Soil erodibility factor is a criterion of soil particle resistance to detachment, transport, and effects of erosivity factors (rain drop, runoff, and wind) during the soil loss processes. In this study, non-linear support vector machines (SVMs) method was used for investigating the effects of some topography, soil physical and mechanical properties on soil erodibility in a part of Northern Karoon watershed at the west of Chaharmahal Bakhtiary province. The obtained results were compared with the outputs of a linear (i.e. multiple linear regression, MLR) method. Furthermore, potential use of the two methods for estimating the soil erodibility factor was investigated in the study area. The correlation coefficient (r), root mean square error (RMSE), model efficiency factor (MEF), and error percentage (ERROR%) were used to evaluate the performance of the models. The results showed that the constructed SVM model had greater performance in predicting the erodibility factor compared to the traditional MLR model. The obtained ERROR% and r values for the developed SVM model were 0.07 and 0.90, respectively. The obtained MEF value for the erodibility factor prediction using the SVM model was 76.79 % while it was -18.04 % for the MLR model. The results from determining the properties influencing the soil erodibility in the investigated points using the two methods revealed that only the sand content, aspect, pods (size between 0.25-0.5 mm), clay content properties were accounted as effective parameters according to the MLR model. Whereas, all investigated properties were determined as effective parameters resulting from the non-linear SVM model. However, the parameters mean weight diameter (MWD) and sand content (size between 0.25-0.5 mm) had the highest importance coefficient and aspect parameter had the lowest coefficient.
Rights and permissions | |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |