Abstract:

In order to explore the influence of different evaluation units on the assessment of regional landslide susceptibility, the landslide susceptibility of Yunyang County in Chongqing, a typical county in the Three Gorges reservoir area, was studied based on grid units and slope units. Twenty-two evaluation factors such as elevation, slope and curvature were selected. Combined with 988 historical landslide data in the study area, slope units were extracted from 30 m×30 m raster data, and geospatial databases of landslide impact factors were established based on slope units and grid units. The landslide susceptibility model was constructed by random forest and Bayesian optimization algorithm to evaluate the landslide susceptibility in the study area. The results indicate that the susceptibility assessment can be divided into five levels: low, moderately low, medium, moderately high, and high. In the landslide susceptibility model based on the grid units, the three factors of elevation, distance from the road and slope contribute greatly to the landslide occurrence. In the model based on slope units, the three factors of INDV (normalized difference vegetation index ), profile curvature and plane curvature contribute greatly to the landslide occurrence. The landslide density of both models increases with the increase of susceptibility level. Compared with grid units, slope units can better explain the relationship between terrain, and the landslide susceptibility model with slope units （AUC=0.744） as the minimum evaluation unit is more accurate than grid units （AUC=0.714）.

Key words: landslide, landslide evaluation units, slope units, grid units, landslide susceptibility, random forest, Three Gorges reservoir area, Yunyang County in Chongqing