Statistical modeling to predict community change for large geographical scale is difficult when community responses to similar climate changes differ among regions. This study attempted to present a solution for this problem by predicting species responses to local environment. Our model for predicting community change consisted of species response models to local environmet parameters, in which the estimated abundance of each species was used to reconstruct the community composition. Our method was verified by data from continuous surveys at Hyangrobong, Odaessan and Taebaeksan in Gangwon-do, in which environmental factors and ground beetle community at each site were continuously recorded. We show our method performes better than an ordinary model that predicts directly a change in the community.