Evaluation of SOILWAT Model for Predicting Soil Water Characteristics in Southwestern Nigeria
Investigations of soil water potential and its variability with soil characteristics are necessary for studying water availability for plants, plant water stress, infiltration, irrigation scheduling, drainage and water conductivity. However, measurements of soil water characteristics are difficult, costly and time consuming. A modified version of the SOILWAT model (2006) of the soil water characteristics was adapted to 50 (replicated thrice) soil samples collected at depths 0-60 cm from upper, middle and lower slopes in 12 dug profile pits in 5 states of Southwestern Nigeria. Samples were assessed for their physical, hydrological and chemical parameters in the laboratory and compared with the simulated values obtained with the SOILWAT model. Coefficient of determination (R2) as a goodness-of-fit index of agreement and the Root Mean Square Error (RMSE) were computed. The soils textural class varied from Sand, Sandy loam, Loamy sand and Sandy clay loam. There was good agreement between the measured and simulated bulk density for Sandy loam texture class (R2 = 0.722; RMSE = 0.476), saturated moisture content (R2 = 0.544; RMSE = 30.135), field capacity (R2 = 0.770; RMSE = 86.877) and available water content (R2 = 0.547; RMSE = 0.940) at p<0.05 probability level. However, a poor fit was observed for saturated hydraulic conductivity and permanent wilting point. SOILWAT had a fairly good prediction of bulk density for a Sand texture class (R2 = 0.460, p<0.05, RMSE = 0.476) but a poor fit for saturated hydraulic conductivity, saturated moisture content, field capacity, permanent wilting point and available water content. However, it had the smallest RMSE when compared to the other texture class.
Received: September 29, 2012;
Accepted: March 08, 2013;
Published: July 22, 2013
Soil water is defined as the infiltrated water shallow enough to be used by
plants (Kern, 1995). Soil water characteristics is
dependent on the soil water retention B soil water potential, which is necessary
for studying water availability for plants, plant water stress, infiltration,
irrigation scheduling, drainage and water conductivity. The distribution of
water within the soil column will be an indispensable factor in understanding
the response of plants and soil water systems to the impacts of climate change
(Walczak et al., 2002).
Measurement and analyses of the soil water characteristics are difficult, costly
and time consuming. Hence, the use of expensive special equipment becomes necesssary.
Several research studies over last three decades has formulated models, which
enables its determination on the basis of measured soil physical and chemical
properties, which serves as inputs (Rajkai and Varallyay,
1989; Williams et al., 1992; Saxton
and Rawls, 2006).
These models are referred to as pedotransfer functions (PTFs) (Bouma,
1989). Soil water modelling is defined as the dynamic simulation of hydrologic
processes by numerical integration of individual processes with the aid of computer
(Saxton et al., 1986). A better understanding
of agricultural water management and hydrological analyses to a form a reliable
predictive soil water characteristics system will be dependent upon simulation
modelling (Saxton and Rawls, 2006).
The Soil and Water Assessment tool (SOILWAT) is a modelling software package
to analyse water, soil, agriculture and nutrient interactions at catchment modelling.
This is a technology used to construct a relatively transparent surrogate (substitute)
for the real soil water, then combined into a more comprehensive results and
analysed by statistics which can be manipulated with far greater ease than the
complex original (Saxton and Rawls, 2006).
Soil water retention as a function of sand and clay textures and organic matter
is described by a set of generalised equations (Rawles and
Brakensiek, 1982; Rawls et al., 1982; Saxton
et al., 1986; Saxton and Rawls, 2006). Gijsman
et al. (2002) reported an extensive review of eight modern estimating
methods applicable to hydrologic and agronomic analyses. They observed significant
discrepancy among the methods due to the regional data basis or methods of analyses
thus creating doubt on the value of lab-measured water retention data for crop
models. They concluded that an analysis with asset of field-measured data showed
that the method of Saxton et al. (1986) performed
the best. Thus, an enhancement of the Saxton and Rawls (2006)
method is an appropriate extension to improve the field applications of soil
water characteristic estimates with improved data basis and supplemented by
recently derived relationships of conductivity and including appropriate local
adjustments for organic matter, density, gravel and salinity.
