salt concentration test
(cont.)
(a)
Figure 5. GWCC determination: (a) capillary rise test, (b) modified pressure plate test,
and (c) salt concentration test
96
(b)
(c)
Figure 5. GWCC determination: (a) capillary rise test, (b) modified pressure plate test,
and (c) salt concentration test (cont.)
97
(a)
(b)
Figure 6. Wicking fabric constant head test: (a) schematic plot, and (b) test setup
98
(a)
(b)
Figure 7. Large-scale direct shear test: (a) test equipment, and (b) test results
99
Inner-Yarn
AEV
(a)
(b)
Figure 8. Performance of the soil-geotextile system: (a) comparisons of SWCC and

GWCCs, and (b) comparisons of K-functions
100
(a)
(b)
Figure 9. Numerical model configurations: (a) model geometry, and (b) mesh generation
1.6
1.4
1.2
Wicking Fabric Location
Elevation (m)
1
0.8
0.6
0.4 With Wicking Fabric
0.2 Without Wicking Fabric
0
0 20 40 60 80 100 120 140
Suction (kPa)
(a)
Figure 10. Simulation results: (a) suction distributions, (b) water content distributions,
and (c) resilient modulus distributions
101
1.6
1.4
1.2
Elevation (m)
1
0.8
0.6
0.4
With Wicking Fabric
0.2
Without Wicking Fabric
0
5 6 7 8 9 10 11 12 13
WC (%)
(b)
1.6
1.4
1.2
Elevation (m)

1
0.8
0.6
0.4 With Wicking Fabric
0.2 Without Wicking Fabric
0
0 100 200 300 400 500 600 700
Resilient Modulus (MPa)
(c)
Figure 10. Simulation results: (a) suction distributions, (b) water content distributions,
and (c) resilient modulus distributions (cont.)
102
REFERENCES
AASHTO (1993). “Guide for Design of Pavement Structures.” American Association of

State Highway Transportation Officials (AASHTO), Washington, DC, USA.
AASHTO (2008). "Mechanicstic-Empirical Pavement Design Guide: A Manual of

Practice." American Association of State Highway and Transportation Office
(AASHTO), Designation: MEPDG-1, Washington, DC, USA.
AASHTO T309-99 (2003). “Standard Method of Test for Resilient Modulus of Subgrade
Soils and Untreated Base/Subbase Materials.” American Association of State
Highway and Transportation Officials (AASHTO), Washington, DC, USA.
Alim, M. A., and Nishigaki, M. (2009). “Shear Strength Behavior of Unsaturated

Compacted Sandy Soils.” Prediction and Simulation Methods for Geohazard
Mitigation: including CD-ROM, 307.
Arnold, G. K. (2004). “Rutting of Granular Pavements.” Doctoral Dissertation,

University of Nottingham.
ASTM 4491/D4491M-17 (2017). “Standard Test Methods for Water Permeability of

Geotextiles by Permittivity.” ASTM International, West Conshokocken, PA,
https://doi.org/10.1520/D4491_D4491M-17.
ASTM 4595-17 (2017). “Standard Test Method for Tensile Properties of Geotextiles by
the Wide-Width Strip Method.” ASTM International, West Conshokocken, PA,
https://doi.org/10.1520/D4595-17.
ASTM C1559-15 (2015). “Standard Test Method for Determining Wicking of Fibrous
Glass Blanket Insulation (Aircraft Type).” ASTM International, West
Conshokocken, PA, https://doi.org/10.1520/C1559-15.
ASTM D1557-12 (2012). “Standard Test Methods for Laboratory Compaction

Characteristics of Soil Using Modified Effort (56,000 ft-lbf/ft3 (2,700 kN-
m/m3)).” ASTM International, West Conshokocken, PA,
https://doi.org/10.1520/D1557-12.
ASTM D4751-16 (2016). “Standard Test Methods for Determining Apparent Opening
Size of a Geotextile.” ASTM International, West Conshokocken, PA,
103
ASTM D5321/D5321M-17 (2017). “Standard Test Method for Determining the

Coefficient of Soil-Geosynthetic or Geosynthetic-Geosynthetic Friction by the
Direct Shear Method.” ASTM International, West Conshokocken, PA,
https://doi.org/10.1520/D5321_D5321M-17.
ASTM D5856-15 (2015). “Standard Test Method for Measurement of Hydraulic

Conductivity of Porous Material Using a Rigid-Wall, Compaction-Mold
Permeameter.” ASTM International, West Conshokocken, PA,
https://doi.org/10.1520/D5856-15.
ASTM D6767-16 (2016). “Standard Test Method for Pore Size Characteristics of
Geotextiles by Capillary Flow Test.” ASTM International, West Conshokocken,
PA, https://doi.org/10.1520/D6767-16.
ASTM D6836-16 (2016). “Standard Test Methods for Determination of the Soil Water
Characteristic Curve for Desorption Using Hanging Column, Pressure Extractor,
Chilled Mirror Hygrometer, or Centrifuge.” ASTM International, West
Conshokocken, PA, https://doi.org/10.1520/D6836-16.
ASTM D6913/D6913M-17 (2017). “Standard Test Methods for Particle-Size Distribution

(Gradation) of Soils Using Sieve Analysis.” ASTM International, West
Conshokocken, PA, https://doi.org/10.1520/D6913_D6913M-17.
Bareither, C. A. and Benson, C. H. (2013). “Evaluation of Bouwer-Rice Large-Particle

Correction Precoedure for Soil Water Characteristic Curves.” Geotechnical
Testing Journal, 36(5), 1-15.
Bouazza, A., Zornberg, J. G., McCartney, J. S., and Nahlawi, H. (2006). “Significance of
Unsaturated Behaviour of Geotextiles in Earthen Structures.” Australian
Bouwer, H. and Rice, R. (1984). “Hydraulic Properties of Stony Vadose Zones.”

Groundwater, 22, 696-705.
Cedergren, H. R. (1987). “Drainage of Highway and Airfield Pavements.” Robert E.

Krieger Publishing Co. Inc., Malabar, FL.
Christopher, B. R., and McGuffey, V. C. (1997). “Pavement Subsurface Drainage

Systems.” Transportation Research Board.
Christopher, B. R., Zornberg, J. G., and Mitchell, J. K. (1998). “Design Guidance of

Reinforced Soil Structures with Marginal Soil Backfills.” Procedings of the Sixth
International Conference on Geosynthetics, Atlanta, Georgia, March, 2, 797-804.
104
Corey, A. T. (1994). “Mechanics of Immiscible Fluids in Porous Media.” Water

Resources Publication.
Currey, J. (2016). “H2Ri Wicking Fabric Experimental Feature Final Report Dalton
Highway MP 197-209 Rehabilitation.” Final Project Report, Alaska Department
of Transportation & Public Facilities (AKDOT&PF).
Delgado, I. E. (2015). “Use of Geotextiles with Enhanced Lateral Drainage in Roads over
Expansive Clays.” Doctoral Dissertation, University of Texas at Austin.
Fredlund, D. G., and Rahardjo, H. (1993). “Soil Mechanics for Unsaturated Soils.” John
Wiley & Sons.
Fredlund, D. G., and Xing, A. (1994). “Equations for the Soil-Water Characteristic
Curve.” Canadian Geotechnical Journal, 31(4), 521-532.
Fredlund, D. G., Rahardjo, H., and Fredlund, M. D. (2012). “Unsaturated Soil Mechanics
in Engineering Practice.” John Wiley & Sons.
Fredlund, M. D., Wilson, G. W., and Fredlund, D. G. (2002). “Use of the Grain-Size
Distribution for Estimation of the Soil-Water Characteristic Curve.” Canadian
Geotechnical Journal, 39(5), 1103-1117.
Giroud, J., Zornberg, J., and Zhao, A. (2000). “Hydraulic Design of Geosynthetic and
Granular Liquid Collection Layers.” Geosynthetics International, 7(4-6), 285-380.
Goldberg, R. N., and Nuttall, R. L. (1978). “Evaluated Activity and Osmotic Coefficients
for Aqueous Solutions: the Alkaline Earth Metal Halides.” Journal of Physical
and Chemical Reference Data, 7(1), 263-310.
Gu, F., Sahin, H., Luo, X., Luo, R., and Lytton, R. L. (2014). “Estimation of Resilient
Modulus of Unbound Aggregates Using Performance-Related Base Course
Properties.” Journal of Materials in Civil Engineering, 27(6), 04014188.
Hagen, M. and Cochran, G. (1996). “Comparison of Pavement Drainage Systems.”

Transportation Research Record: Journal of the Transportation Research Board
1519, 1-10.
Heckel, L. (1997). “Performance Problems of Open-Graded Drainage Layers under

Continuously Reinforced Concrete Pavements in Illinois.” Transportation
Research Record: Journal of the Transportation Research Board 1596, 51-57.
Henry, K. S. (1998). “The Use of Geosynthetics to Mitigate Frost Heave in Soils.”

Doctoral Dissertation, University of Washington.
105
Hibbitt, D., Karlsson, B., and Sorensen, P. (2014). “Abaqus 6.14-2: A Computer
Software for Finite Element Analysis.” Dassault Systems Simulia.
Jin, M. S., Lee, K. W., and Kovacs, W. D. (1994). “Seasonal Variation of Resilient
Modulus of Subgrade Soils.” Journal of Transportation Engineering, 120(4), 603-
616.
Khire, M. V., Benson, C. H., and Bosscher, P. J. (2000). “Capillary Barriers: Design
Variables and Water Balance.” Journal of Geotechnical and Geoenvironmental
Engineering, 126(8), 695-708.
Khoury, N. and Zaman, M. (2004). “Correlation between Resilient Modulus, Moisture

Variation, and Soil Suction for Subgrade Soils.” Transportation Research Record:
Journal of the Transporation Research Board 1874, 99-107.
Koerner, R. M. (2012). Designing with geosynthetics, Xlibris Corporation.
Konrad, J. M., and Morgenstern, N. R. (1980). “A Mechanistic Theory of Ice Lens

Formation in Fine-Grained Soils.” Canadian Geotechnical Journal, 17(4), 473-
486.
Kunze, R., Uehara, G., and Graham, K. (1968). “Factors Important in the Calculation of
Hydraulic Conductivity I.” Soil Science Society of America Journal, 32(6), 760-
765.
Lekarp, F., Isacsson, U., and Dawson, A. (2000). “State of the Art. II: Permanent Strain
Response of Unbound Aggregates.” Journal of Transportation Engineering,
126(1), 76-83.
Li, J. and Qubain, B. S. (2003). “Resilient Modulus Variations with Water Content.” In
Resilient Modulus Testing for Pavement Components. ASTM International, West
Conshokocken, PA.
Li, L., Liu, J., Zhang, X., and Saboundjian, S. (2011). “Resilient Modulus
Characterization of Alaska Granular Base Materials.” Transportation Research
Record: Journal of the Transportation Research Board 2232, 44-54.
Lin, C., Presler, W., Zhang, X., Jones, D., and Odgers, B. (2017). "Long-Term
Performance of Wicking Fabric in Alaskan Pavements." Journal of Performance
of Constructed Facilities, 31(2), D4016005.
Lin, C. and Zhang, X. (2018). “A Bio-Wicking System to Dehydrate Road

Embankment.” Journal of Cleaner Production, 196(20), 902-915.
https://doi.org/10.1016/j.jclepro.2018.06.053
106
McCartney, J. S., Kuhn, J. A., and Zornberg, J. G. (2005). “Geosynthetic Drainage

Layers in Contact with Unsaturated Soils.” Proceedings of the International
Conference on Soil Mechanics and Geotechnical Engineering, AA Balkema
Publishers, 16(4), 2301.
Shoop, S. A., and Henry, K. S. (1991). “Effect of a Geotextile on Water Migration and
Frost Heave in a Large-Scale Test Basin.” Transportation Research Record:
Journal of the Transportation Research Board 1307, 309-318.
Siswosoebrotho, B. I., Widodo, P., and Augusta, E. (2005). “The Influence of Fines
Content and Plasticity on the Strength and Permeability of Aggregate for Base
Course Material.” In Proceedings of the Eastern Asia Society for Transportation
Studies, 5, 845-856.
Stormont, J. C., and Anderson, C. E. (1999). “Capillary Barrier Effect from Underlying
Coarser Soil Layer.” Journal of Geotechnical and Geoenvironmental Engineering,
125(8), 641-648.
TenCate (2015). “Mirafi H2Ri Woven Geosynthetic Specifications.” <

https://www.tencategeo.us/media/39ff27e9-96f5-4450-9aaa-
bd9c1cce3160/yV_goA/TenCate%20Geosynthetics/Documents%20AMER/Produ
ct%20Description%20Sheets/Woven%20Product%20Description%20Sheets/PDS
_H2Ri0417.pdf> (April 26, 2018).
Tripathy, S., Leong, E., and Rahardjo, H. (2005). “Suction of Compacted Residual Soils.”
Unsaturated Soils: Experimental Studies, Springer, 111-122.
Uzan, J. (2004). “Permanent Deformation in Flexible Pavements.” Journal of

Transportation Engineering, 130(1), 6-13.
Witczak, M. (2003). “NCHRP 1-28A: Harmonized Test Methods for Laboratory

Determination of Resilient Modulus for Flexible Pavement Design.” Final Report,
Transportation Research Board of the National Research Council, Washington,
DC, USA.
Witczak, M., Andrei, D., and Houston, W. (2004). “Guide for Mechanistic-Empirical
Design of New and Rehabilitated Pavement Structures.” Transportation Research
Board of the National Research Council, 1-91.
Yang, S. R., Huang, W. H., and Tai, Y. T. (2005). “Variation of Resilient Modulus with
Soil Suction for Compacted Subgrade Soils.” Transportation Research Record:
107
Yang, S. R., Lin, H. D., Kung, J. H., and Huang, W. H. (2008). "Suction-Controlled
Laboratory Test on Resilient Modulus of Unsaturated Compacted Subgrade
Soils." Journal of Geotechnical and Geoenvironmental Engineering, 134(9), 1375-
1384.
Yilmaz, A., and Sargin, S. (2012). "Water Effect on Deteriorations of Asphalt

Pavements." Online J. of Sci. and Tech, 2(1).
Zhang, X. (2005). “Consolidation Theories for Saturated-Unsaturated Soils and

Numerical Simulation of Residential Buildings on Expansive Soils.” Doctoral
Dissertation, Texas A&M University.
Zhang, X., and Presler, W. (2012). “Use of H2Ri Wicking Fabric to Prevent Frost Boils
in the Dalton Highway Beaver Slide Area, Alaska.” Final Report No. 12.23,
Institute of Northern Engineering/Alaska University Transportation Center
(INE/AUTC).
Zornberg, J., Bouazza, A., and McCartney, J. (2010). “Geosynthetic Capillary Barriers:
Current State of Knowledge.” Geosynthetics International, 17(5), 273-300.
108
III. LABORATORY EVALUATION OF DIFFERENT GEOSYNTHETICS FOR

WATER DRAINAGE
Yipeng Guo1, Chuang Lin2, Wuming Leng3, and Xiong Zhang 4
1
Graduate Research Assistant, Department of Civil Engineering, Central South
University, Changsha, Hunan, 410075, China, Email: [email protected]
2
Graduate Research Assistant, Department of Civil, Architectural and Environmental
Engineering, Missouri S&T, 135 Butler Carlton Hall, 1401 N. Pine Street, Rolla, MO
3
Professor, Department of Civil Engineering, Central South University, Changsha,
Hunan, 410075, China, Email: [email protected]
4
(Corresponding Author) Associate Professor, Department of Civil, Architectural, and
Environmental Engineering, Missouri S&T, 135 Butler Carlton Hall, 1401 N. Pine Street,
Rolla, MO 65409-0030, Email:[email protected]
ABSTRACT
In nearly all engineering projects, soils are designed to be compacted at the
optimum water content to achieve the maximum dry density and best performance. Once
laid in the field, compacted soils unavoidably experience increase in water contents due
to capillary rise, rainfall infiltration, and other factors. Conventional drainage designs
using granular materials and geosynthetics rely on materials with large pores to drain
“free” water while capillary water cannot be drained. Numerous studies indicated that a
small variation in water content can lead to significant reductions in soil stiffness and
increments in permanent deformation. Recently two new types of geosynthetics (drainage
belt and wicking geotextile) were used in several field conditions as drainage materials
109
and the results seemed promising. However, most of the field tests are inconclusive due
to complicated site conditions and non-uniform soil distributions in the field. The relative
performances of these drainage geosynthetics and their working mechanisms are largely
unclear. In this study, laboratory model tests were conducted to quantify the cumulative
amount of water drained by four different drainage situations, including poorly-graded
sand as the control case and poorly-graded sand with drainage belt, with wicking
geotextile, and with modified wicking geotextile under controlled laboratory conditions.
The volumetric water content of soils was monitored by moisture sensors installed at
different locations with the testing box, and then the water contents of soils under
different drainage situations were evaluated and compared. Finally, the working
mechanisms of different drainage materials were discussed.
KEK WORDS: Geosynthetics, drainage belt, wicking geotextile, capillary water,
water retention curve, unsaturated soil
1. INTRODUCTION
Water within pavement layers is a principal cause of pavement deterioration
(Cedergren, 1994; Christopher and McGuffey, 1997; Henry and Holtz, 2001). Specific
problems associated with water are stripping of asphalt pavement, joint displacement in
concrete pavements, reduction in pavement strength caused by excess water in the base
course layers, shrinking and swelling of subgrade materials caused by water content
changes, and frost heave and thaw weakening caused by capillary water flow beneath
pavements. Water related problems are thus responsible for decreased pavement life,
110
increased costs for maintenance, and increased pavement roughness, and they occur to
some extent throughout all regions and climates of the United States. A recent NCHRP
study estimated that excess water reduces the life expectancy of pavement systems by
more than half (Christopher and McGuffey, 1997). Government transportation engineers
in cold regions have credited a minimum of half of road maintenance expenditures to the
effects of freezing and thawing (Henry and Holtz, 2001). Figure 1 shows the influences
of water content on the resilient modulus and the permanent deformations for a base
course material (AB3) in Kansas (Lin et al., 2018). The material has an optimum water
content of 8.9% (design value for pavement) and water content of 12% at full saturation.
As shown in Figure 1a, at the confining pressure of 68.8 kPa and water contents of 10.8%
(1.9 % higher than the optimum content), the resilient modulus decreased by 46% of the
design value. On the other hand, at water content of 6.9% (2.0% below the optimum
content), the resilient modulus increased by 250%. Meanwhile, the permanent
deformations increased by 4.8 times compared with the design value if the water content
was 10.8%. In contrast, the permanent deformations reduced to 56% of the design value
when the water content was 2.0 % lower than the optimum content, as shown in Figure
1b. In other words, a small amount of water content variations could lead to significant
reduction in the resilient modulus and increment in the permanent deformation.
Geosynthetics are commonly used for reinforcement and drainage purposes.
Eextensive field and laboratory experimental works have been conducted to investigate
the working mechanism of geosynthetics for reinforcement and drainage purposes. A vast
of geosynthetics products and design guidelines were proposed (Gray and Al-Refeai,
1986; Athanasopoulos, 1993; Ghosh and Madhav, 1994; Gobel et al., 1994; Zornberg and
111
Mitchell, 1994; Cancelli et al., 1997; Perkins, 1999; Som and Sahu, 1999; McGown et
al., 2004 Stulgis, 2005; Pathak and Alfaro, 2010). When used for reinforcement purposes,
geosynthetics are often used with soils compacted at the optimum water content to
achieve the maximum dry density. When used for drainage purposes, conventional
geosynthetics functions under saturated conditions and rely on the large pores to drain
gravitational water (or “free” water). Between the full saturation and optimum water
content, there is a large amount of capillary water which cannot be drained by the
conventional geosynthetics. (Ling et al. 1993; Mitchell and Zornberg 1995; Tan et al.
2001; Iryo and Rowe 2005; Garcia et al., 2007; Benjamin et al. 2007; Noorzad and
Mirmoradi 2010; Raisinghani and Viswanadham, 2011; Portelinha, 2013; Wang et al.,
2017). Most existing drainage systems can only drain “free” water but not capillary water
(Christopher et al., 2010). As shown in Figure 1, increasing in water content can lead to
significant reductions in the resilient modulus and increments in the permanent
deformation. Failing to drain capillary water is the major cause of geotechnical failures
and pavement distresses (Mitchell and Zornberg 1995; Koerner and Soong, 2001; Stulgis
2005; Yoo and Jung 2006).
Recently, some researchers attempted to use new types of geosynthetics to drain
capillary water for better performance. For example, Henry et al. (2002) proposed a
Geocomposite Capillary Barrier Drain (GCBD) to limit moisture changes in pavements.
The GCBD use a hydrophilic fiberglass with a higher specific surface area to absorb
water and transport capillary water under a low suction level. Similar to a conventional
drainage system, the GCBD relies on the gravity to generate hydraulic gradient to drain
the capillary water. It only works for a small suction range which is less than 6 kPa
112
(equivalent to 0.6 m water head). Leng et al. (2017) used plastic sheets with Ω-shaped
grooves (hereafter named as “drainage belt”) to drain water in the roadside slopes along
the railway system in China, as shown in Figure 2a. Field applications indicated that
drainage belt could effectively lower the groundwater table without any clogging. Guo at
al. (2018) demonstrate the drainage efficiency of the drainage belt via laboratory tests.
Recently, a new wicking geotextile with wicking fibers (hereafter named as “wicking
geotextile”) was proposed to dehydrate both gravitational and capillary water under both
saturated and unsaturated conditions, as shown in Figure 2b. The wicking geotextile is a
dual-function woven geotextile for drainage and reinforcement purposes. The wicking
fabric yarns are made of hydrophilic, hygroscopic 4DG TM nylon fibers with
multichannel deep grooves for drainage purposes, as shown in Figure 2b. The
polypropylene fiber yarns have high tensile strength and are used for reinforcement
purposes. Detailed description of the wicking geotextile can be found in the Test Material
section. Zhang and Belmont (2009) reported that the wicking geotextile showed better
drainage performance compared with conventional non-wicking geotextile. In addition,
Lin and Zhang (2018a) carried out wetting and drying tests to evaluate the drainage
performance of the wicking geotextile with different types of soils. found that both
gravitational and capillary water could be drained out by the wicking geotextile.
Furthermore, Lin and Zhang (2018b) also proposed a “bio-wicking” system to further
improve the drainage efficiency. In addition, the wicking geotextile has been successfully
used to mitigate frost heave for Alaskan pavements and the roadway installed with the
wicking geotextile successfully eliminates the frost heave problem (Zhang et al., 2014;
Lin and Zhang, 2017).

113
The results from the above examples seem promising, but most of them are
inconclusive due to complicated site conditions and the nonuniformity of the soil in the
field. The relative performances of these drainage geosynthetics and their working
mechanisms are largely unclear. The objectives of this study were to evaluate and
compare the working mechanisms of different geosynthetics. Laboratory model tests
were conducted to quantify the amount of water drained under four different drainage
situations, including a poorly-graded sand as the control case (without any geosynthetics)
and sand with three different types of geosynthetics, including drainage belt, wicking
geotextile, and modified wicking geotextile. The soil volumetric water contents were
monitored by moisture sensors installed at different locations of a testing box. In
addition, the cumulative amount of water collected from each testing box was also
determined. Finally, the working mechanisms of different geosynthetics as drainage
materials were discussed.
2. TEST MATERIALS AND TEST PROCEDURES
2.1. TESTING MATERIALS
The soil used in the study was Missouri river sand, with the soil particle size
ranged from 1.19 mm to 0.42 mm. The specific gravity (Gs) of the soil is 2.65. Cu, 1.42;
and Cc, 0.93. According to the USCS (Unified Soil Classification System) (ASTM
D2487), the soil can be classified as poorly graded sand (SP). In this paper, sand is
purposely selected to demonstrate that even for coarse-grained soils which are
114
conventionally used as drainage materials, it can still hold large amount of capillary water
under unsaturated conditions which cannot be drained by gravity.
Three different drainage materials were used in this study, including the drainage
belt (Figure 2a), the wicking geotextile (Figures. 2b and 2c), and a modified wicking
geotextile (Figures. 2d and 2e). Detailed descriptions regarding the three geosynthetics
are discussed as follows. For the drainage belt, it is a 2 mm-thick plastic sheet with Ω-
shaped drainage grooves on one side of the sheet as shown in Figure 2a. The drainage
grooves have circular shapes with inner diameters of approximately 1.0 mm and the
spacing between two adjacent grooves is 1.5 mm. The opening size for drainage is about
0.3mm. Due to the relatively small opening size, soil particles with large sizes will not
clog the drainage paths while “free” water can be quickly drained out through the grooves
(Guo et al. 2018). Note that the drainage belt can only drain “free” water while capillary
water cannot be drained by this geosynthetics.
Figure 2b shows the images of the wicking geotextile. The wicking geotextile is a
dual-function woven geotextile for reinforcement and drainage purposes. Similar to other
non-wicking geotextiles, the polyethylene yarns have high tensile strength and are used
for reinforcement purposes. More importantly, the wicking fiber yarns are made of
hydrophilic, hygroscopic 4DG TM nylon fibers for drainage purposes. As shown in
Figure 2b, the average diameter of the wicking fiber is between 30 μm and 50 μm and the
average opening of the grooves is between 5 μm and 12 μm (Lin and Zhang, 2018 a, b).
The multichannel cross-section has a high shape factors and great number of channels per
fiber (specific surface area = 3650 cm2/g), which give wicking fabric great potential for
maximizing capillary action and water transport in an unsaturated environment. Figure 2c

