摘要
为了给延安市志丹县防灾减灾和风险管控提供数据支持,给相似地区危险性评价提供参考依据,补充在崩塌和滑坡灾害危险性评价中未考虑累积降雨方面的影响,在10,20,50和100 a一遇4种不同降雨工况下对研究区进行了危险性评价。以栅格单元作为评价单元,结合灾害特征及区域孕灾背景,通过皮尔逊相关系数法,选取高程、坡度、坡向、曲率、岩土体类型、距河流距离、距道路距离、归一化植被指数8个评价因子,采用信息量模型进行了易发性评价并分析了致灾因子与灾害分布关联性。利用计算机语言,自动处理前期因子数据的分析、转化、管理和出图等流程,改进了信息量(IVM)-随机森林(RF)耦合模型,实现了模型的自动循环迭代对比选择,通过ROC(受试者工作特征曲线,receiver operating characteristic curve)曲线对比了2种易发性模型精度。在耦合模型评价结果的基础上进行危险性评价,用皮尔逊Ⅲ型曲线估算研究区10,20,50和100 a一遇4种不同工况下降雨量,并进行危险性分区。对于易发性分区结果,信息量-随机森林耦合模型评价结果的AUC值(ROC曲线下面积)为0.87,优于IVM模型的评价结果;对于危险性分区结果,从10 a一遇到100 a一遇降雨工况的高和极高危险区面积都逐级增加。研究表明,改进的耦合模型评价方法不仅简化了操作还提高了精度,耦合模型确实拥有更好的评价精度和预测能力。
[Objective]To provide data support for disaster prevention,mitigation and risk management in Zhidan County,while providing reference for risk assessment in similar regions and addressing the research gap in collapse and landslide risk assessment that neglects cumulative rainfall impacts,this study conducted risk assessment under four different rainfall conditions:10,20,50 and 100 years recurrence intervals.[Methods]Using Zhidan County as the study area with grid unit as evaluation element,we selected eight evaluation factors through Pearson correlation coefficient method:elevation,slope,aspect,curvature,rock/soil type,distance to river,distance to roads,and normalized vegetation index.The information model(IVM)was employed to evaluate the susceptibility and analyze the correlations between causative factors and disaster distribution.Through programming implementation,the analysis,transformation,management and drawing of the previous factor data are automatically processed,the IVMRF(random forest)coupled model is improved,and the automatic cycle iteration comparison selection of the model is realized.The accuracy of the two susceptibility models is compared by ROC curve.Based on the evaluation results of the coupled model,the risk assessment was carried out.The Pearson typeⅢcurve was used to estimate the rainfall in the study area under four different conditions of 10,20,50 and 100 years,and the risk zoning was carried out.[Results]For the results of susceptibility zoning,the AUC value of the evaluation results of the information-random forest coupling model is 0.87,which is better than the evaluation results of IVM.For the results of risk zoning,the area of high and extremely high-risk areas have gradually increased from 10 to 100 years.The improved coupled model evaluation method not only simplifies the operation but also improves the accuracy.[Conclusion]According to the survey results,the coupled model does have better evaluation accuracy and prediction ability.
作者
屈鹏鑫
谢婉丽
刘琦琦
王昱琛
QU Pengxin;XIE Wanli;LIU Qiqi;WANG Yuchen(Department of Geology,Northwest University,Xi'an 710069,China;State Key Laboratory of Continental Evolution and Early Life,Northwest University,Xi'an 710069,China)
出处
《地质科技通报》
北大核心
2025年第3期280-295,共16页
Bulletin of Geological Science and Technology
基金
国家自然科学基金项目(42372320,41972292)
陕西省创新能力支撑计划项目(2021TD-54)
陕西省重点研发计划项目(2022ZDLSF06-03)
西安市科技计划项目(24LLRHZDZX0019)。
