| Wudou District,Longnan City,Gansu Province,is positioned in the Qinling Mountain system of the three main geomorphological units of the Qinghai-Tibet Plateau,Loess Plateau,and Sichuan Basin.It is situated at the intersection of the West Qinling-Wudou "mountain" tectonic system.Geological dangers are common in the area because to these factors.The region has some of the most severe geological dangers in the nation in terms of frequency,severity,and scale.Population growth is also accompanied by an increase in the scope and intensity of human activity,a worsening of the tension between human progress and the geological environment,and a rise in the danger that geological hazards provide to people’s lives and property.The investigation and zoning of geological hazards must be done immediately.The components that may produce geological risks are initially screened by analyzing the research area’s environment and sources of geological hazards.For the evaluation of geological hazard susceptibility,the multiple covariance analysis and Relief-F two-step analysis were employed to objectively assess and quantify the predisposing components and establish whether there were any highly associated factors among the predisposing factors.The pre-selected and screened gestation factors were input into three evaluation models,the random forest model,the Bayesian algorithm over-parameter optimized random forest model(Bayes-RF),and the genetic algorithm over-parameter optimized random forest model(GA-RF),respectively.Based on the magnitude of the prediction accuracy and the benefits and drawbacks of each model,the prediction accuracy and model performance of the three evaluation models were compared.Based on the geological hazard susceptibility evaluation maps produced by the three models,the most appropriate model for the research region was chosen after the three models were individually reviewed.Based on this,a plan for improving visualization was put forth,and a better examination of the geohazard susceptibility evaluation was conducted with human participants.The initial findings of the study were as follows:(1)Consider the "collapse-landslide-debris flow" chain as a whole for the geological hazard susceptibility assessment.This study takes these three types of disasters as a whole,considering their mutual influence and synergy,in order to more accurately assess the susceptibility of geological hazards in the study area.(2)Based on the complex occurrence mechanism of geological hazards,this study selected 14 factors that influence these hazards,including slope,aspect,altitude,rainfall,surface coverage,vegetation,geological structure,and lithology.By establishing a more reliable dataset,the susceptibility of geological hazards can be predicted and assessed more effectively.(3)Multiple collinearity analysis and Relief-F method were used to screen the evaluation factors and eliminate poor contributing factors to improve the accuracy and reliability of the model.In addition,this study also combined historical data and previous research to pre-select and grade the evaluation factors,which comprehensively considers the intrinsic and extrinsic factors that cause geological disasters.(4)Two integrated machine learning models,GA-RF and Bayes-RF,were constructed based on Python3.5 for the susceptibility assessment of geological hazards.(5)This study highlights and further explores the impact of human activities,emphasizing the importance of considering the significance of landforms and other potential patterns in susceptibility assessment,and proposes how to incorporate the role of humans into the evaluation model.A comprehensive assessment model is proposed in this paper,aiming to more comprehensively evaluate the susceptibility of geological hazards by considering factors such as human activities and landform significance.The findings of this study can offer a technical and scientific basis for determining the region’s vulnerability to geological hazards,which will be a valuable guide for preventing and safely producing geological hazards.The methodology and model used in this study also have some universal applicability and extension value,which can be used and promoted in other geographical areas and academic disciplines. |
