基于GD-SPCA法的闽西南地区滑坡易发性评价

Assessment of Landslide Susceptibility in Southwestern Fujian Region Based on Geodetector and Spatial Principal Component Analysis

合理选择滑坡易发性评价因子是揭示滑坡易发区空间分布特征的重要前提。通过考虑因子最优离散分类,结合地理探测器(Geodetector,GD)与空间主成分分析法(Spatial Principal Component,SPCA)构建滑坡易发性指数。发现评价因子普遍呈现离散分类数越大,因子解释度越高的趋势,且自然断点法与分位数分类法的离散分类结果较好;因子间的交互作用对因子的解释度有一定影响,归一化植被指数(NDVI)的单因子解释度为0.453,但在交互作用下其耦合解释度范围仅为[0.03,0.226];研究区滑坡易发性指数为0.589,属于中度易发区,高度易发区与极易发区主要分布在研究区北部高植被覆盖的山地丘陵处;研究区的滑坡易发性指数具有显著空间聚集特征,高-高聚集区主要分布在该区北部中低山丘陵处,低-低聚集区主要分布在该区南部低缓河谷处。

Due to the differences in geographical and human conditions among different regions,the importance of evaluation factors for landslide susceptibility also varies dramatically.Therefore,selecting the appropriate evaluation factors is an important prerequisite for revealing the spatial distribution characteristics of landslide-prone areas.By considering the optimal discretization of factors and combining Geodetector(GD)and Spatial Principal Component Analysis(SPCA),a landslide susceptibility index is constructed.It is found that evaluation factors generally exhibit a trend where the higher the number of discrete categories,the higher the explanatory power of the factors,and both the natural breakpoint method and quantile classification method yield good results.The interaction between factors affects the explanatory power of factors for landslides,with the Normalized Difference Vegetation Index(NDVI)having an explanatory power of up to 0.453,but its coupled explanatory power is much lower than that of population density.The landslide susceptibility index of the study area is 0.589,indicating a moderate susceptibility area,with highly susceptible and extremely susceptible areas mainly distributed in the northern mountainous areas with dense vegetation cover.The landslide susceptibility index of the study area exhibits significant spatial clustering characteristics,with high-high clustering areas mainly distributed in the middle and low mountainous areas in the north of the study area,and low-low clustering areas mainly distributed in the gentle valleys in the south of the study area.This study could provide reference for the landslide susceptibility zoning in Southwest Fujian Province.