A soil-water (SOILWAT) model capable of simulating soil hydrological properties
of soil texture will help scientist in providing crucial data set for better
understanding of our soils for better management. These data are also critical
requirements in crop simulation models for decision making aimed at obtaining
optimum results. Xue et al. (1996) compared
soil moisture observations with modeling results, he reported that the soil
hydraulic parameters have a profound impact on the model simulations. The objective
of this study was to compare the prediction made by SOILWAT model with the measured
soil parameters and evaluate the general applicability and prediction accuracy
of SOILWAT model for the predominant soil types in the derived savannah of Southwestern
MATERIALS AND METHODS
Soil and water assessment tool (SOILWAT model): Soil water characteristics
hydraulic properties calculator developed by Keith Saxton in cooperation with
Department of Biological Systems Engineering Washington State University. The
soil water characteristics equations are valid within a range of soil textures
approximately 0-60% clay content and 0-95% sand content. Adjustments to the
solutions have been added to include the effects of bulk density, gravel and
salinity (Tanji, 1990). A programmed texture triangle
as an input screen (Fig. 1) provides a ready solution to the
equations and values for the layer definitions of the soil profile. This methodology
is incorporated in the model and is also available as a stand-alone program.
||Graphical input screen for the soil water characteristic model
Soil sample strategy: Undisturbed samples (50) were collected in triplicates
(150) using a 98.17 cm3 core sampler at depths of 0-60 cm from twelve
dug mini profiles pits of 45x45 cm at the surface and 80 cm deep. Topographic
landscapes at upper, middle and lower slopes in grassland and farmland vegetations
in 5 states (Lagos: Iweka Series, Ogun: Alagba Series, Oyo: Iwo and Apomu Series,
Osun: Itagunmodi Series, Ondo: Owo Series in South-Western Nigeria; Guinea Savannah
zone classified as an Alfisols (USDA) order. The soils are indicated as Typic
Tropoquent (Smyth and Montgomery, 1962) or Euteric Gleysol
(FAO, 1983) and Ferric luvisol and Oxic Rhodustalf (FAO,
1988). Coordinates and elevation above sea level of the sites were obtained
Determination of soil solid and liquid phase of soil water potential-soil water
content characteristics were performed on soil samples using standard procedures
(USDA/SCS, 1982). Measurement of particle size distribution,
bulk densities, gravimetric moisture content, hydraulic conductivities, lower
and upper limit of soil potentials. Chemical analyses of organic carbon content
were also determined. Soil gravel content was separated and weighed.
Soil properties: The physical and chemical properties measured are presented in Table 1. The data presented were of soil properties values ranging from the least to the highest values obtained for each soil characteristics.
Experimental setup and textural class: Particle size analyses of the
soils obtained from the experimental sites indicates a sand, loamy sand, sandy
loam and sandy clay loam textural classes. This variation in textures was used
as the basis for grouping the soils and subsequently, for easy computation of
the data sets for verification by the model; since the texture predominately
determines the water holding characteristics of most agricultural soils reported
by Saxton et al. (1986). A greater number of
the sampled soils were of loamy sand texture with sandy loam, sandy clay loam
and sand textural classes in that order as depicted in Table 2.
|| Map showing the sampled locations (Dept. of Geography, University
of Ibadan, Nigeria)
|| Textural classes and their mean particle contents
|Values are Mean±SD. USDA: United State Department of
Statistical analysis: For comparison of the difference between predicted
soil water characteristic parameters and observed values, coefficient of determination
(R2) as a goodness-of-fit index of agreement and the Root Mean Square
Error (RMSE) were computed. Willmott (1981) described
RMSE as among the best overall measures of model performance, of
which RMSE is more sensitive to extreme values due to its exponentiation; it
therefore can be considered as a high estimate of the actual average error.
The index of agreement is a standardized measure (scale 0-1) of the degree to
which a models predictions are error free. yi denotes the measured
the average of the measured value and N is the total number of observations:
RESULTS AND DISCUSSION
Correlation between measured vs. simulated soil water characteristics:
A comparison is presented between measured and simulated water content for the
50 soil samples using the Saxton and Rawls (2006) model.
