Original Research

Woody Vegetation Increases

Saturated Hydraulic Conductivity
in Dry Tropical Nicaragua
R.J. Niemeyer,* A.K. Fremier, R. Heinse, W. Chávez,
and F.A.J. DeClerck
Land conversion in the tropics from primary forest to agricultural land has
altered soil hydrologic processes. Woody vegetation is known to increase
infi ltration rates and saturated hydraulic conductivity (KS) in primary forests
Land conversion in the tropics compared with agricultural land, but it is less clear if this relationship holds
from primary forest to agriculture for a gradient of woody vegetation. In addition, the mechanisms for the
has altered soil hydrologic effect of woody vegetation on KS have yet to be fully examined. To quan-
tify the effect of woody vegetation structure on vadose zone hydrology, we
processes. We estimated
estimated KS in 15 plots across a dry tropical riparian vegetation gradient in
saturated hydraulic conductivity Nicaragua, taking into account covariates such as soil properties and live-
(KS) across a dry, tropical, riparian stock impact. Using single linear regression, we found that leaf area index
vegetation gradient in Nicaragua, (LAI) had the greatest correlation coefficient of 0.331 to KS, followed by hoof-
taking into account soil properties
print density (0.291) and clay content (0.291). Furthermore, the relationship
between LAI and KS was greater for finer soils than for coarser soils. We found
and livestock impact. We found that a forest soil had eight times more preferential flow paths than a pasture
that leaf area index (LAI) had soil, and most of these were root-initiated flow paths, suggesting a possi-
the greatest correlation to KS, ble mechanism for the positive correlation between LAI and KS. We show
followed by hoofprint density that the KS predictions with a pedotransfer function could be improved by
incorporating LAI. Our findings support the importance of preserving woody
(0.291) and clay content (0.291).
vegetation in key areas on the landscape to maintain hydrologic functions
of tropical soils and ecosystems..

Abbreviations: AIC, Akaike information criterion; BA, basal area; BD, bulk density; LAI, leaf
area index; PTF, pedotransfer function; SOM, soil organic matter.
R.J. Niemeyer, A.K. Fremier, and
F.A.J. DeClerck, College of Natural Widespread forest conversion to agriculture in the tropics alters vadose zone
Resources, Univ. of Idaho, Moscow,
ID 83844; A.K. Fremier, School of the hydrology. Forest removal and land conversion has resulted in reduced water infi ltration
Environment, Washington State Univ., (Mapa, 1995; Bruijnzeel, 2000) and saturated hydraulic conductivity (KS) (Giertz and
Pullman, WA 99164; R. Heinse, Plant,
Soil and Entomological and Sciences,
Diekkrüger, 2003; de Moraes et al., 2006; Mehta et al., 2008), which can lead to significant
Univ. of Idaho, Moscow, ID 83844; W. increases in surface runoff, erosion, peak flow in rivers, and decreased dry-season base flows
Chávez, Instituto de Manejo de Agua (Sandström, 1995; Bruijnzeel, 1989, 2004; Chaves et al., 2008), with far-reaching impacts
y Medioambiente, Gobierno Regional
del Cusco, Cusco, Peru; W. Chávez on human populations.
and F.A.J. DeClerck, Livestock and
Environmental Management Group,
Division of Research and Develop- In searching for the causes of these changes in the hydrology of the vadose zone, surface soil
ment, CATIE, 7170 Turrialba, Costa compaction and increased bulk density following tree removal (Alegre and Cassel, 1996;
Rica; and F.A.J. DeClerck, Bioversity
International, Montpellier Cedex Schack-Kirchner et al., 2007) and intensive grazing (Mehta et al., 2008) have been shown
5, France. *Corresponding author to reduce infiltration rates and KS. While these studies are informative, they often compare
([email protected]).
primary forest and agriculture. Instead of all forests being primary forest, secondary forest
is emerging as one of the dominant land cover types in the tropics (Giambelluca, 2002;
Vadose Zone J. Drigo, 2004; Cuo et al., 2008), and studies have shown that vegetation regrowth follow-
doi:10.2136/vzj2013.01.0025
Received 4 June 2013. ing pasture has significantly higher infi ltration rates and KS than the previous pasture
(Lal, 1996; Zimmermann and Elsenbeer, 2008). In addition, areas in the tropics where
agriculture occurs are complex and can include intensely managed agricultural lands as
© Soil Science Society of America well as seminatural systems such as savannah-like silvopastoral and multistrata agroforestry
5585 Guilford Rd., Madison, WI 53711 USA.
systems that contain native trees amidst pasture or crops (Harvey et al., 2006).
All rights reserved. No part of this periodical may
be reproduced or transmitted in any form or by
any means, electronic or mechanical, including
photocopying, recording, or any information sto- To better understand vadose zone hydrology changes in complex tropical landscapes, sev-
rage and retrieval system, without permission in
writing from the publisher.
eral studies to date have sought to establish a relationship between infi ltration and KS

