GIS-based study on the susceptibility of shallow landslides: a case study of mass shallow landslides in Sanming, Fujian in 2019

Fujian Province is one of the most frequent areas of landslide disasters in China, so the study of landslide susceptibility in this region is of great significance. Based on the mass landslides that occurred in Sanming, Fujian in 2019, the study used DEM data and aerial imagery to sort out the area, elevation, lithology, and road or building proximity of 131 landslides, and calculated the slope, profile curvature, and plan curvature of the landslides after improving the elevation matrix according to the geometry of landslides. Normalized landslide frequency was used to calculate the contribution of each causative factor to landslide disasters, so as to establish logistic regression models of landslide susceptibility on different scales. It was found that the model based on the 500 m * 500 m grid was the most suitable for evaluating landslide susceptibility. However, the logical regression model is deficient, not only the number of samples should reach more than 50, but also the incidence rate should be at least 30–70%, in order to avoid the model being not significant caused of the small number of samples and the sample imbalance, resulting in too many underestimated or overestimated samples.