Impacts of Natural Factors on the Distribution of Disaster-prone Landslide in the Xiangxi Catchment based on Spatial Analysis, China

Since the Three Gorges Reservoir, i.e. TGR, has been impounded in 2003, the landslide problem inner Three Gorges Area, i.e. TGA, is becoming more and more prominent. However, water storage of TGR is only the external cause while the regional geographical and geological conditions are the essence reasons for the local landslide development. Hence,Taking the Xiangxi Catchment as an example, this paper seeks to present the regional disaster-prone landslide distributed pattern, and find the quantitative relationships between the pattern and its main natural impact factors based on field investigation, GIS, general statistic method and Geodetector. The results show that: 1) the stepwise regression analysis indicated that the lithologic frangibility, the fault buffer, the NDVI, the Average of annual extreme value of storm and the lithologic hard degree can explain 80.6% of variation of the pattern, and the contribution rate of each factor in sequence was 28.6% , 19.8%, 18.9%, 16.4% and 16.3%; the Geodetector analysis indicated that the rank of main factors on explained degree of the pattern was the lithologic frangibility (24.5%), the fault buffer (20.9%), the NDVI (18.3%), the soil infiltration (16.3%) and the lithologic hard degree (14.8%), and the interaction between the lithology frangibility and the NDVI, the slope can explained 68.8% and 62.0% of the pattern, respectively; 2) generally, the regional disaster-prone landslide pattern was mainly controlled by lithology, fault and vegetation cover. The harder the rock was, or the farther from the fault, or the higher cover level of vegetation, the less the landslide developed; 3) the Geodetector provides a new perspective for the quantitative analysis of correlation between the nonlinear variables. However, the linear characteristic of independent variable is ignored in this method, which makes it impossible to detect the impact direction of independent variables to dependent variables, and the repeatability among the variables cannot be distinguished, naturally. The classic statistical methods and the Geodetctor are complementary. Therefore, the two methods should be combined in order to understand the results comprehensively.