115
shows the schematic plot of the profile for the wicking geotextile, which comprises a
double layer fabric formed from a single weave. The fabric comprises a weft yarn in the
second layer, a weft yarn in the first layer, wicking fibers woven in the weft direction of
the fabric, and a warp yarn interweaving the first and second weft yarns and the wicking
fibers as shown in Figure 2c. The weft yarn is a high modulus tape comprising an
admixture of polypropylene and a polypropylene/ethylene copolymer for reinforcement
purpose. The specifications of the wicking geotextile are shown in Table 1. Note that the
water is capable of flowing through the wicking geotextile at a rate of 1222 L/min/m2.
Figure 2d shows the images of the in-house fabricated modified wicking
geotextile, in which additional nylon wicking fibers were artificially knitted into the
original wicking geotextile. Figure 2e shows the schematic plot of the profile for the
modified wicking geotextile. The modified wicking geotextile is exactly the same as that
in Figure2b or 2c except that an additional thread of nylon wicking fibers was added to
the middle of the adjacent wicking fibers (as shown in Figure 2e).
2.2. EXPERIMENT SETUP
Figures 3a-3b show the schematic plot and configuration of the experimental
testing box. The testing box frame was built with aluminum alloy to provide strong
support for the testing materials. The sidewalls were made of transparent acrylic to
facilitate the observation of the moisture migrations within the testing box. The
dimensions of the testing box were 132 cm ×60 cm ×20 cm (Length × Height × Width).
In order to compare drainage efficiency of different geosynthetics at the same time, the
test box was divided into two sections, the dimensions for those two sections were 66 cm
116
×60 cm ×20 cm (Length × Height × Width), respectively, as shown in Figure 3a. In all
tests, the soils were compacted to a height of 48 cm with an initial void ratio of 0.64 for
all tests. The corresponding saturated permeability of coefficient of the soils 6.5 ×10-4
cm/s. Two valves were installed at the bottom of the left and right side of the box for
water supply. Meanwhile, there were two openings (20cm (W) × 2 mm (H)) located at
each side of the sidewalls of the test apparatus with an elevation of 10 cm from the
bottom. These openings were designed to install the drainage materials. Rectangular grids
with dimensions of 3 cm × 3 cm were marked on the front acrylic wall to facilitate the
observation of the drainage performance.
2.3. INSTRUMENTATION
METER Environment® EC-5 Moisture sensors were installed in the testing box
to monitor the drainage performance of different materials. In total three layers of soil
moisture sensors were installed at depths of 10 cm, 24 cm, and 38 cm from the bottom,
respectively as shown in Figure 3a. The first layer of sensors (10cm from the bottom) was
installed at the level of opening and immediately above the geosynthetics (if there is). In
each test section, there were three sensors in each layer. As a result, in total 9 sensors
were installed in each test section as shown in Figure 3a. Within each layer, the spacing
between two adjacent sensors was 28 cm, and the sensors at the edges were 5 cm away
from the sidewalls. Note that a ±2% accuracy of the sensor was provided by the
manufacture for non-uniform soils and general field conditions, which represent accuracy
under disadvantageous field conditions. In this study, the sensors were used in a well-
controlled and stable laboratory condition with nearly constant temperature, relative
117
humidity, nearly zero vibration, and very short connection wires. In addition, the soils
used were uniform sand and were carefully prepared. Therefore, the authors assume that
the sensors can provide sufficient accuracy of the soil moisture variations with time
during the tests. In addition, after the test has been completed, a total 54 samples were
taken from each testing box. The volumetric water content was back calculated based
upon the measured gravimetric water content and the specific gravity of the sand. It was
found that the monitored and calculated results matched well with each other, indicating
that the sensor readings were reasonable and accurate. This can be considered as an
indirect calibration. The data acquisition system was composed of a Campbell
Scientific® CR1000 datalogger together with an AM16/32 multiplexer. The soil
volumetric water contents were monitored and recorded every 2 minutes.
2.4. EXPERIMENTAL DESIGN
To evaluate the performance of the four different drainage conditions, drainage
tests were performed in the testing box as shown in Figure 3. Each time two materials can
be compared side by side. Three groups of comparison tests were performed, including
the sand without any drainage material (T1), with drainage belt (T2), with wicking
geotextile (T3), and with modified wicking geotextile (T4). T1 was used as a control case
to evaluate the performance of a drainage system without any geosynthetics. Sand is
purposely selected to demonstrate that even for coarse-grained soils, which are
conventionally used as drainage materials, can still hold large amount of capillary water
under unsaturated conditions which cannot be drained by gravity. T2 was a new drainage
material with an efficient performance in removing “free” water and was commonly used
118
in dealing with poor drainage of existing railways in China (Guo et al., 2018). Lin et al.,
(2018 a, b) reported that the wicking geotextile (T3) can be used to drain both
gravitational and capillary water. T4 was used to demonstrate the influence of the number
of wicking fiber yarns on the drainage efficiency. Each test had three replicates so that
the representative test results can be obtained as reported in the following sections. Since
the test results of the three replicates for each test were consistent, only the representative
set of the test results were presented in this paper to keep the paper concise and avoid
unnecessary over-length.
2.5. TESTING PROCEDURES
All tests were carried out in the temperature and humidity-controlled room. The
relative humidity varied from 48% to 58%, with an average value of 53% and the
temperature varied from 19 °C to 22 °C, with an average value of 21°C. Each test had
three different stages. At the first saturating stage, the soils in the entire testing system
were saturated. During this stage, water was slowly supplied to the testing box through
two valves at the bottom of the box (refer to Figure 3), and gradually saturated soils from
the bottom to the top. The water flow rate was carefully controlled to prevent upward
seepage force from disturbing the tested soils. The soils were considered fully saturated
after a 10 cm water head was ponding above the soil surface. In order to achieve full
saturation, for T1, the opening at the sidewall was blocked in T1 and no water was
allowed to drain out of the system. Since water can be drained out from the openings in
T2 via the opening, the drainage belt was lifted up to minimize the amount of water
flowing out and the tested sand in T2 could be gradually saturated. For T3 and T4, the
119
wicking geotextiles could be left as they were as shown in Figure 3b during the saturation
process.
After the saturation process was completed, the valves at the bottom of the testing
box were shut down and the second stage of free draining started. Water was allowed to
be drained out of the system and the amount of “free” water was collect during the
draining process. Then the top of testing box was covered by plastic wrap to prevent any
possible water loss through evaporation. The drainage materials (if there was), were
naturally hung downwards by gravity. The water collection process started when the
water above the soil surface was at a level of 3 cm. The collected water was weighed
continuously using the digital balance until no free water flowed out from the materials.
This process normally took about six days.
The second stage of the test was completed when there were no visible water
drops could be observed. At the third stage of draining capillary water, the T2, T3, and
T4 (if there was) were elevated to the horizontal direction and exposed to room
atmosphere to allow further water loss through evaporation. The third stage was
completed when all the geotextiles were dried to a distance of 20 cm to the openings.
This process normally took about additional 4.9 days. During the entire three draining
stages, the soil volumetric water contents were monitored every 2 minutes.
After all the three stages were completed, the testing system was disassembled
and soils at different locations were carefully sampled to measure their gravimetric water
contents using the oven method. Solid points in Figure 3a show the locations of the
sampling. From the top to the bottom, a total 9 layers of soils were sampled with an
interval of 6 cm. In each layer, six locations were selected for sampling and the sampling
120
distance between two adjacent locations were 12, 15, 12, 15, and 12 cm (refer to Figure
3a), respectively. As a result, a total of 54 samples were taken for each test section and a
total of 108 gravimetric water contents were measured for the whole testing box for two
test sections.
3. RESULTS AND DISCUSSIONS
In this section, the discharge capacity, the volumetric water content, and
gravimetric water content variations were discussed. The recorded data was also used to
generate the volumetric moisture contours to facilitate the visualizations of water
migrations under each situation.
3.1. GENERAL VISUAL OBSERVATIONS FOR WATER DRAINAGE
During the test, the water drainage process was visually observed and summarized
as follows. In general, water drained out of the system very fast for T1 and T2 within the
first 2 hours when the flow of “free” water dominated. However, after that the water flow
quickly reduced as the soil became unsaturated after 2-3 hours, no measurable amount of
water could be collected from T1 and the draining process ceased at this moment. In
comparison, “free” `water could be discontinuously collected from T2 till 6838 minutes
(4.7 days), indicating that T2 was partially working under unsaturated condition. In
contrast, water could be continuously collected from T3 and T4 for approximately 4.9
days and no visible water could be collected after that. However, the wicking geotextile
121
was still damp 10 days after the test, indicating that the wicking geotextile was still
functional and was able to drain both “free” and capillary water.
Figure 4 shows representative photos for tests on T1, T2, T3, and T4 after 10 days
of drainage before being disassembled. As discussed previously, the testing box allowed
to test two different materials side-by-side. Figures 4a-c show the comparisons on T1-T2,
T2-T3, and T3-T4, respectively. To facilitate the observations, the locations of the
drainage materials and the wetting fronts were also labeled by solid and dotted lines,
respectively. Figure 4 clearly indicates that the geosynthetic drainage materials showed
different performances. When there were no drainage materials (T1 with sand only), the
final wetting front was about 8 cm above the elevation of the opening, indicating that the
capillary rise could be as high as 8 cm. For T2 and T3, the final wetting fronts were 5 and
2 cm above the opening elevation after 10 days of drainage, respectively. For T4, the
final wetting front was 2 cm below the opening elevation, indicating that the modified
wicking fabric can siphon the water from the underlying soil.
3.2. VARIATIONS OF VOLUMETRIC WATER CONTENT
Three sensors (M8, M5, and M2) at the center of the test section were selected to
demonstrate the variations of volumetric water content at different locations with time, as
shown in Figure 5.
Figure 5a shows the variations of volumetric water content with time in the log
scale at sensor M8, which was at 10 cm from the soil surface. As can be seen from
Figure5a, initially the soil was dry with volumetric water contents of nearly zero for all
cases. After water was supplied to saturate the soils, the volumetric water contents of the
122
soil quickly increased to a maximum value of between 0.355 and 0.362. When the soils
were allowed to drain, the volumetric water contents quickly dropped to approximately
0.074 in less than 2 hours. Among the four tests, water was drained faster in T1 (sand
only) and T2 (drainage belt) than those in T3 and T4. It seemed that T4 was the slowest
to drain water. The volumetric water contents in all tests kept dropping slowly after 2
hours and nearly all reached the same value of 0.067 after 10 days of free drainage.
Figure 5b shows the variations of volumetric water content with time at sensor
M5, which was at 24 cm from the soil surface. When soils were allowed to drain, water in
the sand (T1) drained much faster than the other materials. In about 30 minutes, the water
content dropped from 0.369 to 0.15 and remained relatively constant after that. The
draining process after 2 hours was very slow and at the end of 10 days, the final water
content was 0.111. Although water drainage at M5 in the drainage belt (T2) was similar
to than T1, it was clear that more water can be drained by T2. In less than 2 hours the
volumetric water content dropped from saturation of 0.369 to 0.083 and the final water
content was 0.074. In comparison, water drainage in T3 and T4 was much slower. It took
about 0.69 day for the water content to drop from saturation (0.369) to approximately
0.074 and remained nearly constant after that. After 10 days, the volumetric water content
in T4 was 0.082, slightly higher than that in T2 and T3 at the same location.
Figure 5c shows the variations of volumetric water content with time at sensor
M2, which was exactly at the level of openings. If there were geosynthetics, the sensors
were slightly above the geosynthetics. As can be seen from Figure5c, the volumetric
water content in T1 and T2 remained relatively constant during the whole testing periods
except a slight decrease in at the first two hours. Laboratory visual observations during
123
the testing period and disassembling also confirmed that the soils were fully saturated.
The volumetric water contents in T3 and T4 varied with time in a way similar to those in
T1 and T2 for a few days at the beginning and remained saturated. After 5.5 and 2.5 days
respectively, the volumetric water contents in T3 and T4 started to decrease quickly,
indicating the soils were losing saturation. The trends did not flatten out even at the end
of 10 days. Although there was no visible water droplet drainage, visual observations
indicated the exposed T3 and T4 remained damp after 10 days, implying that water can
still be lost if the testing period had been longer.
3.3. VOLUMETRIC WATER CONTENT CONTOURS FOR DIFFERENT TESTS
Figure 5 can only provide information on water drainage at individual points at
different times. However, it is more meaningful to examine the water content changes in
the whole tested soil domain to have comprehensive evaluations of the performance of
different drainage materials. Volumetric water content contours, which connect the points
that share the same value of volumetric water content, were created using Matlab
(MathWorks 2016) for this purpose. A meshgrid was first generated and then the
recorded water contents at different locations were used as controlling points to linearly
interpolate the water contents at the interested points. After that, the points with the same
water contents were connected to generate moisture contours and the moisture contours at
different times were processed into animation movies to dynamically illustrate the
moisture variations. Figure 6 shows the snapshots of the movies at different times for
different drainage materials. However, it should be noted that the 9 sensors only detected
the water contents at 9 discrete points, while the water content contours just illustrated
124
the monitored scope of the entire testing system. Since interpolation unavoidably can
bring errors, therefore it is more meaningful to present the entire variations of water
content than a specific value in the water content contours. The following discussions are
based on the snapshots of the moisture contour animations. Christopher et al. (2010)
classified the quality of drainage into five types (excellent, good, fair, poor, and very
poor) based upon water removal time (2 hours, 1 day, and 1 week, 1 month, and does not
drain). Therefore, in the following discussions, the volumetric water content contours
were also plotted at starting point, 2 hours, 1 day, and 7 days, respectively to evaluated
drainage efficiency of those four different drainage situations.
In Figure 6, the red lines represent locations of the drainage materials (opening
location and T2, T3, T4) and the arrows indicate the direction of water flow. As can be
seen from Figs 6a-d, at the starting point of different testing situations, the volumetric
water contents in all figures were slightly and monotonically increasing from 0.356 at the
top to 0.371 at the elevation of the opening. As discussed previously, at the initial
conditions, the water table was 3 cm above the soil surface and sand in the entire testing
box was saturated. The slightly larger volumetric water contents at the bottom imply
larger void ratio and looser soils. This might be attributed to the fact that the sand was
saturated from bottom to top and upward seepage force did lead to a looser soil structure
at the lower elevations. However, the overall variations were small and the soil can be
considered as uniformly distributed at the initial conditions.
Figure 6a shows the water content contours with time for T1, which was sand
without any types of drainage materials. As can be seen in Figure 6a, the entire soil
volumetric water content decreased quickly in the first two hours. By comparing the
125
volumetric water content contours at 2 hours with those at 1 day and 7 days, it can be
concluded that the volumetric water content contours did not change very much with
time. In other words, most of the “free” water has been drained out of the soil in two
hours, which was also consistent with the visual observations. There was one
phenomenon in Figure 6a deserving special attention. First, as can be seen from Figure
6a, the volumetric water content at the elevation of the openings was largest at the initial
condition (0.369). It decreased to 0.329 after 2 hours, and then slightly increased with
time: 0.33 after 1 day and 0.336 after 7 days. Since soils at the elevation of the opening
were always saturated or near saturation, the only explanation for the decrease in
volumetric water content was that the soils were compacted to a denser state due to
downward seepage force, increasing the vertical effective stress due to water drainage in
the first two hours. The increase in the volumetric water content at the elevation of the
openings after 2 hours was due to internal redistribution of water with time in the vertical
direction since no water was drained out of the opening after 2 hours. In other words,
after two hours, the water in T1 was still moving downwards. However, since no water
was drained out of the testing box in T1 after 2 hours, the downward moving water was
accumulated above the elevation of the opening, causing the water content increased (or
increased in the degree of saturation) in the range from 10 cm to 24 cm above the bottom.
Figure 6b shows the water content contours with time for T2, which was installed
with the drainage belt. The volumetric water content varied from 0.371 to 0.361 at the
starting point, indicating the soil was relatively uniform. After 2 hours, the volumetric
water content at the elevation of the opening decreased to 0.327, which was comparable
to the monitored results for T1 (0.329). In addition, the volumetric water content was
126
0.104 for T2 at an elevation of 22 cm after 2 hours, which was approximately 15% lower
than that for T1. Therefore, the average volumetric water content for T2 was lower than
T1, and more water was drained out T2 within 2 hours due to the installation of the
drainage belt. The Ω-shape drainage grooves shortened the lateral drainage path and were
able to generate higher hydraulic gradient when given the same water head. In
comparison, the water had to flow through the voids among soil particles for T1 which
was not as effective as T2. The drainage belt was more effective in draining “free” water
compared with T1 which was not installed with any drainage materials. After 2 hours, the
“free” water was considered to be all drained out and the volumetric water content
variations were limited during the time period from 2 hours to 1 day. For example, the
volumetric water content at an elevation of 22 cm slightly decreased from 0.104 to 0.103.
Different from the monitored results for T1, at the elevation of the opening, the
volumetric water content continued to decrease from 0.327 at 2 hours to 0.323 at the end
of 1 day. From 1 day to 7 days, the volumetric water contents did not change, indicating
that no excess water could be drained out. In summary, the drainage belt is more effective
in draining “free” water compared with T1, but was not able to drain capillary water.
Figure 6c shows the volumetric water content contours with time for T3, which
was installed with the wicking geotextile. The volumetric water content at the beginning
of the test varied from 0.362 to 0.37, which was consistent compared with that for T1 and
T2. After 2 hours, the volumetric water content at an elevation of 22 cm was 0.148,
which was about 0.02 higher than T1 and 0.04 higher than T2, respectively. This
phenomenon was caused by the reduced opening area at the sidewall. The wicking
geotextile installed at the drainage opening reduced the cross-sectional area for which
127
could have been used for water flow in T1 and T2. The excess water in T3 could flow out
of the system only through the pores within the weaving structure of the wicking
geotextile, which was not effective. However, the volumetric water content at the
elevation of the opening (10 cm) decreased from 0.37 to 0.322 within the first 2 hours,
continuously decreased to 0.314 after 1 day, and further reduced to 0.288 after 7 days.
Note that “free” water was considered to be drained out of the system within the first 2
hours and the decreasing volumetric water content in T3 after 2 hours indicated that the
wicking geotextile was able to drain capillary water out of the system. In addition, the
volumetric water content at an elevation of 22 cm was 0.09 (the lowest among T1-T3),
further validating that the drainage efficiency of the wicking geotextile was better than
that for the drainage belt.
Figure 6d shows the volumetric water content contours for T4. Compared with
T3, additional nylon wicking fibers were artificially knitted into the original wicking
geotextile. Within the first 2 hours, the volumetric water contents for T4 were
comparable with those for T3, with 0.327 at an elevation of 10 cm and 0.09 at an
elevation of 22 cm. This test result indicated that the drainage performance of the
wicking geotextile (T3) and the modified wicking geotextile (T4) were comparable in
terms of draining “free” water. However, the drainage performance of T4 was better than
T3 in draining capillary water. For example, the volumetric water content in T4
decreased to 0.065 at an elevation of 22 cm after 7 days, which was 0.03 lower than that
for T3. Moreover, the volumetric water content in T4 was 0.134 at an elevation of 10 cm,
which was 0.193 less than that for T3 (decreased by 59%). The additional wicking fibers
enhanced the ability of the wicking geotextile to hold and transport water under
128
unsaturated conditions and the volumetric water content was the lowest for T4 after 7
days. Therefore, the overall drainage performance for T4 was also the best among all four
cases.
Comparisons of the drainage performances of T1-T4 concluded that the drainage
belt showed better performance in draining “free” water compared with the control case
while the wicking geotextile was able to drain both “free” and capillary water, resulting
in the best overall performance.
3.4. CUMULATIVE AMOUNT OF WATER
Figure 7a shows the variations of cumulative amount of water collected from
stages 2 and 3. Figure 7b shows an enlargement of Figure 7a for the time range from 0 to
200 minutes and will be discussed later in this section. As can be seen from Figure 7a, the
cumulative amount of water collected from T1 was 17.34 L within 2 hours, then slightly
increased to 17.67 L at 4620 minutes (3.2 days), and no measurable amount of water
could be collected afterward. This phenomenon indicated that the drainage process
stopped at 2 hours and T1 was not able to drain capillary water. For T2 which was
installed with the drainage belt, the cumulative amount of water collected from the
system was 17.72 L within 2 hours, slightly increased to 18.1 L after 1 day, and finally
ended at 19.3 L at 6838 minutes (4.7 days). Note that starting from 5358 minutes (3.7
days), no continuous water flow was observed at the drainage openings. The cumulative
amount of water was discontinuous, implying that the drainage belt was partially
functional in draining water out of the system. The total cumulative amount of water
from T2 was 9.5% higher than that for T1, indicating that the drainage belt was more
129
effective than the control case in draining “free” water. For T3 which was installed with
the wicking geotextile, the cumulative amount of water was 17.2 L within 2 hours,
gradually increased to 19.6 L after 1 day, and eventually reached 20.2 L at 4.9 days. Test
results showed that the wicking geotextile was not as effective as the drainage belt in
terms of draining “free” water. As discussed in Figure 6c, the drainage opening on the
sidewall was partially obstructed by the geotextile and the effective cross-sectional area
for T3 became smaller compared with that for T1 and T2. However, the total cumulative
amount of water from T3 was 14.5% higher than that for T1 and 4.7% than T2. This
means that the wicking geotextile was able to drain out more water if the testing time was
long enough. In other words, the overall drainage performance of the wicking geotextile
was better than the drainage belt. For T4 which was installed with the modified wicking
geotextile, the cumulative amount of water was 17.5 L within 2 hours, increased to 20.8
L after 1 day, and reached 21.5 L at 4.9 days. As discussed in Figure 6d, the additional
wicking fibers did not aim at increasing its drainage efficiency for “free” water, but of
more capable to drain capillary water, resulting in a 6% more water collected from T4
than in T3 after 4.9 days. In summary, the total cumulative amount of water for T2, T3,
and T4 were 9.45%, 14.54%, and 21.51% more than that of T1. The high-to-low rank of
the overall drainage performance was T4, T3, T2, and T1.
As can be seen from Figure7b, drainage of “free” water was very fast at the
beginning of the test for all materials, and the slopes of collected water versus time
curves (flow rates) were very steep. At the beginning of the test (0-25 minutes), the
cumulative amount of “free” water from T1 (17.03 L) was higher than those from T2
(17.0 L), T3 (8.60 L), and T4 (10.01 L). This was mainly attributed to its larger cross-
130
sectional area for “free” water drainage. For all the other materials, part of the openings
was occupied by the drainage materials. As a result, although the area of the opening on
left/right sidewalls of the box was the same, the equivalent drainage areas for all the other
geosynthetics were smaller compared to T1. After about 26 minutes, the water collected
in T2 (17.04 L) was the same as that collected in T1, and exceeded the cumulative
amount of water by T1 after that. At the same time, the cumulative amount of water in T3
and T4 were 8.85 L and 10.24 L respectively. This was mainly attributed to the larger
pores in the drainage belt (compared with T3 and T4) and its direct contact with the soils
from inside (compared with T1), which can facilitate the water drainage. After 128 and
105 minutes, the cumulative amount of water by T3 and T4 were the same as that
collected by T1. T3 and T4 had the same cumulative amount of water as that collected by
T2 at 164 and 141 minutes, respectively.
3.5. GRAVIMETRIC WATER CONTENT AFTER 10 DAYS
Besides monitoring the volumetric water content using the moisture sensors,
gravimetric water contents were determined at the end of the test by sampling the soils
and oven-drying the samples during the disassembling process. The gravimetric water
content contours were to observe the water content distributions below the drainage
openings where the sensors could not cover. For each test, 54 sampling points were
selected (refer to Figure 3a) in each test section with 6 cm per lift and 12 cm-15 cm apart.
Figure 8 shows the gravimetric water content contours for T1-T4 after 10 days.
The dash lines represent the elevation of the drainage openings for T1 or the locations of
the drainage materials for T2-T4. The saturated gravimetric water content of the soil was
131
27% and will be used as a reference for further discussions. As shown in Figure 8a, the
gravimetric water content was lower than 3% for soil with elevations of 24 cm and above,
indicating that this part of the soil was in a relatively dry condition. For soils between 10
cm and 24 cm, the gravimetric water contents decreased with increasing elevations.
Moreover, the soil below 10 cm remained saturated after the test. The gravimetric water
content contour for T2 was shown in Figure 8b. For soils higher than 24 cm, the
gravimetric water content distribution in T2 was comparable to that in T1. Meanwhile,
the soil was also saturated with elevations lower than 10 cm. This phenomenon indicated
that both T1 and T2 were not able to drain the water located lower than 10 cm
(representing groundwater table in the field). However, the soil gravimetric water
contents between 10 cm and 22 cm were lower in T2 than in T1. For example, the 6%
iso-water content line was located at an elevation of 18 cm for T2 while this line was
located at 22 cm for T1. The test results indicated that the average gravimetric water
content within T2 was lower than T1 and more water could be drained out for the system
installed with the drainage belt.
Figure 8c shows the gravimetric water content contour for T3 which was installed
with the wicking geotextile. Different from T1 and T2, the 3% iso-water content line for
T3 moved 6 cm downward to an elevation of 18 cm. In other words, the soils with
elevations of 18 cm and above were drier in T3 than in T1-T2. Moreover, the soils
underlying the wicking geotextile were unsaturated and the highest water content was
24%. This phenomenon indicated that the wicking geotextile was also able to wick the
water from the underlying soils through siphoning effect, transported along the wicking
geotextile, and eventually wicked out of the system. Since all the soil underlying the
132
wicking geotextile was under unsaturated conditions, the effective range of the siphon
effect was at least 10 cm. Figure 8d shows the gravimetric water content contour for T4,
which was installed with the modified wicking geotextile. The 3%-iso water content line
further moved downward to 14 cm, indicating that more soils were in a drier condition
for T4 than for T3. Meanwhile, the gravimetric water content distributions below the
wicking geotextile for T4 was also lower than that for T3 for soils between 6 cm and 10
cm. As for soils lower than 6 cm, the water content was 24% and was the same as in T3.
In conclusion, the average water content in T4 was lower than that for T3 and the
drainage performance of the modified wicking geotextile was better than that of the
wicking geotextile.
To better explore the differences in drainage performance for different types of
drainage materials, Figure 9 plots the relative water content differences for T2-T4 against
T1 based upon the gravimetric water contents at the end of the test. As shown in Figure
9a, the water content differences between T1 and T2 were mainly observed at elevations
between 12 cm and 24 cm. The maximum water content difference was 7% and located at
an elevation of 18 cm. Moreover, due to the existence of the drainage openings on the left
side, the water content difference was not uniformly distributed on the left side
(elevations between 12 cm and 24 cm). Figure 9b shows the water content differences
between T1 and T3. The area with observable water content difference further extended
to 0-24 cm and the maximum water content difference also increased to 19%. This
phenomenon indicated that more water was drained out in T3 than in T2 and the drainage
efficiency of the wicking geotextile was better than that of the drainage belt. In addition,
the distribution of water content difference was not uniform with a higher difference at
133
the right side (corresponding to lower water content), indicating that the water flow was
in a transient state, and the water flowed from right to left. Figure 9c shows the water
content difference between T1 and T4. Due to the additional nylon wicking fibers, the
soil water content difference further increased to 3% from 0-6 cm, indicating the
siphoning effect was more obvious in T4 than in T3. In addition, the maximum water
content difference was 20% and evenly distributed along the location of the modified
wicking geotextile. The relatively uniform distribution of the water content difference in
T4 indicated that the drainage efficiency of the modified wicking geotextile was higher
than the wicking geotextile. In other words, if given a suffice time, the water content
distribution in T3 should be comparable to that in T4.
4. WATER RETENTION CURVES FOR DIFFERENT MATERIALS
The water storage of a material could be demonstrated by the relationship
between water content and suction, referred to as the Water Retention Curve (WRC).
Several techniques have been performed to obtain the WRCs for the materials used in this
study, including capillary rise test (suction ≤ 10 kPa), pressure plate test (suction ≤
1500 kPa), and salt concentration test (suction > 1500 kPa). To reduce the complexities
and variations of the test results, only the drying WRC curves were presented in this
paper. Figure 10 presents the WRCs for the sand (T1), drainage belt (T2), wicking
geotextile (T3), modified wicking geotextile (T4), wicking fibers, and T3 without
wicking fibers. The test results were fitted using Fredlund and Xing (1994)’s equation,
and the results were shown in Figure10 as well. As can be seen in Figure 10, the WRC of
134
sand is flat in the suction ranging from 0-1 kPa, and then shows a sharp reduction in
gravimetric water content to zero for suction between 1 and 4 kPa. The drainage belt
essentially cannot hold any water under suction (negative pore water pressure), and its
water content becomes zero at less than 0.2 kPa. On the other hand, the wicking fibers
can hold 80% of gravimetric water at saturation. Its water content significantly decreases
with suction between 1 and 30 kPa. Due to its multiple micro-channels, the wicking
fibers have water content higher than 11.36% when the suction is as high as 800 kPa. For
comparison purposes, the WRC for the wicking geotextile without the wicking fibers was
also measured by using a similar geotextile made of exactly the same yarns with exactly
the same structure except that the wicking fibers were removed. The WRC for the
wicking geotextile without the wicking fibers is similar to that of the sand with two minor
differences: (1) the water content at saturation was slightly higher, and (2) the water
content quickly reduces to zero between suction of 0.3 and 3 kPa, indicating that the
wicking geotextile without the wicking fibers has a slightly larger pore size and
corresponding air-entry value. The WRCs for the wicking geotextile and the modified
geotextiles have similar shapes as that for the wicking fibers only. Both of them have
much higher ability to hold water at saturated conditions (52.8% and 54.51%
respectively) that sand (27%). Due to the existence of the wicking fibers, both the
wicking geotextile and the modified geotextiles still hold certain amount of water when
the suction is as high as 600 kPa).
Table 2 shows the values of the model parameters for different materials when
fitted using Fredlund and Xing (1994)’s equation. Table 2 also shows the air-entry values
for different materials. The sand, the drainage belt, and the geotextile without wicking
135
fibers all have single air-entry values of 1.65 kPa, 0.13 kPa, and 0.8 kPa respectively,
implying that these materials have relatively uniform pore sizes. In comparison, it seemed
that the wicking fibers, the wicking geotextile, and the modified geotextiles have two
different levels of pore sizes, which corresponding two air-entry values. The lower air-
entry values are believed to be related to the pore sizes in the micro-channels of the
wicking fibers. The air entry values (AEVs) were 0.88 and 0.89 kPa for wicking
geotextile and modified wicking geotextile (suction ≤ 50 kPa) respectively. When the
Suction > 50 kPa, the air entry values were 254 and 300 kPa, respectively. As we can see
in later sections, the difference in the structure of the geosynthetics and their water
retention characteristics has significant influences on their drainage performances.
Figure 11 showed the suction distributions with elevations in T1-T4 after 10 days
calculated from the WRCs in Figure 10 and water content distributions as shown in
Figure 8. The drainage materials were installed at the elevation of 10 cm (red dash line).
The water table for T1 and T2 were at 10 cm and the soil suction should be linearly
distributed at the steady state. As shown in Figure 11, neither T1 nor T2 had reached
steady state within 10 days. The pore water pressure was positive (corresponding suction
was negative) for both T1 and T2, indicating that the soils lower than 10 cm were
submerged in water. For soils between 10 cm and 42 cm, the suction distributions were
deviated from the steady state for both T1 and T2, indicating that the water flow was in a
transient state. For soils with elevations higher than 42 cm, the soils reached steady state.
In comparison, the suction distribution curves for T3 and T4 were positioned at the right
side of T1 and T2, indicating the soils in T3 and T4 were drier. In other words, the
drainage performance of the (modified) wicking geotextile worked better than that of the
136
drainage belt. In addition, the soils underlying the wicking geotextile in T3 and T4 were
under unsaturated conditions. This phenomenon indicated that the wicking geotextile was
able to absorb water from the underlying soil via siphon effect and the effective range
was at least 10 cm. Visual observation showed that the wicking geotextile was still damp
at the end of the test, indicating that the wicking geotextile was still working. Given
longer testing time, the wicking geotextile was expected to drain more water out of the
system and the advantages of the wicking geotextile against the drainage belt would be
more obvious.
5. WORKING MECHANISMS FOR DIFFERENT MATERIALS
Based upon the laboratory test results, the working mechanism of different types
of drainage materials could be schematically explained as in Figure 12. During the
laboratory test, the representative soil profile could be divided into four zones as shown
in Figure 12a: saturated zone, capillary zone, transition zone, and residual zone. Below
the water table, the soil was saturated and the degree of saturation is 100%. Moving up
from the water table entered into the capillary zone where the degree of saturation was
close to 100% but the pore water pressure was negative. This zone corresponded to the
suction range from 0 kPa to the AEV on the WRC (refer to Figure 10) where the
geomaterials were able to hold large amount of water. For the soil at transition zone, air
bubbles were occluded and the air phase was discontinuous. As the soil became further
desaturated, the hydraulic conductivity of the soil significantly decreased because of the
existence of air bubbles within the voids. This zone corresponded to the suction range
137
from AEV to the residual suction on the WRC. Moving upward to the residual zone, the
air phase became continuous while the water phase became coated outside the soil
particles, making it very difficult for water to flow. On the WRC, this zone represented
the suction range that greater than the residual suction.
Figure 12a showed the drainage process for T1 which was the control case
without installation of any drainage material. Water could be drained out via the drainage
opening at the left side of the system. The area of the opening for T1 was the biggest
among all four cases, resulting in the highest cumulative volume of collected water for
T1 within the first 12 minutes (refer to Figure 7b). The soil below the elevation of the
opening remained saturated throughout the test and three distinct zones built up as the
excess water was drained out of the system. From the WRC of sand, the AEV was 1.2
kPa indicating that air bubbles could be easily entered into the voids and block the water
flow. Therefore, the cumulative volume of water for T1 only increased from 17.15 L in
32 minutes to 17.67 L in 4620 minutes (Figure 7). According to Figure 8a, the average
water content for T1 after 10 days was 10.61%, which was the highest among all cases.
In summary, T1 relied on the soil itself to drain the “free” water and was not able to drain
capillary water.
Figure 12b shows the working mechanism for T2, which was installed with the
drainage belt. At the beginning of the test, the cumulative amount of collected water for
T2 was smaller than that for T1 because the opening on the sidewall was partially
blocked by the drainage belt. As the overlying soil became desaturated, the advantages of
the drainage belt were observed. In the lateral direction, the grooves within the drainage
belt became the shortest drainage path for water to flow from right to left. The “free”
138
water would preferably flow into the grooves first and then laterally flowed out of the
system. As shown in Figure 7b, the cumulative amount of collected water from T2
surpassed T1 in approximately 30 minutes after the test. The AEV of the drainage belt
was small (0.2 kPa) and air could easily enter into the grooves to block the water flow
when suction is higher than the AEV. Meanwhile, the water within the transition and
capillary zones gradually flowed downward under the influence of gravitational force.
The pore water pressure for the soil right above the drainage belt would gradually build
up from a negative value (suction) to a positive value as the degree of saturation
continued to increase. To a critical moment when the positive pore water pressure was
higher than the AEV of the drainage belt, the air within the grooves would be drained out.
This explanation was validated by the observation during the test of which measurable
amount of water could be discontinuously collected from T2 from 4569 minutes to 6838
minutes (refer to Figure 7a). It was important to point out that the soil underlying the
drainage belt was still saturated throughout the test and no capillary water could be
drained out by the drainage belt. According to Figure 8b, the average gravimetric water
content for T2 after 10 days was 8.77%.
Figure 12c showed the working mechanism for T3/T4 in which the (modified)
wicking geotextiles were installed. Different from the drainage belt, the wicking fibers
had multichannel cross-sections (refer to Figure 2b) which were able to hold and
transport water under unsaturated conditions. At the beginning of the test, the drainage
performance of the wicking geotextile was not as effective as the drainage belt in terms
of draining “free” water because of the relatively small pore sizes within the weaving
structure. The cumulative amount of collected water for T3 and T4 was lower than that
139
for T1 and T2 at the beginning of the test (Figure 7b). As the soil further desaturated, the
advantages of the wicking geotextile to drain capillary water were observed. Comparison
of the AEVs of the soil (1.65 kPa) and the wicking fibers (254 kPa) (refer to Figure 10)
indicated that the ability of the wicking geotextile to hold water was much stronger than
that of the soil. In other words, the wicking geotextile had much stronger ability to absorb
water from the surrounding soils. This fact also explained the observation (refer to
Figures 8c-d) that the soils underlying the (enhanced) wicking geotextile were
unsaturated. It was the siphon effect (caused by different AEVs of the soil and the
wicking fiber) that enabled the wicking geotextile to absorb water from the surrounding
soil to the grooves within the fibers. In addition, relative humidity (RH) at the exposed
end of the wicking geotextile was lower than 56% (the corresponding suction > 53.7
MPa) while the relative humidity at the other end was close to 100% (suction
approximately 0 kPa). The significant suction gradient was the driving force that
continuously wicked the water of the system in the horizontal direction. Meanwhile, the
exposed end of the wick geotextile was damp throughout the test and the water could be
continuously vaporized to the ambient environment via the evaporation process. Because
the capillary water within the testing apparatus could be drained out, the water tables for
T3 and T4 were much lower than those for T1 and T2. According to Figure 8c and 8d, the
average gravimetric water contents for T3 and T4 after the test were 6.89% and 6.67%,
respectively. In summary, the ambient environment worked as a “pump” and the wicking
geotextile worked as “pipes” that continuously wicked both gravitational and capillary
water out of the soil and made the soil drier with time.
140
6. CONCLUSIONS
This paper studied the drainage performance and working mechanisms of three
types of drainage materials, including drainage belt, wicking geotextile, and modified
wicking geotextile. The major conclusions are summarized as follows:
1. Capillary water within the soils could not be drained out by conventional drainage
systems. For the tested uniform sand, which was considered as a good drainage
geomaterial, the water content was still 10.61% ten days after the test.
2. The microstructures of the geosynthetics influenced the drainage performance
significantly.
3. The drainage belt was able to drain more “free” water out of the system compared
with the control case. However, due to the very small AEV value (0.2 kPa), it was
not able to drain any capillary water. The average water content for the tested sand
could be reduced from 10.61% to 8.77% by installing a layer of the drainage belt.
The average water content was reduced by 17.34%.
4. Different from the drainage belt, the wicking fibers had much higher AEV (254
kPa), which enabled the wicking geotextile to hold and transport water under
unsaturated conditions. The wicking geotextile could drain both “free” and
capillary water, and the average water contents could be further reduced to 6.89%
(with wicking geotextile) and 6.67% (with modified wicking geotextile),
respectively. The corresponding water contents were reduced by 35.06% and
37.13% respectively.
141
5. The wicking geotextile was able to wick water from both the overlying and
underlying soils. The influencing range of the siphon effect was at least 10 cm.
6. Increasing the amount of wicking fibers would enhance its ability to drain capillary
water and therefore increase the total amount of water drained out of the system.
However, the wicking fibers were not aimed at draining “free” water.
7. The working mechanism of the different drainage materials could be explained by
their WRCs and highly depended on their AEVs. Higher AEV value represented a
higher ability of the geomaterial to hold and transport water, and higher capability
to dehydrate the soil.
8. In summary, the drainage belt showed better performance in draining “free” water
but could not drain capillary water. The wicking geotextile could drain both
gravitational and capillary water and showed the best drainage performance among
all the drainage materials.
ACKNOWLEDGEMENTS
This research was supported by the China Scholarship Council, China
(No.201706370160), the Central South University, China (No.2016zzts080), the China
Railway Corporation, China (No. 2015G006-F), the Hunan Science & Technology
department, China (No. 2017SK2212), and the Natural Science Foundation of China
(No.51678572).The authors sincerely thank TenCate Geosynthetics (North American)
for providing the materials used in the study.

142
Table 1. Wicking geotextile specifications

Mininum Average Roll
Mechanical properties Test Method Unit
Value
Apparent Opening Size ASTM D
mm 0.4
(AOS) 4751
ASTM
Pore Size (O50) Microns 85
D6767
ASTM
Pore Size (O95) Microns 195
D6767
ASTM
Permittivity Sec-1 0.24
D4491
ASTM
Flow Rate L/min/m2 1222
D4491
143
Table 2. WRC parameters for different drainage materials

Wicking Geotextile Modified wicking
Wicking Fibers T3
Drainage (T3) Geotextile (T4)
Sand without
Parameter belt Suction
(T1) Suction ≤ Suction > Suction ≤ Suction > Suction ≤ 300 Wicking
(T2) > 300
50 kPa 50 kPa 50 kPa 50 kPa kPa Fibers
kPa
ws (%) 27 9 52.8 5 54.51 8 80.99 11.36 50.58
ψr (kPa) 2.4 1 7 600 8 800 23 800 3

Air Entry
Value 1.65 0.13 0.88 254 0.89 300 1.76 300 0.8
(kPa)
n 7.37 20 2.19 1.62 0.65 1.15 1.8 0.3 3
m 1.69 30 0.99 1.03 3.68 0.86 0.78 0.2 2
1
143
144
(a)
(b)
Figure 1. Influence of soil water content on resilient modulus and permanent

deformation: (a) resilient modulus variations, and (b) permanent deformation variations
(adapted from Lin et al. 2018)
145
(a)
(b)
(c)
Figure 2. Drainage materials used in the laboratory tests: (a) drainage belt (unit: in mm),
(b) wicking geotextile, (c) schematic plot of weaving structure of the wicking geotextile,
(d) modified wicking geotextile, and (e) schematic plot of weaving structure of the
modified wicking geotextile (c and e were modified from King et al., 2014)
146
(d)
(e)
Figure 2. Drainage materials used in the laboratory tests: (a) drainage belt (unit: in mm),
(b) wicking geotextile, (c) schematic plot of weaving structure of the wicking geotextile,
(d) modified wicking geotextile, and (e) schematic plot of weaving structure of the
modified wicking geotextile (c and e were modified from King et al., 2014) (cont.)
147
(a)
(b)
Figure 3. Test apparatus: (a) schematic plot (unit in cm), and (b) configuration
148
(a)
(b)
Figure 4. Wetting fronts after 10 days of drainage: (a) drainage belt versus control case,
(b) drainage belt versus wicking geotextile, and (c) modified wicking geotextile versus
wicking geotextile
149
(c)
Figure 4. Wetting fronts after 10 days of drainage: (a) drainage belt versus control case,
(b) drainage belt versus wicking geotextile, and (c) modified wicking geotextile versus
wicking geotextile (cont.)
150
(a)
(b)
Figure 5. Monitored volumetric water content variations with time under different testing
conditions: (a) top (M8), (b) middle (M5), and (c) bottom (M2)
151
(c)
Figure 5. Monitored volumetric water content variations with time under different testing
conditions: (a) top (M8), (b) middle (M5), and (c) bottom (M2) (cont.)
152
(a)
Figure 6. Volumetric water content contours under different testing conditions: (a)
control case (T1), (b) with drainage belt (T2), (c) with wicking geotextile(T3), and (d)
with modified wicking geotextile (T4) (unit in m3/m3)
153
(b)
with modified wicking geotextile (T4) (unit in m3/m3) (cont.)
154
(c)
155
(d)
156
(a)
(b)
Figure 7. Cumulative amount of water under different testing conditions: (a) 0-8000
minutes, and (b) 0-200 minutes
157
(a) (b)
(c) (d)
Figure 8. Gravimetric water content contours after the test: (a) control case (T1), (b) with
drainage belt (T2), (c) with wicking geotextile (T3), and (d) with modified wicking
geotextile (T4): (unit in %)
158
(a)
(b)
Figure 9. Relative gravimetric water content difference contours: (a) control case versus
drainage belt (T1 vs. T2), (b) control case versus wicking geotextile (T1 vs. T3), and (c)
control case versus modified wicking geotextile (T1 vs. T4)
159
(c)
Figure 9. Relative gravimetric water content difference contours: (a) control case versus
drainage belt (T1 vs. T2), (b) control case versus wicking geotextile (T1 vs. T3), and (c)
control case versus modified wicking geotextile (T1 vs. T4) (cont.)
Figure 10. Water Retention Curves (WRCs) of different testing materials

160
Figure 11. Comparisons of suction distributions with elevations in T1-T4 after the test
161
Figure 12. Working mechanisms for different drainage materials: (a) control case (T1),
(b) drainage belt (T2), and (c) (enhanced) wicking geotextile (T3 and T4)
162
REFERENCES
ASTM D 2487, 2011. Standard Practice for Classification of Soils for Engineering
Purposes (Unified Soil Classification System). ASTM International, West
Conshohocken, PA.
Athanasopoulos, G.A., 1993. Effect of particle size on the mechanical behaviour of sand-
geotextile composites. Geotextiles and Geomembranes, 12(3), 255-273.
Benjamin, C.V., Bueno, B., and Zornberg, J. G., 2007. Field monitoring evaluation of
geotextile-reinforced soil retaining wall. Geosynthetics International, 14(2), 100-
118.
Cancelli, A., Montanelli, F., Rimoldi, P., and Zhao, A., 1997. Full scale laboratory testing
ongeosynthetics reinforced paved roads, in: Ochiai, H., Yaufuku, N., Omine, K.,
(Eds.), Earth Reinforcement, Balkema, 1997, Proceedings of the International
Symposium on Earth Reinforcement, Vol. 1, Fukuoka, Kyushu, Japan, November,
1996, 573-578.
Cedergren, H.R. 1994. America’s pavements: World’s longest bathtubs. Civil

engineering. American Society of Civil Engineering, 64(9): 56–58.
Christopher, B.R., and McGuffey, V.C. 1997. Pavement Subsurface Drainage Systems,
NCHRP Synthesis of Highway Practice 239, Transportation Research Board,
National Research Council, Washington, D.C.
Christopher, B.R., Schwartz, C.W., and Boudreau, R.L., 2010. Geotechnical aspects of
pavements: Reference manual. US Department of Transportation, Federal
Highway Administration.
Fredlund, D.G. and Xing, A., 1994. Equations for the soil-water characteristic curve.
Canadian geotechnical journal, 31(4), pp.521-532.
Garcia, E.F., Gallage, C.P.K., and Uchimura, T., 2007. Function of permeable
geosynthetics in unsaturated embankments subjected to rainfall infiltration.
Geosynthetics International, 14 (2), 8999.
Gobel, C.H., Weisemann, U.C., and Kirschner, R.A., 1994. Effectiveness of a reinforcing
geogrid in a railway subbase under dynamic loads. Geotextiles and
Geomembranes 13 (1), 91-99.
Ghosh, C. and Madhav, M.R., 1994. Reinforced granular fill-soft soil system: membrane
effect. Geotextiles and Geomembranes, 13, 743-759.
163
Gray, D.H. and Al-Refeai, T., 1986. Behavior of fabric-versus fiber-reinforced sand.
Journal of Geotechnical Engineering, 112(8), 804-820.
Guo, Y.P, Leng, W.M, Nie, R.S, Zhao, C.Y., and Zhang, X., 2018. Laboratory evaluation
of a new device for water drainage in roadside slope along railway systems.
Geotextiles and Geomembranes, 46(6), 897-903
Henry, K.S. and Holtz, R.D., 2001. Geocomposite capillary barriers to reduce frost heave
in soils. Canadian Geotechnical Journal, 38(4), 678-694.
Henry, K.S., J.C. Stormont, L.A. Barna, and R.D. Ramos, (2002) “Geocomposite
capillary barrier drain for unsaturated drainage of pavements,” International
Conference on Geosynthetics, Nice, France.
Iryo, T. and Rowe, R.K., 2003. On the hydraulic behavior of unsaturated nonwoven
geotextiles. Geotextiles and Geomembranes, 21(6), 381-404.
Iryo, T. and Rowe, R.K., 2005. Infiltration into an embankment reinforced by nonwoven
geotextiles. Canadian Geotechnical Journal, 42(4), 1145-1159.
Koerner, R.M. and Soong, T.Y., 2001. Geosynthetic reinforced segmental retaining
walls. Geotextiles and geomembranes, 19(6), 359-386.
Leng, W.M., Guo, Y.P., Nie, R.S., Zhao, C.Y., Cui, Y.Q., and Dong, C., 2017. A new
drainage facility and its drainage performance. Journal of the Railway Society,
39(07), 151-158 (in Chinese).
Lin, C., Presler, W., Zhang, X., Jones, D., and Odgers, B., 2017. Long-term performance
of wicking fabric in Alaskan pavements. J. Perform. Constr. Facil. 31 (2),
D4016005.
Ling, H. I., Wu, J. T. H., and Tatsuoka, F. 1993. Short-term strength and deformation
characteristics of geotextiles under typical operational conditions. Geotextiles and
Lin, C. and Zhang, X., 2018. Laboratory Drainage Performance of a New Geotextile with
Wicking Fabric. Journal of Materials in Civil Engineering, 30(11), 04018293.
Lin, C. and Zhang, X., 2018. A Bio-Wicking System to Dehydrate Road Embankment.
Journal of Cleaner Production, 196, 902-915.
Lin, C, Zhang, X., and Han, J., 2018. Comprehensive Material Characterizations for a
Roadway Installed with Wicking Fabric. Journal of Materials in Civil
Engineering, DOI: 10.1061/(ASCE)MT.1943-5533.0002587.
164
MathWorks, 2016. MATLAB user’s guide, version R2016a. South Natick, MA:
MathWorks Inc.
Mitchell, J. K. and Zornberg, J. G. 1995. Reinforced soil structures with poorly draining
backfills, Part II: Case histories and applications. Geosynthetics International,
2(1), 265-307.
McGown, A., Khan, A.J., and Kupec, J., 2004. Determining the design parameters for
geosynthetics reinforcements subject to multi-stage actions using the isochronous
strain energy approach. Geosynthetics International, 11(6), 455-469.
Noorzad, R. and Mirmoradi, S.H., 2010. Laboratory evaluation of the behavior of a

geotextile reinforced clay. Geotextiles and Geomembranes, 28(4), 386-392.
Portelinha, F.H.M., Bueno, B.S., and Zornberg, J.G., 2013. Performance of nonwoven
geotextile-reinforced walls under wetting conditions: laboratory and field
investigations. Geosynthetics International, 20(2), 90-104.
Pathak, Y.P. and Alfaro, M.C., 2010. Wetting-drying behavior of geogrid-reinforced clay
under working load conditions. Geosynthetics International, 17(3), 144-156.
Perkins, S.W., 1999. Mechanical response of geosynthetic reinforced flexible pavements.

Geosynthetics International, 6 (5), 347-382.
Raisinghani, D.V. and Viswanadham, B.V.S., 2011. Centrifuge model study on low
permeable slope reinforced by hybrid geosynthetics. Geotextiles and
Som, N. and Sahu, R.B., 1999. Bearing capacity of a geotextile-reinforced unpaved road
as a function of deformation-a model study. Geosynthetics International, 6 (1), 1-
17.
Stulgis, R. P., 2005. Full-scale MSE test walls. Proceedings of NAGS 2005/GRI-19
Conference, Las Vegas, NV, USA (CD-ROM).
Tan, S. A., Chew, S. H., Ng, C. C., Loh, S. L., Karunaratne, G. P., and Delmas Ph Loke,
K. H. 2001. Large-scale drainage behavior of composite geotextile and geogrid in
residual soil. Geotextiles and Geomembranes, 19(3), 163-176.
Wang, F., Han, J., Zhang, X., and Guo, J., 2017. Laboratory tests to evaluate
effectiveness of wicking geotextile in soil moisture reduction. Geotextile and
Geomembranes, 45 (1), 8-13.
165
Yoo, C. and Jung, H. Y. 2006. Case history of geosynthetic reinforced segmental

retaining wall failure. Journal of Geotechnical and Geoenvironmental
Engineering, 132(12), 1538-1548.
Zhang, X. and Belmont, N., 2009. Use of Mirafi Nylon Wicking Fabric to Help Prevent
Frost Heaving in Alaska Pavement: 1st, 2nd, 3rd, 4th and 5th Progress Reports.
Institute of Northern Engineering (INE), Alaska Univ. Transportation Center
(AUTC), University of Alaska Fairbanks, AK.
Zhang, X., Presler, W., Li, L., Jones, D., and Odgers, B., 2014. Use of wicking fabric to
help prevent frost boils in Alaskan pavements. J. Mater. Civ. Eng.,
10.1061/(ASCE)MT.1943-5533.0000828, 728-740. 385
Zornberg, J.G. and Mitchell, J.K., 1994. Reinforced soil structures with poorly draining
backfills. Part I: reinforcement interactions and functions. Geosynthetics
International, 1(2), 103-147.
166
IV. NUMERICAL SIMULATION OF SEASONAL VARIATIONS OF BASE

COURSE RESILIENT MODULUS IN PAVEMENT STRUCTURES
Chuang Lin1 and Xiong Zhang2
1
Environmental Engineering, Missouri University of Science and Technology, Phone:
(907) 799-9203, Email: [email protected]
2
(Corresponding Author) Associate Professor, M. ASCE, Department of Civil,
Architectural, and Environmental Engineering, Missouri University of Science and
Technology, Phone: (573) 341-6268, Email: [email protected]
ABSTRACT
The seasonal variation of unbound material properties is often significant and
water within a pavement structure is a principal cause of pavement deteriorations. A
small water content increment will result in a significant reduction in soil resilient moduli
and a tremendous increment in permanent deformation. The back calculated resilient
modulus from the falling weight deflectometer test is highly dependent on the ambient
climatic conditions and may not be a representative value. Moreover, the Enhanced
Integrated Climatic Model (EICM) for the MEPDG (Mechanistic-Empirical Pavement
Design Guideline) approach has its own limitations and can be further improved. An
accurate, robust, and efficient numerical model is necessary to accurately predict the
dynamic performance of the pavement structure under the influence of seasonal climatic
conditions. This paper aims at numerically simulating the seasonal variations of the
resilient behavior of a base course under saturated/unsaturated conditions with the
influence of localized climatic conditions. A coupled hydro-mechanical-climatic model

167
was proposed and calibrated. Then, three case studies were conducted to evaluate the
dynamic resilient behavior of the base course under different climatic and working
conditions. The suction, water content, and resilient modulus distributions were presented
to demonstrate the dynamic resilient behavior of the base course and the advantages of
the proposed model against the EICM in terms of simulating the moisture migration
within pavement structures are discussed.
KEY WORDS: subsurface drainage, base course, resilient modulus, unsaturated
soil, numerical simulation, flexible pavement
1. INTRODUCTION
The seasonal variation of unbound material properties is often significant and
water within a pavement structure is a principal cause of pavement deterioration (Miller
1953). Christopher et al. (2010) showed that the initial construction cost of a flexible
pavement may increase by 44% per lane-mile as a consequence of poor drainage. Excess
water can also decrease the resilient modulus and shear strength, resulting in pavement
degradations (Roberson and Siekmeier 2002). For example, when soil water content
increased from 3.3% to 6%, the resilient modulus for Alaska D-1 base course aggregate
reduced by 50% and the permanent deformation was doubled or even tripled (Li et al.
2011). Moreover, government transportation engineers in cold regions credited a
minimum of half of road maintenance expenditures to the effect of freezing and thawing
(Henry and Holtz 2001), which was another common water related hazard occurred in
northern regions of the United States.

168
Both the AASHTO (1993) and MEPDG (Mechanistic-Empirical Pavement
Design Guide) (ARA 2004) design methods include provisions for including seasonal
variations of unbound material properties. ASTM D420-18 (2018) provides detailed
procedures of performing geotechnical site characterizations. Coring and sampling is a
major destructive test which has many limitations, particularly when conducted on
moderate or heavily trafficked highway systems. Moreover, the practical restraints, in
terms of money and time, severely limit the number and variety of destructive tests
conducted on routine pavement evaluation studies (AASHTO 1993; Shahin 1994). In
comparison, the nondestructive test (NDT) becomes more popular since this type of
examination of pavement structure does not include damage or property changes to the
structure. Among all NDT testing techniques, the FWD (Falling Weight Deflectometer)
test is one of the most reliable ways of determining the in-situ moduli of the pavement
system (ARA 2004). It can be used to determine the variations of pavement layer and the
resilient modulus of each layer can be back calculated given the thickness of each layer
(ASTM D4694-09 2015). An accurate prediction of the resilient modulus value is
important since it is a required input to the structural response computation model in both
AASHTO 1993 and MEDPG approaches. However, the resilient modulus values back
calculated from the FWD test are highly dependent on the ambient climatic conditions,
such as the depth of groundwater table, the intensity and duration of precipitation, and
solar radiation, etc. The FWD test is normally conducted on a bi-weekly or monthly basis
(Huang 2004; ARA 2004) and the back calculated resilient modulus values represent the
particular values at that moment of the year, which may not be a representative value for
169
the entire month or season. It is necessary to come up with an accurate approach to
predict the resilient behavior of a base course under localized climatic conditions.
The AASHTOWare Pavement ME Design software is the production version of
the MEPDG and can provide reasonable predictions of seasonal variations of resilient
modulus for different pavement layers based upon the climatic data from the LTPP (Long
Term Pavement Performance) SMP (Seasonal Monitoring Program) test sections
(Witczak et al. 2000). The climatic effects are considered in the software through the
EICM (Enhanced Integrated Climatic Model) (Larson and Dempsey 1997). The EICM
model integrated three separate models addressing different aspects of climatic effects on
the pavement into a single comprehensive package (Richter 2006), including the CMS
(Climatic-Material-Structures) model (Dempsey et al. 1985), the ID (Infiltration and
Drainage) model (Liu and Lytton 1985), and the CRREL Frost Heave and Thaw
Settlement Model (Guymon et al. 1986). However, the EICM model has its own
limitations and still needs to be improved in several aspects. Firstly, the EICM model is a
1D model (ARA 2004), which is not accurate to evaluate the soil moisture content
distributions in a 2D or 3D space. For example, the amplitude of soil moisture content
variations at the center of the road must be lower than that for the soil at the edge of the
road since the soils at the edge is exposed to the ambient environment and will be
significantly affected by the climatic conditions. Secondly, the infiltration water
determined by the EICM model is merely based upon empirical judgement. In the
MEDPG approach, infiltration is assumed to be four levels: none, minor, moderate, and
extreme, corresponding to 0%, 10%, 50%, and 100% of precipitation enters the pavement
(ARA 2004). In reality, the infiltration rate shall be a function of both precipitation and
170
the water storage capacity of the top unbounded layer. For example, if the top of the
unbounded layer is already saturated and cannot absorb any water, the rest of the
precipitation water shall be considered as runoff water rather than continuously enters
into the pavement structure. Thirdly, the drainage model in EICM, also known as DRIP
(Drainage Requirements in Pavements) (Wyatt et al. 1998) model, is based upon the
water flow through saturated soils and is not suitable for unsaturated conditions. The soil
within a pavement structure is often compacted at the optimum water content to achieve
the maximum dry density and the best performance. The soils are under unsaturated
conditions during most of the time in its service life. Obviously, the DRIP model cannot
accurately simulate the moisture migration within the pavement structure. Last but not
the least, the slope of the road embankment is often covered with vegetation for
protection, erosion control, and decoration purposes (Fredlund et al. 2012). The EICM
model is not able to simulate the dynamic interactions between the soil infrastructure and
vegetation under different climatic conditions.
This paper aims at numerically simulating the seasonal variations of the resilient
behavior of a base course under saturated/unsaturated conditions with different climatic
conditions. A coupled hydro-mechanical-climatic model was first proposed. A finite
element method (FEM) software Abaqus (Abaqus 2014) was used to evaluate the
seasonal variations of base course resilient behavior according to daily meteorological
data, such as relative humidity (RH), air temperature, wind speed, solar radiation, and
precipitation. Then three case studies were conducted to predict the dynamic resilient
behavior of a base course and the interactions among the soil infrastructure, vegetation,
and the ambient environment.