Vadose Zone Journal

across a gradient of deforestation and land use. Giambelluca (2002) 2006; Jana and Mohanty, 2011). If woody vegetation is an impor-
showed that in deforested patches, patches that had been defor- tant factor in determining KS , it would be important to include
ested at an earlier date, and thereby had more secondary vegetation, woody vegetation in a PTF.
had higher KS than more recently deforested patches with less sec-
ondary vegetation. In an Amazonian basin, higher KS values were A better understanding of the drivers and mechanisms for differ-
found in primary forests, lower KS in banana (Musa ×paradisiaca ences in KS in complex tropical landscapes is crucial to advance
L.), secondary forest, and teak (Tectona grandis L. f.) plantations, watershed management and the restoration of soil hydrologic
and lowest KS in pastures (Zimmermann et al., 2006). In a com- functions in tropical regions. In particular, there is a need to
plex landscape in the Western Ghats of India, in two of the three further elucidate the complex interactions between woody veg-
dominate soils, KS changed across woody vegetation classes, with etation, soil properties, and land use on KS in tropical areas and
the highest in secondary forest, to degraded forest, to plantations especially in the dry tropics, which lack hydrologic studies com-
(Bonell et al., 2010). pared with the humid tropics (Mbagwu, 1997). To fi ll this gap,
we aimed to: (i) determine the effect of woody vegetation, soil
Despite these informative studies, understanding these changes in properties and livestock on KS and elucidate the interactions
infi ltration and KS in tropical landscapes with complex soil and among these factors; (ii) determine if leaf area index (LAI) could
vegetation patterns remains elusive. Bonell et al. (2010) saw KS be used to improve the prediction of KS with a PTF; and (iii)
decreasing with less woody vegetation in Alfisols and Ultisols but explore root and faunal turbation as a possible mechanism for
not in Vertisols. While Bonell et al. (2010) found that an increas- woody vegetation increasing KS .
ing relationship between KS and vegetation was specific to soil
type, Hassler et al. (2011) and Sobieraj et al. (2002) found that soil
type did not principally determine KS . They found that KS did 6 Materials and Methods
not vary across soil types with a range of clay contents in tropical Study Site
rainforests of Panama and Brazil, respectively. Thompson et al.
(2010) used both observational data from several North Carolina This study was conducted within the dry tropical watershed of Gil
field sites and data from a meta-analysis of studies around the Gonzalez River in the department of Rivas, Nicaragua (Fig. 1).
world to determine if biomass increased infi ltration capacity. They The watershed is 6700 ha and flows into Lake Nicaragua, which
concluded that biomass exerted a primary influence on the infi l- ultimately empties into the Caribbean Sea. The watershed is a
tration capacity for xeric sites, whereas soil type exerted a primary mixture of pasture, crops, and secondary forest of various tree
influence on the infi ltration capacity in mesic and hydric sites. densities. Using remotely sensed data, a classification of land uses
These studies point to the importance of soil type and woody veg- by Chávez (2011) in a portion of the watershed showed that open
etation in determining KS, but clearly the effects and interactions pastures occupy 45% of the total area, followed by silvopastoral
between these factors are still not well understood. systems with trees (31%), forests (12%), crops (5%), and the remain-
ing 7% as municipal areas and other land uses. Agricultural crops
In addition to a need to better understand the strength and include plantain (Musa spp.), rice (Oryza sativa L.), and sugar-
complex interactions between soil type and woody vegetation, cane (Saccharum officinarum L.). Primary and secondary forests
the mechanisms by which these factors affect KS are not well are most commonly found on steeply sloped areas and riparian
established. Several studies in the tropics have cited bioturbation forests along active river channels (D. Saìnchez, personal commu-
as a driving factor in soil morphological and physical properties nication, 2010).
(McDonough et al., 2000; Nkem et al., 2000). Increased macro-
porosity from bioturbation due to roots and faunal activity have Soils in the study area are a part of the Rivas complex derived from
been cited as possible mechanisms that cause high KS in forests Tertiary-age marine parent material composed of clays and sands
in the tropics (Sobieraj et al., 2002; Hassler et al., 2011), yet these of varying thickness. Vertisols, which swell during the wet season
mechanisms are rarely tested in tropical field studies. Similarly, and shrink to form deep cracks during the dry season, dominate
woody vegetation is often not incorporated in KS predictions, the landscape (D. Saìnchez, personal communication, 2010).
hydrologic models, or other predictive tools. For example, pedo- The climate of the region is characterized by a pronounced wet
transfer functions (PTFs) such as the Rosetta soft ware (Schaap season from May to November and a dry season from December
et al., 2001) are used to predict soil hydraulic properties (e.g., to April, with the majority of the 1300 mm of precipitation occur-
KS) with soil physical data but often do not include vegetation ring during the wet season (World Meteorological Organization,
metrics to make predictions. In the past, Wösten et al. (2001) 1996). Pan evaporation in a previous study at Victoria de Julio, 70
suggested that other factors that control soil structure, such as km from Rivas, was 2586 mm/yr averaged across 9 yr (van den
biotic activity, could be used to better predict KS with PTFs. Broek et al., 2001), which is probably comparable to pan evapora-
Some studies in temperate regions have shown that vegetation tion at Rivas.
can improve the ability of PTFs to predict KS (Sharma et al.,

Vadose Zone Journal p. 2 of 11

 Fig. 1. Map of Nicaragua (left) and watershed of Gil Gonzalez River in the department of Rivas, Nicaragua (right), with plot
locations (triangles) and riparian vegetation (green area)

Plot Selection Data Collection

We selected 15 plots to cover a range of woody vegetation densi- We estimated KS using a Guelph permeameter (GP) (Soilmoisture
ties, soil textures, and livestock use intensity. To fi nd plots with Equipment Corp.). To determine the appropriate number of GP
a range in woody vegetation, we processed 10- by 10-m resolu- measurements per plot, we performed a power analysis on KS data
tion Landsat images of the Gil Gonzalez watershed with the from the first three plots. Within each randomly selected 30- by
vegetation index TASSELED CAP (Crist and Cicone, 1984). 30-m plot, we established two transects separated by 10 m. Along
TASSELED CAP values represent a gradient of forest and tree each transect, we measured KS 2 m laterally on either side every
density ranging from pasture to dense, multiple-canopy forests. 4 m. We took 27 measurements in Plot 1 and 28 measurements
Because this study was done in conjunction with a riparian veg- in Plots 5 and 14. The power analysis established that 25 mea-
etation study, TASSELED CAP analysis and subsequent plots surements were sufficient to explain the variability of KS in each
were within 100 m of perennial or ephemeral streams. The plot. In the remaining 12 plots, we took 25 equally spaced (5 m)
TASSELED CAP values were separated into five ordinal bins measurements on a square grid in the 30- by 30-m plot. For each
with an equal number of plots. Areas with 30 by 30 m of the measurement and estimation of KS, we followed standard field pro-
same bin value were deemed potential plots. Eleven plots in each cedure for taking GP measurements at the soil surface (Reynolds,
bin were randomly chosen and then screened based on the fol- 1993; Soilmoisture Equipment Corp., 2008), taking necessary
lowing criteria: (i) accessible within 1 km of a road, (ii) located precautions to reduce compaction of the soil and smearing of the
outside an active river channel, and (iii) on a flat or moderate well wall (Reynolds, 1993). We took our measurements during the
plot slope (slope <10°) for hydrologic measurements. We selected rainy season to reduce the effect on KS of soil shrink–swell cycles
three plots from each bin to stratify soil texture and livestock typical of Vertisols (Messing and Jarvis, 1990). Measurements were
impact. To establish soil texture stratification in each plot, we not taken within 2 d of rainstorms with ≥2 cm precipitation to
performed field texture-by-feel tests and chose plots with a range allow the soils to approach field capacity.
of texture in each bin. While a gradient of soil textures exists in
each bin, due to forests dominating riparian areas lower in the To quantify the woody vegetation structure, we measured the
watershed with coarser soils, Bins 4 and 5 were the only bins with diameter at breast height (DBH) and LAI. The DBH of all trees
coarse soils (sand >30%). To stratify livestock impact in each bin, in the plot with DBH >10 cm were summed for a total plot basal
we chose at least one plot with no or minimal livestock impact area (BA). We took two LAI measurements with a LAI-2000
and one plot with higher intensity livestock impact. Lower bins plant canopy analyzer (Li-Cor, 1991) within each plot between
generally included less woody vegetation and higher livestock the months of March and August 2010 during both the wet and
impact, whereas higher bins included more woody vegetation and dry seasons and averaged the values for the final plot LAI value.
lower livestock impact (see LAI and hoofprint density in Table TASSELED CAP values in our study were highly correlated with
1). Consequently, Bins 1 and 2 included agriculture, pasture, and LAI (p < 0.0001 for all three TASSELED CAP indices) and BA (p
savannah-like silvopastoral systems with sparse woody vegetation < 0.0001 for two TASSELED CAP indices and p = 0.0004 for the
and no ungrazed areas, while Bins 3, 4, and 5 included grazed other); therefore, TASSELED CAP values accurately represented
silvopastoral systems with higher woody vegetation density to increased vegetation biomass, forest stand density, and forest cover
ungrazed forests. (Chávez, 2011). The initial riparian forest study did not contain
plots that fit the criteria for this study for Bin 1, so two pastures

