In planning habitat conservation, potential habitats should be evaluated by elucidating relationships between large-scale environment variables and the real distribution. Species - environment relationships on a broad scale are usually nonlinear and involve effects of interactions. Regression models are not capable for explaining such relationship and no standardized modeling method has been established. We, therefore, constructed a model for predicting potential distribution of forest green tree frog (Rhacophorus arboreus), by detecting thresholds of five environmental variables and applying them to a rule-based tree model. The five variables adapted for the model were forest cover, actual rainfall, maximum depth of snow, gradual slope area and annual maximum temperature. The accuracy of the model was higher when the effect of neighboring habitat was considered in the model (95%) than when it was not (75%). The result of randomized test shows that the model developed in this study is reasonably usable. The actual conservation status in local governments in Japan was evaluated through GAP analysis; comparing actually preserved areas with the potentially distributional areas. In the area along the Pacific-Ocean coast in Tohoku, Chubu and Kinki regions, habitats suitable for the tree frog are scattered and isolated, which requires well-planned and organized conservation effort. None of the local governments concerned, however, particularly Nara, Osaka and Wakayama Prefectures have no appropriate conservation policy. It is urgently required to establish measures for conservation in these prefectures.