Journal of Watershed Management Research

Mendeley  

Zotero  

Gholami V, Borna F, Hadian B. (2021). Estimation of Soil Erosion using Artificial Neural Network (ANN) and Geographic Information System (GIS) on the Rangeland Hillslopes. jwmr. 12(24), 147-158. doi:10.52547/jwmr.12.24.147
URL: http://jwmr.sanru.ac.ir/article-1-1100-en.html

Extended Abstract
Introduction and Objective: Soil erosion is one of the most important problems in natural resources management, especially on the e rangeland hillslopes. Further, soil erosion estimation using field measurement is expensive and time-consuming. Therefore, models can be efficient tool for performing an exact estimation in a short time and a low cost. The aim f this study is to present a methodology to estimate soil erosion on the rangeland hillslopes.
Material and Methods:  In this study, the annual rates of soil erosion have been studied using erosion pins on the rangeland hillslope in the of Kasilian watershed in Mazandaran Province. Annual soil erosion rates were measured using 109 erosion pins (one year after the its establishment) due to changes in soil surface and soil specific gravity. An artificial neural network (ANN) was used in NeuroSolutions software. Soil erosion rates were as the model output and the affecting factors of soil erosion were the inputs. The model inputs rangeland cover percentage, land slope, slope length, slope shape (land curvature) and soil texture (sand, clay and silt percentage). The modeling process was performed using the MLP network. All of the data were separated into three classes included training (65% data), cross-validation (10%), and test stage (25% data). The model was performed and optimized. Further, geographic information system (GIS) was used for mapping soil erosion rates based on the simulated erosion values.
Results: The results of the test stage proved the high performance of the ANN in estimating soil erosion (Rsqr = 0.9). Further, statistical analysis using SPSS software  and the optimum structure of the network and sensitivity analysis showed that the most important factors of soil erosion are vegetation cover, slope shape, land slope, slope length and soil characteristics, respectively. Finally, the optimized network inputs were combined in a GIS environment with a pixel size of ten meters, and annual soil erosion map was generated by coupling the capabilities of ANN and GIS on the studied rangelands.
Conclusion: The proposed methodology can be used as an efficient and alternative method for field measurements of soil erosion in the highlands with a high performance.