Vadose Zone Journal p. 3 of 11

Table 1. Untransformed data for each plot including the TASSELED CAP index (bin), plot description, soil physical properties including sand and clay
contents, bulk density (BD), soil organic matter (SOM), and saturated hydraulic conductivity (KS), livestock impact determined as hoofprint density,
and vegetation variables leaf area index (LAI) and total plot basal area (BA) that were retained for single-variable stepwise regression (variables with
VIF scores <10).
Hoofprint
Plot Bin Description Sand Clay BD SOM density LAI BA KS
—————% ————— g/cm3 % prints/m2 ——m2/m2 —— log(mm/h)
1 1 pasture 6.5 (0.6)† 71.6 (2.9) 0.91 (0.03) 7.9 (0.56) 15.2 0.74 0 −0.73 (0.87)
2 1 pasture 12.5 (1.1) 56.5 (2.7) 1.27 (0.02) 8.0 (0.17) 5.1 0.74 0 −0.09 (0.60)
3 1 plantain field 14.1 (1.1) 51.8 (3.5) 1.09 (0.09) 8.6 (2.3) 0.2 4.60 0 0.75 (0.66)
4 2 pasture with small trees 17.0 (2.1) 59.9 (2.9) 1.20 (0.04) 6.0 (0.54) 22.7 1.81 0.00024 0.02 (0.77)
5 2 pasture with small trees 24.8 (6.5) 44.7 (10.6) 1.04 (0.08) 7.5 (0.93) 1.0 3.94 0.00035 0.91 (0.78)
6 2 corn field (with trees) 8.5 (0.9) 63.4 (2.4) 1.04 (0.04) 7.6 (0.69) 3.2 1.92 0.00025 0.11 (0.68)
7 3 forest-short and tall trees 18.7 (2.0) 41.8 (1.7) 1.18 (0.08) 7.0 (0.49) 2.1 3.13 0.0010 1.26 (0.40)
8 3 forest– short trees 16.0 (1.9) 62.8 (2.4) 1.12 (0.06) 8.0 (0.79) 12.9 2.90 0.00035 −0.12 (0.98)
9 3 forest-short and tall trees 10.3 (1.8) 54.9 (2.5) 1.20 (0.03) 6.8 (0.58) 0.6 3.52 0.0011 0.86 (0.42)
10 4 forest-short and tall trees 37.7 (3.8) 26.8 (1.6) 1.21 (0.04) 5.8 (0.44) 0.0 4.00 0.0016 0.93 (0.69)
11 4 forest-short and tall trees 20.2 (5.8) 55.6 (8.5) 0.87(0.04) 9.3 (0.57) 0.0 3.62 0.0023 0.91 (0.84)
12 4 forest-short and tall trees 33.2 (4.3) 41.5 (2.2) 1.17 (0.07) 5.7 (0.81) 4.2 4.24 0.0013 0.99 (0.68)
13 5 forest-short and tall trees 45.3 (15.1) 26.3 (6.8) 1.12 (0.06) 6.7 (1.2) 0.0 5.42 0.0024 1.35 (0.69)
14 5 forest-short and tall trees 15.4 (3.5) 58.1 (7.8) 0.90 (0.08) 11.0 (1.1) 0.03 5.54 0.0059 1.59 (0.77)
15 5 forest-short and tall trees 62.2 (6.0) 18.8 (2.1) 1.20 (0.04) 4.6 (0.66) 0.0 5.22 0.0031 0.97 (0.71)
† Plot-level standard deviations in parentheses except for plot-level variables (hoofprint density, LAI, and BA).

and one plantain plot were chosen for Bin 1 plots. Due to the lack (the nine 10-m-spaced locations in a square grid) for a total of 36
of woody vegetation in these three plots, the plot BA was 0 cm2 . quadrats per plot. Quadrats were placed either 1 or 2 m (chosen
To estimate LAI for pasture plots (Plots 1 and 2), we used averaged randomly) out from the soil sample location in four perpendicular
literature values for pastures in tropical regions (White and Terry, directions. We analyzed whether there were any spatial correla-
1979; Buschbacher, 1984; McWilliam et al., 1993; Meirelles et al., tions with hoofprints and KS estimates using a simple regression
2011; Kalácska et al., 2004). For the plantain field (Plot 3), we and found no statistical correlation; therefore, we summed the
regressed data from three studies of LAI vs. plantain/banana den- hoofprint data across the entire plot for the rest of the statistical
sity (Turner, 1972; Robinson and Nel, 1986; Cattan et al., 2007) analyses. We counted dung piles inside the plot to get a plotwide
and measured the plantain density in our plot to estimate LAI. dung density value. The hoofprint and dung density values are
referred to as hoofprint and dung henceforth.
We collected two sets of soil samples at each plot, both extracted
with a 100-cm3 bulk density ring (Blake and Hartge, 1986). The Pedotransfer Function and Leaf Area Index
first set contained 18 soil samples from nine equally spaced (10 m) The Rosetta PTF (Schaap et al., 2001) has been widely used to
points on a square grid at two depths (0–5 and 5–15 cm) for bulk translate soil properties into hydraulic parameters based on neural
density (BD), porosity, and void ratio. We additionally collected network analyses that maximize the use of input data (BD, sand,
18 soil samples from six equally spaced locations on a rectangular silt, and clay) to estimate soil hydraulic parameters such as KS. We
grid (5- and 10-m spacing) at three depths (0–5, 5–15, and 15–30 used sand, silt, clay, and BD data at each KS measurement point
cm) for sand, silt, and clay fractions, soil organic matter (SOM), to generate predicted KS values with the Rosetta PTF. To see if
and particle density. We determined soil texture by doing a par- Rosetta could better predict KS by including LAI, we compared
ticle size distribution with the pipette method (Gee and Bauder, predicted KS values from Rosetta with our observed KS values
1986) and the SOM by the loss-on-ignition method (Nelson and across a gradient of LAI.
Sommers, 1996).
Preferential Flow Paths
To estimate the impact of land use on KS in our plots, we counted To characterize preferential flow paths, we performed dye tracer
livestock (cow and horse) hoofprints and dung along three studies in a pasture plot with no woody vegetation (Plot 1) and a
10-m-spaced transects in each plot. We counted hoofprints by plac- forest plot (Plot 14). We used brilliant blue dye to observe prefer-
ing four 1- by 1-m quadrats at the first set of soil sample locations ential flow paths (Flury and Flühler, 1995). Both plots had high

