摘要
滑坡灾害对地区财产和人民安全构成了重大威胁。因此,开展滑坡易发性制图工作对于有效预防和减轻滑坡风险具有极其重要的意义。传统滑坡易发性评估方法大多基于栅格或斜坡单元,未能充分考虑以具体滑坡位置为研究单元的建模需求,且风险因子的选择也缺乏客观性。利用SHAP算法筛选出了10个关键风险因子,依次为:坡度、降水、坡向、到公路的距离、到河流的距离、粗糙度、高程、岩性、到居民点的距离、地震烈度。采用积分嵌套拉普拉斯近似-随机微分方程(INLA-SPDE)方法构建了一个包含空间随机效应的逻辑回归模型,该方法有效避免了空间自相关对滑坡易发性预测的影响。使用该模型对四川省都汶公路沿线区域的滑坡空间分布进行了预测,结果AUC值为0.846,显示出较好的评估性能,为区域滑坡易发性评价提供了一种新颖的研究思路。
Landslide disasters pose significant threats to regional property and human safety.Therefore,conducting landslide susceptibility maps is crucial for effectively preventing and mitigating landslide risks.However,traditional landslide susceptibility assessment methods are mostly based on raster or slope units and fail to fully consider modeling requirements based on specific landslide locations as the study units.Additionally,the selection of risk factors lacks objectivity.This study employed the SHAP algorithm to identify ten key risk factors:slope,precipitation,aspect,distance to roads,distance to rivers,roughness,elevation,lithology,distance to settlements,and seismic intensity.Subsequently,the Integrated Nested Laplace Approximation-Stochastic Partial Differential Equations(INLA-SPDE)method was utilized to construct a logistic regression model incorporating spatial random effects,effectively avoiding the impact of spatial autocorrelation on landslide susceptibility prediction.This model was used to predict the spatial distribution of landslides along the Duwen Highway in Sichuan Province.The results show that the model's AUC value is 0.846,demonstrating good evaluation performance and providing a novel research approach for regional landslide susceptibility assessment.
作者
严国强
唐章英
宋超
郑雪
张雨萌
Yan Guoqiang;Tang Zhangying;Song Chao;Zheng Xue;Zhang Yumeng(School of Geoscience and Technology,Southwest Petroleum University,Chengdu 610500,Sichuan,China;West China School of Public Health and West China Fourth Hospital,Sichuan University,Chengdu 610041,Sichuan,China)
出处
《绿色科技》
2024年第18期245-251,共7页
Journal of Green Science and Technology
基金
国家自然科学基金(编号:42071379,72104159)
中国博士后科学基金(编号:2020M673274)。
