SciELO - Scientific Electronic Library Online

vol.36 issue5Soil as indicator of hillslope hydrological behaviour in the Weatherley catchment, Eastern Cape, South AfricaInvestigation of potential water quality and quantity impacts associated with mining of the shallow Waterberg coal reserves, west of the Daarby Fault, Limpopo Province, South Africa author indexsubject indexarticles search
Home Pagealphabetic serial listing  

Services on Demand



Related links

  • On index processCited by Google
  • On index processSimilars in Google


Water SA

On-line version ISSN 1816-7950
Print version ISSN 0378-4738

Water SA vol.36 n.5 Pretoria Oct. 2010


Soil - water relationships in the Weatherley catchment, South Africa



CW van Huyssteen*; TB Zere; M Hensley

Department of Soil, Crop and Climate Sciences, University of the Free State, PO Box 339, Bloemfontein 9300, South Africa




Soil water content is influenced by soil and terrain factors, but studies on the predictive value of diagnostic horizon type for the degree and duration of wetness seem to be lacking. The aim of this paper is therefore to describe selected hydropedological soil-water relationships for important soils and diagnostic horizons in the Weatherley catchment. Daily soil water content was determined for 3 horizons in 28 profiles of the Weatherley catchment. These data were used to calculate annual duration of water saturation above 0.7 of porosity (AD ), which was correlated against other soil properties. Significant correlations (α = 0.05) were obtained between average degree of water saturation per profile and slope (R2 = 0.24), coarse sand content (R2 = 0.22), medium sand content (R2 = 0.23), fine silt content (R2 = 0.19), and clay content (R2 = 0.38). AD per diagnostic horizon ranged from 21 to 29 d•yr-1 for the red apedal B, yellow brown apedal B, and neocutanic B horizons; 103 d•yr-1 for the orthic A horizons; and from 239 to 357 d•yr-1 for the soft plinthic B, unspecified material with signs of wetness, E, and G horizons. A regression equation to predict AD from diagnostic horizon type (DH), clay to sand ratio (Cl:Sa), and underlying horizon type (DH ) gave: AD = -26.31 + 41.64 ln(Cl:Sa) + 35.43 DH + 13.73 DH (R2 = 0.78).Results presented here emphasise the value of soil classification in the prediction of duration of water saturation.

Keywords: diagnostic horizon, model, slope, soil texture, water saturation




Van der Watt and Van Rooyen (1995) define soil water regime as a qualitative term referring to the state and availability of water in soil, especially in relation to the growth of plants. Another widely used, related, and probably more process-oriented and quantitative term, is soil hydrology. Miyazaki (1993) suggested that soil hydrology could be considered as a bridge between soil physics (where descriptions of water flow tend to be microscopic) and hydrology (where they tend to be macroscopic). Kutilek and Nielsen (1994) defined soil hydrology as the physical interpretation of phenomena which govern hydrological events related to soil, or the uppermost mantle of the earth's crust. Although both terms described above refer to soil-water relationships, neither are completely appropriate for the hydropedological focus of this study or the one by Van Huyssteen et al. (2005). The focus here was specifically on the nature of the long-term water content above the drained upper limit and the marked influence it may have on soil morphology and therefore soil classification.

Soil water content is mainly a function of the ratio between precipitation and evapotranspiration, as described by the soil water balance equation (Hillel, 1980). In general, the higher the rainfall to evapotranspiration (P/ET) ratio, the higher the soil water content. At a given site the soil water content varies depending mainly on variations in the topography, the soil, and the type and density of vegetation (Famiglietti et al., 1998). In a small catchment the rainfall can be considered fairly uniform over the whole area, although ET can vary significantly (Van Huyssteen et al., 2009a). Le Roux et al. (2005) concluded that the vegetation composition in the Weatherley catchment is fairly uniform and that the influence of vegetation on soil water content can therefore also be considered reasonably uniform. Soil and terrain characteristics should therefore be the primary factors that influence soil water content variability in this catchment.

Terrain indices can be grouped into primary and compound attributes. Primary terrain attributes (Famiglietti et al., 1998; Western et al., 1999) include:

• Slope, influencing the hydraulic gradient that drives surface and subsurface flow, surface runoff, infiltration, and drainage

• Aspect, together with slope, which influences the irradiance and thus evapotranspiration

• Plan curvature (perpendicular to the aspect) and profile curvature (parallel to the aspect) are measures of concavity or convexity of the surface. Plan curvature provides a measure of the local convergence or divergence of lateral flow paths, while profile curvature reflects how the hydraulic gradient and flow velocities change. Curvature is expressed as the reciprocal of the radius of curvature of the surface. Broader curves have smaller curvature values than tighter curves. Positive values indicate convex surfaces, zero indicates a straight surface, while negative values indicate concave surfaces.

