Maximum Entropy-Based Model of High-Threat Landslide Disaster Distribution in Zhaoqing, China

Abstract

Landslide disaster that threatened over 100 people in Zhaoqing, China, were taken as samples. Sixteen environmental factors were selected, including altitude, slope degree, slope aspect, lithology, soil texture, normalized differential vegetation index (NDVI), average annual rainfall, distance to developed land, and distance to roads. The Maximum Entropy model was employed for simulation analysis of landslides. The results suggest that: NDVI, lithology, distance to rivers, distance to roads, rainfall variance, and altitude are the leading environmental factors that affect landslide disasters. Of the factors taken into consideration, distance to developed land contributes as much as 43.6% of the AUC (area under the curve) value of the landslide distribution model. In fact, this factor became the absolute leading variable over even the NDVI, indicating that high-threat landslide disasters in the study area are highly correlated with human activities. The closer the landslide location was to developed land, rivers, and roads, the more likely a landslide was to occur. Using the MaxEnt model, the highthreat landslide in Zhaoqing can be favourably simulated. The AUC of the model’s prediction precision reached 0.769 without distance to developed land; whereas, the AUC of the model’s precision reached 0.845 with distance to developed land taken into account.

Keywords

• Asymptotic normality
• conditional hazard quantile function
• functional data
• kernel smoothing
• nonparametric estimator.