Vadose Zone Journal p. 4 of 11

clay contents (72 and 58%, respectively) and were within 0.5 km of stepwise regressions: (i) all remaining variables (LAI, BA, clay, BD,
each other. We allowed a 100-L mix of 1.25 g/L dye to drain into a SOM, hoofprint, and dung), and (ii) the three variables in each
1- by 1-m area. We excavated the plots to a 1-m depth, taking cross- variable category (woody vegetation, soil, and livestock) with the
sections every 10 cm. All dye that reached below the bulk wetting highest single regression R2 value (LAI, clay, and hoofprint) and
front (approximately 20 cm below the surface for each plot) was interaction terms between each of those three variables. Both
documented as a preferential flow path with their total straight- procedures removed the terms with the highest p value until only
line distance from the surface. We documented the cause of the nonsignificant terms remained. We performed a bootstrap analysis
appearance as invertebrate holes (i.e., faunal turbation), fine root with the first stepwise regression to verify that the large sample size
(diameter <1 mm), small root (diameter = 1 mm–1 cm), medium (n = 376) did not influence the significance of the model. Finally,
root (diameter = 1–3 cm), large root (diameter >3 cm), soil struc- to visualize the effect of woody vegetation on KS across gradients
ture characteristic (seam, crack, etc.), or unidentifiable. in soil properties and livestock impact, we generated sets of three
single regressions of KS vs. LAI separated out into tertiles (thirds)
Data Analysis of clay and hoofprint.
Both KS and BA were logarithmically transformed for statisti-
cal analysis. Variables bound by 0 to 1 (clay, sand, silt, SOM, and
porosity) were transformed as X′ = arcsin[√(0.01X)] (Crawley, 6 Results
2007). Dung and hoofprint were transformed by taking their Factors Affecting Saturated
square root (Crawley, 2007). We took soil samples in a consistent Hydraulic Conductivity
gridded fashion: six soil sample points for BD, porosity, and void
and nine soil sample points for sand, silt, clay, and SOM. While Leaf area index had the highest single regression correlation with
each of these soil sample points was at a KS point, to have soil KS of all variables at an R 2 = 0.331 (Fig. 2A; Table 2). Basal area
data for each of the 25 KS measurements, we interpolated the had a slightly lower R2 value of 0.224, yet LAI and BA are known
sand, silt, clay, SOM, BD, void ratio, and porosity data at the plot to be highly cross-correlated. Soil texture measurements of clay
level in ArcGIS 9.3 using inverse-distance weighting. We tested and sand had lower R2 values than LAI, at 0.291 and 0.214, respec-
whether the soil data at different depths or combinations of depths tively. Both hoofprint density and dung were correlated with KS,
better predicted KS with an Akaike information criterion (AIC) with R2 of 0.291 and 0.190, respectively. Interestingly, KS showed
(Burnham and Anderson, 2004), but a depth-weighted average of a nonsignificant trend with BD (R2 = 0.005). Saturated hydraulic
all three soil sample depths (0–5, 5–15, and 15–30 cm) had the conductivity also correlated with increasing TASSELED CAP bin
lowest AIC scores. number (Fig. 2B).

For the regression analyses, we used residual plots and normal Q–Q The importance of LAI in predicting KS is further emphasized
plots to identify and remove outliers and leverage points for statis- by the nonparametric regression tree diagram (Fig. 3). An LAI of
tical analysis (Crawley, 2007), which resulted in removing seven 3.015 separated high and low KS values, and the top branches show
from the initial 383 KS measurements. We then performed initial that LAI separates most of the variability observed in the KS data.
single regressions for each variable to compare the relative effects Clay and sand separate the regression tree in both the high and
of woody vegetation, soil properties,
and livestock on KS . To further com-
pare the complex interactions among
these three factors and their effects on
KS, we performed a multiple regression
analysis. We first eliminated collinear
variables with variance inflation factor
(VIF) scores of 10 or higher (Ott and
Longnecker, 2010), which removed
sand, silt, void ratio, and porosity.
Next, to visualize complex interac-
tions among the variables and their
strength of relationship with KS , we
used a nonparametric regression tree
model to generate a tree diagram to
illustrate complex interactions among
the remaining variables (Breiman et Fig. 2. Single regression of (A) saturated hydraulic conductivity (KS) vs. leaf area index (LAI) and (B) KS
al., 1984). We performed two multiple vs. TASSELED CAP bins. Significant differences (p < 0.05) between bins indicated by letters.

Vadose Zone Journal p. 5 of 11

Table 2. Parameter estimates for predicting the saturated hydraulic con- Table 3. Results of stepwise regressions with the single-term multiple
ductivity (KS) and p and R2 values for each single regression (n = 376). regression with the lowest Akaike information criterion (AIC) and
single (leaf area index [LAI], clay content, and hoofprint density only)
Variable KS estimate R2 p and interaction terms. Parameter estimates for predicting saturated
Leaf area index 0.27 0.331 <0.001 hydraulic conductivity (KS) and p values from stepwise regression with
each variable.
Basal area 0.58 0.224 <0.001
Sand content 1.81 0.214 <0.001 Model Final terms Slope p

Silt content 1.90 0.041 <0.001 Single terms (intercept) 0.83 <0.001
Clay content −2.32 0.291 <0.001 clay −1.09 <0.001
Soil organic matter −2.57 0.015 0.016 hoofprint −0.088 0.002
Bulk density 0.40 0.005 0.164 LAI 0.12 <0.001
Porosity −27.7 0.014 0.021 Single and interaction (intercept) 0.35 0.14
terms
Void ratio −0.23 0.0076 0.091 clay −0.74 0.002
Hoofprint density −0.25 0.291 <0.001 LAI 0.17 <0.001
Dung −0.17 0.190 <0.001 LAI × hoofprint 0.26 <0.001

clay × LAI × hoofprint −0.33 <0.001

low LAI categories, with low clay and high sand being associated for their variable category (Table 2). To decipher interactions
with high KS values and high clay and low sand with low KS values. among woody vegetation, soil properties, and livestock impact,
the second stepwise regression included LAI, clay, and hoofprint
The initial stepwise regression included all seven of the single vari- as well as the two- and three-way interaction terms:
able terms with a VIF <10 (LAI, BA, clay, BD, SOM, hoofprint,
and dung). The model with the lowest AIC (−456.6) had an R 2 K S = LAI + clay + hoofprint + (LAI×clay ) + (clay × hoofprint)
of 0.395 and included the LAI, clay, and hoofprint terms, all with + (LAI× hoofprint) + (LAI×clay × hoofprint )
individual p values of <0.001 (Table 3). The next three models
with the lowest AIC scores all had four terms and included LAI, The model after removing nonsignificant terms in stepwise fashion
clay, and hoofprint in addition to BD, SOM, or dung. The single- was
term model with the lowest AIC was LAI, with a score of −420.6,
underscoring the prominence of LAI in predicting KS. The boot- K S = 0.17LAI*** − 0.74clay** + 0.26(LAI× hoofprint)***
strap method on the lowest overall AIC model (clay, LAI, and − 0.33(LAI×clay × hoofprint)** + 0.35
hoofprint) revealed only a slight deviation from the original R 2
value—from 0.395 to 0.383. Th is means that despite the high where *, **, and *** are used to show significance at the α = 0.05,
number of observations (n = 376), this multiple regression model 0.01, and 0.001 levels, respectively. This final model had an R2 of
is robust (Crawley, 2007). 0.422. In both stepwise regression models, single clay and LAI
terms remained, with LAI having a lower p value (Table 3). Even
For the other multiple stepwise regressions, we used the LAI, clay, though hoofprint as a single variable has a negative slope with KS
and hoofprint variables because they each had the highest R2 value (Table 3), the LAI × hoofprint interaction remained in the model
and had a positive slope (Table 3), indicating that LAI increases KS
despite the negative trend of hoofprint. The three-way interaction
also remained, with a negative slope indicating a complex interac-
tion among the three primary variables.

