Evaluation of pedotransfer functions in predicting soil water characteristics of soils in Zambia

Abstract

Development of models that simulate water flow and solute transport has progress rapidly in recent years. These models require input parameters that can be obtained directly in the field or in the laboratory. However, because of the large spatial and temporal variability of soil hydraulic properties, their direct measurement is both time consuming and expensive. Indirect method e.g pedotransfer functions (PTFs) are therefore increasing being used to provide these estimates. PTFs are functions that estimate soil hydraulic properties from easily measurable soil properties like soil texture and organic matter. Soil exhibit high spatial and temporal variability and this can cause PTFs to fail to predict soil hydraulic properties accurately. A study was undertaken to evaluate the performance of three PTFs (Saxton et al. 2006; Campbell, 1974 and Rawls-Brakensiek, 1985) in predicting soil moisture content at field capacity and wilting point as well as the saturated hydraulic conductivity from four sites at the UNZA. Soils were collected in each soil mapping unit and soil characterization was done as well as the measurement of the soil hydraulic properties like the moisture content at field capacity and wilting point as well as saturated hydraulic conductivity. The coefficient of correlation(R^), Mean Deviation (MD) and the Root mean square error were used compare the measured and predicted values by the three different PTFs and to evaluate the performance of the three different PTFs. From the results obtained, the Saxton et al. (2006) model was the better predictor of moisture content at field capacity with the lowest MD and RMDE of 0.0004 and 0.0359 respectively while having a coefficient of correlation of 0.684. At wilting point, the Rawls-Brakensiek (1985) was a better predictor with the highest coefficient of correlation (0.8636) while having the lowest MD and RMSD of -0.0140 and 0.0249 respectively. For hydraulic conductivity, both Saxton et al., 2006 and Campbell ,1974 proved to be poor predictors of soil saturated hydraulic conductivity with low coefficients of correlation of 0.185 and 0.227 and very high MD of -11.446 and 5.852 respectively. The RMSE was 20.537 and 6.404 respectively. The study showed that in the absence of measured data of the moisture content in the study area, the Saxton et al. (2006) and the Rawls-Brakensiek (1985) models can be used to predict soil moisture content at field capacity and wilting point respectively provided a local calibration factor for the study area is used to take into account any systematic under or over estimations. This cannot be done for saturated soil hydraulic conductivity.