Document Type : Research Paper
Authors
1 Soil Conservation and Watershed Management Research Department, Chaharmahal and Bakhtiari Agricultural and Natural Resources Research and Education Center, AREEO, Shahrekord, Iran.
2 - Soil Conservation and Watershed Management Research Department, Chaharmahal and Bakhtiari Agricultural and Natural Resources Research and Education Center, AREEO, Shahrekord, Iran.
Abstract
Land subsidence, as a serious natural hazard influenced by the overexploitation of groundwater resources and geological factors, causes extensive damage to infrastructure and agricultural lands. This study was conducted with the aim of zoning subsidence risk in Chaharmahal and Bakhtiari province using the powerful AdaBoost machine learning model. In the first step, from among 30 initial effective factors including topographic, hydrological, geological, environmental, and climatic parameters, and after performing correlation analysis and removing collinear variables, 23 final factors were selected for modeling. The AdaBoost model was trained using 2352 training samples (including subsidence and non-subsidence points) and was validated on an independent test set consisting of 772 samples.
The evaluation of the model's performance using valid indicators showed its highly desirable accuracy and efficiency, such that the values for the Area Under the Curve (AUC), Precision, Recall, and Kappa coefficient were obtained as 0.974, 0.936, 0.981, and 0.855, respectively. Based on the model output, the final land subsidence risk zoning map was prepared in five classes: very low, low, medium, high, and very high. The results indicated that the Borujen and Shahrekord plains are at the highest risk of subsidence, and parts of the Lordegan plain also fall into the very high-risk category.
The analysis of variable importance using the SHAP method revealed that the three factors of land slope angle, surface sand percentage, and groundwater level changes have had the greatest impact on the occurrence of land subsidence in the region, in that order. Specifically, an inverse relationship was observed between land slope and subsidence intensity. Furthermore, a drop in groundwater level plays a direct role, while an increase in surface sand percentage plays a mitigating role in the occurrence of this phenomenon.
Keywords
)