171
2. FINITE ELEMENT MODEL
2.1. GOVERNING EQUATIONS
Biot (1941) firstly derived the 3D coupled consolidation theory for unsaturated
soils with occluded air bubbles. Fredlund and Morgenstern (1976) proposed the
constitutive relations for the volume change in unsaturated soils with a continuous air
phase using two stress state variables. Zhang et al. (2005) reexamine the physical
meanings of the parameters used in Fredlund and Morgenstern’s paper and proposed a
new set of differential equations for solving the coupled consolidation problem for
saturated-unsaturated soils by using a thermodynamic analogue. For engineering
applications, the base course was under unsaturated conditions during most of the time in
its service life. However, it might also be fully saturated during heavy rainfall events.
Therefore, this paper adopted Zhang’s equation to evaluate the pavement performance, as
expressed in Equation 1.
 v  ( −uw )
( + G) + G2u − ( 3 + 2G )  +X =0
x x (1a)
 v  ( −uw )
( + G) + G2v − ( 3 + 2G )  +Y = 0
y y (1b)
 v  ( −uw )
( + G) + G2 w − ( 3 + 2G )  +Z =0
z z (1c)
1    ( −uw )   ( −uw )   ( −uw )    ( −uw )

 (k ) + (k ) + (k + 1)  = m1w m + m2w +S
 w  x x y y z z  t t
(1d)
where, u, v, and w = displacements in the x, y, and z directions; X, Y, and Z = body
forces in the x, y, and z directions; G = shear modulus;  = Lame constant;  =

172
coefficient of expansion due to pore water pressure variations;  v = volumetric strain; uw
= pore water pressure;  m = mean stress;  w = unit weight of water; k = coefficient of
permeability; t = time; m1w = coefficient of pore-water volume change due to the change
in mechanical stress; m2w = coefficient of pore-water volume change due to the change in
pore water pressure, m2w = d Cw ;  d = soil dry density; Cw = specific water capacity, or
the slope of the soil water characteristic curve (SWCC); and S = water generation term.
Equations 1a-1c are the equilibrium equations in terms of displacements in the x,
y, and z directions. Equation 1d is the water continuity equation. For Equation 1d, the
value m1w is much smaller compared with the m2w and S terms. Therefore, m1w assumes
to be zero for preliminary simulations. The last term, S, is the water source term which
depends upon the ambient climatic conditions, such as precipitation and
evapotranspiration. The infrastructure-climatic interactions will be discussed in the
following section. For engineering practice, mechanical stress and pore water pressure
are two major factors that will influence the soil volume change, and other factors such as
soil temperature, pore air pressure, and salt concentration can be considered as constants
during the numerical analysis (Zhang and Liu 2008).
2.2. CLIMATIC EFFECT
Unlike classical soil mechanics which deals with water head type of boundary
conditions (or Dirichlet boundary conditions), the ground surface relies on a moisture
flux boundary condition (or Neumann boundary conditions) which interacts with the
ambient atmospheric environment (Wilson et al. 1994). The ground surface boundary
173
conditions must be described in terms of moisture flux so that the moisture exchange
between the saturated/unsaturated soil and the surrounding environment can be
quantified. The roadway embankment is always hydro-seeded after construction to
prevent erosion and dust contamination. Water is either entering into the pavement
structure via precipitation infiltration process or leaving the pavement structure via
evapotranspiration (ET) process. The precipitation infiltration process can be accurately
determined given the infrastructure geometry and the local precipitation data. The
evapotranspiration process can be reasonably simulated using the FAO 56 PM method,
which was proposed by United Nations Food and Agriculture Organization (FAO)
(Allen et al. 1998).
Soil evaporation is strongly influenced by the net radiation from the sun and the
movement of air above the ground surface (Tran et al. 2015). The FAO 56 PM method
requires measurements of air temperature, relative humidity, wind speed, and solar
radiation to determine the reference evapotranspiration, as expressed in Equation 2. For
this paper, the field site was selected to be at Kirkwin, Kansas, and the local
meteorological data of the year 2016 was obtained from WRCC (West Regional Climate
Center) (WRCC 2017), as show in Figure 1a-c. By using Equation 2, the reference
evapotranspiration, ET0 could be determined. Figure 1d shows the hourly ET0 and
precipitation data.
900
0.408 ( Rn − G ) +  u2 ( es − ea )
ET0 = T + 273 (2)
 +  (1 + 0.34u2 )
where, ET0 = reference evapotranspiration;  = slope of the vapor pressure curve; Rn =
net radiation at the crop surface; G = soil heat flux density, which is assume as zero for
174
daily calculations;  = psychometric constant; T = mean daily air temperature at a
height of 2 m; es and ea = saturation and actual vapor pressure, and (es − ea ) is the
saturation vapor pressure deficit (VPD); and u2 = wind speed at a height of 2 m.
The ET0 gives the maximum rate of evapotranspiration under an idealized
condition with a short green crop (grass), completely shading the ground, of uniform
height and with adequate water status in the soil profile. The ET0 value is only
appropriate for open water or fully saturated soil surface (Wilson 1997). However, when
the surface soil is in a relatively dry condition, the actual evapotranspiration (AET) rate is
expected to be much smaller. The challenging part of determination of the net infiltration
is associated with accurately predicting the AET at the ground surface (Wilson et al.
1997) and it is the AET that is required for engineering modeling. Numerous researchers
(Brutsaert 2013; Gray 1970; Morton 1975) have pointed out that the AET began to
decline as the soil surface became unsaturated and the water supply was limited. A water
balance analysis within the grass root zone would be helpful in determining the
reasonable AET values, as shown in Figure 2a.
Figure 2a shows the water balance analysis of a soil element within the grass root
zone. At the top surface of the soil element, the water may leave the soil via
evapotranspiration process or infiltrate into the soil through precipitation process. If the
rainfall intensity was high, not all water will infiltrate into the soil element and most of
the water will be considered as runoff water. Meanwhile, at the bottom of the soil
element, water may also percolate to the underlying soil or wicking to the overlying soil
via capillary action. Therefore, to evaluate the interactions between the soil infrastructure
and the ambient environment, it is necessary to accurately quantify the amount of water
175
that the soil element may lose or gain. There are several critical water contents
representing different soil water states and need to be defined (see in Figure 2a), namely
saturation water content, field capacity (FC), threshold water content, and wilting point
(WP). According to Allen et al. (1998), the total available water (TAW) represents the
capacity of a soil to retain water available to plants. After heavy rainfall events, water in
the soil will be drained under the influence of gravity and the soil water content decreases
from the saturation water content to the field capacity, which is defined as the amount of
water a well-drained soil should hold against gravitational forces. As water continued to
drain and the soil water content becomes lower than the threshold value, the absence of
water supply renders increasing soil stress against the water uptake by the grass and the
remaining water is hold to the soil particles with greater force. At this moment, not all the
TAW will be available for the grass to extract, the actual amount of available water will
be denoted as the readily available water (RAW), which is a fraction of the TAW. If the
soil water content continued to reduce, the grass will permanently wilt if the water
content is lower than the wilting point (WP). In other words, if the soil water content is
lower than the wilting point, the vegetation will not survive.
The TAW is the amount of water that grass can extract from the root zone
whereas the RAW is the amount of water a grass can extract from the root zone without
surfer water stress. The TAW and RAW can be expressed as in Equation 3:
TAW = ( FC − WP )Zr (3a)
RAW = p  TAW (3b)

176
where, TAW = total available water;  FC = water content at field capacity; WP = water
content at wilting point; Z r = grass root zone; RAW = readily available water; and p =
average fraction of TAW that can be depleted from the root zone before soil stress
occurs, [0-1].
Based upon the TAW and RAW, the water stress coefficient, K s , can be
expressed as in Equation 4:
TAW − Dr TAW − Dr
Ks = = (4)
TAW − RAW (1 − p)TAW
where, K s = stress coefficient; TAW = total available water; RAW = readily available
water; Dr is root zone depletion; and p = average fraction of TAW that can be depleted
from the root zone before soil stress occurs.
The AET is not only depends on the water stress coefficient ( K s ), it also depends
on the crop coefficient, K c , which accounts for the differences in evaporation and
transpiration between field crops. For geotechnical applications, the single crop
coefficient approach is appropriate and an average value of 0.6 was used in this paper
based upon the monitored results provided by Romero and Dukes (2016). Therefore, the
AET is eventually a function of both stress coefficient and crop coefficient, and can be
expressed as in Equation 5:
AET = K s  Kc  ET0 (5)
where, AET = actual evapotranspiration; K s = water stress coefficient; K c = crop
coefficient; and ET0 = reference evapotranspiration.

177
To further demonstrate the water exchange between the soil infrastructure and the
ambient environment within the grass root zone, Figure 2b presents the flow chart for the
water balance analysis what is adopted in user subroutine during numerical simulations.
First of all, there are several terminologies need to be defined. Assume the cross-sectional
area, A , of the soil element in Figure 2a is a constant, the volume of water within the soil
element can be expressed as an equivalent water height and can be expressed as in
Equation 6. Meanwhile, the water stress coefficient, K s , can also be rewritten as in
Equation 7. The water source term, S , in Equation 1d can also be expressed as in
Equation 8.
d
hw = w h (6)
w t
where, hw = the equivalent water height corresponding to the current water content; ht =
equivalent height of the soil element; w = gravimetric water content; and  d and  w =
unit weights of soil solids and water, respectively.
hw − hWP
Ks = (7)
(1 − p)(hFC − hWP )
where, K s = water stress coefficient; hw = equivalent water height; p = average fraction
of TAW that can be depleted from the grass root zone before soil stress occurs; and hFC
and hWP = equivalent water height corresponding to field capacity and wilting point,
respectively.
VANWL ANWL  A ANWL

S= = = (8)
V ht  A ht
178
where, S = water source term; VANWL = rate of the volume of water change in the soil
element; V = total volume of the soil element; ANWL = actual net water loss; ht =
equivalent height of the soil element; and A = cross-sectional area of the soil element.
In total, there are three inputs that will be passed to the subroutine at the
beginning of each time step, including the precipitation (Rain), ET (determined by
Equation 2), and the suction value (a field variable that was stored at the end of the
previous time step). Firstly, the actual water stress coefficient, AKS , was determined
based upon Equation 4, and the actual evapotranspiration, AET , was calculated
according to Equation 5. Then, the actual net water loss (ANWL) value was calculated
based upon the water balance analysis shown in Figure 2a. If the ANWL  0, the soil
element is essentially losing water and there shall be no runoff water. The water source
term, S , is determined based upon the comparison of the values for ANWL and (hw − hWP )
. The ANWL term indicates the calculated net water loss while the (hw − hWP ) term
indicates the maximum available amount of water could be lost in the soil. If the ANWL
 (hw − hWP ) , the soil element cannot provide sufficient water to be evaporated and the
water source term shall be determined based upon the value of (hw − hWP ) . On the other
hand, if the ANWL < (hw − hWP ) , the water source term shall be calculated based upon the
ANWL term.
In comparison, if ANWL < 0, water is expected to flow into the soil element and
the rainfall event dominates the water balance analysis. Similarly, the (hsat − hw ) term
indicates the maximum amount of water the soil can absorb and the
( Rain − Percolation − AET ) term represents the calculated amount of water flowing into
179
the soil element. If (hsat − hw )  (Rain – Percolation – AET), the soil element can absorb
all the amount of water and the source term shall be calculated based upon the calculated
value of (Rain – Percolation – AET). However, if (hsat − hw ) < (Rain – Percolation –
AET), the soil element cannot absorb the entire amount of water and the amount of runoff
water shall be the difference between the two terms. Finally, the terms of S , ANWL,
Runoff, AET, Rain, Percolation, and AKS will be stored as solution dependent variables
for future usage.
2.3. MODEL VALIDATION
The laboratory test performed by Leung et al. (2015) was used as a reference to
calibrate the proposed numerical model. Figure 4a shows the schematic plot of the
laboratory test setup. The test box was cubical with dimensions of 0.3 m  0.3 m  0.3
m. The thickness of the soil layer was 0.28 m with four soil suction sensors installed at
elevations of 0.07 m, 0.14 m, 0.20 m, and 0.25 m, respectively. The rainfall simulator
was installed on top of the test box. During the infiltration test, the ponding water on the
soil surface could be drained out through the opening located at the left side of the wall
so that no excess water pressure will be generated. In addition, a fluorescent lamp was
also placed above the test box to provide solar radiation to the plant with an average rate
of potential evapotranspiration of 2.94 mm/day. The soil was classified as clayey sand
with gravel (SC) according to the Unified Soil Classification System (USCS). The plastic
limit and liquid limit of the fines were 26% and 44%, respectively. The target dry density
of the soil was 1.496 Mg/m3 (80% relative compaction) and the saturated permeability
180
was 5.79  10-8 m/s. The vegetation type was selected to be Schefflera heptaphylla (also
known as Ivy tree) with an average root depth of 0.1 m.
For the model validation, only the drying test was used to calibrate the model.
During the drying test, the readings from the four suction sensors at the beginning of the
test were used as the initial condition and a constant ET value of 1.225  10-4 m/hour
(Rain=0 during the drying test) was used as the external input (refer to Figure 2b) for the
water balance analysis. The numerical model simulated the entire 12-hour drying period
and the suction distribution at the end of the simulation was extracted to be compared
with the laboratory test results, as shown in Figure 3b. The dash lines represent the
laboratory test results for the initial and final suction distributions. The solid lines
represent the suction distributions at the end of the drying test based on the proposed
model by the authors and from Leung et al. (2015). At the starting point of the drying test
(right after the wetting test), the soil with elevation greater than 0.15 m had smaller
suction values, indicating that this part of the soil was relatively wet compared with the
underlying soil. After 12 hours of drying process, the evapotranspiration process took out
the water from the soil surface, resulting in higher suction value at an elevation of 0.28
m. Meanwhile, the water continued to percolate to the underlying soil and the soil
suctions continued to decrease for soils with elevations between 0.07 m and 0.20 m.
However, for soils lower than 0.07 m, the water front has not reached this point and the
suction distribution remained intact after 12 hours. For the simulation result provided by
Leung et al. (2015), it significantly deviated from the laboratory test results and their
model could not catch the phenomena of the increasing suction at the top of the soil due
to the imposed solar radiation boundary. In comparison, the simulation results based upon
181
the authors’ proposed model reasonably matched with the laboratory test results. The
effect of the evapotranspiration on the soil suction distribution was observed in both the
laboratory test and simulation results (the top 0.2 m soil). However, the simulated suction
values between 0.07 and 0.20 m were higher than the laboratory test results, indicating
that the provided soil permeability in the reference was lower than the actual value, and
the excess water did not percolate to the underlying soil. In sum, the simulation results of
the proposed FEM model matched well with the laboratory test results and the simulation
results could be more accurate when given a reasonable soil permeability data.
3. MODEL CONFIGURATIONS AND MATERIAL INPUTS
The model configuration is shown in Figure 4a. The flexible pavement is treated
as a two-layered system subjected to static wheel load. To reduce the calculation cost,
half of a two-lane road was simulated and the road was 0.5 m-thick in the longitudinal
direction. The pavement structure is consisted of a 0.1 m-thick Asphalt Concrete (AC)
and a 1.0 m-thick base course. The slope ratio (vertical: horizontal) is 1:3, with grass
hydroseeded on road slope. The grass root depth (H) is important in controlling the water
exchange between the soil and the ambient environment and assumes to be 0.1 m-thick.
An 80-kN (18-kip) single-axle load is often used as the standard vehicle or axle load in
pavement designs and is assumed to be the case for this simulation. Each wheel included
two tire loads (20 kN each) with a spacing of 0.34 m and the contact area assumed to be
circular with a radius of 0.1 m. Pressure was uniformly distributed on top of the AC layer
with a magnitude of 620 kPa. The groundwater table was assumed to be 0.5 m below the
182
ground level. At the top of the base course, there was a 0.1 m thick water infiltration layer
that was used to account for the precipitation infiltration. The water balance analysis of
this layer was similar to that for the soil within the grass root zone, but with the ET term
equals zero throughout the analysis (refer to Figure 2b). As for the boundary conditions,
the x-axis symmetrical boundary condition was applied at ① in Figure 4b. A fixed
boundary condition was applied at ② and a roller boundary condition was applied at ③.
The uniformly distributed wheel load was applied at ④. The displacement in z-axis was
zero for all the nodes at the x-y plane. Zhang (2005) performed the analogue analysis
between thermal-stress and consolidation problems, and used the thermodynamic analysis
in Abaqus software to simulate the consolidation theory for saturated-unsaturated soils.
This paper adopted this method and used 8-node trilinear displacement and temperature
elements (C3D8T), as shown in Figure 4c.
The material properties for AC and base course are summarized in Table 1.
Because this paper focused on the resilient behavior of the base course, the properties of
AC layer assumed to be elastic with constant modulus and Poisson’s ratio of 1,000 MPa
and 0.3, respectively. The AC layer was impermeable and the effect of precipitation
infiltration would be simulated via the water infiltration layer (refer to Figure 4a). To
reduce the complexity of the numerical model, both the base course and subgrade
assumed to be the same type of soil – Aggregate Base Class 3 (AB3) or well graded
gravel (GW-GC) according to Unified Soil Classification System (ASTM 2011). Lin et
al. (2018) performed a series of laboratory tests to characterize the hydraulic and
mechanical properties of the AB3 and here only summarize the test results. The
uniformity coefficient and coefficient of the gradation were 50 and 2.88, respectively.
183
The AB3 contained about 10% of fines with Liquid Limit (LL) and Plasticity Index (PI)
of 20 and 7, respectively. The optimum water content for the AB3 was 8.5% based on
modified proctor test and the corresponding maximum dry density was 2.1 Mg/m3. The
saturated permeability of the AB3 was 5.2  10-6 m/s and the saturation water content
was 12.5%. As for the mechanical property of the base course, the resilient modulus test
was performed according to AASHTO T309-99 (AASHTO 2003). The soil samples were
tested at different water contents (ranged from dry condition (0%) to saturation condition
(12.5%)). After the test, the universal model proposed by Uzan et al. (1992) was used and
modified to account for the influence of water content and stress state, as expressed in
Equation 9.
( -0.9174+0.2058wc ) (0.854-0.5493wc )
M R = ( 30.448 - 2.854  wc ) Pa ( / Pa ) × ( OCT / Pa +1) (9)
where M R = resilient modulus, MPa; Pa = atmospheric pressure (i.e., 101 kPa);  =
bulk stress, 1 +  2 +  3 , kPa; wc = water content, %; and  OCT = octahedral shear
stress, τOCT = 13 (1 −  2 ) + ( 2 −  3 ) + (1 −  3 )

2 2 2
, kPa.
For the hydraulic properties, the abilities of the AB3 to hold and transport water
are two very important characteristics and can be represented by its soil water
characteristic curve (SWCC) and its permeability function (also known as K-function).
The SWCC of the AB3 was determined via pressure plate test and salt concentration test,
and the regression curve can be expressed as in Equation 10. The K-function could be
derived from the combination of the saturated permeability and the SWCC based on the
method proposed by Kunze et al. (1962), as expressed in Equation 11.

184
 
0.412
w = 12.50  C ( ) 1/ ln 2.718 + ( 2.35 ) 
1.241
(10)
 
where, w = soil water content, %; C( ) = 1 − ln(1 + /  r ) / ln(1 + 106 /  r ) ;  = suction,
kPa.; and  r = residual suction, kPa.
k = 10
( −0.0727log
10 ( )4 + 0.6053log10 ( )3 −1.7158log10 ( )2 −0.6322log10 ( )−5.2956 )
(11)
where, k = permeability of the AB3; and  = suction.
4. SIMULATION RESULTS AND DISCUSSIONS
To evaluate the seasonal variations of base course resilient behavior, three
scenarios were simulated, representing different service conditions. Case I aimed at
evaluating the performance of a new pavement, in which the pavement surface was in a
good condition without any cracking. At this stage, the AC layer assumed to be
impermeable and capillary action was expected to be the major detrimental factor
influencing the pavement performance. For Case II, crackings were fully developed and
the pavement was moderate deteriorated with precipitation infiltrating into the pavement
structure. The purpose of this simulation was to quantify the effect of different rainfall
intensities and durations on the resilient behavior of a base course. Finally, Case III
aimed at evaluating the seasonal variations in the resilient behavior of a base course for a
moderately deteriorated pavement. The dynamic performance of the pavement structure
and the interactions between the pavement structure and vegetation under localized
climatic condition was assessed.

185
4.1. CASE I: NEW PAVEMENT WITHOUT CRACKING
Figure 5 shows the mean and shear stress distributions at different locations
beneath the road surface. Three critical locations were selected: pavement centerline
(center of the roadway), tire centerline (center of a single wheel), and axle centerline
(center of the dual wheels). The bottom of the base course was selected to be a datum. At
the pavement centerline, both mean and shear stresses linearly increased with depths
under the influence of the overburden pressure. The wheel pressure had limited influence
on the stress distributions at the pavement centerline due to a relatively long distance
between the two locations. However, the mean stresses at both tire and axle centerlines
first decreased under the influence of applied wheel load. As the wheel load influence
decreased, the mean stress started to increase at an elevation of 0.8 m. The shear stress
distribution at tire centerline decreased with elevation. However, the shear stress
distribution at axle centerline first increased from 24.0 kPa (an elevation of 1.0 m) to 42.0
kPa (an elevation of 0.8 m), and then gradually decreased to 17.8 kPa at the bottom of
base course.
Figure 6 shows the suction, water content, and resilient modulus variations at
three representative locations beneath the tire centerline: top, middle, and bottom of the
base course. The base course was compacted at the optimum water content of 8.5%,
corresponding to a suction value was 15.0 kPa and an average resilient modulus value of
262.0 MPa. These values will be considered as the references and will be used as the
designed value for further discussions. In general, the water flow could be divided into
two phases: free drainage and capillary action. For example, the soil suction at 1.0 m
increased from 15.0 kPa to 18.5 kPa and the corresponding water content decreased from
186
8.5% to 8.2% within 7 days, as shown in Figure 6a. This phenomenon indicated that
water was drained downward under the influence of gravitational force. In comparison,
the soil suction value decreased from 18.5 kPa to 15.0 kPa on Day 52, and remained
constant afterward. The capillary action was the major contributor to this suction
decrement because the groundwater table was only 0.5 m blow the datum. As for soils
between 0.0 m and 0.5 m, the capillary action dominated the water flow process since
suction values kept decreasing and the corresponding water contents continued to
increase. At the equilibrium condition, the resilient modulus values (refer to Figure 6c)
reduced by 30.6% at the top and 76.8% at the bottom of base course, respectively. The
simulation results indicated that even though the base course was designed to work under
the optimum water content of 8.5%, the post-construction water content inevitably
increased with time and the corresponding resilient modulus values were lower than the
designed values. Therefore, the pavement deteriorations would be expected to occur
earlier than expected under repetitive traffic load and the designed pavement service time
would also be shorter than the designed value.
To better demonstrate the moisture migration within the base course, Figure 7
further presents the suction, water content, and resilient modulus distributions at the tire
centerline at different times. At the steady-state condition, the suction distribution should
be linearly distributed and could be expressed as in Equation 12 and in Figure 7a:
 = − w z (12)
where,  = suction;  w = unit weight of water; and z = elevation.
A suction of 15 kPa was the designed value, as shown in Figure 7a. From Day 1
to Day 7, the soil suction with elevations higher than 0.6 m increased due to the free
187
drainage process. In comparison, from Day 7 to Day 52, the soil suction gradually
decreased due to the capillary action and water migrated upward to the base course.
Meanwhile, for soils with elevations lower than 0.6 m, the soil suction continuously
decreased, indicating that only capillary action dominated the water flow in this section.
In addition, the suction variations were significant from Day 1 to Day 30, indicating that
the water flow was transient. From Day 30 to Day 52, the suction values slightly
decreased till the steady-state was achieved on Day 52. All the suction values at the
steady-state were lower than the designed values, indicating that the soil water content
was higher than the designed value.
The water content variations with time were shown in Figure 7b. The designed
water content was its optimum water content which equals to 8.5%. From Day 1 to Day
7, the water contents for soils higher than 0.6 m was lower than the designed value,
indicating that the water flowed downward under the influence of gravity. Then, the
water contents started to increase from Day 7 to Day 30 and remained relative constant
until the equilibrium state was obtained on Day 52. In addition, only the top 0.1 m soil
meet the design value of 8.5%. This fact further proved the authors assertion that no
matter how well the pavement was constructed, the water content of the base course
would inevitably increase with time due to the existence of the shallow groundwater table
and capillary action.
Figure 7c further demonstrates the resilient modulus variations with time. Note
that the resilient modulus of the base course was a function of soil water content, bulk
stress, and shear stress. The designed values in Figure 7c was determined based upon the
same designed water content of 8.5% but with different stress levels. From Day 1 to Day
188
7, the resilient modulus for soils higher than 0.6 m increased with time due to the
decreasing water contents. However, the resilient modulus significantly decreased as the
capillary water migrated upward to the base course. The lower the soil elevation, the
closer the soil to the groundwater table and the lower the resilient modulus value would
be expected. For example, the designed resilient modulus values for soil at 0.0 m was 324
MPa while this value decreased to about 73.5 MPa at the equilibrium condition, which
reduced to about a quarter of the designed value.
In conclusion, the soil resilient modulus was very sensitive to soil water content
variations. The average post-compaction water content of the base course would be
higher than the designed value of 8.5%, and the corresponding resilient modulus would
also be much lower than the designed value with a maximum reduction of 75%.
4.2. CASE II: MODERATELY DETERIORATED PAVEMENT WITH RAINFALL

EVENTS
Precipitation infiltration and capillary action are two major sources that result in
pavement deteriorations (Elsayed and Lindly 1996; Huang 2004). Case I has evaluated
the influence of the capillary action. The infiltration water would penetrate into the
pavement structure via cracks and Van Sambeek (1989) reported that surface water
infiltration could account for as much as 90% to 95% of total moisture in a pavement
system. In practice, infiltration rate cannot exceed a fraction of the precipitation rate and
Ridgeway (1982) proposed Equation 13 to determine the infiltration rate. Case II aimed
at evaluating the influence of rainfall intensity and duration on the resilient behavior of
the base course and three scenarios were simulated. The total amount of cumulative
infiltration water was the same for all three scenarios (0.12 m/day) but with different
189
intensities and durations: (i) 0.005 m/h for 24 hours, (ii) 0.040 m/h for 6 hours, and (iii)
0.120 m/h for 1 hour. The time intervals between two rainfall events were 60 days so that
the influence of the previous rainfall event would not influence the upcoming one.
Wp
q = 0.1( N + 1 + ) (13)
Cs
where, q = infiltration rate; N = number of traffic lanes; W p = width of the pavement
subjected to infiltration; Cs = spacing of transverse cracks.
Figure 8 shows the suction, water content, and resilient modulus variations with
time at different locations beneath the tire centerline. The 1st rainfall event occurred on
Day 60 when the pavement system already reached equilibrium condition. As shown in
Figure 8a, the entire base course was influenced by the 1st rainfall event as the soil
suction at the bottom of the base course slightly decreased from 5.0 kPa to 4.7 kPa three
days after the rainfall event. In addition, the suction variations for the top 0.4 m soil were
more significant because the soil there was closer to the top of the base course where
infiltration water came into the base course. Similar trends were also observed for the 2nd
and 3rd rainfall events. However, it is important to point out that the influence of rainfall
duration was much significant than the intensity. The reasons may be explained in two
folds. Firstly, a longer rainfall duration required longer time for the suction to recover to
the same level. For example, the 1st rainfall event lasted for 24 hours and it took 30 days
for the suction value to recover to the value before the rainfall event (15 kPa). However,
this recovering time was only 20 days and 15 days for the 2nd and 3rd rainfall events,
respectively. Secondly, longer rainfall events would result in higher magnitudes of
suction changes. For instance, the maximum suction variation for the 1st rainfall event
190
was observed at 1.0 m and the suction decreased by 8.8 kPa. In comparison, the
maximum suction variations for the 2nd and 3rd rainfall events were observed at the same
elevation but with lower magnitudes of 8.4 kPa and 5.4 kPa, respectively. In conclusion,
rainfall events with longer durations would result in a higher magnitude of suction
change and a longer recovering time to reach a steady-state again.
Similar observations were observed in the water content, as shown in Figure 8b.
For the 1st rainfall event, the intensity was mild (5 mm/day) but lasted longer (24 hours).
The soil at the top of the base course became wet (but not saturated) and its permeability
significantly increased with the increasing soil water content. As long as the soil was not
saturated, all the infiltration water could be absorbed by the water infiltration layer (refer
to Figure 4a) and would percolate to the underlying soil. However, for the 2nd and 3rd
rainfall events, the soil within the water infiltration layer was saturated and any additional
water would be considered as runoff water and could not infiltrate into the pavement
structure. This is why the water content variations in the middle was less significant in
the 2nd and 3rd rainfall events than in the 1st one. In other words, the influence of the
rainfall duration on soil resilient behavior was more important than the rainfall intensity.
Figure 8c further demonstrates the resilient modulus variations with time during
the rainfall events. As mentioned in the Introduction section, the resilient modulus of a
base course was determined via FWD test on a bi-weekly or monthly basis. However,
those values only represented the resilient modulus at that period of the year (AASHTO
2008) and might not accurate enough to represent the monthly average value. For
example, during the 1st rainfall event, the resilient modulus values decreased from 176
MPa to 59.3 MPa within one day, and increased to 159 MPa after two weeks. Suppose
191
the field engineer performed the FWD test one day after the rainfall stopped, the value of
59.3 MPa was obviously much lower than the monthly average value. Therefore, the
resilient behavior of the base course was significantly influenced by the rainfall intensity
and duration. The numerical simulation of the seasonal variations of the resilient modulus
of the base course should be conducted using hourly climatic data so that the influence of
precipitation infiltration could be reasonably reflected. In addition, the numerical
simulation results can also provide guidance to the field engineers regarding the
appropriate time to conduct field test. For example, if the rainfall duration was relatively
long (e.g. the first rainfall event), engineers should wait at least 2-3 days until the resilient
modulus was recovered to at least 80% of the value before the rainfall event. However, if
the duration of the rainfall was very short, most of the water would be runoff rather than
infiltrate into the pavement structure and the field engineers could perform the test one
day after the rainfall event. It is important to point out that the although EICM provides
environmental data on an hourly basis, the analysis period (design life) was divided into
1-month or 2-weeks periods according to the MEPDG manual (ARA 2004). That is to
say, the simulation of the resilient behavior based on the EICM model was accurate
enough to capture the effect of rainfall intensity and duration on the resilient behavior of
a base course while the proposed model has such capability.
To further quantify the amount of infiltration water that entered into the pavement
structure, Figure 9 shows the amount of precipitation and runoff water during the rainfall
event. The area between the two curves would be the total amount of infiltration water.
As shown in Figure 9a, during the 1st rainfall event, even though the intensity of the
rainfall was not significant (only 5 mm/hr), it lasted for 24 hours. There was no runoff
192
water during the entire rainfall event and all the infiltration water went into the pavement
structure (in total 0.12 m). In comparison, runoff water was observed during the 2nd
rainfall event and the total amount of infiltration water was only 0.0498 m (Figure 8b),
which account for 41.5% of the 1st rainfall event. It is important to emphasize again that
the influence of rainfall duration was much significant than its intensity in affecting the
resilient behavior of the base course. As the rainfall intensity continued to increase to
0.12 m/hour and its duration time continued to decrease to 1 hour (Figure 9c), there were
even less amount of infiltration water. During the 3rd rainfall event, there was only 0.009
m of water infiltrated into the pavement structure, which was only 7.5% and 18.1% of
that for the 1st and 2nd rainfall events. In comparison, the amount of infiltration water that
enters into the top of the unbounded base course was not accurate in the EICM model. In
the MEDPDG approach, the infiltration can assume four values – none, minor, moderate,
and extreme (0%, 10%, 50%, and 100% of precipitation enters the pavement). However,
based on the simulation results in this paper, given the same amount of cumulative
precipitation, the rainfall event with a longer duration would have a much devastating
influence on the soil resilient behavior because it allowed the base course to be exposed
to the wet condition for a longer time. In other words, the proposed model is more
accurate than the EICM model in predicting the resilient behavior of a base course.
Figure 10 shows the resilient modulus contours before and after each rainfall
event. Four representative time frames were selected, including 1day before the rainfall
event, right after, 1 day, and 7 days after the rainfall event. The blue (cold) color
represents wet soil with low resilient modulus values while the red (warm) color
represents dry soil with high resilient modulus values. As shown in Figure 10a, the soil
193
resilient modulus was very high at the road slope with the highest value of 1300 MPa
(upper left image). As the elevation of the soil on the road slope decreased, the soil
suction was expected to decrease due to the existence of the shallow groundwater. A
lower suction value corresponded to a higher soil water content and a lower resilient
modulus. Then the rainfall event lasted for 24 hours and the precipitation water was
allowed to infiltrate from both the grass root zone (road slope) and the water infiltration
layer. As shown in the upper right image in Figure 10a, the maximum soil resilient
modulus on the road slope significantly decreased from 1300 MPa to 520 MPa, reduced
to approximately half of the value before the rainfall event. The resilient modulus value
for the soil within the pavement structure even decreased from 175 MPa to 60.2 MPa,
reduced to one third of the value before the rainfall event. This observation indicated that
the effect of infiltration water had a stronger influence within the pavement than on the
road slope. After 1 day (lower left image in Figure 10a), the soil at the top of the base
course was still in wet condition. The relatively drier soil in the middle had lower
permeability, impeding the excess water from percolating to the underlying soil. The
resilient modulus contour 7 days after the rainfall event showed similar trend as before
the rainfall event.
Figure 10b shows the resilient modulus contours for the 2nd rainfall event.
Because the time lag between the 1st and 2nd rainfall events was 60 days, the influence of
the 1st rainfall event was negligible. Similar to the 1st rainfall event, the resilient modulus
was much higher on the road slope than within the pavement structure (upper left image).
However, due to the higher rainfall intensity, the soils at the top of the base course and on
the road slope were saturated right after the rainfall event (upper right image). The
194
saturated soil impeded additional precipitation water from infiltrating into the pavement
structure, resulting in a relatively drier condition 1 day after the rainfall (lower left
image) when compared with that for the 1st rainfall event. After 7 days, the soil resilient
modulus contours recovered to the condition before the rainfall event. As for the 3rd
rainfall event (Figure 10c), there was only 0.009 m of water infiltrated into the pavement
structure. This fact explained why the resilient modulus values 1 day after the rainfall
event were much higher in the 3rd rainfall event than those in the 1st and 2nd ones.
Moreover, the EICM model is a 1D model that cannot be used to simulate the moisture
migration in the transverse direction. In comparison, the simulation results in this paper
indicated that the soil on the slope of the road embankment was affected by the ambient
environment compared with the soil within the pavement structure at the same elevation.
In summary, the resilient behavior of the base course was sensitive to water
content variations. With the same amount of cumulative precipitation, the rainfall event
with longer duration would cause the most significant resilient modulus reduction. The
rainfall duration played a more significant role than the rainfall intensity in influencing
the base course resilient behavior.
4.3. CASE III: PAVEMENT WITH CRACKS AND REAL TIME CLIMATIC
DATA
To reveal the dynamic base course resilient behavior under the influence of
localized climatic condition, hourly meteorological data presented in Figure 1d was
applied to the model. Cedergren (1974) proposed Equation 14 to correlate the infiltration
rate with rainfall rate, and recommended a constant parameter, C, to be 0.33 to 0.5 for
195
flexible pavement. In this paper, the C value of 0.5, was used in this paper, representing
the scenario of a pavement structure with fully developing cracking.
q = CR (14)
where, q = infiltration rate; C = fraction of infiltration that enters into pavement, equals
to 0.5; and R = rainfall rate.
Figures 11a-c show the suction, water content, and resilient modulus variations
with time. The significant drops in each figure indicated that there was a rainfall event
occurred on that day. For example, the recorded precipitation on Day 10 was 2.54 mm,
and the suction and water content changes during that day were 9 kPa and 0.9%,
respectively. However, the amplitude of variation not necessarily reflected the rainfall
intensity because soil field capacity also played a critical role in determining the amount
of water infiltrating into pavement structure. For instance, a heavy rainfall was observed
on Day 246 (with a precipitation of 117.6 mm). However, the suction and water content
variations were only 1 kPa and 0.15%, which were much smaller than those on Day 10. If
the soil water content was higher than the field capacity, most of the precipitation water
would be considered as runoff water. The difference between in-situ soil water content
and soil field capacity depicts the water storage capacity. The soil water contents at Day
10 and 246 were 8.27% and 8.7%, indicating that the water storage capacity for Day 10
was higher than that for Day 246. In conclusion, the water content variation within base
course was a combined effect of precipitation and soil field capacity.
AASHTO (1993) proposed the effective soil resilient modulus as an equivalent
value which account for the climatic effect on the resilient behavior of a base course.
Since daily modulus value was available, the relative damage (defined in Equation 15a),
196
u f , was determined every day and the effective soil resilient modulus was back
calculated based on Equation 16:
u f = 1.18 108 M R −2.32 (15a)
uf =
u f
(15b)
n
1.18 108
M R = 2.32 (16)
uf
where, u f = relative damage; M R = resilient modulus; n = time interval between
adjacent modulus values are specified; u f = average relative damage; and M R =
effective soil resilient modulus.
In general, two techniques are commonly used in determining the nonlinear
characteristics of a typical material. One way is to divide the nonlinear layer into several
sublayers and analyze the midpoint of each layer (Harichandran et al. 1989; Raad and
Figueroa 1980). The other method is to simply use the midpoint of the nonlinear layer to
conduct the analysis. In this paper, both methods were used to determine the effective soil
resilient modulus. Realizing the conventional base course was 0.2m-0.3 m thick, the
bottom 0.3 m of the base course was divided into three sublayers, and the midpoint of
each layer was used to determine the effective resilient modulus. The average resilient
modulus of the three sublayers was used as the effective resilient modulus of the base
course. The second method is to simply use the resilient modulus value of the midpoint of
the entire base course (at 0.5 m) as the effective value. Table 2 summarizes the effective
soil resilient modulus at the locations mentioned above. Based on the first method, the
197
average water content of the lower 0.3 m base course was 9.87%. The effective resilient
modulus was 198 MPa, which was reduced by 51.2% compared with the average
designed value of 283 MPa. As for the second method, the resilient modulus value was
136 MPa and reduced by 54.3% compared with the designed value. In summary, the
effective resilient modulus of the base course was only half of the designed value.
As shown in Figure 12a, two points were selected within the grass root zone to
demonstrate the water content variations in the vegetated area. Point A was selected to be
at the top of the base course and was relatively far away from the groundwater table.
Point B was selected to be at the ground level which was 0.5 m above the groundwater
table. Figure 12b shows the water content variations with time for the selected two points.
Meanwhile, the precipitation data was also showed in the figure as a reference. The water
generation term (Equation 1d) reflected the actual amount of water infiltrated into the
vegetation area, which not only depends upon the magnitude of precipitation and ET, but
also relied on the current state of water storage. For example, even though the recorded
precipitation data on Day 246 was 117.6 mm, the actual amount of water infiltrated into
the base course was only 2.03 mm. This phenomenon was related to the two adjacent
rainfall events observed on Day 241 and Day244. Due to these two rainfall events, the
water content surpassed the filed capacity (FC) and the water storage capacity had
reached its upper limit. The rest of precipitation water would runoff rather than continued
entering into the pavement structure. Therefore, the total amount of infiltration water was
a function of the precipitation and also the current water state.

198
5. CONCLUSIONS
This paper aimed at predicting the seasonal variations of the resilient behavior of
a base course under localized climatic conditions. A coupled hydro-mechanical-climatic
model was proposed to evaluate interactions between the soil infrastructure and the
ambient environment. Numerical simulations were performed to evaluate the dynamic
performance of a pavement structure under different working conditions. Based on
simulation results, the major conclusions are summarized as follows:
1. The proposed coupled hydro-mechanical-climatic model can be used to predict
the seasonal variations of the resilient behavior of the base course. The proposed
model is suitable for saturated and unsaturated soils and is able to evaluate the
interactions between the pavement structure and the ambient environment.
2. Compared with the 1D EICM model, the proposed model is a three-dimensional
model which is more accurate in predicting the moisture migration within the
pavement structure. Moreover, the proposed model can be used to evaluate the
influence of rainfall intensity and duration on the resilient behavior of a base
course. In addition, the proposed model is also able to simulate the influence of
the vegetation on the resilient behavior of a pavement structure.
3. For a new flexible pavement with shallow groundwater table (Case I), the major
detrimental factor affecting the resilient behavior of a base course was capillary
action. Its influence could last for 52 days until equilibrium condition was
achieved. Meanwhile, the resilient modulus distribution was highly nonlinear,
depending on the stress state and the soil water content. Based on the simulation
199
results, the resilient modulus would reduce to a quarter of the designed value.
Therefore, the resilient modulus of the base course was sensitive to water content
variations.
4. The simulation results for Case II indicated that for a moderately deteriorated
pavement structure, given the same amount of cumulative precipitation (0.12 m),
the influence of rainfall duration was more significant than rainfall intensity in
reducing the soil resilient modulus. For longer duration rainfall events, more
water will be infiltrated into the pavement structure and the corresponding
resilient modulus would be lower. In contrast, if the rainfall intensity was high,
most of the water would runoff rather than flowed into the pavement structure, the
reduction in resilient modulus was not significant as that in rainfall events with
longer durations.
5. The simulation results for Case III indicated that the effective resilient modulus of
the base course material was approximately half of the designed value. The
existing subsurface design methods only considered water flow in saturated soils
and overestimated the drainage ability of the base course. The total amount of
infiltration water was a function of precipitation and current water state.
6. Grass vegetation effectively served as a protective layer to prevent excess water
from infiltrating into the base course. The precipitation and evapotranspiration
data served as the lower and upper limit for soil water storage and played
important roles in keeping the water balance within the grass root zone.
200
Table 1. Material inputs

Mechanical Properties Hydraulic Properties
Material Modulus, E Poisson’s
SWCC K-Function
(MPa) Ratio,ν
0
AC 1,000 0.3 N/A
(impermeable)
AB3 Equation 9 0.35 Equation 10 Equation 11
Table 2. Effective soil resilient modulus

Designed MR
Evaluation Elevation uf Average w MR MR Reduction
Technique
(m) (%) (MPa) (MPa) (%)
0.25 0.0169 9.68 244 282 13.5
Midpoint of
0.15 0.0243 9.86 103 296 65.2
Sublayers
0.05 0.0452 10.08 78.9 314 74.9
Average 0.25 0.0174 9.87 198 283 51.2
Midpoint 0.5 0.0129 9.55 136 260 54.3
201
(a)
(b)
Figure 1. Meteorological data: (a) air temperature, (b) relative humidity, (c) wind speed,
and (d) precipitation and calculated reference evapotranspiration
202
(c)
(d)
Figure 1. Meteorological data: (a) air temperature, (b) relative humidity, (c) wind speed,
and (d) precipitation and calculated reference evapotranspiration (cont.)
203
(a)
(b)
Figure 2. Water balance analysis for soil-climatic interactions: (a) schematic plot of water
balance in a soil element, and (b) water balance analysis flow chart
204
(a)
(b)
Figure 3. FEM model calibration: (a) schematic plot of test setup, and (b) comparisons of
laboratory test and simulation results
205
(a)
(b)
(c)
Figure 4. Model configurations: (a) geometry, (b) boundary conditions, and (c) mesh
generation
206
Figure 5. Stress distributions at different locations
(a)
Figure 6. Simulation results for Case I (a new pavement without cracks): (a) suction
variations with time, (b) water content variations with time, and (c) resilient modulus
variations with time
207
(b)
(c)
Figure 6. Simulation results for Case I (a new pavement without cracks): (a) suction
variations with time, (b) water content variations with time, and (c) resilient modulus
variations with time (cont.)
208
(a)
(b)
Figure 7. Suction, water content, and resilient modulus distributions at tire centerlines: (a)
suction distributions at different times, (b) water content distributions at different times,
and (c) resilient modulus distributions at different times
209
(c)
Figure 7. Suction, water content, and resilient modulus distributions at tire centerlines: (a)
suction distributions at different times, (b) water content distributions at different times,
and (c) resilient modulus distributions at different times (cont.)
210
(a)
(b)
developed cracks): (a) suction variations with time, (b) water content variations with
time, and (c) resilient modulus variations with time
211
(c)
developed cracks): (a) suction variations with time, (b) water content variations with
time, and (c) resilient modulus variations with time (cont.)
212
(a)
(b)
Figure 9. Comparisons of total amount of water infiltrated into the pavement structure:
(a) 1st rainfall event (0.005 m/hr for 24 hrs), (b) 2nd rainfall event (0.040 m/hr for 6 hrs),
and (c) 3rd rainfall events (0.120 m/hr for 1 hr)
213
(c)
Figure 9. Comparisons of total amount of water infiltrated into the pavement structure:
(a) 1st rainfall event (0.005 m/hr for 24 hrs), (b) 2nd rainfall event (0.040 m/hr for 6 hrs),
and (c) 3rd rainfall events (0.120 m/hr for 1 hr) (cont.)
214
(a)
(b)
Figure 10. Resilient modulus contours during rainfall events: (a) 1st rainfall event, (b) 2nd
rainfall event, and (c) 3rd rainfall event
215
(c)
Figure 10. Resilient modulus contours during rainfall events: (a) 1st rainfall event, (b) 2nd
rainfall event, and (c) 3rd rainfall event (cont.)
216
(a)
(b)
Figure 11. Simulation results for Case III (moderately deteriorated pavement with real-
time meteorological data): (a) suction variations with time, (b) water content variations
with time, and (c) resilient modulus variations with time
217
(c)
Figure 11. Simulation results for Case III (moderately deteriorated pavement with real-
time meteorological data): (a) suction variations with time, (b) water content variations
with time, and (c) resilient modulus variations with time (cont.)
218
(a)
(b)
Figure 12. Water content variations with time at selected vegetation areas: (a) selected
locations, and (b) water content variations
219
REFERENCES
AASHTO (1993). Guide for Design of Pavement Structures, 1993, AASHTO.
AASHTO (2003). “Standard Method of Test for Resilient Modulus of Subgrade Soils and
Untreated Base/Subbase Mateirals.” AASHTO T309-99. Washington, DC:
AASHTO.
AASHTO (2008). "Mechanistic-Empirical Pavement Design Guide-A Manual of