The multiple stepwise regressions revealed complex interactions

among LAI, clay, and hoofprint. To further elucidate how woody
vegetation positively affects KS, we compared regressions between
LAI and KS with clay and hoofprint broken out into thirds (ter-
tiles): clay = 15 to 43, 43 to 59, 59 to 75%; hoofprint = 0 to
0.2, 0.2 to 1.8, 1.8 to 4.8 (Fig. 4). The slope of the middle and
Fig. 3. A nonparametric regression tree model of the explanatory last tertile were significant (both p < 0.001), but the first tertile
variables after variables with variance inflation factor (VIF) >10 are
was not significant (p = 0.12). The regression line of LAI onto
removed. Leaf area index (LAI) is clearly a dominant factor correlated
with saturated hydraulic conductivity (KS). High and low LAI values KS increased with increasing clay content (Fig. 4), and the slope
(split at 3.015) separate KS into two clear bins. confidence intervals did not overlap between the 43 to 59 and

Vadose Zone Journal p. 6 of 11

 Pedotransfer Function and
Leaf Area Index

We graphed the predicted KS values from

the Rosetta PTF against the observed KS ,
with color and shape coding to indicate low
(yellow square: LAI < 2), medium (green
circle: 2 < LAI < 4), or high (blue triangle:
LAI > 4) based on LAI values (Fig. 5A).
The KS measurements in plots with low and
medium LAI predominantly overpredicted
K S 75 and 80% of the time, respectively.
Conversely, K S measurements in plots
with high LAI values were only overpre-
dicted 32% of the time, meaning that the
majority were underpredicted. The residu-
als (observed KS − predicted KS), graphed
in Fig. 5B, showed a general negative bias
for K S measurements in low-LAI plots
Fig. 4. Regressions of saturated hydraulic conductivity (KS) and leaf area index (LAI), with data and a positive bias for KS measurements in
divided up into tertiles (thirds) of clay (top) and hoofprint density (bottom). ***Significant at p <
0.001. Slope estimates for each regression are included.
high-LAI plots. The higher percentage of
positive residuals for plots with high LAI
values than for low LAI values suggests that
59 to 75% tertiles but did overlap between the 15 to 43 and 43 with sand, silt, clay, and BD being accounted for, LAI tends to
to 59% tertiles, signifying that the 59 to 75% tertile slope was increase KS .
significantly different than the slopes of the other two tertiles.
Th is shows that the positive relationship between LAI and KS Preferential Flow Path Analysis
varies with clay content. Conversely, the slopes of the KS vs. LAI The dye study revealed a clear difference in preferential flow paths
regression line in the hoofprint tertiles changed minimally across in the forest compared with the pasture (Table 4). In the pasture,
the different tertiles (Fig. 4), and the slope confidence intervals we found six dyed preferential flow paths below the initial wetting
overlapped for all three tertiles. front dye line, while in the forest we found 51 dyed preferential flow
paths past the wetting front (Fig. 6). The pasture plot contained
three preferential flow paths due to soil structure characteristics,

Fig. 5. (A) Predicted saturated hydraulic conductivity (KS) from Rosetta pedotransfer function vs. observed KS and
from our study, with the leaf area index (LAI) of each data point signified by color and shape (yellow squares: LAI < 2;
green circles: 2 < LAI < 4; blue triangles: LAI > 4). Points where Rosetta overpredicted the KS value are above the 1:1
line and where it underpredicted they fall below the 1:1 line; and (B) residuals (residuals = observed KS − predicted
KS) plotted against LAI, with negative residuals below the 1:1 line and positive residuals above the 1:1 line.

Vadose Zone Journal p. 7 of 11

Table 4. Number of preferential flow paths evidenced by dye presence and Diekkrüger (2003) similarly found that soils with woody
past the dyed wetting front and quantities counted of each preferential vegetation can maintain a relatively high KS despite moderate agri-
flow path. cultural use. Contrarily, in the multiple regression the three-way
Flow path count interaction term clay × LAI × hoofprint was significant with a
Cause of dye appearance Pasture plot (Plot 1) Forest plot (Plot 14) negative slope, indicating that with high clay and hoofprint density,
there is a trend to decreasing KS despite increasing LAI. During
—————no. —————
heavy rain events where ponding and overland flow occurred in
Soil structure 3 15
pastures, water ponding was also observed in heavily used livestock
Invertebrate hole 2 1
paths in dense forests (R. Niemeyer, personal observation), suggest-
Fine (<1-mm) root 0 10 ing a threshold of livestock impact where high LAI forest plots do
Small (1 mm–1 cm) root 0 11 not maintain a high KS. Still, it appears that livestock impact on
Medium (1–3-cm) root 1 9 the surface KS may be minimized if woody vegetation remains on
Large (>3-cm) root 0 3 the landscape.
Unidentifiable 0 2
Total 6 51 There was a correlation with KS and increasing clay and decreasing
sand, as has been commonly observed (Saxton et al., 1986; West
et al., 2008; Senarath et al., 2010). However, we found no signifi-
two from invertebrate holes, and one from a medium-sized root. cant relationship between bulk density and KS (Table 2). Other
The forest plot had 15 preferential flow paths from soil structure hydrologic studies in the tropics have showed a positive correlation
characteristics and one insect hole, but the majority of the prefer- between BD and reduced KS and infi ltration (Schack-Kirchner
ential flow paths were from roots: 10 from fine roots, 11 from small et al., 2007; Mehta et al., 2008). This could be due to the unique
roots, nine from medium-size roots, and three from large roots. nature of Vertisols and how land use affects their structure.

Comparing our findings with those presented by Thompson et al.

6 Discussion (2010) suggest some disagreement with the difference in the factors
Our study illustrates that the density of standing woody vegeta- that drive differences in KS. Thompson et al. (2010) concluded that
tion significantly increases KS (Fig. 2 and 5), thereby increasing biomass increases infiltration capacity in xeric sites but not in mesic
the ability of the soil to infi ltrate water and improve the hydro- or hydric sites. Because our study was on a xeric site according to
logic functions of the soil at our sites. We also observed that the their classification (pan evaporation/precipitation > 1), our study
effect of woody vegetation on KS is greater in fine-textured soils confirms their results that biomass will increase the infi ltration
than in coarse-textured soils. The regression slopes of LAI and KS capacity in xeric sites; however, our study also reveals a more nuanced
increased for the two finer soil tertiles (Fig. 4), and the regression relationship between KS–infiltration capacity, woody vegetation–
slopes for those two tertiles were significant. The coarse-soil tertile biomass, and soil texture. In their study, the xeric sites were mostly
was not significant. While we could not find low LAI, high sand fine soils: of the 92 xeric sites with soil data in their study, 43% had
plots, however, our results still suggest that LAI has little
or no effect on KS in coarse soils but does have an increas-
ing effect on finer soils. While coarse soils in general show
a higher KS than fine-textured soils (Saxton et al., 1986;
West et al., 2008; Senarath et al., 2010), our results sug-
gest that woody vegetation is a mechanism in increasing
KS in fine-textured soils where root-macropore formation
significantly increases KS compared with what would be
expected from the soil texture class alone.

The robustness of woody vegetation in maintaining high

KS across livestock impact is less clear. Some evidence sup-
ports woody vegetation maintaining a high KS despite
livestock impact. In the multiple regression, the two-way
LAI × hoofprint term was significant with a positive
slope, indicating that when LAI and hoofprint density
are both increasing, there is still an increasing relationship
with KS. Additionally, some plots (7 and 12) had higher Fig. 6. Soil profiles after dye infiltration. Plots were with 0.5 km of each other. The
LAI, moderate grazing, and still had a high KS . Giertz cause of dye presence (e.g., worm pore, seam, etc.) is identified.

Vadose Zone Journal p. 8 of 11

a high clay content (clay content >40%), and only 24% had a low
clay content (clay content <20%). Contrarily, of the 105 mesic and μm) porosities, preferential porosity had the highest correlation
hydric sites with soil data, 5% had high clay content and 71% had coefficient with KS , followed by mesoporosity. These fi ndings
low clay content. They saw no effect of biomass on infiltration in the were confi rmed by our dye study, where roots larger than fi ne
coarse-soil mesic and hydric sites and an increasing effect of woody were the source of most of the preferential flow paths in the forest
vegetation on the finer xeric sites. Our study showed that woody that had a higher KS . Th is suggests that meso- and preferential
vegetation will increase KS in fine soils but to a lesser extent in coarse porosities due to roots are primary infi ltration pathways in con-
soils (Fig. 4), so we would expect the finer soils of xeric sites to show a ducting water in the soil, thereby increasing the overall KS of
stronger relationship between biomass and infiltration capacity than the soil regardless of soil texture. Bonell et al. (2010) saw that
the coarser soils of the mesic and hydric sites. We suggest that they areas that had previously been degraded for decades to centu-
might not have found a strong relationship between biomass and ries and then converted to tree plantations had KS values that
infiltration capacity in mesic and hydric sites because of the interac- were closer to the background KS (the KS of local forests) at the
tion between soil texture and woody vegetation. Other studies have surface but not at deeper points in the soil profi le. They sug-
confirmed that for fine soils in mesic and hydric sites, woody vegeta- gested that vegetation processes mediate the surface recovery of
tion will be a greater determinant of KS than soil texture. Hassler background KS . Our dye study partially confi rmed that indeed
et al. (2011), in a tropical rainforest (mesic–hydric), saw similar KS biological processes increase preferential flow paths occurring
values across a range of soil types—meaning that woody vegetation in the forest compared with the pasture and could mediate a
was the main driver of KS, not soil texture. Similarly, Sobieraj et al. recovery of degraded landscapes to background KS .
(2002) in a tropical rainforest (mesic–hydric) did not find a correla-
tion with KS and soil type, and attributed this lack of correlation
with soil type to biotic causes increasing the macroporosity and 6 Conclusion
thereby increasing KS. Both of these studies approximated KS across Many tropical areas face a twofold pressure of scarce fi nancial
a range of soil types that were all fine (clay >25%). They confirmed resources and a need to maintain or improve water resources in
that in finer textured soils, woody vegetation can be a greater driver complex landscapes where mismanagement has caused increased
of KS than soil texture. flooding, reduced natural water fi ltration capacity, and decreased
base flows (Bruijnzeel, 2004). Our study illustrates that woody veg-
Further support that woody vegetation does have an effect on KS etation increases KS and thereby the ability of the soils to facilitate
is seen in our PTF analysis (Fig. 5B). Our results indicate that infi ltration and reduce runoff in dry, tropical Nicaragua. The cor-
including woody vegetation density or other vegetation metrics relation between woody vegetation (LAI) and KS was particularly
(e.g., LAI, biomass, etc.) in PTFs to predict KS would improve their present in finer textured soils and was robust to moderate livestock
predictive power. Currently, the Rosetta PTF does not explicitly impact but not to heavy livestock impact. A forested site had eight
include woody vegetation or other biotic parameters to predict times more preferential flow paths than a pasture site, and roots
KS. This finding is further supported by Spaeth et al. (1996), who were the primary source of preferential flow paths in the forest site,
found that the inclusion of vegetation significantly improved pre- suggesting a mechanism for increased KS in sites with woody veg-
dictions of infi ltration capacity compared with predictions solely etation. This information is helpful for managers to predict where
based on soil physical properties. It is important to bring these woody vegetation maintenance and restoration could likely have
findings into use in PTFs, which are increasingly used in research the largest impact on rehabilitating soil hydrologic functions. The
and management to predict how deforestation, agriculture, and incorporation of woody vegetation to predict KS with the Rosetta
climate change will alter hydrologic processes. While more data are PTF showed that KS prediction would be improved by including
needed to determine how woody vegetation structure affects soil woody vegetation.
hydrologic properties in the tropics, future PTFs could incorporate
some measure of woody vegetation with physical soil properties as In our study, LAI served as a good predictor of areas with high
determinants of KS in these models. and low K S based on vegetation metrics and therefore a good
predictor of which areas are more or less vulnerable to sur-
The dye study suggests that preferential flow paths are a prob- face runoff and erosion (e.g., Giertz and Diekkrüger, 2003).
able mechanism for how woody vegetation and forest structure Remote sensing (e.g., TASSELED CAP) could be an inexpen-
increases KS . Bioturbation and the resultant macroporosity from sive method to predict areas of high and low LAI and therefore
roots compared with fauna appear to be a greater source of pref- areas likely to have high or low K S . This information could
erential flow paths in the forest site; however, a more extensive be used to target high-priority areas for management efforts
study is needed to confi rm these results. Even though our study to increase or maintain woody vegetation and thereby reduce
was limited to two plots, other studies confi rm these results. surface runoff and erosion while protecting water resources in
Mbagwu (1997) found that in Nigerian dry tropical soils of tropical areas (Fremier et al., 2013).
micro- (<10 μm), meso- (10–1000 μm), and preferential (>1000

