Landslide susceptibility analysis using remote sensing and GIS in the western Ecuadorian Andes

In this paper we created and validated a predictive model for assessing the susceptibility of landslides along highway E-20 in Ecuador, by measuring the degree of spatial association of a landslide inventory with a set of spatial factors in an empirical way. The main aims of this paper are to: (1) determine what spatial factors are most associated with landslide occurrence, (2) determine whether the E-20 has any type of influence on landslide occurrence and, if so, up to what distance. For this, we created a landslide inventory based on multi-temporal images from different sources and used the Yule coefficient and the distance distribution analysis, which enabled us to determine which spatial factors are more closely related to the occurrence of landslides. The findings support the idea that landslides are not randomly distributed, but are associated (positively or negatively) to the different geo-environmental conditions of the study area; in this case, landslides have shown positive association with areas of active erosive processes, granitic rocks, volcanic sandstone and rainfall ranging from 1500 to 1750 mm. The statistical significance of the model was tested in two different ways; thus, it can be considered as valid, showing that each spatial factor has some influence on the occurrence of landslides.