Field Properties in Surface Irrigation Management and Design

Field Properties in Surface Irrigation Management and Design Theodor s. Strelkoff, M.ASCE 1 ; Albert J. Clemmens, M.ASCE 2 ; and Eduardo Bautista, M.ASCE 3 Abstract: Field properties-topography, hydraulic resistance, and infiltration-play an important role in the performance of surface irrigation systems, and appropriate characterizations of these are required as data input to simulation or design software. The EWRIIASCE Task Committee on Soil and Crop Hydraulic Properties has been charged with preparing a guide for practitioners faced with such data entry. The result is this special section of the Journal of Irrigation and Drainage Engineering in which this paper is the first in the series presented. It describes the characteristics of these field properties and notes a series of caveats to be considered when dealing with them in the course of analyses or designs of surface irrigation systems. 001: 10.1061/(ASCE)IR.1943-4774.0000119 CE Database sUbject headings: Channel beds~ Hydraulic roughnes~ Introduction Field properties such as topography, soil infiltration, and hydraulic resistance (of soil grains and clods, mulch, and plant parts) have a profound effect on the performance of surface irrigation systems, and quantification of these parameters is crucial to good management and design. Surface irrigation systems are often characterized as exhibiting poor performance, with nonuniform application of water and excessive deep percolation and runoff. With the goal of improved design and operation, software has been developed for predicting and optimizing performance with trial values of the controlling variables affecting operation [SIRWinSRFR, USDA Agricultural Research MOD, Walker (204)~ Bautista et al. 2009)]. Based on Service (USDA-ARS) (209)~ laws of physics like conservation of mass and momentum, the software predicts the outcome of a particular set of control variables, such as inflow rate and cutoff time as these interact with the extant set of field variables. It is not immediately evident how to characterize those variables, like topography, system geometry, surface roughness, plant drag, and infiltration, nor how to evaluate them in a given field. To sort through the many available methods of estimating field characteristics, the Environmental and Water Resources Institute of ASCE authorized formation of a Task Committee on Soil and Crop Hydraulic Properties, a child of the parent technical committee for On-Farm Irrigation Systems, a lResearch Hydraulic Engineer, U.S. Arid Land Agricultural Research Center, USDA-ARS, 21881 N. Cardon Ln., Maricopa, AZ 85238 (corresponding author). E-mail: [email protected] 2Research Hydraulic Engineer, Center Director, U.S. Arid Land Agricultural Research Center, USDA-ARS, 21881 N. Cardon Ln., Maricopa, AZ 85238. E-mail: [email protected] 3Research Hydraulic Engineer, U.S. Arid Land Agricultural Research Center, USDA-ARS, 21881 N. Cardon Ln., Maricopa, AZ 85238. E-mail: [email protected] Note. This manuscript was submitted on July 31, 2008; approved on May 12, 2009; published online on May 15, 2009. Discussion period open until March 1, 2010; separate discussions must be submitted for individual papers. This paper is part of the Journal of Irrigation and Drainage Engineering, Vol. 135, No.5, October 1,2009. ©ASCE, ISSN 0733-9437/2009/5-525-536/$25.00. Infiltra o ~ Surface irrigation. member of the Irrigation and Drainage Council. The output of the committee is this special issue of the Journal of Irrigation and Drainage Engineering. In this, the first paper of the group, the various functional forms used to describe surface roughness and vegetative drag and infiltration are reviewed along with some of the features of these field characteristics. Methods for quantifying the parameters of those formulas and sensitivity of simulation and design software to the results comprise the subject of companion papers. Soil-Surface Geometry Field-surface configuration is measured in a level survey. It is subject to change over time, during an irrigation (say, as a freshly tilled soil settles) and over a season or between irrigations. The topography of an irrigated field affects performance at several different scales, in part because departures of the field surface from a theoretical plane occur at various scales, from grain size to field dimensions. Field-scale-average slopes have long been recognized as a major influence in surface irrigation performance. Spatial variations at a scale on the order of field dimensions, for example a gradual decrease in slope with distance along a border strip or furrow, or simply a break in slope at some point in the length of run can be input to computer-simulation models to evaluate the effect of such departures from a plane surface. Fig. 1 shows simulated post-irrigation distributions of infiltrated water under different grading treatments in an Egyptian basin (Clemmens et al. 1999a). The markedly greater total infiltration with traditional leveling stems from the Egyptian practice of cutting off inflow when the stream arrives at the downstream end-the irregularity in the soil surface slows stream advance. The contribution of cross slope to nonuniformity of infiltration can be investigated with two-dimensional simulation models (Playan et al. ~69 1 Strelkoff et al. 2003). Small-scale variations, microtopography, typically random departures from a smooth surface on a spatial scale of, say, a few meters, tend to decrease advance rates and so affect distribution uniformity, but their major influence is through low spots with longer than average recession times, especially if the average field surface is level, or inflow rates are JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / SEPTEMBER/OCTOBER 2009/525 0.50 1.00 2.00 4.00 8.00 16.0 __ 100. t-:r"*'~_.,e= ~. 0 E E N 10.0 --'-~ 0.10 ___I. _____l 10.0 1.00 T (hrs) Fig. 7. Merriam and Clemmens (1985) time-rated intake families. The time to infiltrate 100 mm is shown on each curve (hours) (z: cumulative-infiltration depth-volume per unit infiltrating area). corporated within the Soil Conservation Service (SCS) groupings. If experience justifies describing a soil in this way, the families allow a simple characterization of infiltration, with prescribed values of k, a, and c. Merriam and Clemmens (1985), arguing that the time required to infiltrate a desired depth of water is the key infiltration parameter for surface irrigation management, built upon the NRCS concept of infiltration families-characterized by a general correlation between the Kostiakov k and a-and developed the time-rated intake families. They examined cumulative-infiltration data for many soils and, excluding the cracking clays, empirically found a correlation between the exponent in the Kostiakov formula and the time 'Two (in hours) to infiltrate a characteristic depth of, specifically, 100 mm, namely (10) Thus, for noncracking soils, a single measurement, the time to infiltrate 100 mm, estimates both k and a in the Kostiakov power law. Fig. 7 shows the time-rated families, each family named for 'Two· Fig. 8 contrasts the interdependence between k and a in the two sets of families. Figs. 6-8, by comparing the two, show that 200. . - - - - - - - , - - - - - - - - , . - - - - - - - - - - - , 150. ro..... .e E t- - + - I l- U~ :w. L . I 100. r- + ~I- J! - I S ..::.::: 50. 32 NRCS Intake Families O. 0.30 0.75 0.50 1.00 a Fig. 8. Relationship between Kostiakov k and a in the NRCSIUSDA (SCS) (1984) intake families and the Merriam and Clemmens (1985) time-rated families. The squares identify family members, while the curves that fit them are identified by specific functions, given in the cited references. both sets of families cover roughly the same region of the 'T, z plane, but with the NRCS families including lower k values. The time-rated families exhibit a greater variation of a, dropping to lower values in the tighter soils than the NRCS families. The curious crossing of the time-rated families pictured at the lower end of Fig. 7 stems from extending an empirical expression beyond its intended range. Current research is aimed at relating these purely empirical families to soil texture and other physical characteristics. As with other field properties, the key issues in the empirical estimation of infiltrated depth are the selection of a suitable formula, estimation of its parameters and their spatial and temporal variation, the likely error, and the effect of that error on simulated performance. Different formulations and parameter value combinations will give similar results in simulation, if the infiltrated volume at the average opportunity time is similar. Influence of Wetted Perimeter With all infiltration passing through the irrigation-stream/soilsurface boundary, wetted perimeter plays a role in determining the volume infiltrated per unit length of run. In border strips and basins, infiltration into the berms is typically negligible compared to that into the border-strip surface. Wetted perimeter is then essentially independent of flow depth and is known; volume infiltrated per unit length is directly proportional to the width. In