Effect of landslide spatial representation and raster resolution on the landslide susceptibility assessment

Abstract

The quality of landslide susceptibility assessment is effected by landslide inventory and raster resolution.The point landslide inventory and polygon landslide inventory are usually two spatial representations. In order to explore the effect of spatial representation and raster resolution on the assessment result of landslide susceptibility, two spatial representations (point and polygon) in landslide inventory map and four raster resolution (25m, 50m, 75m and 100m) were introduced in the landslide susceptibility model. A total of 489 landslides with accurate boundary information were obtained in Wanzhou District, Chongqing, China. Five geo-environmental predictors and random forest model as a data-driven model were used in the landslide susceptibility assessment. The research results demonstrate that (i) For two spatial representations, the prediction performance of polygon models was higher than that of point models by comparing reliability and accuracy. (ii) For four different raster resolution, the good prediction performance occurred in the 25m model. (iii) Through the analysis of model robustness, the lowest uncertainty was the 25m resolution model, which has more stable prediction performance. The research is useful for selecting the raster resolution and spatial representations in the data-driven landslide susceptibility assessment.