A Pedotransfer Function for Field-Scale Saturated Hydraulic Conductivity of a Small Watershed

Classical experimental techniques to determine point values of saturated hydraulic conductivity (Ks) are complex and time consuming; therefore, the development of pedotransfer functions, PTFs, to derive Ks from easily available soil properties is of great importance. However, PTFs have been generally developed at the local scale, while hydrological modeling requires Ks estimates at larger scales. A small Austrian catchment, where detailed soil characteristics were available, was selected to address this issue. Values of field-scale saturated hydraulic conductivity (<Ksf>), observed in a number of catchment areas by double-ring infiltrometers, were used to develop two PTFs, one by multiple linear regression (PTFMLR) and one by ridge regression (PTFR). Training and validation of the PTFs in the monitored areas indicate that the PTFR provides better outcomes with smaller average errors. This suggests that the ridge regression is a valid alternative to the classical multiple linear regression technique. Predictions of <Ksf> by the PTFs in the remaining areas, where infiltration measurements were not performed, were also made to obtain a map of <Ksf> for the whole catchment. Two alternative approaches were used: Method A—soil properties were first interpolated and then the PTFs applied; Method B—the PTFs were first applied to sites with available soil properties and then interpolated. The maps of <Ksf> obtained by the PTFMLR are not representative of the <Ksf> spatial variability. On the other hand, the map generated by the PTFR with Method A is consistent with catchment morphology and soil characteristics.