Landslides have emerged as one of the most important disturbance agents in mountainous ecosystems. In mountainous regions in Japan, the locations of old landslides often have heterogeneous vegetation that is peculiar to landslide areas. However, little is known about how and why vegetation patterns differ among landslide sites within a region. In this study, I describe the variations in vegetation at different landslide sites within a region and examine the possible causes of the regional diversity of vegetation at these sites. I classified the vegetation on 1703 landslides by interpreting aerial photos of the 1421 km² study region. Nine types of vegetation were classified. The results show that the heterogeneity of the vegetation increased with the maximum snow depth, and the number of vegetation types on landslide bodies increased with latitude and altitude. Among the 1703 landslides studied herein, 62 included mires (wetlands). Sites with mires were concentrated in areas with heavy snowfall and certain lithologies, such as ultramafic rocks and nonalkaline mafic volcanic rocks. My analysis using Random Forest, a machine-learning algorithm, indicates that maximum snow depth and summer temperatures had the greatest impact on the number of vegetation types and mires on the landslide bodies. The results of this study suggest that snowy conditions have long-term effects on vegetation structure and diversity on old landslide sites.