Practice." AASHTO Officials, Editor.
Abaqus, V. (2014). "6.14 Documentation." Dassault Systemes Simulia Corporation.
Allen, R. G., Pereira, L. S., Raes, D., and Smith, M. (1998). "Crop evapotranspiration-
Guidelines for computing crop water requirements-FAO Irrigation and drainage
paper 56." FAO, Rome, 300(9), D05109.
ARA, I., ERES Consultants Division. (2004). "Guide for Mechanistic–Empirical Design
of New and Rehabilitated Pavement Structures." Final Report, NCHRP Project 1-
37A.
ASTM (2011). "Standard Practice for Classification of Soils for Engineering Purposes
(unified Soil Classification System)." ASTM International.
ASTM D420-18 (2018). “Standard Guide for Site Characterization for Engineering
Design and Construction Purposes.” ASTM International, West Conshohocken,
PA, https://doi.org/10.1520/D0420-18.
ASTM D4694-09 (2015). “Standard Test Method for Deflections with a Falling-Weight-
Type Impulse Load Device.” ASTM International, West Conshohocken, PA,
https://doi.org/10.1520/D4694-09R15.
Biot, M. A. (1941). "General theory of three‐dimensional consolidation." Journal of

applied physics, 12(2), 155-164.
Brutsaert, W. (2013). Evaporation into the atmosphere: theory, history and applications,
Springer Science & Business Media.
Cedergren, H. R. (1974). Drainage of highway and airfield pavements.
Christopher, R.B., Schwartz, C., and Boudreau, R. (2010). “Geotechnical Aspects of

Pavements: Reference Manual.” US Department of Transportation, Federal
Highway Administration, FHWA NHI-05-037.
220
Dempsey, B.J., Herlach, W.A. and Patel, A.J. (1985). “The Climatic-Materials-Structural
Pavement Analysis Program.” Federal Highway Administration (FHWA),
Washington, DC, Final Report, FHWA/RD-84-115, Volume 3.
Elsayed, A., and Lindly, J. (1996). "Estimating permeability of untreated roadway bases."
Transportation Research Record: Journal of the Transportation Research
Board(1519), 11-18.
Fredlund, D., and Morgenstern, N. (1976). "Constitutive relations for volume change in
unsaturated soils." Canadian Geotechnical Journal, 13(3), 261-276.
Fredlund, D.G., Rahardjo, H, and Fredlund, M.D. (2012). “Unsaturated Soil Mechanics
in Engineering Practice.” John Wiley & Sons.
Gray, D. M. (1970). "Handbook on the Principles of Hydrology."
Guymon, G.L., Berg, R.L., and Johnston, T.C. (1986). “Mathematical Model of Frost
Heave and Thaw Settlement in Pavements.” U.S. Army Cold Regions Research
and Engineering Laboratory (CRREL), Hanover, New Hampshire.
Harichandran, R., Baladi, G., and Yeh, M. (1989). "Development of a computer program
for design of pavement systems consisting of bound and unbound materials."
Department of Civil and Environmental Engineering, Michigan State University.
Henry, K. S., and Holtz, R. D. (2001). "Geocomposite capillary barriers to reduce frost
heave in soils." Canadian Geotechnical Journal, 38(4), 678-694.
Huang, Y. H. (2004). "Pavement analysis and design."
Kunze, R. J., Uehara, G., and Graham, K. (1962). "Factors important in the calculation of
hydraulic conductivity." Soil Science Society of America Journal, 32(6), 760-765.
Larson, G., and Dempsey, B.J. (1997). “Enhanced Integrated Climatic Model, Version
2.0, Final Report.” Contract DTFA MN/DOT 72114, Department of Civil
Engineering, University of Illinois at Urbana-Champaign, Urbana, IL.
Leung, A.K., Garg, A., and Ng, C.W.W. (2015). "Effects of Plant Roots on Soil-Water
Retention and Induced Suction in Vegetated Soil." Engineering Geology, 193, 183-
197.
Li, L., Liu, J., Zhang, X., and Saboundjian, S. (2011). "Resilient modulus characterization
of Alaska granular base materials." Transportation Research Record: Journal of
the Transportation Research Board, 2232, 44-54.
221
Lin, C., Zhang, X., and Han, J. (2018). "Comprehensive material characterization of
pavement structure installed with wicking fabrics." Journal of Materials in Civil
Engineering, 31(2), 04018372.
Liu, S.J., and Lytton, R.L. (1985). “Environmental Effects on Pavement Drainage”.
Federal Highway Administration (FHWA), Washington, DC, Final Report, No.
FHWA-DTFH61-87-C-00057, Volume IV.
Miller, E. (1953). "THE WASHO ROAD TEST." Highway Res Abstracts Hwy Res Board.
Morton, F. (1975). "Estimating evaporation and transpiration from climatological

observations." Journal of Applied Meteorology, 14(4), 488-497.
Raad, L., and Figueroa, J. L. (1980). "Load response of transportation support systems."
Journal of Transportation Engineering, 106(1).
Richter, C.A. (2006). “Seasonal Variations in the Moduli of Unbound Pavement Layers.”
Federal Highway Administration (FHWA), Washington, DC, Final Report, No.
FHWA-HRT-04-079.
Ridgeway, H. H. (1982). "Pavement subsurface drainage systems." NCHRP synthesis of

highway practice(96).
Roberson, R., and Siekmeier, J. (2002). "Determining material moisture characteristics for
pavement drainage and mechanistic empirical design." Research Bulletin.
Minnesota Department of Transportation. Office of Materials and Road Research.
Romero, C.C., and Michael D.D. (2016). "Review of turfgrass evapotranspiration and crop
coefficients." Transactions of the ASABE, 59(1), 207-223.
Shahin, M.Y. (1994). “Pavement Management for Airports, Roads and Parking Lots.”
Kluwer Academic Publishers, Dordrecht, Hardbound, ISBN 0-412-99201-9.
Tran, D. T., Fredlund, D. G., and Chan, D. H. (2015). "Improvements to the calculation of
actual evaporation from bare soil surfaces." Canadian Geotechnical Journal, 53(1),
118-133.
Uzan, J., Witczak, M. W., Scullion, T., and Lytton, R. L. (1992). “Development and
validation of realistic pavement response models.” In International Conference on
Asphalt Pavements, 7th, 1992, Nottingham, United Kindom.
Van Sambeek, R. J. (1989). "Synthesis on subsurface drainage of water infiltrating a

pavement structure."
222
Wilson, G. W., Fredlund, D., and Barbour, S. (1994). "Coupled soil-atmosphere modelling
for soil evaporation." Canadian Geotechnical Journal, 31(2), 151-161.
Wilson, G.W. (1997). “Surface Fluxes for Geotechnical Engineering Problems.” Ph.D.
Dissertation, University of Saskatchewan, Saskatoon, SK.
Wilson, G.W., Fredlund, D.G., Barbour, S.L., and Pufahl, D.E. (1997). “The Use of
Ground Surface Moisture Flux Boundary Conditions in Geotechnical
Engineering.” In Proceeding of the International Conference on Soil Mechanics
and Foundation Engineering, AA, Balkema, 3, 1861-1864.
Witczak, M.W., Andrei, D., and Houston, W.N. (2000). “Resilient Modulus as Function
of Soil Moisture – Summary of Predictive Models.” Development of the 2002
Guide for the Development of New and Rehabilitated Pavement Structures,
NCHRP 1-37A, Inter Team Technical Report (Seasonal 1-4), Tempe, Arizona.
WRCC (2017). <https://raws.dri.edu/>. (latest access on June 15, 2018)
Wyatt, T., Barker, W., and Hall, J. (1998). “Drainage Requirements in Pavements User
Manual.” Federal Highway Administration (FHWA), Washington, DC, Final
Report, No. FHWA-SA-96-070.
Zhang, X., and Liu, J. (2008). "Numerical Simulation of Influence of Climatic Factors on
Concrete Pavements Built on Expansive Soil." GeoCongress 2008:
Geosustainability and Geohazard Mitigation, 554-561.
Zhang, X., Lytton, R., and Briaud, J. (2005). "Coupled consolidation theory for saturated-
unsaturated soils." Proc., Proc., 3rd Biot Conf. on Poromechanics (Biot
Centennial), Univ. of Oklahoma, 323-330.
223
V. WORKING MECHANISM OF A NEW WICKING GEOTEXTILE IN

ROADWAY APPLICATIONS: A NUMERICAL STUDY
Chuang Lin1, Xiong Zhang2, Javad Galinmoghadam3 and Yipeng Guo4a, 4b
1
2
(Corresponding Author) Associate Professor, M. ASCE, Department of Civil,
3
Graduate Research Assistant, Department of Civil, Architectural, and Environmental
Engineering, Missouri University of Science and Technology, Phone: (573) 466-1076,
Email: [email protected]
4a
Graduate Research Assistant, Department of Civil Engineering, Central South
University, Changsha, Hunan, China, 410075, Phone: +8615700732101, Email:
[email protected]
4b
Visting Scholar, Department of Civil, Architectural, and Environmental Engineering,
Missouri University of Science & Technology, Rolla, MO 65409-0030, Phone: (573)
ABSTRACT
Geotextiles have been widely used in roadways for reinforcement, separation,
filtration, and drainage purposes. In general, nonwoven geotextiles are too weak for
reinforcement purposes while woven geotextiles are often to be used for reinforcement
purposes due to their high tensile strength. Normally when considering the mechanical
reinforcement of a woven geotextile, the hydraulic properties of the geotextile is not
considered. However, the capillary break effect jeopardizes the reinforcing benefits under
224
unsaturated conditions by impeding water from passing through the geotextile. Excess
water will be accumulated at the soil-geotextile interface. The overall performance of a
soil-geotextile system relies on the soil, the geotextile, and the soil-geotextile interaction.
No matter how strong the geotextile is, the soil-geotextile composite will be still weak if
the soil remains wet. Therefore, the lack of lateral drainage capability under unsaturated
conditions is the major limitation of the conventional woven geotextiles.
A new woven geotextile has been recently developed to drain both capillary and
free water under both saturated and unsaturated conditions. It is a dual-functional woven
geotextile for both reinforcement and drainage purposes. It works as a capillary barrier in
the cross-plane direction meanwhile provides lateral drainage capability in the in-plane
direction. Although both laboratory and field tests have validated its effectiveness in
dehydrating road embankments, there is no existing method to quantify the benefits of the
wicking geotextile. This paper aims at exploring the working mechanism of the wicking
geotextile and quantifying the benefits of the wicking geotextile in term of its drainage
performance in a pavement structure. Firstly, a series of laboratory tests were performed
to characterize the hydraulic properties of the wicking geotextile. Secondly, a numerical
model was established and calibrated to simulate the water flow in both saturated and
unsaturated soils. Thirdly, based on the calibrated model, the drainage performance of the
wicking geotextile under different working conditions was simulated and evaluated.
Finally, the drainage performance of a road embankment installed with the wicking
geotextile was assessed.
KEY WORDS: Woven Geotextile, Unsaturated Soil, Capillary Barrier,
Reinforcement, and Subsurface Drainage

225
1. INTRODUCTION
Geotextiles have been widely used in roadways for reinforcement, separation,
filtration, and drainage purposes. In general, nonwoven geotextiles are too weak for
reinforcement purposes (Han 2015) while woven geotextiles are often to be used for
reinforcement purposes due to their high tensile strength. The improved performance of
pavement due to reinforcement has been attributed to three mechanisms: lateral restraint,
increased bearing capacity, and tensioned membrane effect (Zornberg 2011; Giroud and
Noiray 1981; and Holtz et al. 1998). However, normally when considering the
reinforcing effect of a woven geotextile, the hydraulic behavior of the woven geotextile is
not a major design parameter and the influence of hydraulic properties on the reinforcing
effect is often ignored. This ignorance significantly devastates the reinforcing effect of a
woven geotextile under unsaturated conditions. Since the overall stiffness of a soil-
geotextile composite relies on the strengths of the soil, the geotextile, and the interfacing
friction between those two materials. If the soil remains relatively wet, no matter how
strong the geotextile is, the overall stiffness of the soil-geotextile composite will still be
weak.
Unfortunately, a woven geotextile will cause the buildup of positive pore water
pressure in the overlying soil and deteriorate the pavement structure with time under
repetitive traffic load. Woven geotextiles are predominantly made of polypropylene and
polyester, which are hydrophobic materials (Koerner 2012; and Henry 1995). Figure 1a
demonstrates the ability of a woven geotextile to hold water under unsaturated
conditions. The water droplet will not break through the geotextile until sufficient pore
226
water pressure is built up. In other words, woven geotextiles may act to retard drainage
under unsaturated conditions. The mechanism for the phenomenon in Figure 1a can be
explained via the capillary break effect (Stormont and Anderson 1999). A capillary break
develops when two geomaterials with differing pore sizes are in contact with one another
(Zornberg 2010). In general, a geotextile is a porous material with relatively larger pores
compared to the pores in soils, as schematically plotted in Figure 1b. Under unsaturated
conditions, water within pores of geomaterials is held against gravity by a combination of
both absorptive pressure (osmotic suction) and capillary pressure (matric suction) (Olson
and Langfelder 1965). The osmotic suction is typically considered as a constant and
consequently the capillary phenomenon is only associated with the matric suction
component (Zornberg 2010). The influence of pore sizes on the magnitude of matric
suction can be expressed as in Equation 1 (Fredlund et al. 2012).
2Ts cos 
= (1)
R
where  is suction, Ts is surface tension,  is contact angle, and R is radius of the
pipette.
From Equation 1, the matric suction is inversely proportional to the radius of the
pore size. The geotextile has larger pore sizes (similar to a coarse-grained soil) with a
larger radius of r2 and the corresponding suction value is relatively small compared with
the suction value in the adjacent soil. Therefore, the pore water within the overlying soil
needs higher energy to break into large pores within the geotextile. The geotextile
impedes the water flow from the soil to the geotextile and will cause the buildup of
positive pore water pressure in the overlying soil. Numerous researchers have reported
227
that the moisture storage from a capillary break effect can be detrimental to the long-term
performance of a roadway system (Clough and French 1982; Giroud et al. 2000;
McCartney et al. 2005; Stormont and Anderson 1999; Stormont and Morris 2000). Figure
1c demonstrates that a woven geotextile can impede infiltrating water from percolating to
the underlying soils and will cause water ponding in the overlying soil. To make things
worse, conventional geotextiles can only provide gravity-induced lateral drainage under
saturated (nearly saturated) conditions and do not provide such drainage under the most
typical unsaturated conditions (Zornberg et al. 2017). That is to say, the capillary cannot
be drained out of a pavement structure under unsaturated conditions. On the other hand,
the stiffness of a base course is sensitive to water content variations, especially when the
fine content is higher than 4% (Siswosoebrotho et al. 2005). This means that a
measurable amount of water would be detained in a base course closer to the geotextile
location. The excess water will inevitably soften the base course and the stiffness of the
soil-geotextile composite will be significantly reduced. The excess water accelerates
pavement deteriorations under repetitive traffic load.
In summary, for a woven geotextile, the lack of lateral drainage capability under
unsaturation conditions is a major limitation that will cause the buildup of excess pore
water pressure and will accelerate pavement deteriorations with time.
2. A NEW WOVEN GEOTEXTILE WITH LATERAL DRAINAGE CAPABILITY
A new woven geotextile with wicking fibers (hereafter will be denoted as the
“wicking geotextile”), which has the capability to drain the excess water laterally in a
228
roadway under both saturated and unsaturated conditions, has been recently developed
and has been proved to be effective in dehydrating road embankments. As shown in
Figure 2a, the wicking geotextile is a dual-function geotextile, composing of
polyethylene yarns (black) for the reinforcing purpose and a specially designed
hydrophilic and hygroscopic nylon fiber yarn (white) for the drainage purpose. Figure 2a
also shows the schematic plot of the cross-section of the wicking geotextile. Within a
weft yarn set, the first type of yarns in the cross-machine direction has a larger cross-
sectional area. The wicking fiber yarns are in parallel with the first yarns, and are
distributed at top, middle, and bottom of the wicking geotextile. This weaving structure
ensures that the wicking geotextile can absorb water from both the overlying and
underlying soil. Another warp yarns (in the machine-direction) is interweaving with the
first weft yarns and wicking fiber yarns to maintain the weaving pattern.
Figure 2b shows the microstructure of a single wicking fiber. The first image
shows the schematic plot of a wicking fiber. The average diameter ranges from 30 to 50
μm and the opening of each groove varies from 5 to 12 μm. The second image shows the
SEM (Scanning Electron Microscopy) of a wicking fiber with a magnification of 2000
times. According to Equation 1, the small opening of the deep groove corresponds to a
high suction value and can generate a higher capillary force. In other words, the deep
grooves will maintain saturated under unsaturated conditions and will have high ability to
hold and transport water compared with conventional non-wicking geotextiles. A
laboratory demonstration test illustrates the lateral drainage capability of the wicking
geotextile, as shown in Figure 3a. Firstly, two water droplets were placed on the black
polyethylene yarn and the white nylon wicking fiber yarn, respectively (see the first
229
image). Due to the hydrophobic characteristic of polyethylene materials, the water
droplet on the left did not flow, as shown in the second image in Figure 3a. In
comparison, the wicking fiber yarn was hydrophilic and hygroscopic, and the water
droplet on the right side (see the second image) was quickly drained away along the
wicking fibers. After 10 seconds, the water droplet on the right side was completely
drained away, as shown in the third image in Figure 3a. The desaturation process, or the
lateral water removal capability, is a unique characteristic of the wicking geotextile that
differentiates itself from other non-wicking geotextiles. The detailed explanation and
discussion regarding this unique feature will be presented later in the Material Properties
section. In summary, the wicking geotextile has the ability to laterally transport water
along the wicking fibers under unsaturated conditions.
Figures 3b-c further show the differences between the wicking and non-wicking
geotextiles. As shown in Figure 3b, the excess water may be accumulated in the
overlying soil. However, the wicking geotextile has a higher ability to hold and laterally
transport water along the wicking fibers under unsaturated conditions. Due to the higher
capillary force generated within the wicking fibers, the excess water will be continuously
wicked from the soil to the grooves of the wicking geotextile and laterally transported
long the wicking fibers. In comparison, the pore sizes of within the non-wicking
geotextile are in general larger than those in soils, as shown in Figure 3c. The soil has a
higher ability to hold water compared to the non-wicking geotextile. The excess water
will be detained within the soil pores the excess pore water pressure will build up under
repetitive traffic load.

230
Figure 4a shows the subsurface drainage system with the wicking geotextile. The
wicking geotextile is proposed to be installed within or at the bottom of a base course,
with two ends extended to the surface of the road slope. In the cross-plane direction, the
wicking geotextile impedes capillary water from rising up to the overlying base course
aggregate. Meanwhile, in the in-plane direction, the wicking geotextile can absorb both
free water and capillary water from the surrounding soil, laterally wick the water out of a
road embankment to the surface of the road slopes, and eventually vaporize into the air
through the evaporation process. By creating a relatively dry zone, the resilient modulus
of the base course will be increased and the long-term performance of the roadway will
be improved. In other words, the wicking geotextile works as a “pipe” that connects the
soil inside the road embankment with the outside environment. The surrounding
atmosphere serves as a “pump” that works 24 hours a day and 365 days a year and
continuously dehydrates the road embankment. This conceptual drainage design has
proven to be effective in the field, as shown in Figure 4b. Two layers of the wicking
geotextile were installed in the road embankment to mitigate the frost “boil” (frost heave
and the corresponding thaw weakening) issue (Zhang et al. 2014). The comparison of the
road condition before and after rehabilitation clearly demonstrates the effectiveness of the
wicking geotextile in draining water and reinforcing the road embankment. Lin et al.
(2017) further reported that the frost “boil” problems has been successfully eliminated
and the roadway is still in a good driving condition five years after rehabilitation.
231
Even though the wicking geotextile has been proven efficient in dehydrating road
embankments, there is no existing method to take the full advantage of the wicking
geotextile. It is necessary to quantify the benefits of the wicking geotextile. Therefore, the
objectives of this paper are: 1) to study the working mechanism and the functional range
of the wicking geotextile and demonstrate the advantages of the wicking geotextile over
non-wicking geotextiles in terms of the ability to hold and transport water under
unsaturated conditions, 2) to propose a numerical model that is able to simulate the water
flow through unsaturated geomaterials and quantify the benefits of the wicking geotextile
through elemental level simulations, and 3) to evaluate the drainage performance of a
roadway installed with the wicking geotextile.
To achieve the objectives, the research methodologies are discussed as follow.
Firstly, the macrostructure and microstructures of the wicking geotextile have been
discussed in detail to demonstrate its uniqueness over non-wicking geotextiles. Then, a
series of laboratory tests were performed to characterize the properties of the base course
aggregate, the wicking geotextile, and the non-wicking geotextile. After that, a numerical
model was proposed and calibrated to simulate the water flow in both saturated and
unsaturated soils and calibrated. Moreover, the calibrated model was used to explore the
working mechanism of the wicking geotextile at the elemental level. Finally, the drainage
performance of a pavement structure installed with the wicking geotextile was evaluated.
232
4. NUMERICAL MODEL AND MODEL CALIBRATION
4.1. NUMERICAL MODEL
To quantify the benefits of the wicking geotextile in terms of water removal,
numerical analyses were carried out using a 3D finite element software, Abaqus 6.14-2
(Hibbett et al. 1998). Richards (1931) first derived the governing equation of transient
water flow within an unsaturated soil. For a three-dimensional homogeneous anisotropic
material, this is given by:
 u  u  u K u
( K x w ) + ( K y w ) + ( K z w + z ) = Cw w (2)
x x y y z z z t
where K x , K y , and K z are the permeability in the x-, y-, and z-directions, uw is the pore
water pressure, Cw is the specific water capacity, and t is time.
The left side represents the flow of water through a 3D soil element based on
Darcy’s law, and the right side represents the change in water storage with time. Terzaghi
(1951) used the thermodynamic analogue to explain the uncoupled 1D consolidation
theory for saturated soils. In his theory, the water content was corresponding to heat
energy per unit mass, and the pore water pressure was corresponding to temperature.
Zhang (2005) further extended the theory to describe the consolidation theory for
saturated/unsaturated soils, and this paper was based on Zhang’s consolidation theory.
Due to the space limit in this paper, the detailed discussions are omitted but can be found
in Zhang’s paper (Zhang et al. 2005).

233
4.2. MATERIAL PROPERTIES
Two types of soils, including a fine sand and AB3 (Aggregate Base Class 3), and
two types of geosynthetics, including the wicking geotextile (Mirafi H2Ri) and a non-
wicking filter geotextile (Polyfelt Megadrain 2040), were used in this study. The material
properties for soils and geotextiles are summarized in Table 1. The properties of the sand
and the non-wicking geotextile were obtained from a published paper (Krisdani et al.
2008). AB3 is widely used as a base course material in Kansas and its properties were
used as inputs of an actual base course material for numerical analysis. The properties of
the AB3 and wicking geotextile were obtained from a companion paper (Lin 2018) that
comprehensively discussed the test procedures and test results. For the completeness of
the paper, the hydraulic properties of the tested materials are briefly summarized and
presented.
As shown in Figure 5, the water characteristic curve (WCC) and the hydraulic
conductivity function (or the K-function) are two important hydraulic properties to
describe the ability of a geomaterial in holding and transporting water under unsaturated
conditions. The WCC establishes the relationship between water content and suction, and
is sensitive to the pore size distribution of a soil (Zornberg 2010). The sand used in this
paper was poorly-graded, indicating a relatively uniform pore size distribution. In
comparison, even though the AB3 was categorized as a gravely material, the soil was
well-graded and the particle sizes ranged variously, resulting in relatively smaller pore
sizes in the AB3. The relatively large pores implied that the ability of the sand to hold
water was not as strong as the AB3, as shown in Figure 5a. The water content of the sand
significantly decreased from 26% to 2.5% as the soil suction increased from the air entry
234
value (AEV) of 1.8 kPa to the residual suction value of 4.0 kPa. As the soil suction
exceeded the residual suction, the water within the pores became occluded and the water
content remained constant with increasing suctions. The WCC of the non-wicking
geotextile was similar to that of sand, with a significant decrease within a narrow range of
suction. The AEV of the non-wicking geotextile was only 0.5 kPa. The water content of
the non-wicking geotextile also dramatically decreased from 43% to 0.3% as the suction
increased from 0.0 kPa to 1.5 kPa. As suction continued to increase, most of the pores
within the non-wicking geotextile were occupied with air and the geotextile would work
as a capillary barrier to impede water from passing through.
In comparison, the WCC of the wicking geotextile shows a much stronger ability
to hold water under unsaturated conditions. Due to the existence of wicking fibers in the
in-plane direction, the WCC of the wicking geotextile was described by two sets of
regression parameters (a bimodal), resulting in two AEVs: namely the inter-yarn AEV
and the inner-yarn AEV. The inter-yarn AEV was approximately 1.1 kPa and was mainly
controlled by the relatively large pores between the weaving polyethylene yarns (see the
first image in Figure 2a). In fact, the inter-yarn AEV of the wicking geotextile was
similar to that of the non-wicking geotextile, and air could easily enter into the pores
among the weaving yarns as suction exceeded 1.1 kPa. That is to say, the wicking
geotextile would work as a capillary barrier in the cross-plane direction. However, the
unique feature of the wicking geotextile was the specially designed wicking fibers in the
in-plane direction. Different from other non-wicking geotextiles, the inner-yarn AEV of
the wicking geotextile was 254.0 kPa, which was mainly controlled by the size of the
openings within the deep grooves of the wicking geotextile (see Figure 2b). The deep
235
grooves within the fibers would remain saturated and could serve as water flow channels
under unsaturated conditions. As suction exceeded the inner-yarn AEV, the deep grooves
became desaturated and the ability of the wicking geotextile to transport water would be
significantly decreased. Therefore, the theoretical functional suction range for the
wicking geotextile should be 0-254 kPa. In summary, the wicking geotextile could work
as a capillary barrier in the cross-plane direction meanwhile had lateral drainage ability in
the in-plane direction.
The ability of a geomaterial to transport water under unsaturated conditions can
be explained by its K-function, as shown in Figure 5b. The hydraulic conductivity of an
unsaturated geomaterial with relatively large pores decreased faster than that of fine-
grained soils (Zornberg et al. 2017). This feature led to the hydraulic conductivity of the
non-wicking geotextile to be orders of magnitude smaller than that of the sand or AB3
under unsaturated conditions. The capillary break effect would be maintained until the
“breakthrough” suction was obtained (Ho and Webb 1998). At the breakthrough suction,
the hydraulic conductivities were the same for both the soil and the non-wicking
geotextile. As the suction continued to decrease, the water within the small pores of the
soil was able to penetrate into the large pores of the non-wicking geotextile, and the
capillary break effect would fade away. As for the wicking geotextile, the hydraulic
conductivities of the wicking geotextile were anisotropic in the cross-plane and in-plane
directions. In the cross-plane direction, the k-function of the wicking geotextile was
similar to that of the non-wicking geotextile, with a significant decrease within a narrow
range of suction. Because there were no wicking fibers in the cross-plane direction, as
long as the suction exceeded the inter-yarn AEV of 1.1 kPa, the hydraulic conductivity
236
immediately dropped from saturated values of 6.2×10-4 m/s, to a very small value of
1.1×10-12 m/s (equivalent to the value of water vapor transmissivity and only water vapor
was allowed to pass through in the cross-plane direction).
In comparison, the K-function of the wicking geotextile showed a gradual
decrease with increasing suctions in the in-plane direction. When the suction was lower
than the inter-yarn AEV, both large pores between yarns and the deep grooves within the
wicking fibers were saturated, and the corresponding hydraulic conductivity was the
maximum value (1.1×10-4 m/s). As the suction exceeded the inter-yarn AEV, the water
within the large pores would be replaced with air and could not be used for water flow.
However, as long as the suction was still lower than the inner-yarn AEV of 254.0 kPa,
the wicking fibers could still work effectively to transport water in the in-plane direction.
If applied in the roadway, the wicking geotextile could wick the excess water, especially
the capillary water, out of the road embankment and prevent positive the pore water
pressure from building up in the overlying soil. As the suction continued to increase, the
air would further replace water within the deep grooves and the corresponding hydraulic
conductivity of the wicking geotextile would continue to decrease. When the suction
reached the residual suction of 1500 kPa, most of the water within the fibers was replaced
with air and the hydraulic conductivity was equivalent to water vapor transmissivity of
1.1×10-12 m/s.
4.3. MODEL CALIBRATION
The results of laboratory drawdown test from Krisdani et al. (2008) were used to
calibrate the numerical model. Figure 6a shows the schematic plot and the mesh
237
configuration of the soil-geotextile column. The soil-geotextile column was composed of
a 0.51 m-thick fine sand at the top, a 0.02 m-thick non-wicking geotextile in the middle,
and another 0.47 m-thick fine sand at the bottom. The diameter of the soil-geotextile
column was 0.19 m with tensiometers and TDR sensors installed at different elevations to
measure the soil water content and the corresponding pore water pressures. The
drawdown test was performed by lowering the groundwater table from the top of the soil
column to the bottom. The geometry and the mesh configuration of the numerical model
are also shown in Figure 6a (right image). The DC3D8 (8-node linear brick heat transfer
element) and DS4 (4-node quadrilateral heat transfer shell element) elements were used
for the soil and the geotextile, respectively. Firstly, a constant suction boundary condition
of 0.0 kPa was applied at the top of the soil column and a steady-state analysis was
performed. A linearly distributed suction was used as the initial condition. Secondly,
during the drawdown test, the constant suction boundary condition at the top was
removed while another suction boundary condition was applied at the bottom of the soil-
geotextile column and the suction variations with time was obtained from the recorded
sensor readings located at an elevation of 0.05 m. Then, the simulation lasted for 48 hours
and the suction distributions were plotted at different times.
Figure 6b compares the suction distributions obtained from the simulation and
laboratory test results during the drawdown test. In general, the simulation results
matched reasonably well with the laboratory test results. The capillary break effect was
observed at the soil-geotextile interface starting at six hours after the test. However, the
simulated suction distributions were deviated from the laboratory test results at elevations
greater than 0.75 m. The same problem was observed in Krisdani et al. (2008) simulation
238
results. The simulated suction moved faster towards the steady-state condition. According
to Krisdani et al. (2008), the deviation was caused by the k-function that was used in the
numerical simulation. The hydraulic conductivity of the soil was low enough to sustain
the pressure when the residual suction was obtained, while the k-function used in the
numerical simulation did not account for this phenomenon. In summary, the proposed
numerical model reasonably simulated the capillary break effect observed in the non-
wicking geotextile.
5. THE DRAINAGE PERFORMANCE OF THE WICKING GEOTEXTILE

UNDER DIFFERENT WORKING CONDITIONS
It is well recognized that excess water can accelerate pavement deteriorations
with time. Moisture in a pavement structure may come from many sources, such as
upward seepage via capillary action, rising of water table, laterally water flow through
pavement edges and side ditches, and surface infiltration of rain and meltwater through
defects (AASHTO 1993; and ARA 2004). To comprehensively evaluate the drainage
performance of the wicking geotextile, the numerical simulations divided into two levels.
The first type of simulation was elemental level, which aimed at exploring the working
mechanism of the wicking geotextile and evaluating the performance of the wicking
geotextile at different working and climatic conditions. The soil-geotextile column used
in the model calibration section was used for the elemental level simulation. The water
may flow into the soil-geotextile column from the bottom of the soil via rising of water
table or from the top of the soil via infiltration process. Therefore, the first two cases
simulated the performance of the wicking geotextile when the water table raised to
239
different elevations. Another two cases simulated the performance of the wicking
geotextile during light and heavy rainfall events. As for the full-scale model, the field
performance was evaluated for a pavement structure installed with the wicking geotextile.
The drainage performance of the wicking geotextile was quantified. The simulation
results were compared with field observations to validate the drainage efficiency of the
wicking geotextile. Detailed discussions of case studies are presented as follows.
5.1. RISING OF GROUNDWATER TABLE
The calibration model was carried out in this numerical analysis and the
properties for sand were also adopted. However, the non-wicking geotextile was replaced
with the wicking geotextile. Firstly, the groundwater table (GWT) was assumed at the
bottom of the soil column. A steady-state analysis was performed and the suction
distribution was considered as the initial condition. Secondly, two cases were simulated
with the groundwater table raised by 0.2 m and 0.4 m, respectively. The first case of a 0.2
m rising of the GWT represented the scenario that the GWT was relatively far away (0.3
m) from the wicking geotextile while the second case represented the scenario of a very
shallow GWT with a 0.1 m distance between the wicking geotextile and the GWT.
Finally, a transient analysis was performed for 24 hours for both cases, and the suction
distributions at different times were plotted.
Figure 7a shows the simulation results for the first case. At the initial condition,
the GWT was at the bottom of the soil-geotextile system and the suction was linearly
distributed with 0.0 kPa at the bottom and 10.0 kPa at the top. In general, the suction
distribution in the underlying soil varied with time while the suction distribution in the
240
overlying soil did not change. As the suction boundary condition changed to -2.0 kPa at
the bottom of the soil-geotextile column, the GWT was raised to 0.2 m and the suction
distribution was nonlinear. The negative suction value indicated that the soil was
submerged in water. After 15 minutes, the suction distribution was relatively linear for
soil lower than 0.35 m and approached to the steady-state condition. In addition, the
suction value at the soil-geotextile interface was 3.0 kPa, which was greater than the
inter-yarn AEV of 1.1 kPa. This fact indicated that the wicking geotextile worked as a
capillary barrier impeding water from breaking through the geotextile. This was also the
reason that the suction distribution above the wicking geotextile did not change
throughout the simulation process. The capillary break effect was observed throughout
the simulation process.
In comparison, Figure 7b shows the suction distributions at different times when
the GWT was raised to an elevation of 0.4 m. Similar to the previous case, a linear
suction distribution (0.0 kPa at the bottom and 10.0 kPa at the top) was used as the initial
condition. However, due to the relatively closer distance between the water table and the
wicking geotextile, the suction value at the soil-geotextile interface was 1.0 kPa, which
was lower than the inter-yarn AEV of 1.1 kPa. Due to such a low suction value, the
wicking geotextile was nearly saturated and its hydraulic conductivity significantly
increased. Therefore, the capillary break effect vanished and the wicking geotextile
became permeable for water to pass through. The soil suction of the overlying soil also
redistributed with time because of the relatively shallow GWT. For example, the soil
suctions decreased to 7.9 kPa at 0.8 m and 1.3 kPa at 0.5 m within 5 minutes. After 3
hours, the suction for soils lower than 0.7 m was linear and reached the steady-state.
241
After 6 hours, the entire soil-geotextile column reached the steady state and a linear
suction distribution was obtained with 6.0 kPa at the top and -4.0 kPa at the top. Note that
the negative suction indicated a positive pore water pressure was 4.0 kPa.
In summary, similar to conventional non-wicking geotextiles, the wicking
geotextile could work as a capillary barrier to impede water from passing through in the
cross-plane direction as long as the suction at the soil-geotextile interface was higher than
its inter-yarn AEV of 1.1 kPa. In other words, the installation location of the wicking
geotextile shall be at least 0.11 m above the GWT so that the wicking geotextile would
work as a capillary barrier to control the capillary water from breaking through to the
overlying soil. To be on the conservative side, the recommended minimum distance
between the wicking geotextile and the water table shall be 0.2 m.
5.2. PRECIPITATION INFILTRATION
Water may also penetrate into pavement systems via precipitation infiltration. As
demonstrated in the two cases above, the wicking geotextile worked as a capillary barrier
as long as the suction at the soil-geotextile interface was higher than the inter-yarn AEV
of 1.1 kPa. For the scenarios of precipitation infiltration, two cases were also simulated.
The initial condition was the same for the two cases with zero suction at the bottom of the
column and 10.0 kPa at the top of the column. The GWT was assumed to be at the
bottom of the column. For the first case, a zero suction boundary condition was applied
on top of the soil-geotextile system for 15 seconds. Such a short period of wetting process
would not intense enough to wet the entire overlying soil and the suction at the soil-
geotextile interface was expected to be still higher than the inter-yarn AEV. To
242
demonstrate the lateral drainage capability of the wicking geotextile, the performance of
soil-geotextile systems with wicking and non-wicking geotextiles were simulated
separately and the simulation results were compared. For the second case, the zero
suction boundary condition lasted for 10 minutes so that sufficient water was supplied to
the system to wet the soil. The performance of the wicking geotextile was evaluated.
5.2.1. Light Rainfall Event. Firstly, a linear suction distribution (0.0 kPa at the
bottom and 10.0 kPa at the top) was used as the initial condition. Then, a suction boundary
condition of 0.0 kPa was applied on top of the soil-geotextile system for 15 seconds. After
that, the suction boundary condition at the top of the soil-geotextile system was removed.
Finally, a suction boundary condition of 250 kPa was applied on the right end of the
geotextile. The simulations were performed two times with the non-wicking geotextile as
the first one and the wicking geotextile as the second one. The suction distributions during
and after the rainfall event are presented to demonstrate as follows.
Figure 8 shows the comparisons of the suction distributions for the soil-geotextile
column at different times. As shown in Figure 8a, the initial suction distribution was the
same for both systems (with the non-wicking geotextile and with the wicking geotextile)
with 0.0 kPa at the bottom and 10.0 kPa at the top (red solid line). Due to the same initial
and boundary conditions during the rainfall event, the suction distributions were the same
for both soil-geotextile systems at the end of the light rainfall event (red dash line). For
example, the suction at the top of the soil-geotextile system was 0.0 kPa, indicating that
the soil was saturated at the top. Meanwhile, the wetting front reached to an elevation of
0.9 m at the end of the light rainfall event. However, the suction distributions after the
rainfall stopped showed different variation trends.

243
For the soil-geotextile column with the non-wicking geotextile, the overall suction
distribution in the overlying soils at 1 hour was shifted to the left side compared with the
initial suction distribution. The variation in suction distribution indicated that the
overlying soils became wetter than the initial condition because of the zero suction
boundary condition. In addition, even though the wetting front has reached the soil-
geotextile interface, the suction value at the soil-geotextile interface was 4.8 kPa, which
was still higher than the AEV of the non-wicking geotextile. Due to the capillary barrier
effect, the non-wicking geotextile impede the water from percolating to the underlying
soils and the suction distribution in the underlying soil did not change. After 1 day, the
suction was linearly distributed in the overlying soil but was still left to the initial
condition. This fact indicated that infiltration water could not be drained out of the soil-
geotextile column with the non-wicking geotextile.
In comparison, the suction distribution for the soil-geotextile system with the
wicking geotextile shows different trend. As the zero suction boundary condition was
removed, the suction distribution was the same with 0.0 kPa at the top and the wetting
front at an elevation of 0.9 m. First of all, the suction value at the soil-geotextile interface
was 4.9 kPa, which was higher than the inter-yarn AEV (1.1 kPa). Similar to the non-
wicking geotextile, the wicking geotextile could also work as a capillary barrier and the
suction distribution in the underlying soil did not change throughout the simulation
process. However, due to the lateral drainage capability of the wicking geotextile, the soil
suction distribution in the overlying soil was on the right side to the suction distribution
for the system with the non-wicking geotextile 1 hour after the rainfall stopped. This
simulation result indicated that the wicking geotextile successfully wicked a measurable
244
amount of water out of the system and the overlying soil was in a drier condition. After 1
day, both the suction distributions in the overlying and underlying soils shifted further to
the right side of the initial suction distribution. The results indicated that the wicking
geotextile was able to gradually dry the soil. In other words, the wicking geotextile not
only drained out all the excess water introduced by the light rainfall event, it further
dehydrated the overlying soil to an even drier condition.
Figure 8b further shows the suction distributions for the systems with the wicking
and non-wicking geotextiles at 1 month and 1 year, respectively. For the soil-geotextile
column with the non-wicking geotextile, the suction distributions at 1 month and 1 year
were nearly the same compared with the distribution at 1 day. This result indicated that
the non-wicking geotextile did not have lateral drainage ability and the excess water
could not be drained out. In comparison, the suction distributions for the soil-geotextile
column with the wicking geotextile continued moving to the right, indicating that the soil
was gradually dried out. For example, the average suction for the overlying soil was 75
kPa for the soil with an elevation higher than 0.6 m (10 times higher than that in the
system with the non-wicking geotextile). In addition, the closer the soil approaching the
geotextile, the higher the suction value would be. This simulation result implied that the
drainage of the overlying soil was a bottom-to-top process, rather than a top-to-bottom
process as expected in a conventional subsurface drainage system. Eventually, the
average suction for the overlying soil reached to 252.5 kPa after one year. Note that this
suction value may not reflect the actual field suction condition because the predicting the
actual suction distribution within the road embankment was a more complex problem that
relied on a series of factors, such as the climatic condition, aging of the asphalt concrete
245
layer, and the soil-infrastructure interactions. However, the simulation results clearly
showed the ability of the wicking geotextile to laterally drain the excess water out of the
soil-geotextile system.
5.2.2. Heavy Rainfall Event. For this case, only the soil-geotextile system with
the wicking geotextile was simulated. The initial and boundary conditions were the same
as in the light rainfall event while the duration of zero suction boundary condition
changed from 15 seconds to 10 minutes. The drying process was simulated for one month
and followed by a steady-state analysis to observe the final suction distribution.
Figure 9a shows the suction distributions during the rainfall event. Initially, the
suction was linearly distributed with 0.0 kPa at the bottom and 10.0 kPa at the top. As the
rainfall event was introduced to the model, the suction at the top immediately decreased
to 0.0 kPa. After 1 minute, the waterfront reached to an elevation of 0.8 m and the soil
beneath this elevation was still linearly distributed. Then, the waterfront continued to
migrate downward and reached elevations of 0.7 m and 0.55 m at 2 and 5 minutes,
respectively. Since the soil suction at the soil-geotextile interface was 1.7 kPa, which was
higher than the inter-yarn AEV, the wicking geotextile still worked as a capillary barrier
and the suction distribution for the underlying soil did not change at this moment.
However, as the waterfront continued penetrating through the soil, the soil suction at the
interface became lower than the inter-yarn AEV after 8 minutes. Water started to pass
through the wicking geotextile and the wetting front reached an elevation of 0.4 m. Due
to the excess water flowed to the lower soil, the suction of the soil above the wicking
geotextile slightly increased to 0.9 kPa at 10 minutes, but still lower than the inter-yarn
AEV. This result indicated that the wicking geotextile was still permeable for water to
246
flow and the wetting front further moved downward to an elevation of 0.35 m at 10
minutes. This was beneficial engineering characteristic for the wicking geotextile because
no excess pore water pressure would be expected to build up in the overlying soil and the
excess water would be quickly drained out through the wicking fibers, as demonstrated in
Figure 3. In other words, if the soil was unsaturated and could be maintained at a
relatively low water content, the possibilities for pumping would be expected to be
reduced.
Figure 9b shows the suction distributions 0-4 minutes after the rainfall event. For
the overlying soil, the excess water could be drained downwards to the underlying soil
because the suction of the wicking geotextile was under the inter-yarn AEV. Under the
influence of gravity, water in the overlying soil gradually decreased and the
corresponding suction increased. Moreover, the soil suction increased at 0.5 m because of
the effect of the applied suction boundary (250.0 kPa) at the right end of the wicking
geotextile. The soil lower than 0.3 m did not influence by the wicking geotextile,
indicating that the influencing range of the wicking geotextile in the underlying soil was
approximately 0.2 m. Figure 9c shows the simulation results from 5 minutes to the
steady-state after the rainfall stopped. The wicking effect could be clearly observed with
the maximum suction reaching 29.4 kPa for the soil in contact with the wicking
geotextile 5 minutes after the rainfall. The maximum soil suction further increased to
232.7 kPa at 10 minutes. In addition, the soil overlying the wicking geotextile gradually
dried with time due to the wicking effect. The soil closer to the wicking geotextile firstly
dried, and then absorbed water from the overlying soil, resulting in a nonlinear suction
distribution of the overlying soil. Given a sufficient long drying period, the overlying soil
247
would eventually be in equilibrium with the wicking geotextile and reached an average
suction of 252.0 kPa in the overlying soil.
It is also important to point out that most of the water was drained out within 5
minutes after the rainfall stopped. For example, the average suction for the overlying soil
was 1.3 kPa after 1 minute, corresponding to an average water content of 25% (calculated
based on the SWCC for sand in Figure 5a). In comparison, the average water content
significantly decreased to 13% after 5 minutes, even if the soil suction slightly increased
to 2.1 kPa. This phenomenon was consistent with the characteristic of coarse-grained soil
in which the suction decreases significantly with a narrow range of suction variations.
Furthermore, as the soil suction exceeded the residual suction of 3.5 kPa after one hour,
the rate of suction change was also decreased. For example, the suction at the top of the
overlying soil changed from 11.8 kPa at one day to 42.4 kPa at seven days, and the
corresponding water content only changed from 3.1% to 2.0%. As for the soil underlying
the wicking geotextile, only soil with elevations higher than 0.3 m was influenced by the
wicking geotextile. Due to the wicking effect, the suction distribution from 0.3 m to 0.5
m was highly nonlinear. In summary, the wicking geotextile worked effectively to wick
water out and the suction of the overlying soil could reach 250.0 kPa as long as the
drying process continued.
5.3. DRAINAGE PERFORMANCE OF A ROADWAY INSTALLED WITH THE

