Identification and modeling of effective factors on runoff and sediment production from operated forest stands

In this study, we investigated the effects of forest operation on runoff and sediment using small-scale plots. The runoff and sediment samples were collected from 36 sample plots with dimensions of one by two meters in different areas of the operation. We used the Multi-Layer Perceptron (MLP) for modeling, with 65% of the data for training, 10% for validation, and 25% for testing. We evaluated the accuracy of the model using the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE), and compared our results with optimized models obtained through trial and error. We collected information and produced runoff and sediment maps using GIS. Our results showed that the most important factors affecting runoff production were soil bulk density, rainfall intensity, slope, rainfall values, percentage of grass cover, and canopy cover percentage. For sediment, the most important factors were rainfall intensity, soil bulk density, slope percentage, and surface cover percentage. The MSE and R values for runoff modeling were 0.009 and 0.9 in the training stage and 0.01 and 0.82 in the test stage, respectively. For sediment modeling, the MSE and R values were 0.01 and 0.86 in the training stage and 4.3 and 0.8 in the test stage, respectively. Our results showed that neural networks have high capability in modeling runoff and sediment in forest lands. We also conducted an overlap analysis to measure the accuracy, precision, and efficiency of the results and methods presented in our study. Therefore, the proposed model can be used to combine ANN and GIS in the simulation and modeling of runoff and sediment in forest areas."