The statistical results are presented in Table 3.
|| Linear regression of the texture class as related to the
soil water characteristics
|R2: Coefficient of determination, RMSE: Root mean
square, Ks: Saturated hydraulic conductivity, Sat.MC: Saturated moisture
content, FC: Field capacity, PWP: Permanent wilting point, S: Sand, Lsa:
Loamy sand, SaL: Sandy loam, SaCL: Sandy clay loam, ns: Non significant,
*Significant at 5% probability level
Amongst the four textural classes, the SOILWAT provided a superior estimation
of all the soil water characteristics for the sandy loam than for sand, loamy
sand and sandy clay loam soils. A better fit for bulk density and field capacity
between the simulated and measured soil water characteristics relationships
was observed in sandy loam texture as depicted in Fig. 3.
The SOILWAT model produced a fairly good prediction of bulk density for sand soil with a R2 of 0.460 and RMSE of 0.411. The results showed a lower RMSE for sand soil which ranged from 0.411 to 54.81 when compared to sandy loam which had the better prediction (higher R2) of all the textural class with RMSE values, which ranged from 0.476 to 86.877 as presented in Table 3.
Bulk density: When compared with the measured data, SOILWAT significantly
simulated Bulk Density (BD) for Sandy loam (R2 = 0.722, p<0.05;
RMSE of 0.476 and regression equation given as; Y = 0.754x+0.451) as presented
in Table 3. The relationship between the simulated and measured
values showed that 72.2% of the fluctuation in the measured values was explained
by the model; which indicated a good prediction as shown in Fig.
3. For sandy soil, SOILWAT had a fairly good simulation (R2 =
0.460, p<0.05; RMSE of 0.411). While, for Loamy sand and Sandy clay loam
soils, it had a poor-fit. The concentration of the sand soil in one site i.e.,
non-uniformity of the distribution of sand texture class amongst the sites may
be one among other factors responsible for the fairly good-fit between measured
and simulated data since the variation in the textural class amongst the sites
was defeated. Gijsman et al. (2002) reported
that the SOILWAT, though, performed best among other models compared in his
studies, but this does not apply to all soils. For very sandy soils, no method
performed well. The high amount of coarse-size particles in the sandy soils
is possibly the reason for the fairly good-fit by the model (Fredlund
et al., 2002; Hwang and Powers, 2003).
Saturated hydraulic conductivity: Loamy sand and Sandy loam soils had
a low R5 and a negative regression equations (R5 = 0.025, y = -0.023x+6.939,
RMSE = 166.16 and R5 = 0.135, y = -0.148x+8.532, RMSE = 78.486, respectively)
when simulated with the model as presented in Table 3. For
Sand and Sandy clay loam, it had no correlation (R2 of 0.000 and
RMSEs of 33.25 and 104.69, respectively). Since sample environment (confinement
and overburden) are not represented in laboratory procedures, laboratory data
may not always agree with field data. The disagreement appears more pronounced
at high water contents (Arya and Dierolf, 1992).
Saturated moisture content: The model had a fairly good-fit for sandy loam with R2 of 0.544; RMSE of 30.460. While, sand, sandy clay loam and loamy sand had a poor-fit (R2 of 0.259, 0.228 and 0.191, respectively) as depicted in Table 3.
Field capacity: SOILWAT predicted field capacity significantly for Sandy
loam soil (R2 = 0.770, p<0.05; RMSE of 86.877). This indicted
that 77% of the variation in the measured values is explained by the model as
indicted by the relationship the between simulated and measured values and thus
represented a good-fit (Fig. 3). But it had a poor-fit for
Loamy sand (R2 = 0.181), sandy clay loam (R2 = 0.157)
and R2 = 0.167 for sand soil which also presented a negative regression
as showed in Table 3. Tomasella and Hodnett
(1998) reported that, in many cases, the textures of many tropical soils,
particularly oxisols such as those of Brazilian Amazonia, are outside the range
of validity of these PTFs. As a result, the water retention estimations may
also be significantly in error or they may fail totally, for example by indicating
a water content at field capacity that is higher than at saturation.
||Relationship between simulated vs. measured soil water characteristics
of Sandy loam soil, (a) Bulk density, (b) Field capacity, (c) Saturated
hydraulic conductivity, (d) Permanent wilting point, (e) Saturated moisture
content and (f) Available water content
This occurs using the PTF of Mishra et al. (1989)
for oxisols from Nigeria and Brazil with clay contents of 52 and 85%, respectively.