WICKING GEOTEXTILE.
To better evaluate the drainage performance of the wicking geotextile in the field,
a full-scale model was established, as shown in Figure 10. The wicking geotextile was
assumed to be installed within the road embankment of a two-lane road. Due to the
248
geometrical symmetry, half of the road length was simulated. The thickness of the base
course was 0.5 m and the road embankment had a slope with vertical to horizontal ratio
of 1:3. The wicking geotextile was installed in the middle of the base course with an
elevation of 0.25 m from the bottom. The groundwater table was assumed to be at the
bottom of the base course. The hydraulic properties of the AB3 and the wicking
geotextile were presented in Figure 5. According to the Lin et al. (2018), the optimum
water content of the AB3 aggregate was 8.5% and the corresponding suction was 15 kPa.
Therefore, for the initial condition, a predefined suction of 15.0 kPa was applied to the
entire system. The simulation included two phases. The first phase was the drying
process and lasted for 9 months. A constant suction boundary condition of 250.0 kPa was
applied to the right end of the wicking geotextile to simulate the drying process. The
second phase was the wetting process and a zero suction boundary condition was
imposed to the top of the base course for 15 minutes. Then, the zero suction boundary
was removed and a transient analysis lasted for another 9 months. The following
discussions are based upon the suction distribution at the center of the road.
Figure 11a shows the suction distributions during the drying process. At the initial
condition, the suction of the base course was 15.0 kPa and the corresponding water
content was at its optimum value of 8.5%. The suction at the bottom of the base course
changed to 0.0 kPa due to the existence of the shallow water table. After 5 hours, the
suction for the soil in contact with the wicking geotextile quickly increased to 250 kPa
due to the imposed suction boundary condition at the right end of the wicking geotextile.
The suction values within the overlying soil also gradually increased with time,
indicating that the wicking geotextile gradually wick water out of the roadway. It is
249
important to point out that the suction increased from 15.0 kPa to 47.3 kPa within 1 day,
and the corresponding water content decreased from 8.5% to 7.2%. According to the
laboratory test results (Lin et al. 2018), the resilient modulus of the AB3 at 7.2% of water
content was doubled compared with that at 8.5% water content. Moreover, as the soil
suction continued to increase and surpass the residual suction of 75.0 kPa, water within
the soil pores became occluded and hard to be wicked out. This is the reason why the soil
suction at the top of the overlying soil increased from149.9 kPa to 252.0 kPa (from 1
week to 9 months) but the corresponding water content only decreased from 6.3% to 6%.
However, the rapid water content change within the first day already doubled the resilient
modulus and minimized the possibility of excess water-induced soil softening.
Furthermore, because the water table was 0.25 m below the wicking geotextile, which
was greater than the minimum requirement of 0.11 m, the wicking geotextile worked as a
capillary barrier in the cross-plane direction throughout the numerical simulation.
Figure 11b further shows the suction contours after the rainfall event. The cold
color indicates that the soil is relatively wet while the warm color indicates the soil is
relatively dry. The first image was the suction contour right after the rainfall event. Most
of the overlying soil was relatively wet except for the soil that was closer to the road
slope. It was because the suction boundary condition was only applied on the road
surface and its effect on the road slope was limited. Moreover, the underlying soil was
very wet due to the shallow groundwater table. After 1 hour, the soil that was close to the
wicking geotextile firstly dried due to the wicking effect. Due to the relatively long
distance, it took some time for the water in the middle of the road to be wicked out. After
one day, all the soil that was close to the wicking geotextile was relatively dry, as shown
250
in the third image in Figure 11b. The influencing range of the wicking geotextile was
approximately 0.2 m in the underlying soil. The soil was further dried as time passed by
and the suction of the overlying soil reached 250.0 kPa after 1 month. In summary, the
wicking geotextile has the capability to impede the water from passing through in the
cross-plane direction while can laterally drain the water out of the pavement structure in
the in-plane direction.
6. FIELD VALIDATIONS
Both the elemental level and full-scale simulation results showed the effectiveness
of the wicking geotextile in dehydrating road embankments. Those simulation results can
also be validated via field observations. For example, Figure 12a shows the performance
of the wicking geotextile in a subgrade stabilization project for the Daniel Boone Bridge
on the I-64, MO (TenCate (2013)). The field site was adjacent to the Missouri river
where the subgrade soil was nearly saturated. The wicking geotextile allowed an overall
51 mm (2 in.) reduction in the aggregate base course along with the lateral drainage
ability. After completing the installation process of the wicking geotextile and
compacting the overlying 160 mm (6 in.) aggregate (left image in Figure 12a), a light
rainfall of 6.4 mm (0.25 in.) occurred at the construction site. Several hours after the light
rainfall, the infiltration water in the compacted base aggregate was successfully wicked
out and drained to the edge to the embankment (right image in Figure 12a). The wetted
soil at the edge of the road embankment clearly validated the effectiveness of the lateral
251
drainage ability of the wicking geotextile, which was consistent with the simulation
results presented in Figure 8.
Figure 12b shows another field test section with the application of the wicking
geotextile on the Dalton Highway, AK (Currey 2016). The wicking geotextile was used
to deal with the soft subgrade and the photo was taken in the summer of 2013. The
ponding indicated that the groundwater table in the field was very shallow and the soft
subgrade could not support the overlying road embankment. The wicking geotextile was
installed at the bottom of the base course and the sharp contrast between the wet and dry
surface coincided with the edge of the wicking geotextile. This fact clearly validates the
effectiveness of the wicking geotextile as a capillary barrier to impede water from
migrating to the overlying base course in the cross-plane direction. Meanwhile, all the
base course aggregate was in a relatively dry condition, indicating that the wicking
geotextile has the ability to laterally drain excess water out of the road embankment. In
addition, stuff from AKDOT&PF (Alaska Department of Transportation & Public
Facilities) performed a site inspection in the summer of 2015 and reported that “The
pavement looked very good, and the ride through the project was smooth”. The field
observations were also in accordance with the simulation results in Figures 7a and 11b.
In summary, both the simulation results and field observations show that the
wicking geotextile is effective in reducing soil water content. The wicking geotextile
works as a capillary barrier in the cross-plane direction while can laterally drain excess
water out of road embankment in the in-plane direction.

252
7. CONCLUSIONS
This study demonstrated the working mechanism of a new woven geotextile with
lateral drainage abilities. Both numerical simulation and field observation validated its
effectiveness in dehydrating road embankments. The major conclusions are summarized
as follows:
1. The uniqueness of the wicking geotextile is that it can work as a capillary barrier
in the cross-plane direction meanwhile laterally wick water out of road
embankments. Due to the hydrophilic and hygroscopic characteristics of the
wicking fiber, the wicking geotextile was able to drain the excess water laterally
in a roadway under both saturated and unsaturated conditions.
2. The GWCC of the wicking geotextile showed two AEVs, namely the inter-yarn
AEV of 1.1 kPa and the inner-yarn AEV of 254.0 kPa. The inter-yarn AEV
reflected the relatively large pores between the polyethylene weaving yarns while
the inner-yarn AEV was controlled by the small openings of the deep grooves
within each fiber. The specially designed grooves could remain saturated so that
water could be laterally drained out under unsaturated conditions until the suction
exceeded the inner-yarn AEV. Therefore, the functional suction range of the
wicking geotextile should be 0-254 kPa.
3. In the cross-plane direction, the drainage performance of the wicking geotextile
was mainly controlled by the relatively large pores among the wicking yarns. The
inter-yarn AEV of the wicking geotextile was 1.1 kPa, indicating that the
minimum distance between the wicking geotextile and the groundwater table
253
should be at least 0.11 m (to be conservative, recommend to use 0.2 m as the
minimum distance). If the suction at the soil-geotextile interface was higher than
the inter-yarn AEV, the wicking geotextile would work as a capillary barrier.
However, if the suction at the soil-geotextile interface was smaller or equal to the
inter-yarn AEV, the wicking geotextile would work as a permeable geomaterial
that allowed water to break through. This characteristic was beneficial for the
subsurface drainage of road embankments. During light rainfall events, no excess
pore water pressure would build up in the overlying soil because the excess water
would be drained out along the wicking fibers. Meanwhile, during heavy rainfall
events, the excess water would rapidly percolate to the underlying soil because
the wicking geotextile would work as a permeable geomaterial under such low
suction level.
4. In the in-plane direction, the wicking geotextile was able to drain all overlying
soil if the drying process is long enough. The simulation results of the full-scale
model indicated that the soil water content decreased from 8.5% to 7.2% within
one day during the drying process and the corresponding resilient modulus was
doubled according to the laboratory test results. Moreover, if the soil was initially
saturated (e.g., right after a rainfall event), the soil water content decreased from
12.5% to 8.9% within one hour. The drainage quality of the roadway installed
with the wicking geotextile could be categorized as excellent. Given a longer time
span, the wicking geotextile was able to dehydrate all the overlying soil.
5. The field observations validated the effectiveness of the wicking geotextile as a
drainage material under saturated and unsaturated conditions that worked as a

254
capillary barrier in the cross-plane direction and as a drainage material in the in-
plane direction. The field observations were in accordance with numerical
simulation results.
ACKNOWLEDGEMENTS
This study was partially sponsored by the GSI (Geosynthetic Institute)
Fellowship. The wicking geotextile used in the study was provided by TenCate
Geosynthetics America. The authors appreciate their support.

255
Table 1. Soil and geotextile properties

Soil
Properties Unit Fine Sand AB3
Coefficient of Uniformity, Cu 2.1 54.7
Coefficient of Curvature, Cc 0.89 2.9
N/A Well-Graded
Poorly-Graded
Gravel with Silty
Classification Sand
Clay and Sand
SP GW-GC
3
Maximum Dry Density Mg/m 1.58 2.2
Saturated Hydraulic
m/s 2.70×10-4 5.60×10-4
Conductivity, ks
Geotextile
Properties Unit Wicking1 Non-Wicking2
Thickness mm 1.6 203
Apparent Opening Size
mm 0.425 N/A
(AOS)
O90 mm N/A 0.095
O95 mm 0.195
Permittivity s-1 0.4 N/A
Flow Rate l/min/m2 1222
Water Permeability
l/(m2∙s) N/A 110
(Vertical)
Note:
1
adapted from (TenCate 2015), 2 adapted from (TenCate 2006), and 3 the thickness includes one layer of a
three-dimensional polypropylene monofilaments combined with two layers of polypropylene geotextile
256
(a)
(b)
Precipitation Infiltration
Ponding Water Woven Geotextile
Capillary Action
(c)
Figure 1. Capillary break effect: (a) water droplet detained on a woven geotextile, (b)
mechanism of capillary break effect, and (c) excess water accumulation within pavement
structures (adapted from Lin and Zhang (2018a))
257
(a)
(b)
Figure 2. Macro- and microstructures of the wicking geotextile: (a) macrostructures and
weaving patterns, and (b) microstructures of the wicking fiber
258
Water Droplet
Lateral Drainage
Water Droplets
Desaturated
(a)
(b)
Figure 3. A wicking geotextile with lateral drainage ability: (a) laboratory demonstrative
test, (b) wicking geotextile, and (c) non-wicking geotextile
259
(c)
Figure 3. A wicking geotextile with lateral drainage ability: (a) laboratory demonstrative
test, (b) wicking geotextile, and (c) non-wicking geotextile (cont.)
Precipitation Infiltration
Evaporation Evaporation
Wicking Geotextile
Capillary Water
(a)
May 12, 2010 May 24, 2011
(b)
Figure 4. Successful field applications: (a) a subsurface drainage design with the wicking
geotextile, and (b) field test section at Beaver Slide section on the Dalton Highway, AK
260
AEV (0.5 kPa)
Inter-Yarn AEV (1.3

kPa)
Inner-Yarn AEV (254

kPa)
(a)
In-
Plane
Cross-
Plane
(b)
Figure 5. Hydraulic Properties: (a) SWCCs and GWCCs, and (b) hydraulic conductivity
functions (K-functions)
261
(a)
(b)
Figure 6. Model calibration: (a) test setup and mesh configuration, (b) comparisons of
test result and simulation result
262
Wicking Geotextile
Location
GWT
Wetti
(a)
Wicking Geotextile Location
GWT
Wetting
(b)
Figure 7. Simulation results of rising groundwater table: (a) suction distributions at

different times (a 0.2 m increment of the groundwater table), and (b) suction distributions
at different times (a 0.4 m increment of the groundwater table)
263
GWT
(a)
non-wicking geotextiles): (a) suction distributions at different times (0 min-1 day), (b)
suction distributions at different times (within 1 year)
264
Wicking Geotextile
Location GWT
(b)
non-wicking geotextiles): (a) suction distributions at different times (0 min-1
day), (b) suction distributions at different times (within 1 year) (cont.)
265
Wetting
GWT
(a)
Drying
GWT
(b)
Figure 9. Simulation results of suction distributions during a heavy rainfall event: (a)
during the rainfall event, (b) after the rainfall event (0-5 min), and (c) after the rainfall
event (0 min-steady state)
266
Drying
GWT
(c)
Figure 9. Simulation results of suction distributions during a heavy rainfall event: (a)
during the rainfall event, (b) after the rainfall event (0-5 min), and (c) after the rainfall
event (0 min-steady state) (cont.)
Figure 10. Schematic plot of a road embankment installed with the wicking geotextile
(not to scale)
267
(a)
(b)
Figure 11. Simulation results of a road embankment installed with the wicking geotextile:
(a) suction distributions during the drying process, and (b) suction contours during the
drying process
268
Wet Soil
Base Course
Wicking Geotextile
Base Course
(a)
(b)
Figure 12. Field evidence of the effectiveness of the wicking geotextile: (a) Daniel Boone
Bridge on Interstate 64, MO (adapted from TenCate (2015)), and (b) MP 197-209
Rehabilitation Project on Dalton Highway, AK (adapted from Curry (2016))
269
REFERENCES
Han, J. (2015). “Principles and Practice of Ground Improvement.” John Wiley & Sons,
Inc., Hoboken, New Jersey
Zornberg, G.J. (2011). “Advances in the Use of Geosynthetics in Pavement Design.”

Geosynthetics India’11, India Institute of Technology Madras, Chennai, India,
September 23-24, 2011, Volume 1, 3-21.
Holtz, R.D., Christopher, B.R., and Berg, R.R. (1998). “Geosynthetic Design and
Construction Guidelines.” Federal Highway Administration, Washington, DC,
Final Report, No. FHWA-HI-98-038, 460.
Giroud, J.P. and Noiray, L. (1981). “Geotextile-Reinforced Unpaved Roads.” Journal of

Geotechnical Engineering Division, American Society of Civil Engineers,
Volume 107, No. GT9, 1233-1254.
Morris, C.E. (2000). “Unsaturated Flow in Nonwoven Geotextiles.” In ISRM

International Symposium, International Society for Rock Mechanics.
ARA, I., ERES Consultants Division. (2004). "Guide for Mechanistic–Empirical Design
of New and Rehabilitated Pavement Structures." Final Rep., NCHRP Project 1-
37A.
Clough, I., and French, W. "Laboratory and field work relating to the use of geotextiles in
arid regions." Proc., Proceedings of the 2nd International Conference on
Geotextiles, Las Vegas, Nev, 1-6.
Curry, J. (2016). "H2Ri Wicking Fabric Experimental Feature Final Report-Dalton

Highway MP 197-209 Rehabilitation."Alaska Department of Transportation and
Public Facilities (AKDOT&PF), Juneau, AK.
Fredlund, D. G., Rahardjo, H., and Fredlund, M. D. (2012). Unsaturated soil mechanics in
engineering practice, John Wiley & Sons.
Giroud, J., Zornberg, J., and Zhao, A. (2000). "Hydraulic design of geosynthetic and
granular liquid collection layers." Geosynthetics International, 7(4-6), 285-380.
Hibbett, Karlsson, and Sorensen (1998). ABAQUS/standard: User's Manual, Hibbitt,

Karlsson & Sorensen.
Ho, C. K., and Webb, S. W. (1998). "Capillary barrier performance in heterogeneous

porous media." Water resources research, 34(4), 603-609.
270
Krisdani, H., Rahardjo, H., and Leong, E.-C. (2008). "Measurement of geotextile-water
characteristic curve using capillary rise principle." Geosynthetics International,
15(2), 86-94.
Lin, C., and Zhang Xiong (2018). "Comprehensive Material Characterizations of a

Pavement Structure Installed with Wicking Fabrics."Journal of Materials in Civil
Engineering, 31.2 (2018): 04018372
Lin, C., and Zhang, X. (2018). "A Bio-Wicking System to Dehydrate Road Embankment."
Journal of Cleaner Production, 196, 902-915.
McCartney, J. S., Kuhn, J. A., and Zornberg, J. G. (2005) "Geosynthetic drainage layers in
contact with unsaturated soils." Proc., PROCEEDINGS OF THE
INTERNATIONAL CONFERENCE ON SOIL MECHANICS AND
GEOTECHNICAL ENGINEERING, AA BALKEMA PUBLISHERS, 2301.
Olson, R. E., and Langfelder, L. J. (1965). "Pore water pressures in unsaturated soils."
Journal of Soil Mechanics & Foundations Div, 97(SM1).
Siswosoebrotho, B. I., Widodo, P., and Augusta, E. (2005). “The influence of fines content
and plasticity on the strength and permeability of aggregate for base course
material.” Proc., Proceedings of the Eastern Asia Society for Transportation
Studies, Citeseer, 845-856.
Stormont, J. C., and Anderson, C. E. (1999). "Capillary barrier effect from underlying
coarser soil layer." Journal of Geotechnical and Geoenvironmental Engineering,
125(8), 641-648.
Stormont, J. C., and Morris, C. E. (2000). "Characterization of unsaturated nonwoven

geotextiles." Advances in Unsaturated Geotechnics, 153-164.
TenCate (2006). "Polyfelt Megadrain-Product Data."

<https://www.buildsite.com/pdf/tcmirafi/TenCate-Polyfelt-Megadrain-Product-
Data-B23090.pdf>. (July 17, 2018).
TenCate (2015). "H2Ri Technical Data Sheet."

<https://www.tencategeo.us/media/8a60c7db-3952-4373-bcd0-
9fa7f96875ed/6keJhw/TenCate%20Geosynthetics/Documents%20AMER/Techni
cal%20Data%20Sheets/Woven/Mirafi%20H2Ri/TDS_H2Ri.pdf>. (July 17, 2018).
Terzaghi, K. (1951). Theoretical soil mechanics, Chapman And Hall, Limited.; London.
Zhang, X. (2005). "Consolidation theories for saturated-unsaturated soils and numerical

simulation of residential buildings on expansive soils." Texas A&M University.
271
Zhang, X., Lytton, R., and Briaud, J. "Coupled consolidation theory for saturated-
unsaturated soils." Proc., Proc., 3rd Biot Conf. on Poromechanics (Biot
Centennial), Univ. of Oklahoma, 323-330.
Zornberg, J. (2010). “Geosynthetic capillary barriers.” Proc., 1st International GSI-Asia

Geosynthetics Conference, 16-18.
Zornberg, J. G., Azevedo, M., Sikkema, M., and Odgers, B. (2017). "Geosynthetics with
enhanced lateral drainage capabilities in roadway systems." Transportation
Geotechnics, 12, 85-100.
Koerner (2012). “Designing with Geosynthetics.” Xlibris Corporation, Bloomington, IN.
Henry, K. (1995). “The Use of Geosynthetics Capillary Barriers to Reduce Moisture

Migration in Soils.” Geosynthetics International, 2(5), DOI: 10.1680/gein.2.0040.
Zhang, X., Presler, W., Li, L., and Jones, D. (2014). “Use of Wicking Fabric to Help
Prevent Frost Boils in Alaskan Pavements.” Journal of Materials in Civil
Engineering, 26(4), 728-740. DOI: 10.1061/(ASCE)MT.1943-5533.0000828.
Lin, C., Presler, W., Zhang, X., Jones, D., and Odgers, B. (2017). “Long-Term
Performance of Wicking Fabric in Alaskan Pavements.” Journal of Performance
of Constructed Facilities, 31(2), D4016005. DOI: 10.1061/(ASCE)CF.1943-
5509.0000936.
Richards, L.A. (1931). “Capillary Conduction of Liquids through Porous Mediums.”

Physics, 1, 318-333.
AASHTO (1993). “AASHTO Guide for Design of Pavement Structures.” American

Association of State Highway and Transportation Officials, Washington, DC.
TenCate Geosynthetics America (2013). <https://www.tencategeo.us/en-

us/resources/Case-Studies/Case-study-explorer/UhiT/Subgrade-Stabilization-St-
Louis-County-MO>. (latest access on Febuary 4, 2019)
272
VI. LABORATORY DRAINAGE PERFORMANCE OF A NEW GEOTEXTILE

WITH WICKING FABRIC
Chuang Lin1 and Xiong Zhang2
1
2
Associate Professor, M. ASCE (Corresponding Author), Department of Civil,
ABSTRACT
Geotextiles are commonly used in road design to provide mechanical and
hydraulic functions due to its high strength and porous properties. The long-term
performance of a soil-geotextile system is not solely determined by soil or the geotextile
itself, it also depends upon their interactions. Only emphasizing the geotextile mechanical
functions cannot solve all engineering issues. The performance of a soil-geotextile system
can be disappointing with small soil moisture content increment. Capillary water exists in
a variety of soils and can be difficult to drain under unsaturated condition. Unfortunately,
the existing subsurface drainage design methods can only deal with free water (water
flows under the influence of gravity) and conventional capillary barriers cannot wick
water out of pavement structures, resulting in overestimating the pavement long-term
performance.
A new geotextile with wicking fabric has the potential to solve this issue.
Preliminary laboratory and field tests have proved its effectiveness to wick capillary
273
water under unsaturated condition. This paper further explores some potential concerns
regarding the future extensive applications of the new geotextile. A series of lab tests
were performed to study the new geotextile working mechanism, evaluate its drainage
efficiency, and assess the effects of splice and clogging on the new geotextile drainage
performance. Test results indicated that the new geotextile worked effectively to drain
both gravitational and capillary water out of pavement structures. However, the splices
may reduce the new geotextile drainage efficiency. Clogging effect shall not be a major
concern that influences the new geotextile long-term performance in soils with fine
content lower than 14.5%.
KEY WORDS: geotextile, unsaturated soil, drainage, capillary water, wicking
fabric
1. INTRODUCTION
Geotextiles are part of a large family of geosynthetics and are often used in civil,
geotechnical, environmental, and structural engineering designs. Geotextiles are porous
materials that present the widest range of properties and can be used to fulfill most of the
geosynthetics functions, such as mechanical (protection, reinforcement and separation) or
hydraulic (drainage, filtration, and capillary barrier) (Bouazza et al., 2006). As for their
applications in pavement design, geotextiles together with other geosynthetics materials
are often used to mitigate pavement distresses. Because geotextiles are made of
polyethylene fibers, which have higher tensile strength, they are often used to provide
lateral restraint for base and subgrade materials, to increase the system bearing capacity,
274
and also to provide additional wheel load support (Holz et al., 1998). As separation
materials, geotextiles are placed between two dissimilar materials and maintain the
integrity of both. For confinement function, geotextile and geogrid are utilized together to
prevent aggregate lateral movement and reduce the potential rutting issue (Bueno et al.,
2005). For reinforcement function, geotextiles work effectively to spread the load and
prevent excess load on different components that make up the road (Benjamim et al.,
2007). Berg et al. (2000) showed the benefits of using geosynthetics (geotextile and
geogrid) for reinforcing flexible pavement and concluded that the BCR (Base Course
Reduction) value varied from 30% to over 50%, indicating that the thickness of a base
course can reduce to half if proper geotextile (or other geosynthetics material) is installed.
However, only emphasizing the geotextile mechanical functions cannot solve all
engineering issues and may result in overestimating the pavement long-term
performance. The long-term performance of a geotextile reinforced soil infrastructure is
not solely determined by soil or geotextile itself. It depends upon soil, geotextile, and
their interactions. In fact, soil, as the main body of a soil-geotextile system, is sensitive to
moisture variations. Excess water can cause numbers of detrimental engineering
problems, such as soil expansion and collapsing, soil strength and stiffness reduction,
excess pore water pressure increment, and crack propagation (Han and Zhang, 2014). The
performance of a soil-geotextile system can be disappointing when soil moisture content
increases a small amount. Li et al. (2011) reported that the Alaska D-1 base course would
experience a 50% loss in resilient moduli if the soil moisture content increased from
3.3% to 6.0%. Similarly, Lin et al. (2017) reported that an AB-3 base course also suffered
from a dramatic resilient moduli reduction (from over 600 MPa to about 200 MPa) as the
275
moisture content changed from 6.5% to 8.5%. Moreover, much severe rutting issues of
granular materials were reported in numbers of published papers due to soil wetting
(Arnold et al., 2002; Huurman and Molenaar, 2006; Shoop and Henry, 1991). In addition,
the rise and accumulation of water caused frost heave in cold regions (Konrad and
Morgenstern, 1980).
By investigating the existing subsurface drainage design methods (AASHTO,
1993; Baus and Stires, 2010), only free water are considered drainable. Many previous
studies have discussed the application of geotextiles as drainage materials under saturated
conditions (Koerner, 2012; Palmeira and Fannin, 2002). In contrast, capillary water is
held in the soil against the pull of gravity and cannot be drained out by traditional
drainage systems. In addition, suction (negative pore water pressure) is the technical term
to depict the intrinsic soil property of holding water. Suction tends to increase as a soil
dries out. As the fine content increases in a soil, the capillary action will also increase. A
common treatment of controlling the capillary water is to install a layer of coarse grained
soil (such as gravel or sand) or geotextile, impeding the capillary water from wicking
upward. Unfortunately, the accumulated capillary water flows laterally and gradually
saturates the capillary barrier. As excess water accumulates near the capillary barrier, the
soil suction gradually decreases till the breakthrough point (suction value at which water
can flow through the capillary barrier) is achieved. At this point, the capillary barrier
ceases working as a barrier and the soil water content is expected to increase and
eventually causes severe rutting issues. That is to say, capillary water shall be an
equivalent factor that influences the pavement long-term performance. Moreover, in most
conditions, geotextiles are implemented above the groundwater table and the soil-
276
geotextile system are under unsaturated conditions during most of its service life. The
permeability of both soil and geotextile significantly decreases with decreasing degree of
saturation (Hillel, 2013). The traditional subsurface design methods cannot deal with this
situation and often result in over-estimating the pavement drainage efficiency. In sum,
traditional drainage design methods do not take capillary water into consideration, and
the capillary water treatments suffer from malfunctioning as the soil approaching
saturation. It is urgent to come up with an effective new product or treatment method to
deal with capillary water accumulation within a pavement structure so that its overall
performance will be further improved.
2. A NEW GEOTEXTILE WITH WICKING FABRIC
2.1. PROPOSED DRAINAGE DESIGN WITH WICKING FABRIC
Recently, a newly developed geotextile with wicking fabric (hereafter name as
“wicking fabric”) can potentially be used to laterally transport both gravitational and
capillary water within a pavement structure. Figure 1a shows the weaving texture of the
wicking fabric, which is a dual function geotextile product: the high modulus
polypropylene yarns (black) for reinforcement purpose and the special hydrophilic and
hygroscopic wicking fibers (white) for drainage purpose. Figure 1b shows the Scanning
Electron Microscope (SEM) image of enlarged wicking fabric. As shown in Figure 1b,
the deeply grooved cross section provides larger surface area, thus ensures the channel to
hold and laterally transport water under unsaturated conditions as schematically
illustrated in Figure 1c. The average diameter of the wicking fabric is between 30-50 μm
277
and the average groove spacing is between 5-12 μm. Detailed information for the wicking
fabric hydraulic and mechanical specifications can be found in Table 1. When properly
designed and installed in a road embankment as indicated in Figure 1d, the wicking fabric
has the potential to dehydrate the capillary water and consequently improve the pavement
performance for the long run. The wicking fabric is purposed to be installed in the base
layer. Along the embankment slopes, the wicking fabric is exposed to the air with a
length of 0.3-1.0 m, allowing for water evaporation. The relative humidity (RH) in the air
is often less than 90% and the corresponding suction value can be as high as 14 MPa
(Fredlund and Rahardjo, 1993). The soil surface dries out quickly and forms an air-dried
crust which has very low permeability (nearly impermeable) to prevent the inside water
from getting out. The wicking fabric serves as a “pipe” to transport water and the natural
environment works as a “natural pump”, which will continuously wick the water out of
the embankment by taking advantage of the high suction gradient between the air and the
soil inside. Finally, the water will be evaporated into the air via the exposed end.
Compared with the amount of water needed to saturate the earth’s atmosphere, the
amount of water in the embankment is very small. Therefore, this new type of soil-
geotextile system can be used as an effective and environmentally sustainable drainage
system to reduce the distresses induced by capillary water accumulation and improve the
overall pavement performance in the long run.
2.2. LAB VALIDATION
A series of preliminary lab tests have validated the drainage efficiency of the
wicking fabric. First, it is reported that the wicking fabric could transport water to a
278
distance of 1.83 m within 16.5 hours with zero hydraulic gradient (Figure 2a) and wick
water to a height of 0.25 m within 2 hours during hanging column test (Figure 2b) (under
room temperature and relative humidity of approximately 50%). Test results indicated
that the wicking fabric could not only maintain saturation and transport water under
unsaturated conditions, but also could counteract gravity to some extent.
In addition, researches at University of Alaska Fairbanks (Zhang and Presler,
2012) conducted rainfall infiltration test to evaluate the wicking fabric drainage
efficiency. In total four different types of geotextiles were used, including the wicking
fabric, a high performance (HP) reinforcement geotextile, a geotextile water filter and a
drainage geocomposite. Test results indicated that conventional geotextiles ceased to
transport water within 1 day under unsaturated condition. In contrast, the surface of the
wicking fabric was wet throughout the testing period and soil moisture content was the
lowest among all tested geosynthetics. Test results further proved that the wicking fabric
had advantages to wick capillary water out compared with conventional geotextiles.
Wang et al. (2017) also performed a laboratory rainfall test and indicated that the amount
capillary water drained out of the testing box can be 1.6 times larger than that of free
water.
2.3. FIELD APPLICATIONS
Besides laboratory tests, a few successful field applications of the wicking fabric
are found in the literature. Zhang et al. (2014) reported a successful application of the
wicking fabric to prevent frost boils in Alaska pavements. This project located at the
Beaver Slide of Dalton Highway, AK, which is about 8 km south of the Arctic Circle.
279
The soils were gravel with sand with a fine contents of greater than 6%, which are
considered as Frost Susceptible (FS) soils. The capillary water raised up to the base
course and caused heaving action when soil temperature dropped below 0 °C. During the
next early spring, thaw weakening caused extensive damages to the pavement structures.
Previous rehabilitation by Alaska Department of Transportation and Public Facilities
(AKDOT&PF) with French drain installation and other countermeasures were reported
unsuccessful. Two layers of the new wicking geotextiles were installed at approximately
0.9 m and 1.2 m below the existing road surface. In total 22 pairs of Time-Domain
Reflectometer (TDR) and temperature sensors were used to monitor the temperature and
moisture content changes of the 18.3 m-long road section. Figure 3 shows the comparison
of the field test section before and after rehabilitation. One year after installation, all soft
spots disappeared and no thaw weakening was observed in the spring time. In addition, a
clear road surface difference was visualized for sections with and without wicking fabric.
Field visit indicated that the soil at shoulder was damp, implying that water flowed along
the wicking direction. The wicking fabric successfully eliminated the frost boils and thaw
weakening to a depth of 1.2 m. Lin et al. (2017) continuously monitored the test section
and the results showed a promising performance for a period of 5 years.
3. RESEARCH OBJECTIVES
Although both laboratory and field test results indicated that the application of the
wicking fabric for capillary water removal was very promising, concerns were raised
regarding the use of the geotextile for more extensive future applications. For instance,
280
will the wicking fabric be effective for other types of soils, say dirty sand or gravel with
higher fines contents? Will the wicking fabric continue to work when length requirements
exceed the width of the geotextile? Also, will clogging effect influence the wicking fabric
drainage efficiency?
In order to answer those questions, two types of soils were used in this paper,
including well-graded river sand and E-1 aggregate with about 14.5% of fines. The sand
material is used to provide an understanding of how well the wicking fabric works in
permeable soil and the E-1 aggregate is used to assess the geotextile performance for FS
soil. Two testing flumes with different lengths were constructed to evaluate the soil-
geotextile system drainage efficiency. One was 6.5 m in length filled with sand and the
other was 22.25 m in length filled with E-1 aggregate. The moisture content variations
with time during the wicking and wetting processes were evaluated.
4. TESTING FLUME CONSTRUCTION
Figure 4 shows the schematic plot of the design for the testing flume for sand and
the construction procedures. The dimensions of the testing flume were 6.50 m × 0.41 m ×
0.30 m (Length × Width × Height). A layer of plastic wrap was placed within the testing
flume to prevent unexpected leakage. The wicking geotextile was installed at a depth of
2.54 cm from the bottom. Because the length of the wicking fabric was shorter than the
length of the testing flume, two pieces of the wicking fabrics were spliced together with
about 1 m overlapping starting at 1.52 m from left end of the testing flume (red line
shown in Figure 4a). One end of the wicking fabric on the left was exposed to the air for
281
evaporation purpose. Within the testing flume, in total three layers of soil moisture
sensors were installed at elevations of 2.54 cm, 12.70 cm, and 22.86 cm from the bottom
of the testing flume, respectively. The spacing between the adjacent two sensors was 0.76
m.
The construction process started with the testing flume, as shown in Figure 4b.
The frame of testing flume was made of wood stud and the side walls were made of
plywood. The first picture shows the configuration of the testing flume. The testing flume
seated on the supporting wooden frame which was made of wood stud. The vertical wood
studs were attached to the side walls to provide additional confinement. After the testing
flume was ready to use, a layer of 6-mil plastic wrap was put within the testing flume.
Then, a layer of 2.54 cm-thick sand was hand compacted, leveled and saturated (second
picture). After that, a layer of wicking fabric installed on top the saturated sand layer and
the third figure shows the location of the splice area. The other end of the wicking fabric
was exposed to the air, as shown in the fourth picture. The fifth picture shows the sensor
installation process. Two types of sensors were used, including soil volumetric moisture
content sensor (left side) and water potential sensor (right). The sand was then backfilled
to the testing flume with 2.54 cm-height per lift to ensure sufficient the compaction
effort. The sixth picture demonstrate the testing flume condition after all the sand was
backfilled in the testing flume and sensors were installed in location. The top of the
testing flume was then covered by another layer of plastic wrap to minimize water
vaporization. All the sensors were connected to a CR1000 data logger with an AM 16/32
multiplexer (for volumetric moisture content sensors) and a bread board (for water
282
potential sensors), as shown in the seventh picture. The data logger was pre-programmed
to record data hourly.
The design and sensor layout were similar for both testing flumes. For the testing
flume for E-1 aggregate, the length of the testing flume was 22.25 m and the spacing
between two adjacent sensors was 1.37 m. The locations of the wicking fabric splices
were centered at 1.62 m, 5.33 m, 8.99 m, 12.65 m, 16.31 m and 19.96 m from the left
side of the testing flume, respectively.
To obtain the moisture content distributions within the monitored area, a
continuous moisture content distribution field is needed and only a discrete moisture
content field can be obtained from the recorded data. Therefore, a numerical interpolation
technique was used to generate a continuous moisture content field. Contour plots were
used to facilitate visualization of the moisture content migrations within the testing
flume. It is important to point out that the 24 sensors could not cover the entire area of the
testing flume and the moisture content contours could only represent the moisture
migration with in the monitored area. Moreover, due to the limited sensors and potential
mathematical error in numerical interpolation, it is more meaningful to read the pattern of
a whole contour plot than to focus on the value of a single point in the contour plot.
Firstly, the recorded data from 24 sensors was used as controlling points to linearly
interpolate the moisture contents between sensors. Secondly, a mesh grid was created
upon interpolated locations by using the Delaunay triangulation method (MathWorks,

283
2016). Thirdly, the contours connected the points with the same moisture contents.
Finally, the moisture contour was generated hourly and the moisture contours at different
times were processes into movies to continuously illustrate the dynamic water movement
and soil moisture variations. The following discussions are based on the snapshots of the
moisture contour movies.
5.1. TESTING FLUME FOR SAND
In this section, the test results for the testing flume for sand will be discussed and
demonstrated. First, the total amount of water variation with time within the testing flume
will be discussed to demonstrate the wicking fabric overall performance. Secondly, the
wicking fabric performance during the wicking and wetting (left end and right end) test
will be presented to further evaluate the wicking fabric performance under different
conditions.
5.1.1. Wicking Fabric Overall Performance in Sand. To better understand the
moisture migration within the testing flume, the total amount of water and its variation
with time are analyzed and shown in Figure 5. The iso-moisture content lines could be
first determined based upon each moisture contour frame. After that, the area between two
adjacent iso-moisture content lines was calculated. Then, the average volume of water
within each area could be determined by multiplying the average volumetric water content
with the covered volume (area × testing flume width (0.41 m)). Finally, the total amount
of water within the monitored area would be the summation of the volume of water within
each area. Since the testing time was also known from the recorded data, the average flow
rate for the monitored area could also be determined.

