Soil erosion is among the most important environmental problems in the world resulted from unsustainable ecosystems. It could also accelerate degradation of the ecosystems. Therefore, it is firstly vital to determine soil erosion-critical locations for the erosion-ameliorating practices. Several soil erosion models have been developed to predict soil erosion, which need accurate inputs for good estimation of soil erosion. Soil surface shear strength is an important input parameter in the soil erosion models, but its direct measurement is difficult, time-consuming and costly in the watershed scale. This study was done to predict soil surface shear strength using artificial neural networks (A), multiple linear regression (MLR) and adaptive fuzzy system based on neural networks (ANFIS). It was also aimed to determine and rank the factors most effective on soil surface shear strength. A direct shear box was designed and constructed to measure in situ soil surface shear strength. The device can determine two soil shear strength parameters i.e. cohesion ( c ) and angel of internal friction ( ? ). The study area (3500 km 2 ) was located in Semirom region, Isfahan province. Soil surface shear strength was determined using the shear box at 100 locations under three land uses of grassland, irrigated farming and dryland farming. Soil samples were also collected from 0?5 cm layer of the same 100 locations. Particle size distribution, fine clay content, organic matter content (OM), carbonate content, bulk density and gravel content were determined on the collected soil samples. Normalized difference vegetation index (NDVI) was also calculated using satellite images. Multiple linear regression (MLR), artificial neural networks (A) and adaptive fuzzy system based on neural networks (ANFIS) were used to model/predict soil surface shear strength ( c and ? ) using two groups of inputs: 1) easily-available soil properties (pedotransfer functions, PTFs) and 2) easily-available soil properties and NDVI (soil spatial prediction functions, Fs). A strong negative correlation was found between soil c and ? in the studied area. Positive correlations were also obtained between c and total clay content (especially fine clay content) and between ? and sand or gravel contents. The results showed that NDVI is an important factor affecting soil shear strength parameters ( c and ? ). LSD mean comparisons were used to investigate the effect of land use on soil surface shear strength and showed that dryland farming with maximum mean of clay content, fine clay content and c , had also minimum mean of sand content, gravel content and ? . Also irrigated farming with maximum mean of sand content and minimum mean of clay content had minimum mean of c . However, grassland and irrigated farming did not have significant differences in terms of soil c and ? . Prediction models of shear strength derived in group 2 (Fs) were more accurate than those derived in group 1 (PTFs). In order to compare the modeling methods, efficacy indices were used for evaluation of the best result of each method. The results showed that ANN models were more feasible in predicting soil shear strength parameters than MLR and ANFIS models. The A, MLR and ANFIS accounted for 90, 43 and 37 percents of variations ( R 2 ) of c , respectively. Corresponding values for ? prediction were 92, 57 and 52 percents, respectively. Therefore, all three models were more successful in predicting frictional part than cohesive part of soil shear strength. Moreover, A were the best models for predicting/modeling soil surface shear strength in the region because they can extract patterns and detect nonlinear trends that are too complex. Results of sensitivity analysis for ANN models indicated that NDVI, bulk density and fine clay content are more effective parameters on c , However, land use, carbonate, sand and organic matter contents were less effective on this parameter. Only two predictors of OM/clay ratio and silt content had sensitivity coefficients of less than 1 and were ineffective in soil c in the studied region. Also it was found that sand content, bulk density and NDVI were more effective parameters and OM/clay ratio and organic matter content were less effective parameters on ? in the region.