Spatial autocorrelation modeling to assess geohazard susceptibility assessment in the mountainous Shennongjia area of China.

Geohazards are a serious concern in mountainous areas, especially in China. Substantial efforts have been made to evaluate the geohazard susceptibility in different areas via quantitative analysis. However, such efforts have focused mainly on the spatial heterogeneity of disasters, ignoring both their spatial autocorrelation and the effects of human activities connected to rapid urban development. To address these limitations, we develop a spatial autocorrelation regression (SAR) modeling framework for geohazard susceptibility assessment using human and natural activity data from the typical mountainous Shennongjia area of China. Moreover, we compared different evaluation models using cross-validation and receiver operating characteristic (ROC) curves. The key findings of this work are as follows: (1) SAR is more suitable for geohazard susceptibility assessment compared with other models, as demonstrated by the ROC values of 0.829, 0.789, and 0.527; (2) the human activities of road construction, building construction, and agricultural activity make the highest contributions to geohazards in the area, at 42.78%, 27.84%, and 11.41%, respectively; and (3) there are three high-risk areas in the case area, revealing obvious spatial aggregation. The evidence reported here can be used to accurately identify the risk of geohazards, guide the government emergency departments in disaster prevention, and avoid risky areas to carry out follow-up urban construction. [ABSTRACT FROM AUTHOR]