Vadose Zone Journal p. 9 of 11

References Hassler, S.K., R.M. Lark, A.E. Milne, and H. Elsenbeer. 2011. Exploring the
variation in soil saturated hydraulic conductivity under a tropical
rainforest using the wavelet transform. Eur. J. Soil Sci. 62:891–901.
Alegre, J.C., and D.K. Cassel. 1996. Dynamics of soil physical properties
doi:10.1111/j.1365-2389.2011.01400.x
under alternative systems to slash-and-burn. Agric. Ecosyst. Environ.
58:39–48. doi:10.1016/0167-8809(95)00654-0 Jana, R.B., and B.P. Mohanty. 2011. Enhancing PTFs with remotely sensed
data for multi-scale soil water retention estimation. J. Hydrol. 399:201–
Blake, G.R., and K.H. Hartge. 1986. Particle density. In: A. Klute, editor,
211. doi:10.1016/j.jhydrol.2010.12.043
Methods of soil analysis. Part 1. SSSA Book Ser. 5. SSSA and ASA, Madi-
son, WI. p. 377–382. doi:10.2136/sssabookser5.1.2ed.c14 Kalácska, M., G.A. Sánchez-Azofeifa, B. Rivard, J.C. Calvo-Alvarado,
A.R.P. Journet, J.P. Arroyo-Mora, and D. Ortiz-Ortiz. 2004. Leaf area in-
Bonell, M., B.K. Purandara, B. Venkatesh, J. Krishnaswamy, H.A.K. Acha-
dex measurements in a tropical moist forest: A case study from Costa
rya, U.V. Singh, et al. 2010. The impact of forest use and reforesta-
Rica. Remote Sens. Environ. 91:134–152. doi:10.1016/j.rse.2004.02.011
tion on soil hydraulic conductivity in the Western Ghats of India: Im-
plications for surface and sub-surface hydrology. J. Hydrol. 391:47–62. Lal, R. 1996. Deforestation and land-use effects on soil degradation
doi:10.1016/j.jhydrol.2010.07.004 and rehabilitation in western Nigeria: I. Soil physical and hydrologi-
cal properties. Land Degrad. Dev. 7:19–45. doi:10.1002/(SICI)1099-
Breiman, L., J.H. Freidman, R.A. Olshen, and C.J. Stone. 1984. Classifica-
145X(199603)7:1<19::AID-LDR212>3.0.CO;2-M
tion and regression trees. Wadsworth, Monterey, CA.
Li-Cor. 1991. LAI-2000 plant canopy analyzer: Instructional manual. Li-Cor,
Bruijnzeel, L.A. 1989. Deforestation and dry season flow in the tropics: A
Lincoln, NE.
closer look. J. Trop. For. Sci. 1:229–243.
Mapa, R.B. 1995. Effect of reforestation using Tectona grandis on in-
Bruijnzeel, L.A. 2000. Forest hydrology. In: J.C. Evans, editor, The forests
filtration and soil water retention. For. Ecol. Manage. 77:119–125.
handbook. Blackwell Scientific, Oxford, UK. p. 301–343.
doi:10.1016/0378-1127(95)03573-S
Bruijnzeel, L.A. 2004. Hydrological functions of tropical forests: Not seeing
Mbagwu, J.S.C. 1997. Quasi-steady infiltration rates of highly permeable
the soil for the trees? Agric. Ecosyst. Environ. 104:185–228. doi:10.1016/j.
tropical moist savannah soils in relation to landuse and pore size dis-
agee.2004.01.015
tribution. Soil Technol. 11:185–195. doi:10.1016/S0933-3630(96)00138-9
Burnham, K.P., and D.R. Anderson. 2004. Multimodel inference: Under-
McDonough, C.M., M.J. DeLaney, P. Quoc Le, M.S. Blackmore, and W.J.
standing AIC and BIC in model selection. Sociol. Methods Res. 33:261–
Loughry. 2000. Burrow characteristics and habitat associations of
304. doi:10.1177/0049124104268644
armadillos in Brazil and the United States of America. Rev. Biol. Trop.
Buschbacher, R.J. 1984. Changes in productivity and nutrient cycling fol- 48:109–120.
lowing conversion of Amazon rainforest to pasture. Ph.D. diss. Univ. of
McWilliam, A.L.C., J.M. Roberts, O.M.R. Cabral, M. Leitao, A.C.L. De Costa,
Georgia, Athens.
G.T. Maitelli, and C. Zamparoni. 1993. Leaf area index and above-
Cattan, P., F. Bussiere, and A. Nouvellon. 2007. Evidence of large rainfall ground biomass of Terra Firme Rain Forest and adjacent clearings in
partitioning patterns by banana and impact on surface runoff gen- Amazonia. Funct. Ecol. 7:310–317. doi:10.2307/2390210
eration. Hydrol. Processes 21:2196–2205. doi:10.1002/hyp.6588
Mehta, V.K., P.J. Sullivan, M.T. Walter, J. Krishnaswamy, and S.D. DeGloria.
Chaves, J., C. Neill, S. Germer, S.G. Neto, A. Krusche, and H. Elsenbeer. 2008. Impacts of disturbance on soil properties in a dry tropical forest
2008. Land management impacts on runoff sources in small Amazon in southern India. Ecohydrology 1:161–175. doi:10.1002/eco.15
watersheds. Hydrol. Processes 22:1766–1775. doi:10.1002/hyp.6803
Meirelles, M.L., A.C. Franco, S.E.M. Farias, and R. Bracho. 2011. Evapo-
Chávez, W. 2011. Functional diversity and the absorption capacity of transpiration and plant–atmospheric coupling in a Brachiaria brizan-
riparian forests of the Gil Gonzalez watershed in the department of tha pasture in the Brazilian savannah region. Grass Forage Sci. 66:206–
Rivas. Centro Agronómico Tropical de Investigación y Enseñanza, Tur- 213. doi:10.1111/j.1365-2494.2010.00777.x
rialba, Nicaragua.
Messing, I., and N.J. Jarvis. 1990. Seasonal variation in field-saturated
Crawley, M.J. 2007. The R book. John Wiley & Sons, Chichester, UK. hydraulic conductivity in two swelling clay soils in Sweden. J. Soil Sci.
41:229–237. doi:10.1111/j.1365-2389.1990.tb00059.x
Crist, E.P., and R.C. Cicone. 1984. Application of the tasseled cap con-
cept to simulated thematic mapper data. Photogramm. Eng. Remote Nelson, D.W., and L.E. Sommers. 1996. Total carbon, organic carbon, and
Sens. 50:343–352. organic matter. In: D.L. Sparks, editor, Methods of soil analysis. Part 3.
SSSA Book Ser. 5. SSSA and ASA, Madison, WI. p. 961–1010. doi:10.2136/
Cuo, L., T.W. Giambelluca, A.D. Ziegler, and M.A. Nullet. 2008. The roles
sssabookser5.3.c34
of roads and agricultural land use in altering hydrological processes
in Nam Mae Rim watershed, northern Thailand. Hydrol. Processes Nkem, J.N., L.A. Lobry de Bruyn, C.D. Grant, and N.R. Hulugalle. 2000.
22:4339–4354. doi:10.1002/hyp.7039 The impact of ant bioturbation and foraging activities on surround-
ing soil properties. Pedobiologia 44:609–621. doi:10.1078/S0031-
de Moraes, J.M., A.E. Schuler, T. Dunne, R.O. Figueiredo, and R.L. Victoria.
4056(04)70075-X
2006. Water storage and runoff processes in plinthic soils under for-
est and pasture in eastern Amazonia. Hydrol. Processes 20:2509–2526. Ott, R.L., and M. Longnecker. 2010. An introduction to statistical methods
doi:10.1002/hyp.6213 and data analysis. 6th ed. Brooks/Cole, Belmont, CA.
Drigo, R. 2004. Trends and patterns of tropical land use changes. In: M. Reynolds, W.D. 1993. Saturated hydraulic conductivity: Field measure-
Bonell and L.A. Bruijnzeel, editors, Forests, water and people in the ment. In: M.R. Carter, editor, Soil sampling and methods of analysis.
humid tropics: Past, present and future hydrological research for inte- Lewis Publ., Boca Raton, FL. p. 599–613.
grated land and water management. UNESCO Int. Hydrol. Ser. Cam-
Robinson, J.C., and D.J. Nel. 1986. The influence of banana (cv. Williams)
bridge Univ. Press, Cambridge, UK. p. 9–39.
plant density and canopy characteristics on ratoon cycle interval
Flury, M., and H. Flühler. 1995. Tracer characteristics of brilliant blue FCF. Soil and yield. Acta Hortic. 175:227–232 (ISHOS).
Sci. Soc. Am. J. 59:22–27. doi:10.2136/sssaj1995.03615995005900010003x
Sandström, K. 1995. Forest and water: Friends or foes? Hydrological impli-
Fremier, A.K., F.A.J. DeClerck, N.A. Bosque-Pérez, N.E. Carmona, R. Hill, cations of deforestation and land degradation in semi-arid Tanzania.
T. Joyal, et al. 2013. Understanding spatiotemporal lags in ecosystem Ph.D. diss. Linkoping Univ., Linkoping, Sweden.
services to improve incentives. BioScience 63:472–482. doi:10.1525/
Saxton, K.E., W.J. Rawls, J.S. Romberger, and R.I. Papendick. 1986. Estimat-
bio.2013.63.6.9
ing generalized soil-water characteristics from texture. Soil Sci. Soc.
Gee, G.W., and J.W. Bauder. 1986. Particle-size analysis. In: A. Klute, edi- Am. J. 50:1031–1036. doi:10.2136/sssaj1986.03615995005000040039x
tor, Methods of soil analysis. Part 1. 2nd ed. SSSA Book Ser. 5. SSSA and
Schaap, M.G., F.J. Leij, and M.Th. van Genuchten. 2001. Rosetta: A com-
ASA, Madison, WI. p. 383–411. doi:10.2136/sssabookser5.1.2ed.c15
puter program for estimating soil hydraulic parameters with hierarchi-
Giambelluca, T.W. 2002. Hydrology of altered tropical forest. Hydrol. Pro- cal pedotransfer functions. J. Hydrol. 251:163–176. doi:10.1016/S0022-
cesses 16:1665–1669. doi:10.1002/hyp.5021 1694(01)00466-8
Giertz, S., and B. Diekkrüger. 2003. Analysis of the hydrological processes Schack-Kirchner, H., P.T. Fenner, and E.E. Hildebrand. 2007. Different re-
in a small headwater catchment in Benin (West Africa). Phys. Chem. sponses in bulk density and saturated hydraulic conductivity to soil
Earth 28:1333–1341. doi:10.1016/j.pce.2003.09.009 deformation by logging machinery on a Ferralsol under native forest.
Soil Use Manage. 23:286–293. doi:10.1111/j.1475-2743.2007.00096.x
Harvey, C.A., A. Medina, D.M. Sanchez, S. Vilchez, B. Hernandez, J.C.
Saenz, et al. 2006. Patterns of animal diversity in different forms of Senarath, A., A.S. Palmer, and R.W. Tillman. 2010. Soil spatial variability of
tree cover in agricultural landscapes. Ecol. Appl. 16:1986–1999. drainage properties in relation to phosphate retention and mineral-
doi:10.1890/1051-0761(2006)016[1986:POADID]2.0.CO;2 ogy on a river terrace of northern Manawatu, New Zealand. Aust. J.
Soil Res. 48:69–76. doi:10.1071/SR09042