• Specific upslope area, also called upslope contributing area, is the area above the contour segment divided by the length of that contour segment. It is a measure of the potential area that can contribute to lateral flow.

Terrain indices can be computed from digital elevation models (Mitasova and Hofierka, 1993). Compound indices, for example the wetness index (WI) defined by Beven and Kirkby (1979), can be calculated from primary topography attributes, and are often adapted, for example, the 'new index' (NI) for drier environments (Gomez-Plaza et al., 2001). The predictive power of terrain indices has been researched extensively, for example by Western et al. (1999), who did correlation analyses between terrain indices and soil water content in Australia. They defined the potential solar radiation index as the ratio of potential solar radiation on a sloping plane to that on a horizontal plane. During dry conditions cos(aspect) explained up to 12%, the potential solar radiation index explained up to 14%, and a combination of ln(α) and the potential solar radiation index explained up to 22% of the variance.

The influence of terrain indices on soil water content variability is often combined with other soil properties. Canton et al. (2004) and Lin et al. (2006) noted that it was difficult to identify the relative importance of the soil and topographic factors because of their mutual and multiple influences on soil water content. Soil factors include infiltration rate, water storage capacity, hydraulic conductivity, and the presence or absence of an impeding layer (Burt and Butcher, 1985). Gomez-Plaza et al. (2001) state that the influence of the soil factors studied on soil water content was different in the vegetated and non-vegetated zones and that the predictive power of WI was significantly increased by incorporating aspect (i.e. NI) and removal of slope, especially in unburnt catchments.

The relationship between soil properties and soil wetness has been studied in South Africa (Donkin and Fey, 1991; Le Roux, 1996; Van Huyssteen and Ellis, 1997; Van Huyssteen et al., 1997; Van Huyssteen et al., 2005) and in other countries (Rabenhorst et al., 1998; Blavet et al., 2000). Soil morphology contains significant information on the spatial and temporal distribution and circulation of water, both at profile and at landscape levels (Lin et al., 2005). Researchers used this information to predict soil wetness from soil morphology (Van Huyssteen and Ellis, 1997; Van Huyssteen et al., 1997; Rabenhorst and Parikh, 2000; He et al., 2003; Tassinari et al., 2002). These studies indicate that variability in soil water content is influenced by a combination of soil and terrain factors, and that the two are often inseparable. What has not yet been studied is the possible improvement to hydrological modelling that could be obtained by including the diagnostic horizon in the prediction of degree and duration of wetness.

The objective of this paper was to describe selected hydropedological soil-water relationships of selected soils and diagnostic horizons in the Weatherley catchment.


Materials and methods

Daily soil water contents were calculated using the procedure described by Van Huyssteen et al. (2009b). This included daily soil water contents for two to five, 300 mm layers of 28 soil profiles in the Weatherley catchment, for 6 years from 01 July 1997 to 30 June 2003. Profile numbers were defined to be distinctive within this catchment and also between other research catchments in the area (Hensley and Anderson, 1998). Profile numbers are used here in their original form to facilitate comparison. Soil water contents were expressed in terms of the degree of saturation (s), which is the fraction of the porosity that is occupied by water. Expressing soil water content as s enables comparisons between different soils and facilitates pedological interpretation. This point can be illustrated by considering 2 soils, say A and B, with bulk densities of 1.3 Mg•m-3 and 1.7 Mg•m-3 respectively. The volumetric water content (θv) for Soil A and Soil B at s = 0.7 would be 36% and 25%, and at complete saturation (s = 1.0) θv would be 51% and 36%, respectively. Reporting the θv values (51% and 36%) might give the impression that Soil A is wetter than Soil B, unless the bulk density (or porosity) values are also reported, while in reality both of them are completely saturated.

Detailed morphological soil profile descriptions, with chemical and physical analyses, are given by Van Huyssteen et al. (2005). Results for particle size analyses (Soil Classification Working Group, 1991) were averaged for the 0-900 mm depth and therefore no not necessarily add to 100%.

The s s value for the different soils in the Weatherley catchment, as used here was defined as the mean daily s value of the 0-900 mm layer over the 6-year measurement period. There were only 2 profiles, P207 and P209, with measurements for the top 2 layers only. For these profiles the ss values were taken over the 2 top layers. The ss values of the different soils were compared with soil and topographic variables that might affect soil wetness. Mean daily s values for each month were also calculated for specific diagnostic horizons. These values aided in demonstrating the influence of the soil and topographic variables on s during different seasons.

