LIU Shuai, HE Bin, WANG Tao, LIU Jiamei, CAO Jiawen, WANG Haojie, ZHANG Shuai, LI Kun, LI Ran, ZHANG Yongjun, DOU Xiaodong, WU Zhonghai, CHEN Peng, and FENG Chengjun
Objective On December 18, 2023, an MS 6.2 earthquake occurred in Jishishan County, Gansu Province, China. Coseismic geological hazards induced by the earthquake crucially threatened the safety of personnel and property. Existing research is mainly concentrated in the vicinity of active faults and the concentrated distribution area of hidden danger points. Moreover, no special susceptibility assessment studies have been carried out on coseismic geological hazards in the administrative area of Jishishan County, making it challenging to meet the needs of the county's post-disaster recovery and reconstruction planning. Hence, the development laws of coseismic geological hazards must be summarized and analyzed crucially, and county susceptibility must be analyzed in time to support post-earthquake recovery and reconstruction. Methods The development characteristics of coseismic geological hazards are analyzed and summarized through emergency investigations, field surveys, and result analysis. Using the newly added and exacerbated coseismic hazard points identified during post-earthquake investigations as analysis samples, influencing factors were selected using the Pearson correlation coefficient and random forest Gini coefficient analysis methods. Then, a machine learning-random forest model was applied to assess the susceptibility of coseismic geological hazards in Jishishan County. Results In analyzing the development characteristics of coseismic geological hazards, we identified 134 instances of increased and exacerbated hazards in Jishishan County. Overall, the degree of development of these hazards was relatively low, with primarily small-scale occurrences. These hazards were categorized into three main types and eight sub-categories: ① Collapse (including cut slope loess collapse, high loess collapse, and high rock collapse); ② Landslide (encompassing loess landslide, secondary sand/mudstone landslide, and potential landslide); and ③ Debris flow (comprising gully debris flow and slope debris flow). In terms of factor selection, 15 influencing factors were screened. Regarding the susceptibility assessment results, the AUC value of the susceptibility assessment results of coseismic geological hazards in most Jishishan counties was 0.961, and the results showed that the areas of extremely high susceptibility accounted for approximately 8.67 %, mainly distributed in Hulinjia, Xuhujia, Liugoujia, and other townships. The statistical results of the proportion of susceptibility zones in 17 townships in Jishishan County showed that the top three townships with the largest proportions of extremely high-susceptibility areas are Hulinjia (24.67%), Xuhujia (21.24%), and Biezang (20.94%). Conclusion (1) Most coseismic geological hazards in Jishishan are distributed in the loess hilly area, with few occurrences in the Jishishan area and the right bank terrace of the Yellow River. (2) The influence of elevation and peak ground acceleration (PGA) on hazard occurrence is notably greater than that of other factors, playing a predominant role in developing coseismic geological hazards. (3) Utilizing the random forest model, the susceptibility assessment of coseismic geological hazards in Jishishan County demonstrates high accuracy, with hidden danger points clustered in highly susceptible areas. This alignment between susceptibility assessment results and the spatial distribution of hidden dangers underscores the reliability of the assessment outcomes. Significance In addition to identifying existing hidden danger points, this study offers predictive insights into slope deformation and potential landslides significantly affected by seismic cracking. The assessment results exhibit high accuracy and reliability, offering valuable geological safety support for post-disaster recovery and reconstruction planning in the county.