The predicted moisture retention curve often falls to zero volumetric water
content before the experimental data are complementary desaturated (Fredlund
et al., 1997).
Permanent wilting point: A poor-fit with SOILWAT was achieved in all
the texture classes (Table 3). (Nagpal
and De Vries, 1976; Arya and Dierolf, 1992) reported
that laboratory data may not always agree with field data, probably as a result
of sample size as observed in the laboratory measurements. The results of the
study showed that the model had a poor-fit for the soil types at low soil water
content. However, Arya and Dierolf, 1992 reported that
the disagreement appears more pronounced at high water contents.
Available water content: Sandy loam recorded a fairly good-fit when simulated (R2 of 0.547 with RMSE of 0.940). A poor-fit of the model was observed for sandy clay loam, loamy sand and sand texture classes. Comparison of the measured to the simulated values shows a wide variation in the textural classes analysed as presented in Table 3.
Validation of the SOILWAT goodness-of-fit for soil water characteristics:
For the 50 samples, the soil water characteristics were predicted using SOILWAT
(Saxton and Rawls, 2006) model and the goodness-of-fit
measurement obtained from the R2 value. SOILWAT was found to provide
a reasonable estimate of the soil water characteristics for sandy loam soils.
There appears to be greater difficulty in estimating the soil water characteristics
for loamy sand soils, sandy clay loam soils and sand soils.
It has been previously noted that it is particularly difficult to estimate
the soil water characteristics from particle-size distribution for some texture
class. The general soil categories include: (i) soils that have high amount
of coarse-size particles, mixed with few fines and (ii) soils that have high
amount of clay size particles (Fredlund et al., 2002;
Hwang and Powers, 2003). The same trend was found to
be true for the experimental/sampled soils. Sampling error or bias in field
sampling e.g., sample size, presence of roots and gravel particles which serves
as obstacle in obtaining undisturbed samples. The model had a tolerable range
of 0-60% for gravel content.
The SOILWAT model resulted in higher coefficient of determination for Sandy loam soil which expresses the goodness-of-fit between the simulated and measured values. However, the poor-fit measurement of the model for sand, loamy sand and sandy clay loam might be as a result of the sensitivity of the model in terms of location or site-specific and the high gravel content of the sampled soils which made up this textural class. The SOILWAT model has a tolerable range of 0-60%. The ability of SOILWAT to simulate soil water characteristics for sandy loam soil demonstrates the potential of the model when properly initialised and field measurement accurately taken. SOILWAT has shown the potential of serving as tool that would enable decision makers to explore the future of sustainable agriculture, even in developing countries where soil water extraction apparatus have become a limitation in determining soil water availability. Despite the optimistic position of system modelling, realisation of the full potential depends considerably on availability and quality of inputs for running the model, taken into consideration location or site-specific information in developing the model.
1: Arya, L.M. and T.S. Dierolf, 1992. Predicting soil moisture characteristics from particle-size distribution: An improved method to calculate pore radii. Proceedings of the International Worhshop on Indirect Methods of Estimating the Hydraulic Properties of Unsaturated Soils, 0ctober 11-13, 1989, Riverside, CA., USA., pp: 115-124
2: Bouma, J., 1989. Using soil survey data for quantitative land evaluation. Adv. Soil Sci., 9: 177-213.
3: FAO., 1983. Guidelines: Land evaluation for rainfed agriculture. FAO Soils Bulletin No. 52, Food and Agriculture Organization of the United Nations, Rome.