Other hydropedological characteristics which were calculated for diagnostic horizons included the mean annual duration of s>0.7 (AD ), the mean annual frequency (Fs>0.7 ), and the mean duration (Ds>0.7 ) of periods with s>0.7 (Van Huyssteen et al., 2005). The 0.7 threshold was chosen as an estimation of where reduction would set in, because 100% water saturation is seldom reached (Hillel, 1980). The duration of s>0.7 for a particular year is the total number of days in that year on which s was above 0.7 of porosity. Whenever it occurs, the value s>0.7 could occur for periods of 1 day or longer. These periods are referred to as events. The frequency term (Fs>0.7 ) refers to the number of such events. The mean duration of events (Ds>0.7 ) is thus the ratio of the total number of days per year to the number of events per year. Due to their annual nature, a single value of each of ADs>0.7 , Fs>0.7 and Ds>0.7 could be calculated per year during each of the 6 years of study for each diagnostic horizon. Their means were then calculated from the 6 annual values.

Only measurements that represented at least 80% of the volume of a particular diagnostic horizon were considered for this analysis. This was necessary to avoid possible erroneous conclusions introduced by measurements taken over transitions between diagnostic horizons. Terrain indices for the Weatherley catchment were calculated using ArcView 3.1 (ESRI, 1999). A 20 m grid-based digital elevation model was calculated, using kriging interpolation, from a 2 m interval contour map of the Weatherley catchment (BEEH, 2003).

Linear correlation analyses were done for diagnostic horizons between soil chemical, physical, and morphological properties and topographic indices as independent variables, with AD and mean monthly s for the 12 months as dependent variables. This was done in the form of a correlation matrix, with 70 observations of each variable. Prior to this, pair-wise scatter plots of independent and dependent variables were drawn to test the linearity of the relationships. In some cases it was necessary to transform the independent variable data to linearise the relationships. Qualitative data (diagnostic horizon and soil structure) were assigned dummy variables (Cody and Smith, 1997) to enable representation in the regression analyses. Eight different types of diagnostic horizons were available for the analyses. An integer between 1 and 8 was assigned to each of them in increasing order of their relative wetness based on their definitions (Soil Classification Working Group, 1991) and experience (Van Huyssteen et al., 2005). The numbering assigned was neocutanic B = 1, red apedal B = 2, yellow apedal B = 3, orthic A = 4, soft plinthic B = 5, E = 6, unspecified material with signs of wetness = 7, and G = 8. Saprolite, which was present underlying a diagnostic horizon in a single instance, was assigned 9. Soil structure was divided into 2 classes: massive, crumb, and granular were assigned 0, whereas subangular blocky, angular blocky, columnar, and prismatic were assigned 1.

The correlation matrix results helped to identify the strength of relationships between independent and dependent variables. These results also assisted in determining if relationships existed between independent variables, a situation that might lead to multi-collinearity problems and was therefore considered undesirable (NCSS, 1998; Fekedulegn et al., 2002).

Independent variables that correlated with ADs>0.7 were further analysed using multiple linear regression analysis (NCSS, 1998) to identify their combined effect on ADs>0.7. To reduce the number of variables to a manageable level, while at the same time keeping a sufficient number in the model, a procedure known as All Possible Regressions (NCSS, 1998) was used. This procedure helps to select the best number and combination of independent variables from the available options.

The most commonly used selection criteria (NCSS, 1998) are:

• Coefficient of determination (R2) - the larger the R2 value, the better the model

• Root mean square error (RMSE) - the smaller the RMSE, the better the model

• Mallow's statistic (Cp) - in the optimum model the Cp value is close to p+1, where p is the number of independent variables. Deviation from this indicates that the model contains either too many or too few variables.

Multiple regression analysis was run on the independent variables selected according to the above-mentioned criteria. To use the resulting regression equation for prediction, the residuals (predicted - observed) must be independent of one another (Webster, 1997). This assumption was verified visually by plotting the residuals against each of the independent and dependent variables.

A test of the predictive capacity of the fitted regression model was conducted using a cross-validation procedure. This procedure, described by Wilks (1995), uses all observations (n) of the dependent variable (Y) for validation in a way that allows each observation to be treated, one at a time, as independent data. It regresses n-1 observations of Y on the corresponding observations of the predictors (X) to develop a regression equation. The equation is then used to predict the remaining ith observation of Y from the X on row i. This is repeated for each observation (n times), producing n equations and therefore n predicted values. The final reported equation is fitted using all n observations. The advantage of this type of cross-validation procedure is that it avoids splitting the data into model development and validation data, thus yielding higher degrees of freedom. The statistical model evaluation procedures of Willmott (1981) were used to test the cross-validation results.


