Soil organic matter (SOM) is an important soil property that affects physical, chemical, and biological properties of soil. Accurate estimation of SOM variability could provide critical information for understanding nutrients cycling and sediment. In the current study, artificial neural networks (ANNs) were developed to predict SOM variability based on topographic variables (topographic wetness index, relative position index, slope length and elevation) in hilly areas. A total of 265 soil samples collected from a depth of 0-20 cm were used to calibrate and validate the models. The best performed ANN model was compared with multiple linear regression (MLR) equation. The performance accuracy was evaluated by Pearson's correlation coefficient (r), mean error (ME), mean squared error (MSE), root mean squared error (RMSE), and coefficient of determination (R-2). In terms of MSE and r, the ANN model with topographic wetness index, relative position index, and slope length outperformed other ANNs. The best performed ANN model was also superior to the MLR equation. Values of ME, RMSE, and R-2 were -0.0337 g/kg, 1.0919 g/kg, and 0.8714 for ANN model, and were 0.1574 g/kg, 1.3296 g/kg, and 0.8172 for MLR equation, respectively. The results of ANN and MLR suggested that topographic wetness index was the most important topographic indicator affecting SOM variability in the current study area.
