Functional behaviour of pedotransfer functions in soil water flow simulation

Soil hydraulic parameter values for large-scale modelling cannot be obtained by direct methods. Pedotransfer functions (PTFs) that relate soil hydraulic properties (SHPs) to generally available soil texture data may provide an alternative. A considerable number of PTF models has been developed, the application of three recent PTFs is evaluated. As a first step sets of SHPs derived from the PTFs are compared with measured sets of SHPs for three sites. No good agreement was found statistically between measured and PTF results or between PTF results. As a second step and from a practical point of view results for three hydrologically functional variables were compared and evaluated. The three selected functional variables are saturated hydraulic conductivity, k(0), in relation to infiltration excess runoff, available soil water amounts for evapotranspiration and water table depth for a specified upward flux or capillary rise. Derived k(0) distributions from PTFs show substantially less variance than from the measured data at all three sites. This can have a considerable impact on infiltration excess runoff, depending on the actual rainfall regime. Simulated available soil water amounts for evapotranspiration for some combinations of PTFs and sites are close to those obtained for measured SHPs, however, no consistency in results can be detected. Water table depths for specified upward flux densities using PTF derived SHPs are generally deeper than those based on measured SHPs and means a potentially higher water availability. Overall, differences in capillary rise among the selected PTFs and between measured and PTF based results are again inconsistent and show no clear relationship with soil texture. Finally, as a third step; effective SHPs were calculated by using spatially averaged texture as PTF input representing areal average behaviour. For these effective SHPs the calculated effective values for the three selected functional variables appear to be close to the areally averaged values obtained with step 2. The selected functional variables thus appear to depend linearly on the PTFs over the range for which the data are representative. This suggests that for our specific PTFs areal mean or effective values for the three functional variables can be obtained fairly accurately from a single measurement of a bulk collection of soil samples as input. Copyright (C) 2003 John Wiley Sons, Ltd.