Results and discussion

The values of ss display a trend that was more-or-less consistent with the pedologically expected relative wetness levels of these soils (Fig. 1). At the drier end of the spectrum were the Tukulu, Hutton, Avalon, Pinedene, and Bloemdal soil forms, followed by Longlands and Westleigh soil forms and, towards the wet end, by the poorly drained Kroonstad and Katspruit soil forms.



The increase in ss (Table 1) was accompanied by a discernible change in topographic and soil characteristics. The terrain unit changed from primarily 3 to primarily 3(1), 4, and 5, while the slope angle generally decreased. Particle size distribution changed from primarily coarser for the drier soils to primarily finer for the wetter soils. Correlation analysis of the numeric values of the terrain and soil factors with ss across the soils (n = 28) resulted in significant correlations (α = 0.05) for slope (R2 = 0.31), coarse sand (coSa) content (R2 = 0.28), medium sand (meSa) content (R2 = 0.24), fine silt (fiSi) content (R2 = 0.28), and clay (Cl) content (R2 = 0.18). Other factors did not show significant correlation.

Monthly means of daily s values for each month of the year, averaged from the 6-year daily data and grouped per diagnostic horizon, are presented in Fig. 2. E horizons had the highest mean s, followed by G, unspecified material with signs of wetness, and soft plinthic B horizons.



Pedologically, one would expect the E horizons to be drier than the G horizons and unspecified material with signs of wetness (Soil Classification Working Group, 1991). However, this was not the case in this study. The discrepancy could be due to error introduced by the small number of E horizons (3), as compared to the G horizons (21) and unspecified material with signs of wetness (18), possibly causing bias. One of the E horizons was from P226 (Kd) and the other two were from P229 (Kd), both of which were very wet soils (Table 1). In both cases, the E horizons were underlain by G horizons which were wetter than the overlying E horizons. On average, the G horizons were, however, drier than the E horizons (Fig. 2). A significant contributing factor was probably the G horizons located in drier soils (Table 1) in the catchment e.g., P205 (Kd), P208 (Kd), and P232 (Ka).

Orthic A horizons were considerably wetter than the yellow brown apedal B, red apedal B and neocutanic B horizons during the summer (rain) months (December, January, February, and March). During the drier months of the year, however, the orthic A horizons were drier than the yellow brown apedal B and red apedal B horizons. This was expected, because orthic A horizons, being the upper horizons, have the greatest net water loss due to evapotranspiration (ET) during the rain-free winter months, while receiving most of the rainfall in summer months.

The yellow brown apedal B, red apedal B, and neocutanic B horizons decreased in wetness in the order: yellow brown apedal B > red apedal B > neocutanic B. These 3 horizons, as well as orthic A and soft plinthic B horizons, displayed considerable variations through the year, as opposed to the fairly constant wetness of the E and G horizons and unspecified material with signs of wetness. This reflected the extent of lateral and vertical drainage and losses due to ET from the orthic A and soft plinthic B horizons. The deeper and less permeable horizons lost less water through drainage and ET, while receiving more interflow water from higher-lying soils.

The ADs>0.7 , Fs>0.7 and Ds>0.7 values for the diagnostic horizons are presented in Table 2. Summary statistics, averaged over 6 years and grouped by diagnostic horizon (Table 3), were done to facilitate comparisons of the variability of the values within and between diagnostic horizons. Mean ADs>0.7 values ranged from 21 to 29 d•yr-1 for the red apedal B, yellow brown apedal B, and neocutanic B horizons; 103 d•yr-1 for the orthic A horizons; and from 239 to 357 d•yr-1 for the soft plinthic B, unspecified material with signs of wetness, E, and G horizons. All 3 red apedal B horizons in this analysis (Table 3) were located in P221, which had a relatively wet orthic A horizon (ADs>0.7= 115 d•yr-1) and very dry red apedal B horizons (ADs>0.7 < 50 d•yr-1). The reason for this anomaly was not clear. The relatively high standard error for the neocutanic B horizons can be attributed to the relatively wet neocutanic B horizon (ADs>0.7 = 167 d•yr-1) at P222, compared to the drier neocutanic B horizons at P203 and P212 (ADs>0.7 < 5 d•yr-1). The ADs>0.7 values in the yellow brown apedal B horizons ranged from 0 at P202 to 32 d•yr-1 at P233 and 50 d•yr-1 at P234, resulting in the high standard error.

The means of Ds>0.7 values (Table 3) were consistent with the morphology of the diagnostic horizons. All the diagnostic horizons which normally show considerable signs of wetness (i.e. soft plinthic B, unspecified material with sig