furrows, on the other hand, wetted perimeter varies significantly with flow depth. During typical stream advance, runoff, and recession, the depth and wetted perimeter at any section of a furrow rise rapidly when the stream arrives, hold more or less steady, and eventually, after cutoff, return to zero. During runoff, discharge and flow depths at the tail end of a long furrow can be significantly smaller than at the head end. Under cutback or cablegation inflows, additional changes in wetted perimeter occur because of changes in flow rate. In the early stages of wetting of a furrow section, after the stream arrives, infiltration through the furrow wall is entirely dependent on the wetted perimeter and the soil properties-which may themselves change upon contact with the irrigation water and exhibit some surface sealing, or crusting. Infiltration is not, at first, dependent on coincident infiltration from neighboring furrows, but later, as the wetted soil regions around each furrow expand and approach each other, lateral moisture flow is inhibited, and furrow spacing becomes an important factor. In the event of extensively cracked soils, the transverse cracks quickly convey infiltrated water (water that has crossed the nominal furrow boundary) laterally to meet with water spreading from the neighboring furrow. In this case, the effective wetted perimeter for infiltration is the furrow spacing from the start. In contrast to the previous theoretical considerations on the role of wetted perimeter in furrow infiltration, a recent review of extensive field data (Walker and Kasilingam 2004) indicated that measured wetted perimeter failed to show any consistent variation with either time or distance during the course of the irrigations. Expected decreases in wetted perimeter with distance down the furrow in many cases failed to materialize, possibly as the result of significantly greater measured roughness in the lower reaches of a furrow and erosive changes in the furrow-boundary geometry. These circumstances lead to doubts that incorporating theoretical changes based upon the assumption of constant cross sections (in the dry) in the course of a simulated irrigation provides any enhancement of accuracy. Nonetheless, despite the incompleteness of theoretical descriptions of furrow infiltration, something can be learned from solu- 530/ JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / SEPTEMBER/OCTOBER 2009 tion of the Richards unsaturated-soil water flow equations in homogeneous isotropic soils. Infiltration from furrows is, strictly speaking, a three-dimensional phenomenon, as some longitudinal soil-moisture movement occurs during an irrigation. However, much of the phenomenon is captured by viewing it in two dimensions, in a plane normal to the furrow length. The two-dimensional version of the HYDRUS software (Simunek et al. 1999), with coordinate directions vertical and lateral, transverse to the furrow length, provides numerical solutions to the Richards equation for variably saturated water flow (a) t=600 min (b) t=18oo minutes m aO m- a [(ah K -+ 0 at - aXi i} ax} iy )] (11 ) in which Om=volumetric water content [L 3L -3]; hm=pressure head [L] in the soil matrix; Xi (i = 1, 2) = spatial coordinates [L]; t=time [T]; Ki}=components of the unsaturated hydraulic conductivity anisotropy tensor [LT- 1] (a simple scalar in the cases described here), and oiy=Kronecker delta with y the vertical direction (equal to one If i corresponds to the vertical and zero otherwise). Unlike the sudden step change at the wetting front assumed for development of the Green-Ampt equation, the soil water content changes gradually, with soil suction and hydraulic conductivity both a function of water content. The Richards equation assumes that the porous medium is rigid and that the air phase plays an insignificant role in the liquid flow process. In a special version of HYDRUS-2D, the assigned boundary conditions in the furrow were made to change dynamically depending on the position of the water level in the furrow. Positive hydrostatic equilibrium pressures were assumed below the water level and a potential seepage face was established on the furrow wall above water level. In this variant, water is made available to the soil matrix at all points of the furrow wall below the free surface. A seepage face is a special type of a system-dependent boundary condition through which water can flow out of the soil from the saturated part of the flow domain (such as during a drawdown of the water level in the furrow). Without these special considerations, a negative pressure