AN R SCRIPT FOR LANDSLIDE SUSCEPTIBILITY ASSESSMENT USING RANDOM FOREST: CASE STUDIES FROM EL SALVADOR AND HONDURAS

This study presents an R script designed to calculate landslide susceptibility in a given area by using a landslide inventory, a Digital Elevation Model (DEM), and a lithology map. The script exploits Whitebox Geospatial tools for geographic information system (GIS) analyses, including the extraction of a set of terrain attributes from the DEM, which serve as predictors for the spatial distribution of landslides in addition to lithology. The Random Forest (RF) model, implemented through the R library "randomforest," is employed as modeling technique to identify potential relationships between landslides spatial distribution and geographical variability of predictor variables. The R script is applied to two test areas where landslide inventories, DEMs, and lithology layers are available. The first test area is in El Salvador, north of the Ilopango caldera. The second corresponds to the Tegucigalpa area, the capital of Honduras. The results underscore the importance of integrating multiple datasets and advanced statistical methods to achieve efficient landslide prediction. The script's application to these diverse regions demonstrates its robustness and versatility in different geological and topographical settings. In conclusion, this approach provides a valuable and reproducible methodology for landslide susceptibility assessment to researchers and specialists involved in landslide risk management and mitigation efforts, which is reliable and simple to use wherever the necessary input data is available. The research was developed in the framework of the CASTES project, funded by the Italian Agency for Development Cooperation (AICS).