284
Figure 5 shows the overview of all the tests performed on the testing flume for
sand. In total three types of tests were performed, including one wicking (drying) test,
two wetting tests (left end and right end). Before performing the wicking test, the soil
was resaturated. The total amount of water that required to restaurate the entire testing
flume was only 4 L, which did not make sense. This phenomenon is mainly caused by
three reasons: 1) the monitored area was limited and could not cover the entire testing
flume area; 2) the permeability for sand was relatively large (10-4 m/s) and water could be
quickly drained downward to the bottom of the testing flume where the monitored area
could not cover; and 3) the data logger scanned every 5 minutes but stored data hourly. It
was highly possible that the first recorded data was not the peak moisture content that the
sensors experienced, resulting such a low amount of water required to saturate the testing
flume. Within 6 hours, the total amount of water almost dropped back to the condition
before testing, which indicated that at this stage, free water was drained under the
influence of gravity. Starting from this point till the end of the test (October 29, 2014), an
extra 4 L of capillary water was drained out by the wicking fabric and the average flow
rate was 0.08 L/day.
Then the left end wetting test started on October 29, 2014. The wicking fabric was
lifted to 0.6 m above the bottom of the testing flume and submerged in water. This test
was used to simulate an extreme situation when there is long lasting rainfall. As shown in
Figure 5, limited amount of water flowed back to the testing flume due to the existence of
the geotextile splice. The testing flume experienced cyclic water increasing and
decreasing processes. The fluctuation of water content in the testing flume during the
wetting test can be explained as follows. If the splice area was fully saturated, there was
285
no suction gradient between the two ends of the wicking fabric. Under this condition,
water could not further transport into the testing flume. However, the exposed end of the
geotextile could wick capillary out of the testing flume through evaporation process. As
the degree of saturation in the splice area continued to decrease, the suction gradient
between the wicking fabric two ends increased. Increment in suction gradient, in turn,
resulted in water flowing back to the testing flume. This was a dynamic water balance
and the drying and wetting cycles continued repetitively.
After the first wetting test was completed, a period of 4 months was taken to
observe the moisture migration within the testing flume. Unfortunately, due to the
malfunction of the data logger, part of the data was missing (dash line in Figure 5).
During this period, the wicking fabric was exposed to the air again so that the wicking
fabric drainage efficiency could be determined for a relatively longer time span. The flow
rate was 0.08 L/day, which was consistent compared with wicking (drying) test. For the
second wetting test (right end), each time approximately 38 L of water was introduced to
the system but only 19 L was observed in the monitored area. Again, the difference was
caused by the same reasons as discussed in the wicking test. The corresponding flow
rates were 0.16 L/day and 0.26 L/day, respectively, which were higher than the first
wetting test (left end).
In sum, the wicking fabric could work effectively to drain capillary water out of
the system. The splice would be a potential concern regarding the geotextile drainage
efficiency. Extra caution shall be paid to the installation process and the top piece should
correspond to the upper stream of the water flow direction. Moreover, the tested sand was
considered as good drainage material and should have good drainage properties. Even for
286
such a material, large amount of capillary water could not be drained out, let alone other
worse construction materials. Therefore, it is equally important to take capillary water
drainage into consideration and the wicking fabric indeed has the potential to wick both
gravitational and capillary water out of pavement structures.
5.1.2. Wicking (Drying) Test. Figure 6a shows the wicking test for sand during
the wicking test. The volumetric moisture content contours were plotted at the starting
point, 2 hours, 1 day, and 1 month. The dash line represents the geotextile location and the
colormap ranges from white (lowest moisture content, 0.1) to black (highest moisture
content, 0.35). The test started at 2 pm on September 6, 2014. The saturation moisture
content for sand was 0.3, which could be used as an indication of any saturated zone in the
testing flume. The saturation process was not captured because the contour could not
cover the entire testing flume area and the data was scanned every 5 minutes but stored
hourly. The top of the testing flume already experienced an unsaturated condition at the
starting point. That was the reason why the moisture content at the starting point were all
below 0.30. However, the moisture contents were relatively uniformed distributed with a
nearly saturation zone (0.29) extended from 1.8 m to 3.6 m in the horizontal direction,
where was the location of wicking fabric splice.
The next plot demonstrates the moisture content contours 2 hours after the test,
when free water was already drained out (Barber and Sawyer, 1952). The moisture
content on top of the flume further decreased to 0.21. However, the drainage efficiency
significantly decreased at the splice area. As more water drained downward and could not
be further drained out laterally, the saturation zone was observed at 2.2 m-3 m in the
horizontal direction, exactly at the splice location. The testing flume was now artificially
287
divided into two sections: the left section started from the left edge to the saturation zone,
and the right section was from the saturation zone to the right end. Firstly, on the left
side, due to the existence of high suction gradient, the excess water could be easily
drained out and the moisture content on the left side was much lower than that on the
right. In contrast, the suction gradient was limited due to the existence of the saturation
zone, flow velocity on right side would be significantly decreased and the overall
moisture content was relatively high one day after the test (as shown in the third picture
in Figure 6a). As water flowed downward and accumulated at the bottom, an extra
saturation zone was observed at 5 m-6 m in the horizontal direction. The fourth picture
shows the moisture contour 1 month after the test. It is worth to note that the overall
moisture content on the right side decreased and the saturation area increased in size.
This phenomenon indicated that the water within the testing flume gradually migrate to
the left size, but at a very slow rate. Compared with the previous moisture contours, the
area of the saturation zone at 3 m increased while the saturation zone at 5.5 m was
replaced by a drier area than the surrounding soil. In sum, even though the existence of
geotextile splice significantly reduced its drainage efficiency, the geotextile showed its
ability to drain capillary water out of the system.
5.1.3. Wetting Test. Two types of wetting test were performed: one was dipping
the exposed geotextile directly into water and letting water flow back to the testing flume
through the geotextile by slightly raising the geotextile to an elevation higher than the
testing flume; and the other test was to dig out part of the sand on the right side and
directly introduce water to the system. Each time, 38 L of water was added to the upper
288
level of the testing flume, flowing through the whole testing flume, and drained out at the
exit end of the wicking fabric.
Figure 6b gives the moisture contours for the first wetting test starting point, 2
hours and 1 day. At the starting point, a saturation area was observed from 2.1 m-3.4 m
over the splice area. Two hours after the test, an extra saturation zone was observed at 4.7
m and the average moisture content on the right side increased. This phenomenon
indicated that the water flowing back to the testing flume. However, the amount of water
flow into the test flume was limited. In addition, the overall moisture content decreased
after 1 day, this phenomenon will be discussed in detail later in this section.
Figure 6c shows the moisture contours for the second wetting test where in total
38 L of water was introduced to the system on March 13 and 28, 2015, respectively. The
wetting test on March 13, 2015 is taken as an example to explain the moisture migration
process. At the starting point, the moisture content closer to the right bottom side
experienced increment due to the excess water introduced to the system. However, the
moisture content in the upper part of the test flume did not increase too much, indicating
that the added water was quickly drained through the wicking fabric at the bottom under
saturated conditions. This was also consistent with the visual observations that water was
also flowing out of the wicking fabric during the testing process at the exit. One day later,
the area of the nearly saturation zone on the right side slightly decreased. After 10 days,
the moisture content on the right side continuously decreased and a saturation zone was
observed at 3 m, which was an indication that the water was flowing from the right side
to the left side. It is worthwhile to point out that the flow rate of right end wetting test
289
(second wetting test) was higher than that of the left end wetting test (first wetting test),
due to existence of the splice area.
5.1.4. Influence of Wicking Fabric Splice Area. Figure 7 further demonstrates
how the geotextile splice area influences the system drainage efficiency. Figure 7a
schematically shows the profile when two pieces of directional wicking fabrics (refer to
Figure 1a) are spliced together under saturated conditions. The white arrows represent the
nylon wicking fibers, while the black colors represent the reinforcement component. As
shown in Figure 7a, when two pieces of wicking fabrics are spliced together, there are
inter-geotextile pores with much larger pore sizes than those in the wicking fabric and
even more than those in the wicking fibers. It is well known that the air-entry values of
porous media are inversely related to the pore size. In other words, the larger pore sizes
between the two pieces of wicking fabrics have much lower air-entry values than the
wicking fabric and will be firstly desaturated when the suction increases as shown in
Figure 7b. Assume water flows from right side (top piece) to the left side (bottom piece).
Due to the geotextile interwoven structure, the two pieces are point to point contact within
the splice area. The special multichannel structure ensures that the inner drainage paths
remain saturated under unsaturated conditions (please refer to Figure 1c). In other words,
the inner drainage channels are unobstructed for water transportation. When the system is
under saturated condition (Figure 7a), the space between two geotextile pieces is filled
with water and water can easily flow from the top piece to the bottom piece under the
influence of gravity. This was the reason why in Figure 6c, when water was added to the
testing flume, it was quickly drained through the wicking fabric. However, when the
system is under unsaturated conditions (Figure 7b), the space between two pieces is
290
(partially or fully) fill with air, which impedes the water transportation from the top piece
to the bottom one. Because the unsaturated hydraulic conductivity significantly decreased
with degree of saturation, the drainage efficiency of the geotextile will be reduced
dramatically.
5.2. TESTING FLUME FOR E-1 AGGREGATE
In this test, the length of the testing flume increased to 22.25 m to evaluate the
wicking fabric performance when length requirements exceed the width of the geotextile.
The E-1 aggregate was used, which contained about 14.5% of fines. Due to such high
fine content, E-1 aggregate was classified as frost susceptible soil and were not
recommended to be used as base course. Therefore, this type of soil can be a
representative to demonstrate the wicking fabric drainage efficiency in a relatively bad
field condition. In addition, the numbers of the wicking fabric splice sections may
increase for wider road sections and the drawbacks of the application for this type of road
needs to be further explored.
5.2.1. Wicking Fabric Overall Performance in E-1 Aggregate. Figure 8 shows
the total amount of water within the 22.25 m-long testing flume for E-1 aggregate. In total
two types of tests were performed, including wicking test and wetting test. The wicking
(drying) test started on June 12, 2015 and ended on August 8, 2015. The wetting test
started on August 8, 2015 and ended on September 3, 215. Again, because the E-1
aggregate was compacted with several lifts and saturated after compaction of each lift, it
did not require too much water to resaturate it. Moreover, similar to the sensor
distributions for the testing flume of sand, the monitored area was also not able to cover
291
the entire testing flume area. The total amount of water increased by approximately 50 L
within the monitored area during the saturation process. Because the permeability of E-1
aggregate was much lower than the tested sand, the total amount of water gradually
reduced with time and no sudden drop was observed. There were three distinct stages of
water flow, as shown in Figure 8. From the starting point to Day 3 (June 12-15), the
degree of saturation for E-1aggregate was relatively high, resulting in a higher water flow
rate of 14.3 L/day. Then from June 15 to July 3, the degree of saturation kept decreasing
and the corresponding flow rate further reduced to 1.83 L/day. After that, limited amount
of water could be drained out from the system due to relatively high fine content in the
aggregate. The total amount of water drained out from the monitored area was 12 L within
36 days (0.3 L/day).
For the wetting test, about 0.3 m-wide E-1 aggregate on the right side of the
testing flume was removed and water was introduced to the system. In total 105 L of
water was poured into the system. Because the water flow path was 22.25 m, which was
much longer than that for sand, the water flow rate was smaller than that for the wicking
test. The average flow rate during the first three days was 10 L/day and then reduced to
2.1 L/day (from Aug 12 to Aug 25). After that, the flow rate further decreased to 0.57
L/day. In sum, the geotextile worked effective in reducing the capillary water for E-1
aggregate. However, considering the high fine content and the extremely long transfer
distance, the wicking fabric drainage efficiency was much slower and it would take
longer time to achieve the same performance.
5.2.2. Wicking (Drying) Process. Figure 9a shows the moisture content contours
for the wicking test at starting point, 4 weeks and 8 weeks. The saturation moisture
292
content was 0.40. Similar to the moisture contour for the testing flume of sand, the
moisture contour could not capture the saturation process. In total six saturation or nearly
saturation areas were observed, all at the wicking fabric splices. This phenomenon
indicated that splice was indeed an important factor that might influence the wicking
fabric performance, regardless of soil types. Then the soil moisture content consistently
dropped over time. After 4 weeks, the number of saturation areas reduced to three and
only one of them was saturated (located at 17 m). This indicated that wicking fabric
worked effectively to wick water out of the testing flume.
The next plot shows the moisture content contour 8 weeks after the wicking test.
The moisture content distribution did not change significantly compared with the
previous plot. This implied that the geotextile did not work effectively to wick water out
during the past 4 weeks. However, the total amount of water within testing flume kept
decreasing (refer to Figure 8), but at a relatively low rate. This implies that the wicking
fabric was still functional, but the drainage efficiency decreased. Visual observation
during the test also validated the authors’ assertion. There was always at least 0.30 m of
the exposed wicking fabric remained wet throughout the test, which indicated that the
wicking fabric was still working. There were two major reasons that resulted in the
reduction of wicking fabric drainage efficiency. Firstly, the E-1 aggregate contained
about 14.5% of fine content. Croney and Jacobs (1967) indicated that the smaller the pore
size distribution of a soil, the greater the driving force (capillary action) and the greater
the capillarity. In other words, the high fine content within the E-1 aggregate increased its
ability to hold water and made it even harder for wicking fabric to wick water out of the
aggregate. Secondly, the hydraulic conductivity of both the soil and the wicking fabric
293
significantly reduced under unsaturated conditions. Corey (1957) found out that the
hydraulic conductivity is heavily dependent on the connectivity of it pore size. As soil
and wicking fabric became unsaturated, they contained many tortuous pores that abruptly
end and water would have less path to completely pass directly through the material.
5.2.3. Wetting Process. Figure 9b shows the moisture contours for wetting test at
the starting point, 1 week and 3 weeks. At the starting point, the excess water on the right
side could easily flow horizontally with a relatively large saturation are observed from 16
m to 22 m. Two days after the wetting test, the bucket at the exposed end collected about
19 L of water and no liquid water could be collected since then. One week after the
wetting test (second picture), the saturation area further transferred to the middle section.
The saturation or near saturation zone was distributed relatively even from 5 m to 20 m in
the horizontal direction, indicating that the wicking fabric redistributed the water within
the testing flume. After 3 weeks (third picture), the connected saturation area became
separated, indicating the water was gradually drained out by the wicking fabric and.
compared with the moisture content in the second picture, the moisture content from 0.0 -
1.5 m from left side increased, also implying the overall water within the testing flume
was gradually migrating from right to left. In summary, the geotextile did gradually wick
the water out of the system, but the drainage efficiency was lower than that for sand due to
the existence of high fine content.
5.3. CLOGGING EFFECT
Except for the macroscopic study of the geotextile wicking ability, another
equally important factor that might influence its drainage efficiency is clogging. Due to
294
the special multichannel cross section structure, the average opening spacing was about
5-12 μm. Soils with particle size smaller than this value may cause clogging issue. To
evaluate the wicking fabric clogging severity level in a microscopic scope, wicking fabric
samples were collected from both Beaver Slide field section and from the testing flume
for E-1 aggregate. The reason to collect samples from the field section was that the in-situ
soils would be more representative to evaluate the clogging severity on the wicking
fabric. Due to a relatively larger soil particle size (greater than 75 μm), the well-graded
sand is not considered as a factor causing clogging issue.
Figure 10a gives the SEM (Scanning Electron Microscope) images of the
geotextile samples collected from field section. Because the base course was composed
of degraded granite, which was classified as silt with gravel according to USCS
classifications. The geotextile was in direct contact with the surrounding soil, the surface
was contaminated, and the deep grooves of the surface fibers were all covered with fine
soil particles (left image in Figure 10a). However, with the same wicking fabric yarn, the
wicking fibers laid under the surface fiber did not suffer from clogging effect as indicated
in the right image in Figure 10a. This phenomenon inferred that the wicking fabric as an
integrity was still effective to wick capillary water out of the road, even though part of
the surface fibers were affected by the clogging of the surface fiber. Field observation
also proves this assertion since the wicking fabric was still functional five years after
rehabilitation and no soft spot was observed ever since.
Figure 10b shows the SEM images of the geotextile collected from the testing
flume for E-1 material. As indicated in the left image, most of the wicking fibers on
surface were relatively clean. Even though the E-1 material contained about 14.5% of
295
fines with medium plastic property, the small soil particles were either formed as a soil
cluster or coated on the surface of gravel, rendering a less severe clogging issue on the
wicking fabric surface. In addition, the right image in Figure 10b shows the wicking
fibers beneath the surface, no clogging issue was observed. In conclusion, clogging effect
shall not be a major concern that will influence the wicking fabric long-term drainage
efficiency.
6. CONCLUSIONS
This paper studied the wicking fabric drainage efficiency with different soil types.
Series of laboratory wicking and wetting tests were performed and samples were
collected to evaluate the clogging severity. The conclusions are summarized as follows:
1. The wicking fabric had the ability to wick both gravitational and capillary water
out of pavement structure by taking advantage of the suction gradient generated
between the two ends. For conventional good drainage material such as sand, the
wicking fabric could further reduce the moisture content and continuously
dehydrate the soil. The wicking process of capillary water required much longer
time compared with free water flow, the water content could be further reduced if
longer testing time was allowed.
2. The splice area significantly reduced the wicking fabric drainage efficiency. Point
to point contact made the hydraulic conductivity of the splice area significantly
decreased under unsaturated conditions. However, if properly designed and

296
installed with the top piece located at the upper stream of water flow, the effect of
the splice area could be minified.
3. The wicking fabric worked effectively for E-1 aggregate with 14.5% of fines.
Even though the drainage efficiency would be expected to reduce due to soil
increasing capability to hold water, the capillary water could still be drained out
by the geotextile, but at a lower rate.
4. Clogging effect shall not be a major concern for the tested soils. The wicking
fibers in direct contact with the surrounding soils serve as a protective layer to
keep the fibers beneath from clogging.
5. The amount of water drained out by the wicking fabric was much higher than that
flows into the roadway considering the limited rainfall events within a year. Field
observation clearly indicated that no soft spots were observed after installing the
new geotextile. Therefore, very limited amount of water is expected to be wicked
into the embankment.

297
Table 1. Geotextile specifications

Mechanical Average Roll
Test Method Unit Tested Value
Properties Value
Tensile Modulus
ASTM
@ 2% Strain kN/m 657
D4595
(CD)
ASTM
Permittivity Sec-1 0.24
D4491
ASTM
Flow Rate l/min/m2 611
D4491
ASTM
Pore Size (O50) microns 85
D6767
ASTM
Pore Size (O95) microns 195
D6767
Apparent
ASTM
Opening Size mm 0.43
D4751
(AOS)
Wet Front 6.0
ASTM
Movement inches (Vertical
C1559
(24 minutes) Direction)
Wet Front
73.3
Movement ASTM
inches (Horizontal
(983 minutes) C1559
Direction)
Zero Gradient
298
(a) (b) (c)
(d)
Figure 1. New geotextile structure and conceptual application: (a) geotextile weaving
texture, (b) SEM image of wicking fiber, (c) wicking fiber cross section, and (d)
conceptual drainage design with the new geotextile
(a)
Figure 2. Wetting front movement tests: (a) horizontal wicking test (zero hydraulic
gradient), and (b) vertical wicking test (anti-gravitational force)
299
(b)
Figure 2. Wetting front movement tests: (a) horizontal wicking test (zero hydraulic
gradient), and (b) vertical wicking test (anti-gravitational force) (cont.)
Figure 3. Wicking fabric field application at Beaver Slide on Dalton Highway, AK

300
(a)
Figure 4. Testing flume design and construction: (a) schematic plot, and (b) construction
process
301
(b)
Figure 4. Testing flume design and construction: (a) schematic plot, and (b) construction
process (cont.)
302
Figure 5. Total amount of water (within the monitored area) in the testing flume for sand
302
303
(a)
(b)
Figure 6. Moisture content contours of the testing flume for sand: (a) wicking (drying)
test, (b) wetting test (left end), and (c) wetting test (right end)
304
(c)
Figure 6. Moisture content contours of the testing flume for sand: (a) wicking (drying)
test, (b) wetting test (left end), and (c) wetting test (right end) (cont.)
(a)
Figure 7. Effect of wicking fabric splice area: (a) saturated condition, and (b) unsaturated
condition
305
(b)
Figure 7. Effect of wicking fabric splice area: (a) saturated condition, and (b) unsaturated
condition (cont.)
306
Figure 8. Total amount of water (within the monitored area) for the testing flume for E-1 aggregate
306
307
(a)
(b)
Figure 9. Moisture content contours of the testing flume for E-1 aggregate: (a) wicking
(drying) test, and (b) wetting test
308
(a)
(b)
Figure 10. SEM images of wicking fabric samples: (a) collected from Beaver Slide test
section (left: on surface; and right: beneath surface), and (b) collected from testing flume
for E-1 aggregate (left: on surface; and right: beneath surface)
309
REFERENCES
AASHTO (1993). “AASHTO Guide for Design of Pavement Structures, 1993”, Volume 1,
American Association of State Highway and Transportation Officials (AASHTO).
Arnold, G., Dawson, A., Hughes, D., and Robinson, D. (2002). “The Application of
Shakedown Approach to Granular Pavement Layers”. Paper presented at the
Proceedings of the 9th International Conference on Asphalt Pavements,
Copenhagen.
Barber, E., and Sawyer, C. (1952). “Highway Subdrainage”. In Highway Research Board
Proceedings, Volume 31.
Baus, R., and Stires, N. (2010). “Mechanistic-Empirical Pavement Design Guide

Implementation”. No. FHWA-SC-10-01. University of South California,
Department of Civil & Environmental Engineering.
Benjamim, C., Bueno, B., and Zornberg, J. (2007). “Field monitoring evaluation of
geotextile-reinforced soil-retaining walls”. Geosynthetics International, 14(2),
100-118.
Berg, R. R., Christopher, B. R., and Perkins, S. (2000). “Geosynthetic Reinforcement of

the Aggregate Base/Subbase Courses of Pavement Structures”. No. GMA White
Paper II.
Bouazza, A., Zornberg, J. G., McCartney, J. S., and Nahlawi, H. (2006). “Significance of
Unsaturated Behaviour of Geotextiles in Earthen Structures”. Australian
Bueno, B. S., Benjamim, C. V. S., and Zornberg, J. G. (2005). “Field Performance of A

Full-Scale Retaining Wall Reinforced with Nonwoven Geotextiles.” Slopes and
Retaining Structures under Seismic and Static Conditions, 1-9.
Corey, A. T. (1957). “Measurement of Water and Air Permeability in Snsaturated Soil”.

Soil Science Society of America Journal, 21(1), 7-10.
Croney, D., and Jacobs, J. (1967). “The frost susceptibility of soils and road materials”. Rrl
Reports, Road Research Lab, UK.
Fredlund, D. G., and Rahardjo, H. (1993). “Soil Mechanics for Unsaturated Soils”: John
Wiley & Sons.
310
Han, J., and Zhang, X. (2014). “Recent Advances in the Use of Geosynthetics to Enhance
Sustainability of Roadways”. Paper presented at the Invited Keynote Lecture,
Conference on Advances in Civil Engineering for Sustainable Development.
Hillel, D. (2013). “Introduction to Soil Physics”: Academic Press.
Holz, R., Christopher, B. R., and Berg, R. R. (1998). “Geosynthetic Design and
Construction Guidelines”. No. FHWA HI-95-038.
Huurman, R., and Molenaar, A. (2006). "Part 2: Rutting Performance Prediction of

Aggregate Layers: Permanent Deformation in Flexible Pavements with Unbound
Base Courses". Transportation Research Record: Journal of the Transportation
Research Board 1952, 31-38.
Koerner, R. M. (2012). “Designing with Geosynthetics”. Volume 1, Xlibris Corporation.
Konrad, J.M., and Morgenstern, N. R. (1980). “A Mechanistic Theory of Ice Lens

Formation in Fine-Grained Soils”. Canadian Geotechnical Journal, 17(4), 473-486.
Li, L., Liu, J., Zhang, X., and Saboundjian, S. (2011). “Resilient modulus characterization
of Alaska granular base materials”. Transportation Research Record: Journal of
the Transportation Research Board 2232, 44-54.
Lin, C., Presler, W., Zhang, X., Jones, D., and Odgers, B. (2017). “Long-Term
Performance of Wicking Fabric in Alaskan Pavements”. Journal of Performance
of Constructed Facilities, 31(2), D4016005.
Mathworks, MATLAB User’s Guide (2016). “Mathwork Inc.” South Natick, MA.
Palmeira, E., and Fannin, R. (2002). “Soil-Geotextile Compatibility in Filtration”. Paper

presented at the Geosynthetics: State of the Art-Recent Developments. Proceedings
Of The Seventh International Conference On Geosynthetics, 22-27 September 2002,
Nice, France.
Shoop, S. A., and Henry, K. S. (1991). “Effect of a Geotextile on Water Migration and
Frost Heave in a Large-Scale Test Basin”. Transportation Research Record:
Wang, F., Han, J., Zhang, X., and Guo, J. (2017). “Laboratory tests to evaluate
effectiveness of wicking geotextile in soil moisture reduction”. Geotextiles and
Zhang, X., and Presler, W. (2012). “Use of H2Ri Wicking Fabric to Prevent Frost Boils in
the Dalton Highway Beaver Slide Area, Alaska.” Final Report, No. INE/AUTC
12.23, Alaska University Transportation Center.
311
Zhang, X., Presler, W., Li, L., Jones, D., and Odgers, B. (2014). “Use of Wicking Fabric
to Help Prevent Frost Boils in Alaskan Pavements”. Journal of Materials in Civil
Engineering, 26(4), 728-740.
312
VII. LONG-TERM PERFORMANCE OF WICKING FABRIC IN ALASKAN

PAVEMENTS
Chuang Lin1, Wendy Presler2, Xiong Zhang3, David Jones4, and Brett Odgers5
1
Graduate Research Assistant, S.M. ASCE, Department of Civil and Environmental
Engineering, University of Alaska Fairbanks, Phone: (907) 799-9203, Email:
[email protected]
2
Graduate Research Assistant, Department of Civil and Environmental Engineering,
University of Alaska Fairbanks, AK 99775-5900.
3
Associate Professor, M. ASCE, Department of Civil Engineering, Architectural
Engineering, and Construction Management, College of Engineering and Applied
Science, University of Cincinnati, Cincinnati, OH 45221. Email:
[email protected] (Corresponding Author)
4
Director of Product and Application Development, TenCate Geosynthetics, 365 South
Holland Dr., Pendergrass, GA 30567.
5
Director of Roadway Reinforcement, TenCate Geosynthetics, 365 South Holland Dr.,
Pendergrass, GA 30567.
ABSTRACT
Beaver Slide is near 177.8 kilometer (110.5 mile) on the Dalton Highway and it is
downhill when heading north. The road gradient is approximately 11% and the road
prism is on a side hill. Each year, there are soft spots that usually appear in late April and
remain all summer, which are also called “frost boils”. The frost boils have resulted in
extremely unsafe driving conditions and frequent accident occurrences. Conventional
repair methods have not worked. A newly developed geosynthetic wicking fabric was
installed in the road structure in August 2010. The fabric has a high specific surface area
313
(consequently high wettability and high capillary action) and high directional
permittivity. Test results over the initial two years had proved the effectiveness of
wicking fabric to mitigate frost heave and the subsequent thaw weakening issue.
However, there were still some concerns regarding its long – term performance, such as
clogging of the microscopic drainage channels and mechanical failures. The data
collected during the past five years were analyzed to evaluate the effectiveness of the
wicking fabric. A scanning electron microscope (SEM) was used to explore the
interaction between the wicking fabric and in situ soils, and to determine the condition of
the fabric five years after installation.
KEY WORDS: frost heave, thaw weakening, geosynthetic, wicking fabric, and
long-term performance
1. INTRODUCTION
Beaver Slide is located at 177.8 kilometer (110.5 mile) on the Dalton Highway
and it is about 8.0 kilometers (5.0 miles) south of the Arctic Circle. The road is downhill
when heading north, at approximately 11% gradient. The road is constructed on a side
slope, where shallow groundwater drains down the slope starting each spring and lasts
until early winter. The excess water comes up into the road embankment, causing soft
spots and subsequent road damage, which are called “frost boils”. Heavy truck drivers
tend to brake when encountering the soft spots and make the condition even worse. The
frost boils have resulted in an extremely unsafe driving conditions and frequent accident
occurrences. Zhang et al. (2014) concluded that the frost boil issue was caused by two
314
mechanisms: (a) frost heave and subsequent thaw weakening in early spring
(Chamberlain 1987), and (b) upward pressurized water flow during lengthy rainy period
during mid-summer and fall.
For frost heave, during periods of freezing, water in large void space freezes into
ice crystals as the freezing front is moving downward into the road. As water is drawn to
the freezing front by capillary movement through the frost susceptible soils, the ice
crystals continue to grow (Casagrande 1947; Csathy and Townsend 1962), causing the
road surface to heave. During the spring, the ice lenses start to melt and the fine soil
particles are separated from the matrix, which causes depressions and soft areas at the
road surface (Taber 1930, 1978 and 1980).
Engineers from the Alaska Department of Transportation & Public Facilities
(AKDOT&PF) tried several conventional methods to mitigate the frost boil issue but
none have eliminated the issue. There are three necessary elements in the formation of
frost heave (Holtz and Kovacs 1981): (1) frost susceptible soil, (2) subfreezing
temperature, and (3) water. Therefore, removal or minimizing of the three conditions will
mitigate the frost heave and thaw weakening potential. One way is to remove the existing
road section and replace with better materials. However, this method is not feasible due
to the extremely high cost and long transportation distance to the remote site. It is also
not feasible to artificially alter the environmental condition, which is the source of water.
The most practicable way is to reduce the water content in the pavement structure.
AKDOT&PF has tried to install French drains to remove water out of the road section.
This conventional repair method does not work well because the French can only drain
free water or runoff water from the road surface and drainage ditch. The capillary water
315
in the pavement structure cannot be drained by conventional drainage methods. In order
to break the capillary flow path, a capillary barrier can be an alternative way to stop frost
action. A capillary barrier is a layer of coarse-grained soil or geotextile placed in a frost
susceptible soil. Taber (1929) indicated that one effective way to eliminate the frost
heaving issue is to place a layer of coarse sand above the water supply in frost-
susceptible soil specimens. He also noted that frost heaving requires substantially more
water than is naturally available in the soil pores. Casagrande (1938) and Beskow (1946)
described placing a layer of sand or gravel above the water table in road construction to
reduce frost heave of overlying fine-grained soil. Later, Rengmark (1963) and Taivenen
(1963) documented using a sand layer above the water table to help prevent frost heave in
overlying frost susceptible soil. However, the capillary barrier only stopped the capillary
water from moving upward. The excess water would be accumulated beneath the
capillary barrier, and with time would finally reduce the stiffness of the pavement
structure.
An effective way is needed to mitigate the frost heave and subsequent thaw
weakening issue. A new type of woven wicking fabric was recently developed and had
the potential to solve this issue. The basic physical, mechanical and hydraulic properties
of the wicking fabric are presented in Table 1. The wicking fabric is made of hydrophilic
and hygroscopic 4DGTM fiber, which has a high specific surface area (consequently high
wettability and high capillary action) and high permittivity. This type of wicking fabric
can laterally transport water under unsaturated conditions. Preliminary laboratory tests
indicated it had great promise as a cost-effective mean to solve the frost boil problem.
Zhang and Belmont (2009) compared four different types of geotextiles to evaluate their
316
effectiveness to drain water under unsaturated conditions. Test results indicated that the
soils installed with wicking fabric obtained the lowest water contents after the test, which
validated the advantages of wicking fabric to drain water out of soils comparing with
conventional geotextiles.
In order to further analyze the performance of the wicking fabric to mitigate frost
heave and subsequent thaw weakening issue, a test section installing two layers of
wicking fabrics was built at the Beaver Slide area of the Dalton Highway (Zhang et al.
2014). In total 22 pairs of sensors were used to monitor the temperature and moisture
content change in the 18.1 meter (60 foot) long road section. In addition, other useful data
such as air temperature and relative humidity was also recorded. Performance of the
wicking fabric was monitored under different climate conditions, such as rainfall events,
freezing processes and thawing processes. The first two years of monitoring indicated
good overall performance and field observation showed a remarkable road surface
difference between the test section and sections without the wicking fabric. No soft spots
were observed during early spring and soil at the road shoulder of the down slope side
was damp, while the other section without installing wicking fabric still had frequent
“frost boils” occurrences. This indicated that water flowed along the wicking fabric and
out of the road structure and the wicking fabric successfully eliminated the frost boil
problem to a depth of 1.07 meters (3.5 feet). Even though soil 1.37 meters (4.5 feet)
beneath the surface and lower showed the existence of excess water, it had limited effect
on roadway performance.
Although both laboratory and field test results proved that the wicking fabric was
a very promising drainage material to remove water from the pavement structure, the
317
long-term performance of the wicking fabric still needed to be further evaluated. In
addition, there were still some concerns regarding the extensive application of the
wicking fabric. Firstly, the clogging effect might influence the long-term performance of
the wicking fabric. Because the wicking fabric was directly in contact with the soil, there
was a potential that the drainage paths (or deep grooves) might become blocked by the
finer soil particles. Secondly, permanent deformation might also influence the wicking
fabric long-term performance. During the construction process, the soil above the
wicking fabric was compacted and introduced with relatively high loading pressure. The
permanent deformation might be further developed with time due to heavy truck traffic
on the road surface during its service life. Permanent deformation of the wicking fabric
could also reduce the amount of water held in the deep grooves and might reduce the
effectiveness of the fabric's wicking ability. Thirdly, aging and mechanical failure could
influence the long-term performance of the wicking fabric. Both the hydrophobic and
hydroscopic yarns of the wicking fabric would suffer from physical and chemical aging,
but the rate and severity of aging were unknown.
This paper focuses on the long-term performance of the wicking fabric to mitigate
the frost boil issue in Alaskan pavements in the past five years (2010-2015). It was found
that water entered into the data acquisition station in August 2014 and caused
malfunction of the datalogger. Consequently the pavement moisture and temperature
variations for only four years (2010-2014) were recorded and analyzed. Macroscopic
analyses of the wicking fabric's performance in various climatic conditions (such as
rainfall events, freezing and thawing processes) were first discussed. Field samples of the
fabric were collected at the end of the five-year period (September 2015) to evaluate the
318
wicking fabric performance at a microscopic level, using a scanning electron microscope
(SEM).
2. TEST SECTION CONSTRUCTION AND INSTRUMENTATION
The description of the test section construction process can be found in Zhang et
al. (2014). The following is a summary of the test section construction and sensor
installation processes. The test section was selected because AKDOT&PF identified it as
the section of road on the Dalton Highway with the most soft spots (frost boils) observed
in the spring of 2010. Figure 1 shows the profile of the test section. The road section was
originally built directly on the tundra on the hill side, using the degraded granite. Sieve
analyses indicated that the soil was classified as gravel with sand, according to USCS
classification, and contained about 6% or more fines (material passing the #200 sieve).
The original tundra was found at about 0.91 meter (3.0 feet) below the ground surface at
the west edge of the road, and about 1.36 meters (4.5 feet) below the centerline of the
road section. The buried vegetation was degraded into a dark yellow layer which was
about 0.05-0.1 meter (1-2 inches) thick. In situ crushed rocks and sand were encountered
below the degraded vegetation. Ground water was found 0.15 meter (6 inches) below the
tundra surface once the tundra was removed. Additionally, water was found during the
construction process in the existing drainage ditch along the west (up-slope) side of the
road.
In total, 22 pairs of sensors were installed in the pavement structure. Each pair of
sensors consisted of a Campbell Scientific 107–L temperature sensor and a CS616–L

319
moisture content reflectometer. An HMP45C air temperature/relative humidity sensor
was also installed at the site to monitor the air temperature and relative humidity. Four
layers of sensors were installed at depths of 0.45, 0.76, 1.06 and 1.97 meter(s) (1.5, 2.5,
3.5 and 6.5 feet) below the road surface. Two layers of wicking fabric were installed at
depths of 0.76 meter and 1.06 meters (2.5 feet and 3.5 feet) below the road surface. Since
water transportation in the wicking fabric was directional, care was taken to make sure
the direction of the wicking fabric was along the transverse direction of the road section,
so that water was transported horizontally to the road shoulder. On the east side of the
roadway, the two wicking fabric layers were left exposed to the air at 1.21 meters (4 feet)
off the shoulder. Sensor 22 was installed at the location closest to the drainage ditch, and
could be used as a representation of the saturated moisture content in the pavement
structure in summer time. All of the sensors wires were protected using aluminum
conduit to prevent damage from the traffic load. The aluminum conduits were grouped
together, buried in a small ditch in the transverse direction, and connected to a Campbell
Scientific CR1000 data logger. All of the data acquisition devices were organized into an
ENC14/16-NC-NM weather-resistant enclosure which was installed on the tundra about
6.1 meters (20 feet) from the west edge of the road.
3. RESULTS AND ANALYSIS
3.1. GENERAL CLIMATIC CONDITIONS
Figure 2 presents the hourly air temperature and relative humidity data, for the
test section from August 2010 through August 2014. Figure 2a indicates that in general,
320
the average value of summer air temperature increased from 2011 to 2013, followed by a
decrease in 2014. The average value of winter air temperature increased each year during
the same period. Within year, air temperature dropped below zero in late September and
rose above zero in mid to late April. The lowest temperatures recorded at the site were on
February 23, 2010 (-36.8 °C), January 29, 2011 (-39.8 °C), December 17, 2012 (-35.2
°C) and January 13, 2014 (-34.8 °C). The highest temperatures recorded were on May 27,
2011 (24.4 °C), June 24, 2012 (22.9 °C), June 19, 2013 (26.2 °C) and July 6, 2014 (24.9
°C). The daily temperature variations in the summer times were smaller than those in
winter times.
Figure 2b presents the monitored hourly relative humidity data for the four year
period from August 2010 through August 2014. In winter months, the relative humidity
at the site was between 70% and 90%, due to relatively low air temperatures. However,
the relative humidity varied from 20% to 90% in summer months. The relative humidity
during daytime hours was lower than that at night. During significant rainfall events, the
relative humidity increased rapidly over 95% in the test section, and then decreased
below 85% very soon after the rain stopped.
3.2. SOIL TEMPERATURE CHANGES
Figure 3 presents the soil temperatures of the monitored 22 sensors from August
2010 through August 2014. Figure 3a shows the temperature variations for sensors 10,
11, 12 and 13, which were located at the center of the embankment. The sensors were
buried at depths of 0.45, 0.76, 1.06, and 1.97 meter(s) (1.5, 2.5, 3.5, and 6.5 feet),
respectively. In general, the trend of temperature changes in the soil followed the air
321
temperature trend. The temperature change in the soil decreased in magnitude as depth
increased because of the soil insulating effect. For instance, sensor 10 was the closest
sensor to the road surface, and its temperature variations followed the air temperature
changes very closely during the summer. During the winter months, the temperatures at
sensor 10 were higher than the air temperatures. In comparison, the temperature
variations at sensor 13, which was installed 1.97 meters (6.5 feet) below the road surface,
ranged from -12 °C to 3 °C for the entire year. The variations at sensor 13 were much
smaller than air temperature changes, which were as much as 30 °C in winter and 16 °C
in summer. It was also observed that the soil temperature 1.97 meters (6.5 feet) below the
road surface only experienced temperatures above 0 °C for less than 3 months each year
(i.e. July 20, 2012 to October 30, 2012). This indicates that the soil at this depth could be
considered as a permeable layer for approximately 3 months, and as an impermeable
layer for the rest of the year.
Figure 3b-3e show the temperature changes at the sensor locations 0.45, 0.76,
1.06 and 1.97 meter(s) (1.5, 2.5, 3.5, and 6.5 feet) below the road surface, respectively,
for the past four years.
Figure 3b-3c present temperature data for the sensors located at 0.45 and 0.76
meter (1.5 and 2.5 feet). In general, the amplitudes of temperature changes at 0.76 meter
(2.5 feet) were smaller compared with those at 0.45 meter (1.5 feet). However, the soil
temperature changes in both layers followed the air temperature trends during the
summer, and were warmer than the air temperature in winter. Thus, soils 0.76 meter (2.5
feet) below the road surface could be considered a permeable layer during the summer,
which were able to drain the melting snow from the road surface. Additionally, the soil
322
temperatures observed in the center of the road were lower than temperatures at the edges
during winter months; and soil temperatures at the west side of the road were lower than
those at the east side. There are two reasons to explain this phenomenon: (1) snow was
routinely removed and piled on the shoulders, insulating the shoulder to make it warmer
than the center of the road; and (2) the roadway on the east side received more solar
energy than the west side, resulting higher temperature at the east side of the roadway.
Figure 3d-3e shows the temperature change for sensors at 1.06 and 1.97 meters
(3.5 and 6.5 feet) below road surface. The insulation effect became more obvious as
depth below the surface increased. As can be seen in Figure 3d, soil temperatures at 1.06
meters (3.5 feet) experienced approximately a 1 month time lag compared with the air
temperature change (time difference for the starting dates of soil and air temperatures
above 0 °C). As for soils at a depth of 1.97 meters (6.5 feet), this time lag could be as
large as 3 months, as shown in Figure 3e. This phenomenon indicated that during the
early spring (late April or early May), the soils at 1.06 meters (3.5 feet) and below were
still frozen and could not be considered as a drainage layer until 1-3 months later. In
other words, since the second layer of wicking fabric was installed at 1.06 meters (3.5
feet) below the road surface, it would not be able to drain the water out of the
embankment until early June. The first snowfall at the site was expected in early October
each year, theoretically allowing the second layer of wicking fabric to remain functional
until early November, when the soils at this depth became thoroughly frozen.
323
3.3. SOIL MOISTURE CHANGES
Figure 4 shows the soil moisture changes for the installed 22 sensors during the
four-year period from August 2010 through August 2014. As can be seen in Figure 1,
sensor 22 was buried at about 1.2 meters (3.9 feet) below the road surface, and was 1.5
meters (4.9 feet) away from the up-slope drainage ditch. Its elevation was 0.1 meter (0.3
foot) below the drainage ditch. The drainage ditch had water flow all year around except
during winter months when everything was frozen. Since the moisture content at sensor
22 was controlled by the drainage ditch and maintained saturated or nearly saturated, it
was reasonable to use sensor 22 as a reference for comparison purposes in all figures. As
shown in Figure 4a, moisture contents at sensor 22 were relatively constant in the
summers and winters between 2010 and 2014, and independent of the daily weather
conditions. The recorded average volumetric moisture content continuously decreased
from 0.38 in 2010 to about 0.32 in 2013, and slightly increased to 0.35 in 2014. However,
the unfrozen water in winter months barely changed and was maintained within the range
of 0.07-0.12. It is also worth noting that it took nearly 2 months to thoroughly freeze the
soil at this depth in winter months, but only took about 2 weeks to thoroughly thaw the
frozen soil in subsequent early spring. Taking the 2014 thawing season as an example,
the unfrozen moisture content for sensor 22 was approximately 0.09 on April 10, 2014
when the average daily air temperature was -15.5 °C. However, this value increased to
0.14 on April 24, 2014, when the average daily air temperature changed to -1.1 °C. This
phenomenon indicated that solar radiation is capable to increase the unfrozen water
content in frozen soil even if the air temperature is still below zero.
324
Figure 4a shows the soil moisture changes at 0.46 meter (1.5 feet) below road
surface. It is obvious that the moisture content at this depth was far below the moisture
content at the reference location (sensor 22). On one hand, the infiltration water could be
easily runoff due to the existence of both longitudinal and transverse slopes. On the other
hand, the evaporation process at the road surface was much faster owing to an easy
access to the open air. Because the elevation of the ground adjacent to the roadway at the
west side of the road was higher than the east side, it was reasonable that the moisture
contents of the soils on west side were higher than soils on the east side. Figure 4a was
consistent with the observations made by AKDOT&PF maintenance and operation
personnel who reported no soft spots were observed at the test section and were very
satisfied with the performance of the wicking fabric in the past five years.
Figure 4b shows the soil moisture changes for sensors at 0.76 meter (2.5 feet)
below road surface, where the first layer of wicking fabric was installed. In general, the
soil moisture contents were not higher than the referencing sensor (sensor 22), except for
some long and intensive rainfall events. Any sudden, large variation in soil volumetric
moisture content change indicated a rainfall event. Compared with Figure 4a, soils at this
depth, 0.76 meter (2.5 feet) were more affected by the rainfall events. However, the soil
moisture contents dropped back quickly after the rainfall event stopped, which indicated
that the drainage condition at this depth was favorable. Since the soils at 0.45 meter (1.5
feet) remained unsaturated in the four years, the excessive water at the depth of 0.76
meter (2.5 feet) was from the horizontal direction.
Figure 4c shows soil moisture changes at 1.06 meters (3.5 feet) below road
surface, where the second layer of wicking fabric was installed. Sensor 1 was buried
325
fairly shallow on the east side of the road shoulder, and the moisture content was much
lower than that for the reference sensor 22. Similarly, the moisture contents for sensors
on the east side were lower and the moisture contents for sensors on the west side were
all higher compared with the referencing sensor. The amplitudes of moisture content
changes after intensive rainfall events were also higher at 1.06 meters (3.5 feet) than the
previous two depths discussed. This phenomenon could be the results of (1) accumulated
water at the drainage ditch, and (2) the one-month time lag to melt the frozen soil at this
depth. Since the snow started to melt in early May, and the elevation of the ground
adjacent to the road on the west side was higher than on the east side. Soils at 1.06 meters
(3.5 feet) and below would not start to melt until late May or early June and the water
source was therefore not from the drainage ditch. A large amount of water was probably
trapped in the drainage ditch, which provided a large quantity of water to the roadway
structure. The freezing temperatures and excess moisture resulted in a hard, frozen core at
the center of the road, impeding the drainage path. The higher moisture content in west
side of the road was caused by the trapped water in the pavement structure.
Figure 4d shows the moisture content changes for sensors at 1.97 meters (6.5 feet)
below road surface. The moisture content distribution followed the trend presented in
Figure 4c. All of the sensors on the west side had moisture contents higher than at the
east side. It is noteworthy that sensor 13 (located at the centerline of the embankment) did
not fully melt until mid-August, which was nearly 3 months after snow melting began. As
discussed previously, the frozen soil impeded the natural water flow and caused excess
water to become trapped on west side of the road. Moreover, because the melting process
took such a long time, soils on the west side of the road embankment could hold more
326
water during the summer time, allowing the unfrozen moisture contents to remain
approximately 4% higher than the moisture contents at the reference sensor in winter
months, and intensifying the frost heaving process. In contrast, the moisture contents on
the east side were much lower than at the reference sensor location, except for during
some intensive rainfall events.
3.4. PERFORMANCE OF WICKING FABRIC AT DIFFERENT CLIMATIC

CONDITIONS
The monitored hourly temperature and moisture data at the 22 sensor locations, as
shown in Figure 1, were used as controlling points to generate temperature and moisture
contour maps with time. The meshgrid in Matlab was firstly used to generate the
interpolation locations among the sensors and then the Delaunay triangulation was used
to generate the mesh upon which linear interpolations were used to compute the moisture
and temperature values at the desired locations. After that, contour maps for temperature
and moisture were generated for a specific time. The contour maps were then displayed
with time to generate videos to show the energy and moisture movements in the
embankment in the past four years. The long-term performance of the wicking fabric
could be visualized via different climatic conditions: during rainfall events and during
freezing and thawing processes.
3.4.1. During Rainfall Events. Table 2 summarizes all of the major rainfall
events for the four-year period monitored. Since the relative humidity in the air during the
summer time was about 50% without rainfall, the water evaporation rate was faster than
the water infiltration rate during light rainfall events. Moreover, the water could easily
runoff via the longitudinal and transverse slopes if the rainfall events were not intensive.
327
Therefore, it was reasonable to assume that light rainfall events were not able to raise the
relative humidity above 95%. Table 2 only summarizes the duration of rainfall events in
which the recorded relative air humidity was greater than 95%. The total amount of
rainfall hours were thoroughly recorded for three years: 530 hours in 2011, 617 hours in
2012, and 376 hours in 2013, respectively.
In a summary, the effect of rainfall intensity was limited due to the good drainage
condition of the base course (gravel with sand). Therefore, only the effect of rainfall
duration is discussed in this section. Figure 5 shows the comparison of two rainfall events
to demonstrate this effect: one is a short duration event lasting several hours, and the
other is a long duration rainfall lasting for several days. As shown in Figure 5a, the first
recorded rain fall occurred at 10 pm on August 28, 2013 and lasted for 5 hours. By
looking up the recorded data, it was determined that no other significant rainfall events
occurred within a week prior to this event. The 3 moisture contour figures show the soil
moisture distribution before the rainfall, 1 hour after the rainfall and 1day after the
rainfall. It was apparent that the soil moisture distribution did not change significantly as
a result of this event. This phenomenon indicated that a 5-hour rainfall was not long
enough to change the water moisture distribution within the pavement structure.
In comparison, another rainfall occurred at 11 pm on July 8, 2013 and continued
on the following day. The total rainfall duration was about 27 hours from July 8 to July 9,
ending around 2 pm on July 9. Figure5b shows the soil moisture distribution at the
beginning of the rainfall event, 7 hours after the rainfall and about 1 day after the rainfall.
The soils in the east side of the roadway were significantly drier than the soils in the west
side of the roadway prior to the rainfall. 7 hours after the rainfall event, more water had
328
accumulated in both the east and the west side of the road structure. The saturation zone
in the west side was larger because water flowed into the pavement structure via a
drainage ditch up-slope and adjacent to the west side of the embankment. Meanwhile, on
east side of the road, the saturated zone was observed at the location of the wicking fabric
layers. This phenomenon indicated that the wicking fabric was able to suck the water
from the surrounding soils and laterally transport it to the shoulder. The third contour
figure shows that 1 day after the rainfall event, the saturation zone was smaller than
before in the west side of the roadway. The soils near the wicking fabric were
comparatively drier than the rest of the soils on the east side of the embankment, where
no apparent saturation zone was observed.
3.4.2. During Freezing Process. For purposes of this discussion, the moisture
contours when the air temperature dropped below zero during the recorded four year
period were summarized and compared, as shown in Figure 6. It was critical to determine
the moisture content distribution before the freezing front moved downward, because the
severity of the thaw weakening in the following spring was directly related to the amount
of water stored in the pavement structure before the freezing process started in the
previous year. In other words, the soft spots in the following spring, if frost boiling was
observed, would be expected where the saturation zones were observed before the freezing
process started in the previous year. It should be noted that the areas of the saturation
zones decreased with time. For instance, there were two saturation zones that were
connected together in 2010 and 2011 (Figure 6a-6b); however, the saturation zones
became separated into two smaller zones on October 11, 2012 (Figure 6c). The less the
amount of water stored in the pavement structure, the less severity of the frost heave
329
would be expected during the freezing process (less negative pore water pressure would be
generated due to the water phase changing from liquid state to solid state). Moreover,
since the soil moisture contents were lower than the previous years, it took less energy to
move the freezing front downward. In comparison, the freezing front had already
penetrated to 0.9 meter (2.95 feet) on October 11, 2012, which was about 0.3 meter (0.98
foot) deeper than the previous year. Furthermore, the saturation zone continued to
decrease at the bottom of the roadway in year 2013, as shown in Figure 6d. This
phenomenon could be apparent because: (1) precipitation variation may cause such
variations in the soil moisture content distribution, and (2) the wicking fabric worked
effectively to reduce the moisture content in the soil. Because the rainfall event summary
presented in Table 2 indicated that there were no significant rainfall events that occurred
right before the selected days, precipitation was not the reason that caused the decrease in
soil moisture contents. Therefore, the wicking fabric did reduce the water content in the
east side of the roadway embankment, and reduced the size of saturation zone in the west
side of the road. However, the performance of the wicking fabric to drain the water out
needed to be further evaluated during the spring thawing process to validate its efficiency.
3.4.3. During Thawing Process. Figure 7 shows the moisture contours on May
25 of each year. Firstly, the unfrozen water contents for sensor 22, which located nearest
to the drainage ditch, remained below 0.1 within the monitored four years. This
phenomenon indicated that the drainage ditch was still frozen at this time, and that there
was no water supply from melting snow. Therefore, the water source that caused thaw
weakening issue was major resulted from the melting of frozen soil within the pavement
structure.
330
Secondly, it is important to point out that the mean monthly temperature for May,
2013 was lower than in previous years, so the thawing front only penetrated to 1.22
meters (4 feet) on the east side and 0.76 meter (2.5 feet) on the west side. For other years
monitored, the distance between the thawing front and the 0 °C isothermal curve
increased each year. This phenomenon can be explained by referring to the moisture
contour during freezing process. Since the saturation zones in the pavement structure was
decreasing during the monitored four years, the total amount of water stored in the
pavement structure (including capillary water extracted from shallow groundwater) was
also decreasing. The distance between thawing front and the 0 °C isothermal curve was
expected increasing with time.
Thirdly, the highest moisture content areas were all located on the west side of the
embankment, but no saturation zone was observed during the thawing process in the
monitored four years. This phenomenon proved that the wicking fabric successfully
eliminated the frost boiling issues. Moreover, the thawing front on east side of the
embankment was deeper than that on west side and reached to the elevations where the
two layers of wicking fabrics were buried. This phenomenon indicated that the wicking
fabric on east side of the pavement structure was partially functional and started to
laterally drain the water out of the pavement structure in late May, 2014, as shown in
Figure 7d.
3.5. CLOGGING EFFECT
Figure 8a presents the woven structure of an intact sample at Beaver Slide with
×55 magnification. Large amounts of soil particles were detained on the surface of the
331
wicking fabric. Because the soil contained approximately 6% of fines, the clogging effect
was obvious at this level, and the deep grooved drainage paths were blocked by the fine
materials. Figure 8b shows a closer view of the wicking fabric at the surface with ×350
magnification. It further illustrated the fact that the deep grooved drainage paths were
completely filled with fine soil particles. In comparison, Figure 8c shows the wicking
fabric fibers just beneath the surface layer. The deep grooves beneath the surface were
much cleaner than those above, and there were very few particles detained in the drainage
paths. In other words, the fibers of the wicking fabric at surface served as a protective
layer, preventing the fine soil particles from penetrating deeper into the fabric structure.
Figure 8d shows the comparison of the wicking fibers on the surface and the fibers just
beneath the surface. It could be seen from the figures that even though the wicking fabric
fibers on surface were filled with fine soil particles, the wicking fibers beneath the
surface still were able to effectively drain water out of the pavement structure. It was
worth noting that it was not fair to evaluate if the wicking fabric was clogged based upon
the fibers on the surface since “surface” was a theoretic term and was difficult to define
during the SEM analyses. If too many soils were left on the surface of the wicking fabric,
there was no doubt that the wicking fabric would be covered by the soils. On the other
hand, if we took all the surface soil away, the evaluation for the clogging effect was not
objective. It seemed more reasonable to evaluate the clogging effect based upon the
wicking fibers below the surface.

