Evaluation of geological hazard susceptibility of collapse and landslide in Yuanyang County using slope units and random forest modeling

Abstract

The model based on grid unit and qualitative and quantitative method has the disadvantages of low prediction accuracy and frequent use in the evaluation of geological hazard susceptibility, was utilized to evaluate the potential for collapse and landslide in Yuanyang County. Using ArcGIS, 7851 slope units were divided via the curvature watershed method. Through a large number of statistical study and analysis of geological environment condition, nine evaluation factors were selected, including engineering geological petrofabric, landform type, elevation, gradient, slope direction, curvature, ups and downs, rivers, distance and fault distance. These factors were analyzed and their weights determined using SPSS software, in conjunction with data on the development characteristics of disaster points. The random forest model was then applied to establish a vulnerability evaluation model, which categorized landslide geological disaster in Yuanyang County into four types: low, medium, high and extremely high, occupying an area of 410.06 km2, 470.21 km2, 550.02 km2 and 776.87 km2 respectively. These areas correspond to 18.58%, 21.30%, 24.92% and 35.20% of Yuanyang County’s total area. The evaluation results were compared with the detailed investigation results and were found to be highly consistent. The accuracy of ROC curve was calculated at 92.7%, indicating a high level of accuracy. The central and southwest parts of Yuanyang County were found to be highly susceptible to geological disasters.

Keywords

• slope element
• random forest model
• susceptibility evaluation of geological disasters
• roc curve
• yuanyang county