Abstract This study is pertaining to an evaluation of landslide occurrence on natural terrain due to snowmelt in Japan, using a probabilistic model based on multiple logistic regression analysis. The evaluation concerns several physical parameters such as hydraulic parameters, geographical parameters and geological parameters which are considered to be influential in the occurrence of landslides. A Snow Water Equivalent model (SWE) is utilized to estimate snowmelt and associated infiltration in light, heavy and normal snow years. Using the constructed spatial data-sets, we apply a multiple logistic regression model to produce landslide susceptibility maps showing the spatial–temporal distribution of landslide hazard probabilities throughout Japan using 1 km × 1 km resolution grid cells. The results have revealed that, over 95% landslide hazard probability exists in the mountain ranges on the western side of Japan (the Japan Sea side). In particular, this study is dealing with the Aizu region of Fukushima prefecture in order to verifying the landslide hazard probability. Verification proved that, the areas identified as high risk areas (having over 90% landslide hazard probability in numerical modeling) show 87% agreement with observed landslides in the Aizu region. Also we evaluated the relationship between landslides and snow melting process giving special concern to change of temperature in the spring.