332
3.6. PERMANENT DEFORMATION AND MECHANICAL FAILURE
Figure 9 presents the SEM images of samples that suffered permanent
deformation and mechanical failure. Figure 9a presents image of new wicking fabric,
which was never used before. It was apparent that the wicking fabric fibers under the
woven polypropylene yarns had already experienced some permanent deformation, and
that the deformation was in the vertical direction. This deformation might have been
caused by the pressure applied in manufacturing process, or it might have occurred
during the transportation process. Figure 9b shows the image of the wicking fabric that
was collected from the field. The permanent deformation observed in the new materials
had further increased. Due to additional vertical pressure, the wicking fabric fibers were
nearly flattened, and the deep grooves were not able to hold water under unsaturated
conditions. Furthermore, Figure 9c presents the front view of the wicking fabric. Deep
grooves were seen not only in the vertical direction, but also tended to close in the
horizontal direction.
Another mechanical failure known as “puncturation” is illustrated in Figure 9d.
“Puncturation” refers to the puncturing of the soil fibers by the large soil particles that are
detained on the wicking fabric surface. The large soil particles, especially those with
sharp edges, acted as a cutting edge that severred the deep grooves of the wicking fabric.
This likely occurred due to the high overburden soil pressures and the dynamic traffic
loads applied to the road surface. The drainage paths were broken and became unable to
continue to laterally transport water; however, this phenomenon was only observed in 5
out of 30 samples. According to the observed macroscopic results at the Beaver Slide, it
seemed that neither permanent deformation nor puncturation were major concerns,
333
possibly for two reasons (1) there were relatively less percentage of the wicking fabric
having permanent deformation or puncturation, and (2), surrounding fine soil particles
might have “bridging effect” for water transport at locations where permanent
deformation or puncturation occurred.
3.7. AGING
Because the wicking fabric is buried under the soil, another concern involves the
wicking fabric’s physical and mechanical aging issue, as shown in Figure 10. Figure 10a
shows the aging severity of the wicking fabric under the woven polypropylene yarns.
Because the fibers on the surface were directly in contact with the soil particles, the aging
phenomena were usually observed at this location. Figure 10b shows the fibers at the
surface without the woven polypropylene yarns. As believed, the aging phenomenon was
likely due to direct contact with the soil particles. The aging effect at the bottom of the
deep grooves was more severe than in the other areas of the wicking fabric. In
comparison, Figure 10c-d show the wicking fabric beneath the surface. No obvious aging
effect was observed below the surface layer, and the deep grooves were much cleaner
than those of the fibers on surface.
Table 3 summarizes the SEM analyses results. In general, all wicking fabric fibers
on the surface suffered from the clogging effect. Clogging and permanent deformation
were observed in every scanned sample. Therefore, these two aspects become the major
potential issues that needs to be taken into consideration in evaluating the wicking fabric
long-term performance. However, only 6.67% of the wicking fabric fibers beneath the
surface suffered from the clogging effect. This indicates that even though the surface was
334
contaminated and the drainage paths were blocked, the wicking fibers beneath the surface
were well protected and worked effectively as a drainage material to transport water
laterally under unsaturated conditions. Additionally, the permanent deformation was
observed in every sample under the polypropylene woven area. The permanent
deformations resulted from one, or both of the following two processes: (1) high pressure
during the manufacturing process, and (2) high vertical overburden soil pressure and
dynamic traffic load during its service life. The permanent deformation would likely
affect the wicking fabric’s long-term performance, since the drainage paths were either
cutoff or narrowed down, and the deformation would continue to develop with time. The
aging effect and mechanical failure were not considered to be major concerns that would
influence the long-term performance of the wicking fabric.
4. DISCUSSIONS
4.1. FACTORS INFLUENCING THE OCCURRENCE OF FROST BOILS/SOFT

SPOTS
As previously mentioned (Zhang et al. 2014), the frost boils often occurred during
the end of April through May each year. Based on the comparisons of the pavement
performance during the monitored four years, the water sources that were available to
form the frost boils came from the thawing of in situ ice lenses that developed in the
pavement structure due to frost heave in the previous winter. The moisture content of the
pavement structure at the beginning of the freezing process was one of the major factors
that determined the intensity of the frost heave and the subsequent thawing in the next
year. The larger the fully saturated zones were in the pavement structure, the more
335
suction or negative pore water pressure (due to water expansion during the freezing
process) it would generate. Since the freezing process penetrated the pavement structure
from the top to the bottom, the only water source must be from the shallow water table
beneath the pavement structure. Higher suction values would further increase the
moisture content in the pavement structure and cause a zone of over saturation. The
melting water from the over-saturated zone would provide sufficient water during the
following spring to create soft spots at the surface, because the water was forced to the
road surface when the soil beneath was still frozen.
Another factor influencing the severity of the frost boils was when the thawing
front penetrated down to the bottom of the pavement structure. The thawing front
penetrated to the bottom of the pavement structure in late July or early August, which
was almost three months after the thawing season began. Because the frozen soil in the
west side of the roadway held a large amount of frozen water, it took a larger amount of
solar energy to melt the frozen soil. The only drainage path for the melting snow and
runoff water was to flow through the pavement structure. This would further reduce the
soil stiffness and intensify the frost boil issue. Furthermore, the center of the pavement
structure formed a hard, frozen core during the melting season. The frozen core altered
the water flow direction and trapped a large amount of water in the west side of the
pavement structure, which intensified the frost heave action during winter time.
It was also worth noting that a large amount of rainfall would cause another issue
called pressurized water overflow, which might also have generated soft spots on the road
surface in summer time. Rainfall duration served as a more deteriorating factor to the
pavement performance than rainfall intensity. The soft areas would heal up if there were
336
periods of no rain. The moisture contents in the pavement structure beneath 0.47 meter
(1.5 feet) experienced short periods of time of overly saturated. By carefully examining
the rainfall events summary in Table 2, it was seen that there were several days of rainfall
before the sudden increases in moisture content. Since the road prism was built on a side
hill, the water naturally flows from west to east. Also, the 11° downhill slope made the
hydraulic gradient the highest at the test section. These factors were evidence that the
sudden increases in moisture content were due to pressurized water overflow to the road
surface. Although the two issues presented the same superficial phenomena, the
mechanisms causing the phenomena were different.
4.2. WICKING FABRIC LONG-TERM PERFORMANCE
Data collected during four years of monitoring indicates that the wicking fabric
has worked effectively to prevent the frost heave and thaw weakening issues previously
observed at Beaver Slide. During the rainfall events, the water could be drained out 1-2
days after the rainfall stopped. Moreover, the moisture contents were gradually
decreasing within the pavement structure before the freezing process (Figure 6). This
indicated that the water supply for frost heave was decreasing, and that the moisture
contents during the next thawing process were also decreasing. As shown in Figure 7,
there were no saturation zones present in the pavement structure during the thawing
process. Additionally, the wicking fabric successfully drained water laterally out of the
pavement structure within several hours, even after heavy rainfall events, as shown in
Figure 5b. The moisture contents near the wicking fabric were much higher than moisture
contents in other areas, demonstrating that the wicking fabric worked effectively to
337
transport the water out of the embankment. The moisture contents on east (dry) side of
the pavement structure reached equilibrium generally within 1 day of significant rainfall
events.
In addition to macroscopic field observations of the improved road performance,
the microscopic SEM analyses showed the interaction between the wicking fabric and the
soil. One of the major concerns was whether the fines in the soil would be retained in the
deep grooves and potentially restricted the drainage path. Figure 8 clearly shows that
even though the wicking fabric fibers at surface were covered by fine-grained soils, the
fibers beneath them were relatively clean and no-clogging effect existed. The permanent
deformation due to high vertical pressure might narrow down or even cut off the deep
grooves, blocking their ability to laterally transport water in unsaturated conditions. This
deformation might be induced by the high pressure during manufacturing process or
traffic load. More research is needed in this direction. Aging and mechanical failure was
observed in the SEM analyses. However, the occurrence percentage was relatively low
and was not a major concern for the long-term performance of the wicking fabric.
5. CONCLUSIONS
The Beaver Slide project has been monitored for more than five years, and the
results indicate that the wicking fabric has successfully eliminated the “frost boils” at the
site. The following conclusions were summarized based on the previous analyses:
1. The soft spots observed during the early spring were caused by the ice formation
and thaw weakening of the soils; however, the soft spots observed after heavy
338
rainfall resulted from pressurized water flow. Although the phenomena were
similar, the mechanisms were entirely different.
2. The severity of the thaw weakening in spring was relative to the moisture content
present in the pavement structure before freezing began at the start of the previous
winter. The monitoring data shows that the moisture contents were decreasing, and
that the saturated zones were smaller each year after the fabric was installed.
Moreover, the moisture contents in the pavement structure were not observed to
exceed the saturation moisture contents. This indicates that the wicking fabric
worked successfully to eliminate the ice formation and subsequent thaw weakening
issue during the past five years. The wicking fabric exhibited promising long-term
performance results. However, additional monitoring and data analysis should be
performed to establish longer-term performance.
3. Clogging was only observed in the surface fibers of the wicking fabric. The wicking
fabric fibers beneath the surface layer were relatively clean. The clogging effect
was not considered to be a major issue for the application of the wicking fabric.
4. The permanent deformation might be an issue that would affect the long-term
performance of the wicking fabric. The deformed deep grooves would reduce the
amount of water that the wicking fabric could laterally transport. The permanent
deformation might develop further over with time due. The wicking fabric
implementation depth and its influence on long-term performance should be studied
further.
5. Mechanical failure and aging of surface fibers were observed in only a limited
number of samples. Mechanical failure might be caused by compaction during the

339
construction process and high vertical pressure during its service life. Aging was
observed in the surface fibers of the wicking fabric, where fibers were directly in
contact with the surrounding soils.
6. In conclusion, visual observations made by AKDOT&PF Maintenance and
Operation personnel and the authors through field trips as well as the measurements
of moisture and temperature indicated that after five years, the wicking fabric is
still working effectively to remove the water from the embankment, which has
eliminated the frost boil problem in the test section.
ACKNOWLEDGEMENTS
The research presented in this paper was sponsored by TENCATE
GEOSYNTHETICS (North America). The authors gratefully acknowledge them for their
financial support. Jeff Currey from AKDOT&PF helped facilitate the field construction,
installation design and sampling of the wicking fabric used for the SEM analyses. His
help during the process is greatly appreciated.

340
Table 1. Geotextile specification
Mechanical Properties Test Method Unit Average Roll Value

Tensile Modulus @ 2% Strain
ASTM D4595 kN/m 657
(CD)
Permittivity ASTM D4491 Sec-1 0.24
Flow Rate ASTM D4491 l/min/m2 611
Pore Size (O50) ASTM D6767 microns 85
Pore Size (O95) ASTM D6767 microns 195
Apparent Opening Size (AOS) ASTM D4751 mm 0.43
Tested Value
ASTM C1559 inches
(24 minutes) Vertical Direction
ASTM C1559 inches
(983 minutes) Zero Gradient Horizontal Direction
341
Table 2. Rainfall events summary

Year 2010 Year 2011 Year 2012 Year 2013 Year 2014
Duration Duration Duration Duration Duration
Event Date Time Event Date Time Event Date Time Event Date Time Event Date Time
(hrs) (hrs) (hrs) (hrs) (hrs)
1 8/18/2010 10 3 1 5/23/2011 7 2 1 5/13/2012 4 4 1 6/4/2013 4 3 1 5/17/2014 13 4
2 8/18/2010 23 12 2 6/8/2011 21 8 2 5/26/2012 6 27 2 6/4/2013 12 1 2 5/22/2014 6 3
3 8/23/2010 3 6 3 6/9/2011 15 1 3 5/29/2012 17 1 3 6/4/2013 18 5 3 5/31/2014 19 38
4 8/24/2010 3 7 4 6/14/2011 7 4 4 5/30/2012 1 2 4 6/9/2013 22 6 4 6/5/2014 5 5
5 8/24/2010 21 12 5 6/15/2011 2 2 5 6/2/2012 6 8 5 7/1/2013 2 4 5 6/6/2014 13 17
6 9/4/2010 15 23 6 6/21/2011 2 11 6 6/10/2012 1 12 6 7/8/2013 3 4 6 6/9/2014 2 14
7 9/6/2010 22 2 7 6/22/2011 7 3 7 6/27/2012 3 37 7 7/8/2013 23 12 7 6/11/2014 0 14
8 9/7/2010 11 23 8 6/22/2011 12 3 8 6/29/2012 6 28 8 7/9/2013 23 15 8 6/11/2014 18 16
9 9/9/2010 1 10 9 6/22/2011 21 11 9 7/4/2012 2 5 9 7/16/2013 7 5 9 6/18/2014 1 7
10 10/2/2010 15 1 10 6/24/2011 0 6 10 7/4/2012 19 11 10 7/16/2013 20 12 10 6/19/2014 0 3
Total Rainfall Hours 99 11 6/26/2011 22 9 11 7/17/2012 4 2 11 7/18/2013 4 1 11 6/22/2014 4 9
12 6/27/2011 22 2 12 7/23/2012 4 8 12 7/19/2013 15 20 12 6/25/2014 18 54
13 6/28/2011 7 3 13 7/24/2012 12 16 13 7/20/2013 17 1 13 7/8/2014 1 13
14 6/30/2011 3 6 14 7/30/2012 23 8 14 7/24/2013 20 1 14 7/8/2014 22 13
15 6/30/2011 20 12 15 8/2/2012 4 13 15 7/30/2013 3 3 15 7/11/2014 12 15
16 7/11/2011 20 39 16 8/3/2012 9 6 16 7/31/2013 6 2 16 7/15/2014 18 44
17 7/16/2011 5 7 17 8/4/2012 0 13 17 8/5/2013 4 2 17 7/18/2014 23 10
18 7/18/2011 5 7 18 8/5/2012 0 8 18 8/10/2013 18 9 18 7/19/2014 20 17
19 7/24/2011 17 31 19 8/6/2012 23 16 19 8/12/2013 7 4 19 7/22/2014 5 31
20 7/29/2011 19 43 20 8/7/2012 22 5 20 8/14/2013 11 3 20 7/23/2014 18 16
21 8/1/2011 19 41 21 8/25/2012 21 39 21 8/14/2013 21 13 21 7/25/2014 2 9
22 8/5/2011 0 4 22 8/28/2012 0 3 22 8/16/2013 4 6 22 8/1/2014 0 13
23 8/5/2011 8 3 23 8/29/2012 0 63 23 8/18/2013 0 11 23 8/2/2014 0 9
24 8/6/2011 0 2 24 9/1/2012 7 30 24 8/19/2013 19 15 24 8/7/2014 9 2
25 8/7/2011 2 10 25 9/3/2012 17 15 25 8/21/2013 6 4 25 8/7/2014 15 1
26 8/8/2011 1 8 26 9/4/2012 23 11 26 8/24/2013 3 1 26 8/14/2014 6 3
27 8/10/2011 1 12 27 9/5/2012 19 41 27 8/28/2013 22 5 27 8/15/2014 13 43
28 8/12/2011 3 4 28 9/8/2012 0 14 28 9/2/2013 18 16 Total Rainfall Hours 423
29 8/13/2011 0 13 29 9/15/2012 17 12 29 9/3/2013 18 14
30 8/13/2011 20 15 30 9/18/2012 13 27 30 9/5/2013 17 15
31 8/20/2011 17 23 31 9/23/2012 16 10 31 9/7/2013 1 13
32 8/23/2011 21 16 32 9/28/2012 20 60 32 9/7/2013 20 4
33 9/1/2011 3 8 33 10/2/2012 5 32 33 9/9/2013 0 13
34 9/3/2011 5 8 34 10/5/2012 7 30 34 9/11/2013 0 7
35 9/3/2011 17 19 Total Rainfall Hours 617 35 9/11/2013 22 10
36 9/4/2011 23 1 36 9/13/2013 14 6
37 9/6/2011 19 16 37 10/13/2013 20 9
38 9/7/2011 20 19 38 10/15/2013 1 101
39 9/8/2011 21 15 Total Rainfall Hours 376
40 9/10/2011 9 29
41 9/12/2011 0 7
42 9/15/2011 20 10
43 9/16/2011 19 14
44 9/18/2011 1 23
Total Rainfall Hours 530
Table 3. SEM analyses summary

Clogging Mechanical
Sample Total Observation
Surface Beneath Failure
Count 30 2 5
Percentage
100 6.67 16.67
(%)
Beaver Permanent
30 Observation Aging
Slide Deformation
Count 30 7
Percentage
100 23.33
(%)
342
Figure 1. Profile of the test section (Zhang et al., 2014)
(a)
Figure 2. Hourly climatic data at beaver slide test section: (a) hourly air temperature data,
and (b) hourly relative humidity data
343
(b)
Figure 2. Hourly climatic data at beaver slide test section: (a) hourly air temperature data,
and (b) hourly relative humidity data (cont.)
(a)
Figure 3. Soil temperature changes: (a) soil temperature versus depth, (b) 0.45 m below
road surface, (c) 0.76 m below road surface, (d) 1.06 m below road surface, and (e) 1.97
m below road surface
344
(b)
(c)
(d)
m below road surface (cont.)
345
(e)
m below road surface (cont.)
(a)
Figure 4. Soil moisture changes: (a) 0.45 meters below road surface, (b) 0.76 meters
below road surface, (c) 1.06 meters below road surface, and (d) 1.97 meters below road
surface
346
(b)
(c)
surface (cont.)
347
(d)
surface (cont.)
348
(a)
Figure 5. Soil moisture contours during rainfall events: (a) short-time rainfall events, and
(b) long-time rainfall event
349
(b)
Figure 5. Soil moisture contours during rainfall events: (a) short-time rainfall events, and
(b) long-time rainfall event (cont.)
350
(a)
(b)
(c)
Figure 6. Moisture contours before freezing: (a) moisture contour on September 20, 2010,
(b) moisture contour on September 25, 2011, (c) moisture contour on October 11, 2012,
and (d) moisture contour on October 5, 2013
351
(d)
Figure 6. Moisture contours before freezing: (a) moisture contour on September 20, 2010,
(b) moisture contour on September 25, 2011, (c) moisture contour on October 11, 2012,
and (d) moisture contour on October 5, 2013 (cont.)
(a)
Figure 7. Moisture contours during thawing process: (a) moisture contour on May 25,
2011, (b) moisture contour on May 25, 2012, (c) moisture contour on May 25, 2013, and
(d) moisture contour on May 25, 2014
352
(b)
(c)
(d)
Figure 7. Moisture contours during thawing process: (a) moisture contour on May 25,
2011, (b) moisture contour on May 25, 2012, (c) moisture contour on May 25, 2013, and
(d) moisture contour on May 25, 2014 (cont.)
353
(a) (b)
Beneath Surface
Surface
(c) (d)
Figure 8. SEM images of clogging effect: (a) intact sample (surface), (b) fabrics on
surface, (c) fabrics beneath surface, and (d) wicking fabrics comparison
354
(a) (b)
(c) (d)
Figure 9. SEM images of mechanical failure: (a) new wicking fabric (surface), (b)
permanent deformation (surface), (c) permanent deformation (surface), and (d)
puncturation failure (surface)
355
(a) (b)
(c) (d)
Figure 10. Aging effect: (a) Beaver Slide sample 3, (b) Beaver Slide sample 11 (surface),
(c) Beaver Slide sample 21 (beneath surface), and (d) Beaver Slide sample 26 (beneath
surface)
356
REFERENCES
Beskow, G. (1946). “Supplement: some results of Scandinavian soil frost research 1935–
1946.” CRREL Special Report 91-23, U.S. Army Corps of Engineers, Cold
Regions Research and Engineering Laboratory (CRREL), Hanover, 161–169.
Casagrande, A. (1947). “Classification and identification of soils.” Proc., American

Society of Civil Engineers, 73(6), 283.
Casagrande, L. (1938). “Examination of the sub-soil of roads.” Proc., 8th International

Road Congress, The Hague, 1–27.
Chamberlain, E. J. (1987). “A freeze-thaw test to determine the frost susceptibility of

soils.” U. S. Army Corps of Engineers, Cold Regions Research and Engineering
Laboratory (CRREL), Hanover, Special Report: 87-1.
Csathy, T. I., and Townsend, D. L. (1962). “Pore size and field frost performance of
soils” Highway Research Board Bulletin, 331, 67-80.
Holtz, R. D., and Kovacs, W. D. (1981). An introduction to geotechnical engineering,

Englewood Cliffs, Prentice-Hall, NJ.
Rengmark, F. (1963). “Highway pavement design in frost areas in Sweden.” Highway

Research Record, Pavement Design in Frost Areas II. Design Considerations, ,
137–157.
Taber, S. (1929). “Frost heaving.” Journal of Geology, 38, 303–317.
Taber, S. (1930b). “Freezing and thawing of soils as factors in the destruction of road
pavements.” Public Roads, 11(6), 113-132.
Taivenen, O.A. (1963). “Preventive measure to reduce frost action on highways in

Finland.” Highway Research Record, 33, 202–216.
Takagi, S. (1978). “Segregation freezing as the cause of suction force in ice lens
formation.” U. S. Army Corps of Engineers, Cold Regions Research and
Engineering Laboratory (CRREL), Hanover, Report No. 78-6, 12.
Takagi, S. (1980). “The absorption force theory of frost heaving.” Cold Regions Science
and Technology, 3, 57-81.
357
Zhang, X., and Belmont, N. (2009). “Use of mirafi nylon wicking fabric to help prevent
frost heaving in Alaska pavement: 1st, 2nd, 3rd, 4th and 5th progress reports.”
Progress Reports to TENCATE GEOSYNTHETICS (North America), Institute of
Northern Engineering (INE)/Alaska University Transportation Center (AUTC),
University of Alaska Fairbanks.
Zhang, X., and Presler, W. (2012). “Use of H2Ri wicking fabric to prevent frost boils in
the Dalton highway beaver slide area, Alaska.” Alaska University Transportation
Center (AUTC), Project Report, INE/AUTC 12.23.
Zhang, X., Presler, W., Li, L., Jones, D. and Odgers, B. (2014). “Use of wicking fabric to
help prevent frost boils in Alaskan pavements.” Journal of Materials in Civil
Engineering, 26(4), 728-740.
358
VIII. A BIO-WICKING SYSTEM TO DEHYDRATE ROAD EMBANKMENT
Chuang Lin1 and Xiong Zhang 2
1
Chuang Lin, Graduate Research Assistant, Department of Civil, Architectural and
Environmental Engineering, Missouri University of Science and Technology, Rolla, MO
65409-0030; Tel: 907-799-9203; Email: [email protected]
2
(Corresponding author) Xiong Zhang, Ph.D., P.E., Department of Civil, Architectural
and Environmental Engineering, Missouri University of Science and Technology, Rolla,
MO 65409-0030; Tel: 573-341-6286; Email: [email protected]
ABSTRACT
Water within pavement layers is the major cause of pavement deteriorations. A
small moisture content increment will result in significant reduction in both base course
and subgrade resilient behavior and increment in permanent deformation. Conventional
drainage systems can drain gravitational water, but not capillary water. An economically
feasible, energy saving and environmentally friendly alternative is required to deal with
the excess water induced distresses. Both lab and field tests have proven the effectiveness
of a newly developed geotextile with wicking fabric in dealing with such problems as
frost heave, thaw weakening, and moisture content induced differential settlement.
However, the geotextile is exposed to the open air in the original design, raising several
potential application concerns, such as ultraviolet degradation, mechanical failure,
malfunction due to high suction in the air, and clogging issues.
This paper aims at studying the possibility of using a bio-wicking system to
address the potential concerns and further dehydrate the moisture content of base course
359
material for the long run. Two types of tests, elemental-level and full-scale tests, were
performed to evaluate the moisture migration in a typical aggregate with 14.5% of fines.
Test results indicated that the bio-wicking system successfully addresses the concerns in
the original design and is a more effective drainage system to dehydrate a base course
compared with the original design.
Key Words: Unsaturated soil, pavement, resilient modulus, moisture content,
geotextile
1. INTRODUCTION
Water within pavement layers is a major cause of pavement deteriorations
(Cedergren, 1994). A recent NCHRP study estimated that excess water reduced the life
expectancy of pavement systems by more than half (Christopher and McGuffey, 1997).
Government transportation engineers in cold regions have credited a minimum of half of
road maintenance expenditures to the effects of freezing and thawing (Henry and Holtz,
2001). When a road is built, both the base course and subgrade are compacted at their
optimum moisture contents to achieve the best performance. After construction,
aggregates inside the pavement structure tend to reach equilibrium with the ambient
environment. The surface aggregate can be quickly air dried since the suction (negative
pore water pressure) in the air can be as high as 14 MPa (Fredlund and Rahardjo, 1993).
Under such a high suction level, the hydraulic conductivity for the surface aggregate will
be very small (nearly impermeable) under unsaturated conditions (Brooks and Corey,
1964) and the water exchange between the aggregates and the ambient environment will
360
be impeded. On the other hand, the moisture content within the pavement structure will
gradually increase due to capillary action, precipitation infiltration, and water
condensation. Numerous studies (Barksdale, 1972; Haynes and Yoder, 1963; Li et al.,
2011) have indicated that the pavement performance can be significantly influenced by a
small increment in the moisture content. Conventional drainage systems rely on gravity
to drain water out of pavement systems (AASHTO, 1993; ARA, 2004; Henry and
Stormont, 2002), which cannot drain capillary water or prevent the above scenarios from
happening. Beskow (1991) found that conventional drainage systems are not wholly
effective at reducing water-related problems in partially saturated soils. However,
pavement is under unsaturated conditions during most of its service life. Consequently,
no matter how well the road is constructed, it will inevitably experience distresses with
time due to the increasing moisture content.
Even though numerous techniques have been developed to mitigate pavement
damages caused by excess water, current engineering practices indicate that
improvements are still expected in a more cost-effective way. In general, the state-of-
practice can be divided into three categories: edge drain, open graded base courses
(permeable base), and the use of geotextiles for drainage purposes (Ariza and Birgisson,
2002). Pavement edge drains are designed to collect and remove water within and under
the pavement structure (Ridgeway, 1982). Edge drains must have the necessary hydraulic
capacity to handle the discharged water without getting clogged. Therefore, the
applications of edge drains are limited to relatively clean base materials (FHWA, 1994).
Another treatment is to use open graded base course (OGBC) in the pavement structure
(Fassman and Blackbourn, 2010; Lin et al., 2014). The high permeability of the OGBC
361
allows water to freely flow to the road edge. However, not all places can find such good
construction materials and the cost of producing and/or hauling may be unaffordable.
Firstly, OGBC are mainly composed of crushed stone with limited fines (FHWA, 1994).
The manufacturing of OGBC requires a large amount of energy during the excavation,
screening, and sizing processes. According to the DOE (Department of Energy) (DOE,
2002), 33,775 kilojoules of energy are required to mine and process one ton of aggregate.
This is not an environmentally friendly and sustainable way of producing large quantity
of construction materials and the price for OGBC is more expensive than conventional
base course. Schmitt et al. (2010) reported that the initial construction cost of asphalt-
treated OGBC was 27% higher than that for dense-graded base course (without OGBC).
Secondly, such a good material is not available near all construction sites and the cost for
hauling may be not affordable. For example, the geological survey of the greater
Fairbanks Area, Alaska (Mulligan, 2004) indicated that most of the landscape is covered
with finer sediments and organic material of varying thickness for about 2 m, and
permafrost covers one-third to one-half of the survey area (1043 km2). Even though
specification (Jeffers, 2017) requires the fine content (soil particles passing No. 200 sieve
(sieve opening of 0.075 mm)) for D-1 base course to be lower than 6%, it is not
economically feasible to get such material within a reachable distance considering the
large quantity required. Therefore, OGBC is a neither environmentally friendly nor
universally reachable construction material. In addition, geotextiles and geocomposites
are also frequently used as capillary barriers (Doré and Zubeck, 2009) to prevent water
from rising to the base course. Although the geotextiles and geocomposites impede the
capillary water intrusion, they also result in excess water accumulation near the barrier.
362
Numerous researchers have reported that capillary barriers will lead to an increase in
water content of the overlying soil (Clough and French, 1982; Richardson, 1997;
Zornberg et al., 2010). In summary, none of the treatments can effectively solve the
problem. Improvements are still expected to reduce or eliminate the impact of moisture
accumulation in a more effective, energy saving, environmentally friendly, and
sustainable way.
A newly developed geotextile with wicking fabric has the capability to solve this
issue. The geotextile is a dual functional product: the high modulus polypropylene yarns
(black) for reinforcement purpose and the special hydrophilic and hygroscopic wicking
fibers (white) for drainage purpose, as shown in Figure 1a. The key to this type of
specially designed fiber material is its high wettability (to maintain saturation under
unsaturated conditions) and high unsaturated hydraulic conductivity (to laterally transport
water). The cross section has a high shape factor and great numbers of channels per fiber
(specific area is 3650 cm2/g) (Tencate, 2015), which gives the wicking fabric great
potential for maximizing capillary action and water transport under unsaturated
conditions. Most importantly, it can maintain saturation under low relative humidity (RH)
conditions. The original drainage design is also presented in Figure 1a. By installing a
layer of the new geotextile (hereafter the new geotextile will be denoted as “wicking
fabric”) horizontally, both gravitational and capillary water in the pavement structure can
be absorbed from the base course, transported along the wicking fabric to the shoulder,
and eventually vaporized into the surrounding atmosphere. In this way, the wicking fabric
serves as a “pipe” and nature serves as a “natural pump”, which can work 24 hours a day
and 365 days a year to dehydrate the roadway. When the water is removed and the base
363
course is kept relatively dry, the pavement performance will be significantly improved.
By doing so, one can use worse materials to achieve the same performance, or use the
same materials to achieve a better performance. This design has the potential to become
an economically feasible, environmentally friendly, and sustainable alternative to deal
with excess water induced pavement deteriorations. Moreover, this concept has been
validated by several field applications (Azevedo and Zornberg, 2013; Delgado, 2015; Lin
et al., 2017; Zhang and Belmont, 2011). For example, the wicking fabric has been used to
prevent frost heave and subsequent thaw weakening issues at the Dalton Highway Beaver
Slide, Alaska (Zhang and Presler, 2012). After 7.5 years of field observation, the wicking
fabric has successfully eliminated the frost heave issue. In addition, the applications of
the wicking fabric also extended to the treatment of differential settlement in expansive
subgrades (Delgado, 2015).The monitored results indicated that the wicking fabric was
effective in uniformly distributing moisture of the pavement subgrade.
However, the original design (Figure 1a) may have some potential concerns. In
the original design, the wicking fabric was exposed at the roadside so that water can be
vaporized to the ambient environment. This design may cause issues when considering
the long-term performance of the wicking fabric during the pavement’s service life
(usually 20-30 years). Such potential concerns are: (1) the wicking fabric may degrade
over time due to sunlight exposure; (2) routine grass mowing maintenance may cause
mechanical damage to the wicking fabric; (3) the wicking fabric may lose function under
high suction conditions due to air intrusion into the drainage channels; and (4) clogging
and salt concentration may influence the drainage efficiency of the wicking fabric.
364
This paper aims at further improving the original design by investigating the
possibility of a bio-wicking system to address the above concerns. The proposed drainage
design of the bio-wicking system is shown in Figure 1b. The wicking fabric is buried at
certain depth below the topsoil at the road shoulder. The road shoulder is then hydro-
seeded to establish vegetation so that evapotranspiration will occur at the grass leaves,
instead of directly evaporating from the wicking fabric. The vegetation works as a
“pump” to vaporize the water while the wicking fabric serves as a pipe that maintains
saturation under unsaturated conditions. By doing so, the wicking fabric will not become
overly dried since vegetation wilts at a suction of 1500 kPa (Kramer and Boyer, 1995).
This bio-wicking system also maintains the benefits of wicking fabric while simplifying
maintenance. By adding a layer of vegetation protection, the wicking fabric will be
protected from direct UV (Ultraviolet) deterioration and mechanical failure. The
objectives of this paper is to (1) compare the drainage efficiency of the original design
and the proposed bio-wicking system; (2) evaluate the long-term performance of the bio-
wicking; and (3) explore the working mechanism of the bio-wicking system.
2. TEST MATERIALS AND TEST METHODOLOGIES
Two types of tests were performed to characterize the short-term and the long-
term drainage efficiency of the bio-wicking system. The elemental level test lasted for 19
days and aimed at comparing the drainage efficiency of the conventional drainage system
(no wicking fabric and no vegetation), the original design (with wicking fabric and no
vegetation), and the bio-wicking system (with wicking fabric and with vegetation). The
365
other type of test was full-scale test and lasted for nearly one year. The full-scale test
included ten testing flumes to monitor the moisture migration through the base course
material. Each testing flume was 3.6 m long (12 ft. long) to simulate a half of two-lane
pavement. Detailed information regarding the test material, design, and schedule will be
discussed in this section.
2.1. TEST MATERIALS
2.1.1. E-1 Aggregate Properties. To better evaluate the performance of the bio-
wicking system under extreme conditions, the authors selected E-1 aggregate (Jeffers,
2017) and its gradation curve is shown in Figure 2. Compared with the recommended D-1
base course (fine content lower than 6%), E-1 aggregate contained a much higher fine
content (14.5%). The authors aimed at evaluating the performance of the bio-wicking
system under bad construction conditions where high quality construction materials were
not available. According to Unified Soil Classification System (USCS) (ASTM D2487),
the aggregate was defined as well-graded gravel with sand and silt (GW-GM). The
constant head test was performed (ASTM D2434), and the aggregate saturated hydraulic
conductivity was 6.5×10-5 m/s.
2.1.2. Vegetation Properties. When the road construction has been finished, the
road embankment is often revegetated for erosion control purpose. For the bio-wicking
system, it is highly possible that the wicking fabric will be buried under the hydroseeded
grass layer. To best simulate the field condition, typical grass seeds and seed blend are
selected according to the recommendations by Alaska DOT&PF (Department of
Transportation & Public Facilities) (Wright and Hunt, 2008). Based on the selection
366
criteria of state regions, aggregate moisture conditions, and soil types, the final seed blend
included 60% of nortran tufted hairgrass, 30% of arcared red fescue, and 10% of annual
ryegrass. The nortran tufted hairgrass can tolerate acidic soils, cold and wet conditions,
and low fertility locations (Mitchell, 1986). The arcared red fescue has deep root system
and resistant to wear and drought conditions. The annual ryegrass can be used to provide
quick vegetation coverage at the early stage. The selected seed mix is a typical blend for
cold regions.
2.2. TEST DESIGNS AND CONSTRUCTION PROCESSES
2.2.1. Elemental Level Test. The elemental level test included three testing boxes
(B1-B3), as shown in Figure 3. Each testing box was made of plastic tote with dimensions
of 0.91 m × 0.45 m × 0.45 m (36 in. × 16 in. × 16 in.) (Length × Width × Height). The
testing box was covered with a layer of 0.15 mm thick plastic wrap to prevent water
leakage from unexpected locations. Two layers of moisture sensors were installed at
elevations of 0.04 m and 0.15 m (1.5 in. and 6 in.) from the bottom to monitor the
moisture content variations within the testing box. In the horizontal direction, the spacing
between the sensor and the sidewall was 0.15 m (6 in.). The total height of the E-1
aggregate was 0.3 m (12 in.) and the top of the testing box was sealed with plastic wrap to
prevent water evaporation.
The three testing boxes represent three different type of drainage systems in a
pavement structure. As shown in Figure 3, the top box (B1) served as a control case to
simulate the water flow with conventional drainage system. Neither wicking fabric nor
vegetation was installed for this testing box. The test box B1 had several pre-drilled
367
drainage holes at the left bottom corner to allow excess water flowing out of the system
under the influence of gravity. In comparison, the middle box (B2) represents the base
course with the original drainage design, where the wicking fabric was installed but
without vegetation. In this box, a layer of wicking fabric was buried at 0.04 m (1.5 in.)
from the bottom. The wicking fabric extended out to the open air through a 0.33 m ×0.01
m (13 in. × 0.5 in.) opening located at the left bottom of the testing box. The extension
part of the wicking fabric was 1.2 m (4 ft.) in length and was directly placed on top of a
layer of E-1 aggregate. For the bottom box (B3), it represented the base course installed
with the proposed bio-wicking system. The wicking fabric was installed at the same
location as described for B2, but the extension part was buried underneath a layer of 0.05
m thick topsoil. The relative shallow thickness of the topsoil is to ensure a good contact
between the grass root and the wicking fabric.
Figure 4 shows the testing box construction process and here take B2 as a
demonstration. Firstly, the testing box was covered with a layer of plastic sheet, as shown
in Figure 4a. Two cross beams were added to the top of the box to provide additional
confinement. Secondly, a layer of 0.04 m (1.5 in.) thick E-1 aggregate was backfilled and
then the wicking fabric was installed in place. After that, the first layers of sensors were
installed at the target locations and a shovel of E-1 aggregate was piled on top of the
sensors to secure them in positions (Figure 4b). Thirdly, the other end of the wicking
fabric was extended out of the testing box through the opening and placed on top of a
layer of E-1 aggregate (Figure 4c). The opening might subject to closure due to heavy
overburden soil pressure and closure of the opening would significantly reduce the
wicking fabric drainage efficiency. Therefore, two wood beams were used to provide
368
additional support to the testing box sidewall where the opening was cut. Then, another
0.11 m (4.5 in.) thick E-1 aggregate was backfilled and the second layer of sensors was
installed. Finally, the last 0.15 m (6 in.) of E-1 aggregate was backfilled to the target
height and the top of the testing box was sealed with plastic sheet.
Figure 4d gives the final configuration of the testing boxes (from left to right: B1,
B2, and B3). The construction process for B1 and B3 were similar as for B2 and detailed
procedures will not be discussed. Note that there were no supportive wooden beams for
testing box B1 because only drainage holes were prefabricated for B1. As for the testing
box B3, the extended portion of the wicking fabric was buried 0.05 m (2 in.) beneath the
topsoil and then the topsoil was hydroseeded with the selected grass seed blend. The
grass took roughly 5 weeks to fully mature and the test did not start until the grass was
matured.
2.2.2. Full-Scale Test. Figure 5 shows the full-scale test design. The aim of the
test was to evaluate the drainage efficiency of half of a two-lane roadway installed with
the bio-wicking system. Figure 5a shows the top view of the tested area. In total, ten
testing flumes were built and labeled from T1 to T10, respectively. Eight out of ten testing
flumes were designed to use the bio-wicking system for drainage purpose. Each testing
flume had dimensions of 3.65 m × 0.38 m × 0.30 m (12 ft. × 15 in. × 12 in.) (Length ×
Width × Height), with 2.74 m (9 ft.) long installed with wicking fabric and the other 0.91
m (3 ft.) without. The extended part of the each wicking fabric strips for the bio-wicking
system was buried within a vegetation area with dimensions of 2.74 m × 2.44 m (9 ft. × 8
ft.). On the north side of the vegetation area, there was a drainage ditch to collect excess
water during heavy rainfall events. For the rest two testing flumes, the wicking fabric was
369
buried under E-1 aggregate for testing flume 5 and was exposed to the air for testing flume
10. Due to limited project budget, four out of ten testing flumes were installed with
moisture sensors to monitor the moisture migration. The red dots represent sensor
locations in Figure 5a. Each sensor location within testing flume represents three sensors
with various buried depths. However, each sensor location in vegetation area indicates one
sensor because the wicking fabric was buried at relatively shallow depth (0.05 m). Testing
flume 7 (red dash line in Figure 5a) will be discussed in detail to demonstrate the
configuration of a single testing flume.
Figure 5b gives the front view of testing flume 7, which installed with bio-
wicking system. The bottom of the testing box sat on the ground level. A layer of plastic
sheet was placed into the testing flume to prevent unexpected water leakage (blue line in
Figure 5b). In total 12 moisture sensors were distributed into three layers with depths of
0.04 m, 0.15 m and 0.26 m (1.5 in., 6 in., and 10.5 in.) from the bottom, respectively. The
spacing between two adjacent sensors was 1.22 m (4 ft.). The wicking fabric was
installed at a depth of 0.04 m (1.5 in.) from the bottom. Part of the wicking fabric was
extended to the vegetation area. The gap between the vegetation area and the testing
flume was also connected and hydroseeded with a smooth slope. The vegetation section
was 0.15 m (6 in.) in depth and was also covered with plastic sheet to impede excess
water flowing into the bio-wicking system. The wicking fabric was buried at a depth of
0.05 m (2 in.) from the ground surface.
Figure 6 shows the construction process of the vegetation area. The test section
was first cleared and excavated to the designed depth of 0.15 m (6 in.) and the surface
was leveled. A layer of 0.15 mm plastic sheet was then installed in place. After that, the
370
top soil was blended with the original soil (1:2 by volume), backfilled and compacted, as
shown in Figure 6a. A lawn roller weighted with water was used to compact and level the
blended soil. Before backfilling with another layer of soil, the surface of the compacted
soil was scratched with a rake to ensure good contact between the adjacent soil layers.
Then eight wicking fabric strips were installed in place as shown in Figure 6b. Because
the grass would take several weeks to mature, part of the wicking fabric strips were
wrapped and covered with plastic bags to prevent from contamination. The wicking
fabric strips that were covered with plastic bags were further buried under the ground to
prevent from UV (ultra-violet) deterioration and would be dug out right before the start of
the test. Care was taken to ensure that no wicking fabrics were in contact with each other
and there was about a 0.15 m (0.5 ft.) spacing between two adjacent wicking fabric strips.
When the soil compaction and the wicking fabric installation processes were finished, the
soil surface was raked again to create shallow grooves for grass seeds to set (Figure 6c).
Then the grass seed and the fertilizer were evenly applied via a hand spreader. The
construction of the vegetation area was completed in July 2015. The field test was not
performed until summer 2016 because the authors would like to ensure that the grass
roots were well established and the contacts between grass root and wicking fabric were
fully developed. However, large amount of melting snow flooded the vegetation area in
spring 2016 and the authors had to excavate a drainage ditch on the north side of the
vegetation area, as shown in Figure 6d. The vegetation area was ready for testing in June
2016.
The construction of the testing flumes started in July 2016. The frame of the
testing flume was made of wood stud and the sidewalls were made of plywood board.
371
The components of the testing flumes were first prefabricated, then sent to the
construction site, and assembled on site. All ten testing flumes first placed at the designed
locations, as shown in Figure 7a. On east and west sides of the test section, four testing
flumes on each side faced to the vegetation area. The rest two testing flumes were laid
down on south side of the vegetation area. After the testing flumes were in place, each
testing flume was covered with two layers of plastic sheet, as shown in Figure 7b. The
plastic sheet was long enough to cover the top of the testing flume and prevent
unexpected water leakage. Crossbeams made of wood stud were used to fix the plastic
wrap in position and provide additional confinement to the sidewalls. After the testing
flumes were ready, a layer of 0.04 m (1.5-in.) thick E-1 aggregate was first backfilled
into the testing flume. Then, the buried wicking fabric strips were unwrapped, cleaned
with tap water, and spread out in the testing flume. After that, the first layer of sensors
was installed, and one shovel of E-1 aggregate was gently piled on top of each sensor to
secure them in position (Figure 7c). The rest of the E-1 aggregate was backfilled into the
testing flume with two extra lifts, as shown in Figure 7d. Finally, the entire testing flume
was sealed and the final configuration was shown in Figure 7e. All the sensors were
connected to a CR1000 data logger with an AM 16/32 multiplexer and the data
acquisition system was powered by a lead-acid storage battery. The data logger was pre-
programmed to record the air temperature and the soil moisture content every 15 minutes.
2.3. TEST SCHEDULE
The short-term, elemental level test was performed from June 13, 2016 to August
1, 2016, lasting for 19 days. The test did not start until the grass was fully matured. The
372
test started with oversaturation of each testing box until excess water flowing out. Then
the air temperature, soil temperature, and soil volumetric moisture content was monitored
every 15 minutes. After the test was finished, samples from the vegetation area were
collected for SEM (Scanning Electron Microscope) analyses to evaluate the interaction
between the grass and the wicking fabric.
For full-scale test, the construction of the vegetation area was completed in July
2015. To ensure a good contact between the grass and the wicking fabric, the test did not
conducted until July 2016. Two saturation processes were executed on July 25, 2016 and
August 15 2016. After that, the air temperature and soil volumetric water content were
monitored until June 15, 2017.
To monitor the moisture content migrations within the tested area, a continuous
moisture content distribution field was required. However, due to the limited numbers of
sensors, a numerical interpolation technique was necessary to facilitate the visualization
of the moisture content migrations within the test area. Moreover, the sensors could not
cover the entire area of the testing flume and the created moisture contours could only
represent the moisture migration within the monitored area. Therefore, it is more
meaningful to read the pattern of a whole contour plot than to focus on the value of a
single point in the contour plot. Detailed discussions of the tested results are
demonstrated in the following sections.

