This repository reports the data input and output used to evaluate the regional landslide susceptibility in West Tajikistan. The analysis has been conducted subdividing the study area into Unit Condition Units (UCU) intersecting catchments and lithology. The average and standard deviation of 9 geo-environmental variables per UCU have been collected in the dataset: slope degree, relative relief, plan curvature, profile curvature, Peak Ground Acceleration, annual precipitation, land use, lithology and area. The target variable is instead a binary label assigned to the UCU depending of the presence or absence of landslides. The model to estimate the susceptibility corresponds to a Bayesian version of a binomial Generalized Additive Model implemented in R-INLA. As a result the posterior mean of the susceptibility was estimated together with the uncertainty around it. The experiment involves testing the variation induced by an increasing incomplete landslide inventory. The covariates per UCU and the landslide susceptibility of Tajikistan are reported in the spatial dataset published as vector format.

The file is a spatial vector dataset loaded in .gpkg (GeoPackage). It can be managed by GIS software.