Modeling soil erosion and runoff requires accurate estimates of input parameters such as surface soil shear strength (), soil saturated hydraulic conductivity (K s ) and infiltration (I). Three MLR models and three ANN structures: multilayer perceptron (MLP), generalized feed-forward (GFF), and modular neural networks (MNN were tested and investigated for , K s and I in this study. Routinely measured (available) parameters included soil surface and subsurface attributes in addition to topographic and vegetation attributes: particle size distribution, soil organic carbon, bulk density, initial soil moisture, gravel, CaCO 3 , slope, aspect, elevation and normalized difference vegetation index (NDVI) were used. The performances of the different models were evaluated using spearman’s correlation coefficient (r) between the observed and the estimated values, normalized mean square error (NMSE), mean absolute error (MAE), geometric mean of error ratio (GMER), geometric standard deviation of the error ratio (GSDER) and Nash- Sutcliffe efficiency criterion. Results showed that land degradation significantly reduced , K s and I. Addition of slope, aspect, elevation and NDVI to soil attributes as input parameters improved the performance criteria for these three-parameters. The differences between the MLP with GFF structure were statistically insignificant, and MLP and GFF were significantly different from MNN structure. Improvements were achieved with all ANN-based pedotransfer functions (PTFs) and soil spatial prediction functions (Fs) over a MLR PTFs and Fs. Keywords Surface soil shear strength, saturated hydraulic conductivity, Infiltration, Artificial neural network, Soil erosion.
