Comparison of applying multi linear regression analysis and artificial neural network methods for simulating topographic factors effect on soil organic carbon

Soil organic carbon is one of the most important soil characteristics, and any changes in its content and composition, affects soil physical, chemical, and biological characteristics. Enhancing soil organic carbon improves soil structure, increases water and nutrients in soils, reduces soil erosion and degradation and thus greater productivity of plants and water quality are expected in watersheds and ultimately soil and ecosystem reclamation happens. Climatic, topographic and managerial factors affect soil organic carbon content. In local scale, climatic factors have not high efficiency on soil organic carbon and topographic factors play more important role compared to climate on soil organic carbon variability. The objective of this study was to predict and evaluate the effects of topographic factors such as elevation, slope percent, aspect, hill shade, and curvature on the soil organic carbon content of a rangeland in Mereg watershed, Kermanshah, Iran. Stepwise Multi Linear Regression (MLR) and Artificial Neural Network (ANN) were employed to develop models to predict soil organic carbon. A Multi-Layer Perceptrons (MLP) ANN with back-propagation error algorithm was applied to this research. The result showed that the multi linear regression and ANN models explained 53 and 77 percent of the total variability of soil organic carbon, respectively. The calculated RMSE and MBE were 0.40 and 0 for the MLR and 0.16 and 0.003 for MLP models. Results indicated that designated ANN model with 5-9-1 arrange was more feasible than multi linear regression for predicting soil organic carbon. Elevation with 0.79, hill shade with 0.64 and slope percent with 0.24, were identified as the important factors that explained the variability of soil organic carbon.