Fig. 2

Flow diagram of the study. a The snail distribution dataset in the form of a spatial shapefile including ‘presence–absence’ records (left); re-assessment to correct the spatial bias (middle); environmental and geographic predictor variables prepared as raster files of the area (right). b Random forest model first fitted with the training set followed by testing by the validation set. c Probability of snail occurrence (left) and risk output based on thresholds (right)