373
3.1. CLIMATIC DATA
Figures 8 shows the temperature and precipitation data during the test period. As
shown in Figure 8a, the maximum and minimum temperature observed were 32.8 ºC and
-46.7 ºC, respectively. The mean temperature dropped below 0 ºC on October 15, 2016
and raised above 0 ºC on March 30, 2017. The accumulated rainfall was 11.55 in. (from
the start of the test to the end of year 2016) and significantly dropped to 3.74 (from the
beginning of 2017 to the end of the test period). During the springtime, the maximum
temperature for might exceed 0 ºC even though the average temperature was still below 0
ºC on that day. Therefore, a small amount of precipitation might be observed during cold
days (March 30, 2017).
3.2. ELEMENTAL LEVEL TEST
3.2.1. Test Boxes Net Water Gain. Figure 9a shows the calculated net water gain
variations for B1 and the precipitation data during the testing period. The net water gain
value represents the difference in total amount of water in the testing box between the
current state and the state before the test. Therefore, a positive net water gain value
indicated that excess water flowed into the testing box while a negative value implied that
excess water flowed out of the testing box. The procedures of determining the net water
gain will be explained as follows. The iso-water content lines could be first determined
based upon each moisture contour frame. After that, the area between two adjacent iso-
water content lines was calculated. Then, the average volume of water within each area
could be determined by multiplying the average volumetric water content with the covered
volume (area × testing flume width (0.41 m)). The total amount of water within the
374
monitored area would be the summation of the volume of water within each area. Finally,
the difference between the total amount of water between the current state and the
condition before the test would be the net water gain value.
Because the moisture sensors were temperature dependent, the calculated net
water gain was also temperature dependent, as shown in Figure 9a. In general, the
aggregate temperature fluctuation followed the trend of air temperature variation, but
with small amplitude due to soil insulation effect. Whenever there was a rainfall event,
the air and soil temperature fluctuations significantly decreased. During the saturation
process, a 5 L water increment was observed within the monitored area. Again, it was
important to point out that this was not the actual amount of water introduced to the
testing box. Within 2 hours, 3.7 L (76%) of the water was drained out mainly under the
influence of gravity. The hydraulic conductivity of an unsaturated soil significantly
decreases as the degree of saturation decreased (Dirksen, 2000). Since E-1 aggregate
contains about 14.5% of fines, its ability to hold water under unsaturated conditions is
relatively high. The average net water gain for B1 was relatively constant (1±0.1 L)
starting from 1 day after the test to the end of the test. It is expected that the actual
amount of water within the monitored area will be higher than the monitored value. This
phenomenon indicated that at least 24% of the water was capillary water and this portion
of the water cannot be drained out by conventional drainage system.
Figure 9b shows the comparison of net water gain and temperature fluctuation for
B1, B2 and B3. To compare the efficiency of different system to drain capillary water,
the net water gain 2 hours after saturation (when most of the gravitational water was
drained out) was selected as the reference point for each testing box. Again, negative
375
value for net water gain indicated that the total amount of water within the testing box
was decreasing. At the end of the test, the amount of water drained out within the
monitored area for B1, B2, and B3 were 0.4 L, 0.7L and 1.5 L, respectively. Therefore,
the drainage efficiency for bio-wicking system (with wicking fabric and vegetation) was
2 times higher than that for original design (with wicking fabric and without vegetation),
and about 4 times higher than that for conventional drainage system (without wicking
fabric and without vegetation). In sum, the bio-wicking system is the most effective
system among all three cases.
3.2.2. Moisture Contours in Testing Boxes. AASHTO (1993) indicated that the
base course could be considered as an excellent (good) drainage material if all the water
could be drained within 2 hours (1 day) after heavy rainfall events. Therefore, Figure 10
shows the plots of moisture contours for the three tested boxes at the start of the test (when
saturation process was completed), 2 hours after saturation and 1 day after saturation,
respectively. The first column shows the volumetric water content (VW) distributions
within each testing boxes. The saturation VW was about 0.5 according to the moisture
sensor specification but no saturation VW was observed in each testing box. This
phenomenon could be explained as follows. Firstly, the excess water flowed faster under
the influence of gravity and the soil at the top of the test box immediately became
unsaturated at the starting point. Secondly, the soil water content was monitored every 15
minutes and the saturation process followed the sequence from B1 to B3, which indicated
that not all peak VW values would be captured during the saturation process. Thirdly, the
installed sensors could not cover the entire testing box and water flowed to the bottom of
the testing box where was outside the monitored range.

376
After two hours (second column in Figure 10a), the maximum VW for B1, B2,
and B3 decreased to 0.21, 0.23 and 0.19 respectively. The obvious VW decreases in the
first 2 hours of the test indicated that the excess water was drained at a faster speed under
the influence of gravity. However, after one day (third column in Figure 10c), the VW
values for B1 and B2 only decreased by 0.02 and 0.03 respectively. It is important to
bring the reader’s attention that excess water was allowed to freely flow out thought the
drainage hole for B1. In contrast, the water flow path for B2 was through the wicking
fabric. In this sense, the drainage efficiency for B2 was much better than B1 even though
the VW values were very close. At the end of the testing period, the maximum VW
values for B3 was 0.11, almost half of that value compared with B1 and B2, which
indicated that the bio-wicking system was the best among all three cases.
3.3. FULL-SCALE TEST
3.3.1. Testing Flumes Net Water Gain. Figure 11 gives the calculated net water
gain for T6, T7, T8 and T10, respectively. The starting point of the test was selected as the
reference condition to evaluate the net water gain for each testing flume. On July 25,
2016, the monitored results indicated that the net water gain was 35 L due to the first
saturation process. Based on field observations, no excess water flowed out of the testing
flume. To ensure that each testing flume was fully saturated, the authors decided to
resaturate each testing flume on August 15 2016 until excess water was flowed out of the
testing flume. As indicated in Figure 11, the net water gain for the second saturation
process was about 55 L within the monitored area. During the testing period of 2016, T6
was the most sensitive to ambient environmental fluctuations due to excess water
377
accumulation in the vegetation area but the excess water could be wicked out very fast
during sunny days. As a result, even though the fluctuation of net water gain for T6 was
significant, the net water gain was still lower than T10 (representing the original design).
The effect of the vegetation area will be discussed in detail in the following section.
Moreover, comparisons of T6 through T10 also indicated that testing flumes installed with
bio-wicking system all showed relatively lower net water gain. When winter came, the
water became frozen and the unfrozen water content dropped close to zero, all net water
gain experienced a significant drop in early October. The snow did not start to melt until
mid-April when excess water was observed flowing back to T6 during the snow melting
process. The melting snow could be considered as another saturation procedure, which
resulted in significant increment in net water gain. After the melting process, the net water
gain for T6 became the lowest among all monitored testing flumes as excess water was
quickly wicked out. However, the net water gain for T7, T8, and T10 were very close,
varied from 7.8 L to 9.5 L. Two reasons might contribute to this phenomenon. Firstly, the
plastic sheet covered in the vegetation area impeded excess water flowing out of the
system. During the snow melting process, the hydroseeded portions of the wicking fabric
for T7 and T8 were close to saturation (even though not full saturated compared with T6)
and part of the melting snow could be wicked back to the testing flumes. Meanwhile, the
extended wicking fabric for T10 was directly exposed to the air (outside the vegetation
area) and no excess water was introduced to T10. The biased initial condition contributed
to the similar performance for T7, T8, and T10 in 2017. Secondly, the soil permeability
decreases significantly as degree of saturation decreases. It might take longer time for
capillary water to be drained out under unsaturated condition. Comparisons of T7, T8, and
378
T10 indicated that for a relatively dry season with limited amount of precipitation, the
drainage efficiency of the bio-wicking system was at least the same as the original design.
3.3.2. Effect of Rainfall Event. The bio-wicking system relied on the suction
gradient at the ends of the wicking fabric strip. One end of the wicking fabric was buried
within the testing flume and the other end was buried within the vegetation area.
Therefore, Figure 11 gives the moisture content distributions within the vegetation area
and within the testing flume, respectively. A heavy rainfall event occurred on September
5, 2016 with 20.0 mm (0.8 in.) precipitation. The figures show the moisture contours
before, right after, and 2 days after the rainfall event. The dash lines indicated the
centerlines of the buried wicking fabrics for T6, T8, and T10, respectively. The elevation
on the east side of the vegetation area (with higher VW values) was lower than that on the
west side, the general VW on the east side water higher than west side (first figure). Due
to the relatively low elevation on the northeastern side, the runoff water flowed and
accumulated near the northeastern side (close to T6), resulting in the highest VW value
(second figure). Two days after the rainfall event, the moisture contours showed similar
distributions as before the rainfall event (third figure).
The moisture content distributions within the vegetation area affected the testing
flume performance significantly. Figure 11b shows the moisture variations during the
same rainfall event for the three representative testing flumes, including T6, T8 and T10.
T6 and T8 were bio-wicking systems but with T6 closer to the drainage ditch. T10 was
the system with wicking fabric directly exposed to the air. The moisture contours before
the rainfall event, right after the rainfall event and 2 days after the rainfall event are
presented. For T6, because the bio-wicking system was closer to the drainage ditch, its
379
moisture distributions was significantly affected by the ambient environment. In other
words, when there was a heavy rainfall event, the wicking fabric outside the testing flume
was saturated due to excess water accumulation near the drainage ditch. Water flowed
from outside back to the testing flume. The highest VW for T6 increased from 0.23
(before rainfall event) to 0.32 (after the rainfall event) (second figure first row). After 2
days, the highest VW value decreased to 0.28 and an obvious VW decrement was
observed (third figure first row). However, the authors would like to emphasize that in
field conditions, the wicking fabric will be buried under the road shoulder where excess
water would runoff to the edge drain system. The phenomenon observed in this paper
will not happen as long as the wicking fabric is not in direct contact with ponding water.
Comparisons of T6 and T8 (both with bio-wicking system) indicated that a good
drainage condition is necessary for the bio-wicking system to be effective in capillary
water removal. For T8, due to a relatively higher elevation, no excess water was observed
and the wicking fabric was in drier condition for the portion buried in the vegetation area
(Figure 12a). Therefore, the overall moisture condition for T8 was lower than T6 within
the same testing period. The comparison of T8 (bio-wicking system) and T10 (original
design) showed that the bio-wicking fabric was more effective in reducing the capillary
water within the base course.
3.4. BIO-WICKING SYSTEM WORKING MECHANISM
As demonstrated in the elemental level and full-scale tests, the bio-wicking
system showed the best drainage performance among all three cases. In this section, the
380
authors would like to explore the working mechanism of the bio-wicking system on both
macroscopic and microscopic levels.
3.4.1. Macroscopic Analysis. Figure 13 is the profile of the bio-wicking system
after the elemental level test was completed. It clearly showed that the grass roots grew
downward and penetrated the wicking fabric layer (Figure 13a). The wicking fabric was in
good contact with the grass roots and good water transporting channels between the
wicking fabric and the grass roots would be maintained. Figure 13b shows the reverse side
of the wicking fabric after the topsoil was carefully removed. This figure further proved
that the woven texture between the wicking fabric and the grass roots was well
maintained. . A good weaving structure is critical for continuous water transportation and
grass roots grew randomly without preference to weft or warp directions. The grass root
continued to grow for about 0.08 m-0.13 m (3 in.-5 in.) after penetrating through the
wicking fabric and the buried depth should not be deeper than 0.18 m (7 in.) when applied
to the typical type of seed blend used in the project.
3.4.2. Microscopic Analysis. Part of the sample in Figure 13 was used to further
examine the interactions of the grass and the wicking fabric using a scanning electron
microscope (SEM), as shown in Figure 14. The analyzed grass stem was selected closer to
the grass root. The similarities between the deep grooved structure of the grass and the
artificially designed multi-channel of the wicking fabric indicated that they shared the
same working mechanism. The transpiration process of the grass was driven in part by
capillary action and water potential difference (Zhu and Zhang, 2015). If the water
potential in ambient air is lower than the water potential in the grass leaf airspace of the
stomatal pore, water travels down the gradient and moves from the leaf airspace to the
381
atmosphere via the evaporation process. Therefore, evapotranspiration is the sum of
evaporation and transpiration of the entire water cycle within a plant. Similarly, water
transportation within the wicking fabric is driven by capillary action or by water potential
difference (suction gradient) between the two ends of the geotextile. The water was
absorbed from the E-1 aggregate, transported along the wicking fabric, wicked along the
grass root to the leaves, and finally vaporized into the open air.
It is also important to evaluate the equivalent radius of the deep groove, which
determines the surface tension and the potential capillary force. The average radius of the
grooves in the grass ranged from 5 μm to 10 μm, which was smaller than that for the
wicking fabric (average value of 12 μm). Since surface tension is inversely proportional
to the radius of curvature (Wohlfarth, 2008), the grass root was expected to have a
relatively higher suction (surface tension) compared with that for wicking fabric. This
fact ensured that once the water was wicked out from the soil, the grass root had the
ability to continuously transport the water to the leaves for evapotranspiration process.
Essentially, the bio-wicking system is a combination of two drainage materials that share
the same working mechanism of transporting water. The spacing or pore size differences
among soil particles, wicking fabric and grass roots kept the water wicked out from the
soil to the wicking fabric, laterally transported to the shoulder, and eventually vaporized
at the grass leaves. By reducing the excess water within pavement structure in a cleaner,
energy saving, more effective and more sustainable way, one is expected to use worse
material to achieve the same performance, or use the same material to achieve better
performance.
382
4. CONCLUSIONS
This paper focused on studying the possibility of using a bio-wicking system to
reduce the base course moisture content for the long run. Both elemental level and full-
scale tests proved that the bio-wicking system was much efficient in reducing both
gravitational and capillary water. The major conclusions are summarized as follows:
1. Capillary water consisted of at least 24 % of the total water for the tested E-1
aggregate. This part of the water was considered undrainable in conventional
subsurface drainage designs, but now can be drained by using the bio-wicking
system. Capillary water is an important factor in determining the pavement long-
term performance and shall be considered in the drainage design.
2. The bio-wicking system is a more effective drainage design in mitigating
capillary water within the base course compared with the original design. The
total amount of capillary water drain out by the bio-wicking system was 2 times
higher than original design, and 4 times higher than the control case.
3. The microscopic analysis of the grass-wicking fabric interactions further validated
the effectiveness of the bio-wicking system. The microstructure similarities of the
grass and the wicking fabric indicated that the essential working mechanism for
grass and wicking fabric is the same. Both of them take advantage of capillary
action and water potential difference (suction gradient) to transport water. The
smaller radius of the opening in grass root ensures water to be absorbed from the
soil, laterally transport through the wicking fabric, wicked from the wicking
fabric to the grass, and eventually vaporized to the air through grass leaves.
383
In addition to these conclusions, the following are several recommendations for
future applications of the wicking fabric:
1. The wicking fabric can be buried deeper than described in this paper. The buried
depth in this project was 0.05 m (2 in.) to make sure a good contact between the
grass and the wicking fabric. However, the grass roots penetrated through the
geotextile and kept growing downward. The maximum buried depth for the
wicking fabric would be 0.13 m (5 in.).
2. A good drainage condition is critical to maintain the bio-wicking system
efficiency. In other words, the wicking fabric shall not be directly in contact with
ponding water to prevent excess water from flowing back to the pavement
structure.
3. The process of capillary water drainage is a slow process compared with
gravitational water flow. Free water can be quickly drained out of base course
within two hours, while capillary water drainage takes several weeks or even
months. The efficiency of the bio-wicking system cannot be observed in a
relatively short period. Therefore, elongate the monitoring time is necessary for
future study.
4. It is recommended to use numerical methods to quantify the effectiveness of the
bio-wicking system. However, the difficulties of numerical simulations may exist
in the nonlinearity of material properties and the complicated climatic effects
(combining the factors of wind speed, air temperature, relative humidity,
precipitation and solar radiation, etc.). A hydro-mechanical-climatic model needs
to be developed to evaluate the interactions between infrastructure and climate.

384
ACKNOWLEDGEMENTS
This study was funded by the Center for Environmentally Sustainable
Transportation in Cold Climates (CESTiCC) and TenCate Geosynthetics Americas. The
authors gratefully acknowledge for their financial supports.

385
(a)
(b)
Figure 1. Comparisons of the original and the bio-wicking system designs: (a) Original
design of the wicking fabric in road embankment, and (b) proposed bio-wicking system
386
Figure 2. Gradation curve for E-1 aggregate

387
Figure 3. Schematic plot of elemental level test designs (not to scale)

388
(a) (b) (c)
(d)
Figure 4. Elemental level testing box construction process: (a) box with plastic wrap, (b)
sensor installation, (c) wicking fabric outside box, and (d) final configuration
389
(a)
(b)
Figure 5. Full scale test design (not to scale): (a) top view, and (b) front view
390
(a) (b)
Drainage Ditch
(c) (d)
Figure 6. Vegetation area construction process: (a) soil compaction, (b) wicking fabric
installation, (c) grass seed spreading, and (d) drainage ditch excavation
391
(a) (b)
(c) (d)
(e)
Figure 7. Testing flumes construction process: (a) testing flumes layout, (b) plastic sheet
installation, (c) sensor and wicking fabric installation, (d) soil backfilling, and (e) final
configurations
392
(a)
(b)
Figure 8. Climatic data during testing period: (a) air temperature, and (b) precipitation
393
(a)
(b)
Figure 9. Elemental level test results: (a) net water gain for B1, and (b) comparison of net
water gain for B1, B2, and B3
394
Figure 10. Moisture contours for small box test

395
Figure 11. Comparisons of net water gain for testing flumes
395
396
(a)
(b)
Figure 12. Moisture contours for full-scale test: (a) moisture contours in vegetation area,
and (b) moisture contours in testing flume.
397
(a) (b)
Figure 13. Grass root and wicking fabric interactions: (a) grass root zone, and (b) back
side of the wicking fabric
(a) (b)
Figure 14. SEM images of grass and wicking fabric: (a) grass stem, and (b) wicking
fabric
398
REFERENCES
AASHTO, 1993. AASHTO Guide for Design of Pavement Structures, 1993. AASHTO.
ARA, 2004. Guide for mechanistic-empirical design of new and rehabilitated pavement
structures. Part.
Ariza, P., Birgisson, B., 2002. Evaluation of water flow though pavement systems.
Azevedo, M., Zornberg, J., 2013. Capillary barrier dissipation by new wicking geotextile,
Panamerican Conference on Unsaturated Soils. pp. 20-22.
Barksdale, R.D., 1972. Laboratory evaluation of rutting in base course materials, Presented
at the Third International Conference on the Structural Design of Asphalt
Pavements, Grosvenor House, Park Lane, London, England, Sept. 11-15, 1972.
Beskow, G., 1991. Soil freezing and frost heaving with special application to roads and
railroads. Cold Regions Research & Engineering.
Brooks, R., Corey, T., 1964. HYDRAU uc Properties Of Porous Media. Hydrology Papers,
Colorado State University.
Cedergren, H.R., 1994. America's pavements: world's longest bathtubs. Civil Engineering
64(9), 56.
Christopher, B.R., McGuffey, V.C., 1997. Pavement subsurface drainage systems.

Transportation Research Board.
Clough, I., French, W., 1982. Laboratory and field work relating to the use of geotextiles
in arid regions, Proceedings of the 2nd International Conference on Geotextiles,
Las Vegas, Nev. pp. 1-6.
D2434, 2000. Standard Test Method for Permeability of Granular Soils (Constant Head).
Annual Book of ASTM Standards, American Society For Testing and Materials,
West Conshohocken, PA 4.
D2487, A., 2011. Standard practice for classification of soils for engineering purposes
(Unified Soil Classification System). ASTM International, West Conshohocken,
PA www. ASTM. org.
Delgado, I., 2015. Use of geotextiles with enhanced lateral drainage in roads over
expansive clays.
399
Dirksen, C., 2000. Unsaturated hydraulic conductivity. Soil Analysis: Physical Methods,
second ed., Marcel Dekker Inc., New York, 183-237.
DOE, 2002. Energy and environmental profile of the US mining industry. US Department
of Energy Office of Energy Efficiency and Renewable Energy. Energiebilanz der
Nuklearindustrie.
Doré, G., Zubeck, H.K., 2009. Cold regions pavement engineering.
Fassman, E.A., Blackbourn, S., 2010. Permeable pavement performance over 3 years of
monitoring, Low impact development 2010: Redefining water in the city. pp. 152-
165.
FHWA, 1994. Drainable Pavement Systems-Participant Notebook. Demonstration Project

87.
Fredlund, D.G., Rahardjo, H., 1993. Soil mechanics for unsaturated soils. John Wiley &
Sons.
Haynes, J., Yoder, E., 1963. Effect of Repated Loading on Gravel and Crushed Stone Base
Course Materials Used in the AASHO (American Association of State Highway
Officials) Road Test. Highway Research Record(39).
Henry, K.S., Holtz, R.D., 2001. Geocomposite capillary barriers to reduce frost heave in
soils. Canadian Geotechnical Journal 38(4), 678-694.
Henry, K.S., Stormont, J.C., 2002. Geocomposite capillary barrier drain for limiting
moisture changes in pavement subgrades and base courses. No. NCHRP-IDEA
Project 68.
Jeffers, J., 2017. Standard Specifications for Highway Construction, in: Facilities,
A.D.o.T.P. (Ed.).
Kramer, P.J., Boyer, J.S., 1995. Water relations of plants and soils. Academic press.
Li, L., Liu, J., Zhang, X., Saboundjian, S., 2011. Resilient modulus characterization of
Alaska granular base materials. Transportation Research Record: Journal of the
Transportation Research Board(2232), 44-54.
Lin, C., Presler, W., Zhang, X., Jones, D., Odgers, B., 2017. Long-Term Performance of
Wicking Fabric in Alaskan Pavements. Journal of Performance of Constructed
Facilities 31(2), D4016005.
400
Lin, W., Ryu, S., Cho, Y.-H., 2014. A case study of flow characteristics of permeable
pavements by time and space model. Canadian Journal of Civil Engineering 41(7),
660-666.
Mitchell, W.W., 1986. Notice of release of 'nortran' tufted hairgrass. Agroborealis Alaska
Agricultural and Forestry Experiment Station, Univeristy of Alaska Fairbanks.
Mulligan, D., 2004. Soil survey of the Greater Fairbanks area, Alaska. Nat. Resour.
Conserv. Serv., US Dep. of Agric., Washington, DC.
Richardson, G.N., 1997. Fundamental mistakes in slope design. Geotechnical Fabrics

Report 15(2), 15-17.
Ridgeway, H.H., 1982. Pavement subsurface drainage systems. NCHRP synthesis of

highway practice(96).
Schmitt, R., Owusu-Ababio, S., Crovetti, J., 2010. Performance Evaluation of Open
Graded Base Course with Doweled and Non-Doweled Transverse Joints.
Tencate, 2015. Mirafi H2Ri Technical Data Sheet.

http://www.tencate.com/amer/Images/TDS_H2Ri%20170111_tcm29-16704.pdf.
(Accessed 11/09/2017 2017).
Wohlfarth, C., 2008. Surface tension of water, Supplement to IV/16. Springer, pp. 16-21.
Wright, S.J., Hunt, P., 2008. A revegetation manual for Alaska. Alaska Plant Materials
Center, Division of Agriculture, Alaska Department of Natural Resources.
Zhang, X., Belmont, N., 2011. Use of wicking fabric to help prevent differential
settlements in expansive soil embankments, Geo-Frontiers 2011: Advances in
Geotechnical Engineering. pp. 3915-3924.
Zhang, X., Presler, W., 2012. Use of H2Ri wicking fabric to prevent frost boils in the
Dalton Highway Beaver Slide area, Alaska.
Zhu, H., Zhang, L.M., 2015. Evaluating suction profile in a vegetated slope considering
uncertainty in transpiration. Computers and Geotechnics 63, 112-120.
Zornberg, J., Bouazza, A., McCartney, J., 2010. Geosynthetic capillary barriers: current
state of knowledge. Geosynthetics International 17(5), 273-300.
401
SECTION
2. CONCLUSIONS AND RECOMMENDATIONS
2.1. CONCLUSIONS
The use of wicking geotextile to dehydrate road embankment has been
comprehensively evaluated. The dissertation can be divided into three major sections:
material characterizations, numerical simulations, and field applications. The remarks for
each section have been summarized as follows:
2.1.1. Laboratory Characterizations. For laboratory characterizations of WRCs
of woven geotextiles, the testing techniques for the determination the WRCs of soils or
nonwoven geotextiles have been reexamined and proper testing techniques have been
proposed for woven geotextiles. Due to the directional distribution of the pores within a
woven geotextile, the test samples shall be place in such a way that the direction of the
water flow is in parallel with the direction of the highest continuity and connectivity for
the pores. Because of the existence of two distinct pore sizes in the wicking geotextile, a
bimodal shall be used to depict the WRC of the wicking geotextile. There were two AEVs
for the wicking geotextile, including the inter-yarn AEV (1.3 kPa) and inner-yarn AEV
(254.0 kPa). The Based on the laboratory tests, the water storage ability of the wicking
geotextile was much higher than other non-wicking geotextiles and its water storage
ability is highly dependent on the number of wicking fiber yarns. By establishing the
relationship between the number of wicking fiber yarns and the regression parameters, the
water storage ability of the geotextile can be predicted. Regarding the soil-geotextile
402
interactions, the performance of a geotextile reinforced pavement structure relies on the
performance of the soil, the geotextile, and their interactions. Therefore, a series of
laboratory tests were performed to comprehensively characterize the material properties of
the soil-geotextile system. Based on the test results, the soil resilient modulus was very
sensitive to water content variations while the interface frictional angle was not sensitive
to water content change. The saturated and unsaturated hydraulic properties (WRCs and
K-Functions) of the soil and the geotextile, and the equivalent unsaturated hydraulic
conductivity of the soil-geotextile interface were determined in the lab. Based on the
laboratory test results, the working mechanism of the soil-geotextile system was discussed
and validated via numerical simulations. Theoretically, the wicking geotextile was able to
reduce the water content of the AB3 aggregate by 2% from the optimum value and the
corresponding resilient modulus can be increased by three times.
To further evaluate the drainage performance of the wicking geotextile, physical
model tests were performed in a well-controlled laboratory environment. In total, three
drainage materials were used, including the drainage belt, wicking geotextile, and the
modified wicking geotextile. Test results indicated that capillary water could not be
drained out by conventional non-wicking drainage material. In comparison, the wicking
geotextile was able to drain both “free” water and capillary water. In addition, the
wicking geotextile can also absorb water from the underlying soil and is the only
drainage material that has lateral drainage ability.
2.1.2. Numerical Simulations. To accurately predict the dynamic performance of
a pavement structure under different climatic conditions, a coupled hydro-mechanical
model with considerations of climatic effect was proposed to evaluate the soil-vegetation-
403
climate interactions. The proposed model was suitable for both saturated and unsaturated
soils. The simulation results indicated that the model was able to accurately predict the
seasonal variations of the base course aggregate. The climatic factors, such as solar
radiation, air temperature, precipitation, wind speed, and relative humidity, were
considered in the model via an integrated boundary condition. The numerical model was
also able to accurately predict the resilient modulus variations and the back-calculated
effective resilient modulus could be more representative.
By incorporating the proposed hydro-mechanical model, the working mechanism
of the wicking geotextile was simulated via elemental level model. The simulation results
indicated that the wicking geotextile could work as a capillary barrier in the cross-plane
direction while could laterally drain the excess water in the overlying soils out of the soil-
geotextile system. However, during heavy rainfall events, the wicking geotextile could
also work as a permeable layer for excess water percolating to the underlying soil. This is
a beneficial engineering characteristic because no positive pore water pressure would
build up and the severity of excess water induced pavement deterioration could be
reduced. After that, a full-scale model was established to quantify the benefits of the
wicking geotextile to dehydrate road embankment. Engineering Applications.
2.1.3. Engineering Applications. To evaluate the drainage performance of the
wicking geotextile in the real word, a full-scale model was first built up in the laboratory
under a relatively well-controlled condition. Two types of soils were used in the full-scale
model to evaluate the performance of the wicking geotextile in good and bad drainage
conditions. The wicking geotextile was able to drain both “free” and capillary water.
However, the splice area significantly reduced the drainage efficiency of the wicking
404
geotextile. This fact could become a beneficial engineering characteristic if the wicking
geotextile could be properly installed. The overlying piece of the wicking geotextile
should be located at the upper stream of water flow so that excess water would not flow
back into the road embankment.
The field test section was selected at Beaver Slide on the Dalton Highway,
Alaska. Two layers of the wicking geotextiles have been installed in 2010 to mitigate
frost heave and the subsequent thaw weakening issues. In total 22 pairs of sensors were
installed to monitor the soil temperature and moisture variations within the road
embankment. Both field observations and monitoring results indicated that the wicking
geotextile successfully eliminated the “frost boil” problem. In addition, samples were
collected from the field and SEM analyses were performed to evaluate the potential
issues that might influence the long-term performance of the wicking geotextile. SEM
analyses results indicated that mechanical failure, clogging, and salt concentration shall
not be major issues while permanent deformation was a major issue because it would
cutoff the drainage paths.
To further improve the drainage performance of the wicking geotextile, a bio-
wicking system was proposed and the drainage performance was evaluated via elemental
level and full-scale tests. Test results indicated that the bio-wicking system was more
effective than the original design. In addition, the bio-wicking system addressed the
potential concerns of the original design in which the wicking geotextile was exposed to
the air. By growing a layer of grass on top of the wicking geotextile, the bio-wicking
system could protect the wicking geotextile from UV deteriorations, mechanical failure,
and malfunction due to extremely high suction level.

405
In summary, the wicking geotextile is a promising drainage material to dehydrate
road embankment. Different from conventional drainage materials which can only drain
“free” water, the wicking geotextile is able to drain both “free” water and capillary water
under saturated and unsaturated conditions. The drainage efficiency of the wicking
geotextile has been validated via laboratory tests, numerical simulations, and field
observations. By using the wicking geotextile, one can built a more cost-effective and
sustainable soil infrastructure.
2.2. RECOMMENDATIONS
The following researches are recommended in the future:
1. Quantify the drainage efficiency of the wicking geotextile in microscopic level.
Further research is required to determine the effect of the types of materials and
the shapes of the deep grooves on the drainage performance of the wicking
geotextile.
2. Further research needs to extend the numerical model to be able to simulate the
freezing and thawing cycles. So far, the numerical model is only suitable for
unfrozen soils.
3. This study focuses on the application of the wicking geotextile in pavement
engineering. However, there are other fields, such as slope stability and
foundation issues, that the wicking geotextile can be applicable for. Future
research is recommended to extend the application to other related fields.
4. Future research can also focus on quantify the mechanical, hydraulic, and
combined reinforcing effect respectively. In this way, the benefits of drying the
406
base course will be more quantifiable. Future research is also required to optimize
the installation location for the wicking geotextile so that field engineers will have
more specific guidance when using the wicking geotextile.

407
BIBLIOGRAPHY
AASHTO (1993). “AASHTO Guide for Design of Pavement Structures.” American

Association of State Highway and Transportation Officials (AASHTO),
Washington D. C., USA.
Anderson, D. A., Huebner, R. S., Reed, J. R., Warner, J. C., and Henry, J. J. (1998).
“Improved Surface Drainage of Pavements.” National Cooperative Highway
Research Program (NCHRP), Project No. 1-29.
ARA, INC. (2004). “Guide for Mechanistic-Empirical Design of New and Rehabilitated
Pavement Structures.” Final Report, NCHRP Project 1-37A, Transportation
Research Board of the National Academies, Washington, D.C..
Aravin, V. I. and Numerov, S. N. (1953). “Theory of Fluid Flow in Undeformable Porous

Media.” Gostekhizdat, Moscow, USSR.
Barber, E. S. and Sawyer, C. L. (1952). “Highway Subdrainage.” Proceedings, Highway

Research Baording,Vol. 31.
Brown, S. A., Stein, S. M., and Warner, J. C. (2001). “Urban Drainage Design Manual,
Hydraulic Engineering Circular 22, Second Edition.” Federal Highway
Administration, Publication No. FHWA-NHI-01-021.
Cedergren, H. R. (1974). “Drainage of Highway and Airfield Pavements.” John Wiley and
Sons, New York.
Cedergren, H. R. (1994). “America’s Pavements: World’s Longest Bathtubs.” Civil

Engineering, pp. 56–58.
Cedergren, H. R., Arman, J. A., and O'Brien, K. H. (1973). “Development of Guidelines

for the Design of Subsurface Drainage Systems for Highway Pavement Structural
Sections, Final Report.” Federal Highway Administration, Washington, D, C.,
February, 1973.
Christopher, B. R., and McGuffey, V. C. (1997). NCHRP Synthesis of Highway Practice

239: Pavement Subsurface Drainage Systems. TRB, National Research Council,
Washington, D.C., 1997.
FHWA (1980). “Highway Subdrainage Design. Publication.” Federal Highway

Administration (FHWA), FHWA-TS-80-224, U.S. Department of Transportation.
408
Han, J. and Zhang, X. (2014). “Recent Advances in the Use of Geosynthetics to Enhance
Sustainability of Roadways.” 20th International Conference on Advances in Civil
Engineering for Sustainable Development, Suranaree University of Technology,
Nakhon Ratchasima, Thailand, 29-39.
Henry, K. S., and Holtz, R. D. (2001). “Geocomposite Capillary Barrier to Reduce Frost
Heave in Soils.” Canadian Geotechnical Journal, 38(4), 678–694.
Kochina, P. and Ya, P. (1952). “Theory of the Motion of Ground Water.” Gostekhizdat,
Moscow, USSR.
Lane, K. S. and Washburn, D. E. (1946). “Capillary Tests by Capillarimeter and by Soil

Filled Tubes.” Proceedings, Highway Research Board, Vol. 26.
Muskat, M. (1946). “The Flow of Homogeneous Fluids Through Porous Media.” J. W.

Edwards, Publisher, Ann Arbor, Michigan.
Rechart, F. E. (1957). “Review of the Theories for Sand Drains.” Transactions, American
Society of Civil Engineers, Vol. 124.
Rutledge, P. C. and Johnson, S. J. (1958). “Review of Uses of Vertical Sand Drains.”

Bulletin 173, Highway Research Board, Wahsington, D. C..
409
VITA
Chuang Lin was born in Changchun, China. He received his Bachelor’s degree in
Civil Engineering in Jilin University in 2011. Then, he went to the University of Alaska
Fairbanks to pursue master’s degree in Geotechnical Engineering and graduated in 2014.
His thesis title was “Applications of Shallow Anchors in Ice-Rich Silt”. He continued his
Ph.D. study in Civil Engineering (Geotechnical Engineering) and transferred to Missouri
University of Science and Technology in August 2016. His dissertation title was “Use of
Wicking Geotextile to Dehydrate Road Embankments”. He received his Ph.D. degree in
Civil Engineering from Missouri University of Science & Technology in May 2019.

Use of Wicking Geotextile To Dehydrate Road Embankments

Uploaded by

Copyright:

Available Formats

Use of Wicking Geotextile To Dehydrate Road Embankments

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Use of Wicking Geotextile To Dehydrate Road Embankments

Uploaded by

Copyright:

Available Formats

Scholars' Mine

Doctoral Dissertations Student Theses and Dissertations

Use of wicking geotextile to dehydrate road embankments

Follow this and additional works at: https://scholarsmine.mst.edu/doctoral_dissertations

Presented to the Faculty of the Graduate School of the

MISSOURI UNIVERSITY OF SCIENCE AND TECHNOLOGY

In Partial Fulfillment of the Requirements for the Degree

Xiong Zhang, Advisor

All Rights Reserved

PUBLICATION DISSERTATION OPTION

used by the Missouri University of Science and Technology:

Geotextiles and Geomembranes.

Paper IV (pages 166-222) is under review for publication in the Journal of

Paper V (pages 223-271) is under review for publication in the International

Excess water within a pavement structure is the major cause of pavement

sensitive to water content variations. Unfortunately, conventional drainage system can

are evaluated. Eventually, a new-bio-wicking system is proposed to further improve the

dehydrate road embankments.

his professional guidance, support, and encouragement. It is my pleasure and privilege to

cannot express how grateful I am to have such an amazing advisor.

Part of my study was sponsored by TenCate Geosynthetics. I would like to express my

Fellowship for the 2017-18 academic year.

Thanks further extended to the geotechnical research group at Missouri S&T,

John Bullock, and Greg Leckrone.

Their love and unconditional support encourage me to overcome any difficulties in my

without their love.

PUBLICATION DISSERTATION OPTION ................................................................... iii

LIST OF ILLUSTRATIONS .............................................................................................xv

LIST OF TABLES .............................................................................................................xx

1.1. PROBLEM STATEMENT ......................................................................... 1

1.1.1. Adverse Effect of Excess Water in Road Embankments. .............. 1

1.1.2. Limitations of Conventional Subsurface Drainage Systems. ........ 2

1.2. A NEW DRAINAGE SYSTEM WITH WICKING GEOTEXTILE ......... 4

1.3. RESEARCH OBJECTIVES AND METHODOLOGIES ........................... 6

1.4. DISSERTATION OUTLINE ...................................................................... 7

I. CHARACTERIZATIONS OF WATER RETENTION CURVES FOR SEVERAL

2. RESEARCH OBJECTIVES AND METHODOLOGIES ........................................ 13

3. TESTING MATERIALS ......................................................................................... 15

4.TESTING PROCEDURES ....................................................................................... 18

4.1. CAPILLARY RISE TEST ........................................................................ 19

4.2. PRESSURE PLATE TEST ....................................................................... 21

4.3. SALT CONCENTRATION TEST ........................................................... 24

5. TEST RESULTS AND DISCUSSIONS .................................................................. 25

5.1. CAPILLARY RISE TEST ........................................................................ 25

5.2. PRESSURE PLATE TEST ....................................................................... 29

5.3. SALT CONCENTRATION TEST ........................................................... 31

6. COMPARISONS OF TESTING TECHNIQUES .................................................... 33

7. WATER RETENTION CURVES (WRCs) ............................................................. 36

II. COMPREHENSIVE MATERIAL CHARACTERIZATIONS OF PAVEMENT

2. A NEW GEOTEXTILE WITH WICKING FIBERS............................................... 62

3. RESEARCH OBJECTIVES AND METHODOLOGIES ........................................ 65

4. MATERIAL CHARACTERIZATIONS .................................................................. 66

4.1. SOIL PROPERTIES.................................................................................. 66

4.1.1. Physical Properties. ...................................................................... 66

4.1.2. Resilient Modulus. ....................................................................... 66

4.1.3. Soil Water Characteristic Curve (SWCC). .................................. 70

4.1.4. Saturated Hydraulic Conductivity. .............................................. 73

4.2. WICKING FABRIC PROPERTIES ......................................................... 73

4.2.1. Initial Tangent Modulus............................................................... 73