We used classification tree analysis to develop a climate-based distribution model for Fagus crenata forests in Japan. Four climatic variables judged likely to affect the distribution of the species (summer and winter precipitation, minimum temperature of the coldest month and Kira's warmth index) were chosen as independent variables for the model. Latitudinal and longitudinal information was also used to examine effects of spatial autocorrelation on the model. The climatic factors associated with the distribution of the forests were analysed using a classification tree to devise prediction rules. Predicted areas of high probability for forest occurrence lay mainly on the Sea of Japan side of northern Honshu and southern Hokkaido. This is consistent with actual forest distribution. Some areas with high predicted probabilities of F. crenata forest occurrence were beyond the current natural northern range limits of these forests. Since these areas were widely scattered, it was assumed that the species has been hindered from colonizing them due to dispersal limitations. Deviance-weighted scores, used to compare magnitudes of the contributions of predictor variables, revealed winter precipitation as the most influential factor, followed by the warmth index, the minimum temperature of the coldest month and summer precipitation. Attempts were made to generate ecological explanations for the effects of the four climatic factors on the distribution of F. crenata forests.
Abbreviations: CA = Classification accuracy; DWS = Deviance weighted score; JMA = Japan Meteorological Agency; MER = Misclassification error rate; TMC = Minimum temperature of the coldest month; NSNE = National Survey on the Natural Environment; OE = Omission error; PRS = Summer precipitation; PRW = Winter precipitation; RMD = Residual Mean Deviance; WI = Warmth index.
Nomenclature: Ohwi & Kitagawa (1992).