Vadose Zone Journal p. 10 of 11

Sharma, S.K., B.P. Mohanty, and J.T. Zhu. 2006. Including topography and West, L.T., M.A. Abreu, and J.P. Bishop. 2008. Saturated hydraulic con-
vegetation attributes for developing pedotransfer functions. Soil Sci. ductivity of soils in the southern Piedmont of Georgia, USA: Field
Soc. Am. J. 70:1430–1440. doi:10.2136/sssaj2005.0087 evaluation and relation to horizon and landscape properties. Catena
73:174–179. doi:10.1016/j.catena.2007.07.011
Sobieraj, J.A., H. Elsenbeer, R.M. Coelho, and B. Newton. 2002. Spatial
variability of soil hydraulic conductivity along a tropical rainforest cat- White, L.D., and W.S. Terry. 1979. Creeping bluestem response to pre-
ena. Geoderma 108:79–90. doi:10.1016/S0016-7061(02)00122-2 scribed burning and grazing in South Florida. J. Range Manage.
32:369–371. doi:10.2307/3898018
Soilmoisture Equipment Corp. 2008. 2800K1 Operating instructions. Soilm-
oisture Equipment Corp., Santa Barbara, CA. World Meteorological Organization. 1996. Climatological normals (CLI-
NO) for the period 1961–1990. WMO, Geneva, Switzerland.
Spaeth, K.E., F.B. Pierson, M.A. Weltz, and J.B. Awang. 1996. Gradient
analysis of infiltration and environmental variables as related to range- Wösten, J.H.M., Y.A. Pachepsky, and W.J. Rawls. 2001. Pedotransfer func-
land vegetation. Trans. ASAE 39:67–77. doi:10.13031/2013.27481 tions: Bridging the gap between available basic soil data and miss-
ing soil hydraulic characteristics. J. Hydrol. 251:123–150. doi:10.1016/
Thompson, S.E., C.J. Harman, P. Heine, and G.G. Katul. 2010. Vegetation–
S0022-1694(01)00464-4
infiltration relationships across climatic and soil type gradients. J. Geo-
phys. Res. 115:G02023. doi:10.1029/2009JG001134 Zimmermann, B., and H. Elsenbeer. 2008. Spatial and temporal variability
of saturated hydraulic conductivity in gradients of disturbance. J. Hy-
Turner, D.W. 1972. Banana plant growth: 2. Dry matter production, leaf
drol. 361:78–95. doi:10.1016/j.jhydrol.2008.07.027
area and growth analysis. Aust. J. Exp. Agric. Anim. Husb. 12:216–224.
doi:10.1071/EA9720216 Zimmermann, B., H. Elsenbeer, and J.M. de Moraes. 2006. The influence
of land-use changes on soil hydraulic properties: Implications for
van den Broek, R., L. Vleeshouwers, M. Hoogwijk, A. van Wijk, and W.
runoff generation. For. Ecol. Manage. 222:29–38. doi:10.1016/j.fore-
Turkenburg. 2001. The energy crop growth model SILVA: Description
co.2005.10.070
and application to eucalyptus plantations in Nicaragua. Biomass Bio-
energy 21:335–349. doi:10.1016/S0961-9534(01)00032-0

Vadose Zone Journal p. 11 of 11