4: FAO, 1988. Soil Map of the World. Revised Legend, Reprinted with Corrections. World Soil Resources Report 60, FAO, Rome
5: Fredlund, M.D., D.G. Fredlund and G.W. Wilson, 1997. Prediction of the soil-water characteristic curve from grain-size distribution and volume-mass properties. Proceedings of the 3rd Brazilian Symposium on Unsaturated Soils, April 22-25, 1997, Rio de Janeiro, Brazil, pp: 13-23
6: Fredlund, M.D., G.W. Wilson and D.G. Fredlund, 2002. Use of grain size distribution for estimation of the soil water characteristics curve. Can. Geotech. J., 39: 1103-1117.
Direct Link |
7: Gijsman, A.J., S.S. Jagtap and J.W. Jones, 2002. Wading through a swamp of complete confusion: How to choose a method for estimating soil water retention parameters for crop models. Eur. J. Agron., 18: 75-105.
8: Hwang, S.I. and S. Powers, 2003. Using particle-size distribution models to estimate soil hydraulic properties. Soil Sci. Soc. Am. J., 67: 1103-1112.
9: Kern, J.S., 1995. Evaluation of soil water retention models based on basic soil physical properties. Soil Sci. Soc. Am. J., 59: 1134-1141.
10: Mishra, S., J.C. Parker and N. Singhal, 1989. Estimation of soil hydraulic properties and their uncertainty from particle size distribution data. J. Hydrol., 108: 1-18.
11: Nagpal, N.K. and J. De Vries, 1976. An evaluation of the instantaneous profile method for in situ determination of hydrologic properties of layered soil. Can. J. Soil Sci., 56: 453-461.
Direct Link |
12: Rajkai, K. and G. Varallyay, 1989. Estimating soil water retention from simpler soil properties by regression techniques. Proceedings of the International Workshop on Indirect Methods for Estimating the Hydraulic Properties of Unsaturated Soils, October 11-13, 1989, Riverside, CA., USA -
13: Rawles, W.J. and D.L. Brakensiek, 1982. Estimating soil water retention from soil properties. J. Irrig. Drain. Div. ASCE, 108: 166-171.
Direct Link |
14: Rawls, W.J., D.L. Brakensiek and K.E. Saxton, 1982. Estimating soil water properties. Trans. Am. Soc. Agric. Eng., 25: 1316-1320, 1328.
15: Saxton, K.E., W.J. Rawls, J.S. Romberger and R.I. Papendick, 1986. Estimating generalized soil water characteristics from texture. Soil Sci. Soc. Am. J., 50: 1031-1036.
Direct Link |
16: Saxton, K.E. and W.J. Rawls, 2006. Soil water characteristics estimates by texture and organic matter for hydrologic solutions. Soil Sci. Soc. Am. J., 70: 1569-1578.
17: Smyth, A.J. and R.F. Montgomery, 1962. Soils and Land Use in Central Western Nigeria. Government of Western Nigeria, Ibadan, Nigeria, Pages: 265
18: Tanji, K.K., 1990. Agricultural Salinity Assessment and Management. American Society of Civil Engineers, New York, ISBN-13: 9780872627628, Pages: 619
19: Tomasella, J. and M.G. Hodnett, 1998. Estimating soil water retention characteristics from limited data in Brazilian Amazonia. Soil Sci., 163: 190-202.
Direct Link |
20: USDA/SCS, 1982. Procedures for collecting soil samples and methods of analysis for soil survey. Soil Survey Investigations Report I, Washington, DC.
21: Walczak, R., B. Witkowska-Walczak and C. Slawinski, 2002. Comparison of correlation models for the estimation of the water retention characteristics of soil. Int. Agrophys., 16: 79-82.
22: Williams R.D., L.R. Ahuja and J.W. Naney, 1992. Comparisons of methods to estimate soil water characteristics from soil particle size, bulk density and limited data. Soil Sci., 153: 172-184.
23: Willmott, C.J., 1981. On the validation of models. Phys. Geogr., 2: 184-194.
Direct Link |
24: Xue, Y., F.J. Zeng, and C.A. Schossler, 1996. SSiB and its sensitivity to soil properties-Acase study using HAPEX-Mobilhy data. Global Planet. Change, 13: 183-194.