Mapping soil erosion rates using self-organizing map (SOM) and geographic information system (GIS) on hillslopes

Abstract

Assessment of soil erosion is necessary for any long-term soil conservation plan, yet, the procedure is costly and time-consuming when applied to a large domain. Using a portion of the Alborz mountains in the north of Iran as a test site, a methodology was developed and tested based on a self-organizing map. First, annual soil erosion rates were measured on a hillslope in the study area through the installation of 120 erosion pins. Soil erosion controlling factors (slope gradient, slope length, slope shape, vegetation canopy, and the percentage of clay, silt, and sand) were determined through analysis of a digital elevation model (DEM) and field studies. Then, the data were normalized and divided into three subsets of training, cross-validation, and testing subsets. Then, a self-organizing map (SOM) was constructed to establish a relationship between soil erosion and its controlling factors. The SOM network was trained using the training and cross-validation subset and was evaluated on the testing subset using statistical coefficients (NMSE and R-squared). The evaluation of the SOM on the testing subset showed its high performance in the soil erosion modeling (NMSE = 0.1 test R-squared = 0.9). Next, the tested SOM was fit to the input variables to model the annual soil erosion rate across the study area. Finally, the modeled values were exported to the geographic information system (GIS) to generate the final map. The generated soil erosion map was verified by comparing the estimated soil erosion rates on the map with the recorded values of 11 erosion pins.