head would be specified on the boundaries in the furrow above the water level. This type of boundary condition does not allow controlling the flux through the boundary and both inflow and outflow into or out of the soil profile could occur depending on pressure heads in the soil. This modification has subsequently been made a standard feature of the HYDRUS-2D upgrade called HYDRUS (2D/3D) and is described in detail in its technical manual and online help (Simunek et al. 2006). Fig. 9 illustrates the case of steady furrow flow, at a constant depth of 100 mm, in a trapezoidal furrow formed in a silty loam soil. The corresponding unsteady flow of soil moisture as calculated by HYDRUS-2D is shown by the distribution of moisture after 600 and 1,800 min. The rise of moisture above the water level in the furrow as well as lateral penetration into the soil is clearly evident. The character of the solution changes over time as the initial unfettered penetration of water into the soil in all directions is increasingly constrained by the lateral advance of water from adjacent furrows (modeled in HYDRUS-2D as an impenetrable boundary condition). This theoretical solution supports the observation by the NRCS/USDA (SCS) (1984), quantified in a subsequent section of this paper, that in furrows, the infiltration per unit area of infiltrating surface is greater than in a border strip. This phenomenon has led to two schools of thought emerging around the empirical assessment of furrow infiltration. Some Fig. 9. Cross section of furrow and surrounding silt loam soil. Contours of moisture content computed by HYDRUS-2D (Simunek et al. 1999) stemming from a constant IOO-mm depth of irrigation water in the furrow. Initial moisture content: 0.24; saturated: 0.45. (a) time =600 min; (b) time= 1,800 min. models of the surface irrigation process [e.g., SIRMOD, Walker (2004)] expect as data input the parameters of empirical infiltration formulas directly expressing cumulative volume per unit length of furrow, A z, as a function of time (in the literature, this is often given the symbol, Z, with the corresponding formula parameters then, K, Q, B, and C; the use of A z here underscores the fact that the units of volume per unit length are area units, i.e., length squared). This implies that at some point in time, inflow and outflow measurements were taken in the furrow to yield a time rate of growth of infiltrated volume, thus incorporating the ramifications of whatever wetted perimeter was present when inflow and outflow were being measured. Infiltration-formula parameters are often obtained by measuring the increasing infiltrated volume in a section of furrow containing water at some constant depth and then fitting the per-unit-Iength data with the formula. Thus, the measured infiltration is viewed as a consequence of both soil properties (including initial moisture content) and the details of the irrigation system (furrow size, shape, spacing, flow depth, etc.) But how these data should be adjusted to apply to water at, say, a different depth-a different wetted perimeter-even a constant one, is not obvious [however, see Alvarez (2003) and Langat et al. (2007) for a suggested approach]. In models from the other school of thought, such as SRFR (Strelkoff et al. 1998), or in the furrow design procedures of the NRCS/USDA (SCS) (1984), the infiltration-data input is considered a function of soil properties alone, with the contribution of the irrigation system then calculated by the model. The data input, then, consists of formula parameters for volume infiltrated per unit infiltrating area (z, cumulative depth of infiltration), basically, one-dimensional (ID) infiltration, as in a basin. The volume infiltrated per unit length of furrow, A z , is then calculated for the varying flow depths and wetted perimeters of the simulated irrigation stream-in the case of SRFR, through a convenient but heuristic algorithm (A. J. Clemmens, personal communication, March 2, 1990) described in the following paragraph. Accounting for Wetted Perimeter It is assumed that the infiltration rate pertinent to the entire current wetted perimeter P of some cross section in the flow is given by an empirical function of the time of wetting of the furrow JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / SEPTEMBER/OCTOBER 2009/531 bottom, 'T. This is simply current time minus the stream-advance time to the given location. The rate of growth of A z reflecting this infiltration rate, i, is then (12) This expression does not account for the possibility of a constant term c in the formula for cumulative depth z reflecting essentially instantaneous accumulation of water in cracks as the irrigation stream arrives. Clemmens (personal communication, 1990) adds an additional term to the right hand side of Eq. (12), as a reasonable description of such a cracking soil dP - > d'T for a f (14) i· P(t)d7 For implementation in a simulation that proceeds over a series of time steps, the increase in A z over a time step is f T 8A z = A Z2 - A Z1 = 2 dz d'T P(t)d'T = j2 P(t)dz 2 (15) 21 Tj in which the subscripts 1 and 2 refer to the beginning and end of the time step, respectively. Following the rules for integration by parts, we obtain (16) With 8P=P 2 - Ph and assumed minor linear variation of z over the time interval such that f 2 1 zdP b= 3.77 -!i Zl+Z2 == Z8P = - - 8 P 2 I- a.. w ~ a ( 17) it follows that 8A z is given by the increase in depth 8z multiplied by the average wetted perimeter over the time step, i.e., combining Eqs. (16) and (17) (18) Thus, the SRFR algorithm operates on the data describing 1D infiltration to allow it to be used to approximate the results of an essentially two-dimensional process. In the following, this approach is compared with theoretical solution of the two- 10.0 f-~ - +'_. ~-io" - f -+ l ~0::: ~ (a) u:: z 1.00 10.0 100. 1000. TIME (t min) l (13) For example, with an extensively cracked tight clay soil, which exhibits virtually no additional infiltration once the cracks have been filled, Eq. (12) would predict no infiltration at all as the depth and wetted perimeter at a furrow section increase with time, while Eq. (13) would respond reasonably. However, it must be noted that Eq. (13) can be valid only for increasing depth and wetted perimeter with time-A z does not decrease at a location as the irrigation-stream depth there reduces to zero in the recession process. Eq. (13) is the basis for the algorithm incorporated in SRFR, and in WinSRFR when the local wetted-perimeter option is selected, for dealing with cracking soils. To simplify the exposition, no further consideration is given herein to the contribution of a possible constant term in the cumulative-infiltration formula, which is straightforward. Eq. (12) leads to the following expression for the growth of A z over time Az = 100. E I dA z -=i·P d'T dA z dP -=i·P+cd'T d'T E 10.0 . - - - - - - - - - , - - - - - r - - - - - . - - - - - - - , - - - , K= 1 ts il 1 ~- 0= 4'/H :,j I I 5.0 1/"i -- ., , . ,f 0::: ./' 0 / -5. 0 1·~.f,H'/ 7..O~ ..~ ..E. I I U I ffi ~ I~\ -j _""\ uc.=:, j/ ; _ _-_ + j .f .f _10.0'----_ _·i---.J. 0.10 1.00 --'- 10.0 ---'-- 100. --'----l 1000. TIME (min) Fig. 10. Modified Kostiakov formula fitted to solution of Richards equation (HYDRUS-2D: Simunek et al. 1999) in a border strip (1D flow). Silt loam soil at initial water content of 0.24. (a) fit; (b) error. dimensional Richards equation applied to a single cross section to yield cumulative volume infiltrated per unit length of furrow as a function of time (Perea et aL 2003). First, an initial HYDRUS-2D run was performed to define the 1D infiltration into the soiL In this case infiltration time 'T is simply equal to total time t. Depth of infiltration z was sought as a function of time t-to be fitted by the modified Kostiakov formula (19) For this scenario, in a border strip 1,000 mm wide, the soil was assumed a silt loam (from a HYDRUS-2D soils database) with a uniform initial moisture content 8m =0.24. Water depth on the border strip was assumed constant at h= 100 mm. HYDRUS-2D then calculated the moisture changes in the soil for a period of 30 h. Cumulative flux at the soil surface (in mm2 ) output by HYDRUS-2D was converted to depth (in millimeters) through division by the 1,000-mm-width of the border strip. Formula parameters, k, a, and b in Eq. (19) were selected interactively. Fig. 10(a) illustrates the fit that proved possible for the given conditions, while Fig. 1O(b) shows the difference between the two functions, in millimeters and as a percent. In this case, during the time period of interest, a fit within 1/2 mm proved possible everywhere, with the values k= 18.71 mm/h a , a=0.426, b =3.77 mm/h in Eq. (19) [see also Furman et al. (2006)]. Fig. 11 shows the results of the SRFR algorithm in a trapezoidal furrow with base width 100 mm and side slopes 1/1 H:V compared to HYDRUS at two different spacings, 1,000 and 500 mm. Initially, the SRFR algorithm fits the HYDRUS results, but at large infiltration times it inexorably underpredicts the infiltration, the effect being greater at the wider spacing. 532/ JOURNAL OF IRRIGATION AND DRAINAGE ENGINEERING © ASCE / SEPTEMBER/OCTOBER 2009