We investigated distributions of 3 species of crabs,Uca lactea, Macrophthalmus japonicus and Helice tridens, and soil conditions at random 364 plots in the mouse of the Onosato River, Osaka in September 2008. We also surveyed microtopography using a terrestrial laser scanning system in this area and made digital elevation model (DEM). Classification tree models were developed using distribution data of crabs and soil and topographic condition dataset. Deviance weighted scores (DWS) of explanatory variables at these models suggested that the most influential variables for distributions of crabs was the ratio of sediment particles which diameter under 0.250 mm. DWSs of explanatory variables at other classification tree models without using soil condition dataset were indicated that influential variables were distance to vegetation, distance to tidal flat and elevation. Prediction accuracies of distributions of crabs at later models were almost same compared to prior models. Mapping predicted suitable habitats of crabs on later models generally worked well, but had some mispredictions, because these models could not represent soil conditions adequately.
