2005 Volume 60 Issue 5 Pages 439-444
Two different predictive distribution models, the one for forests and the other for individuals, of Fagus crenata were developed in central Japan. Four climatic variables were used as predictor variables and two different types of vegetation data sets were respectively used as response variables. Tree-Based Models were adopted for the model development. The dataset for the F. crenata forests was 1-km2 vegetation data based on an actual vegetation map, and the other dataset for F. crenata individuals was obtained from over 4,500 Phytosociological Relevé Data Base (PRDB). Predicted presence/absence of F. crenata forests and individuals for every 1-km2 grid square were classified into four groups. Grid squares classified as ‘F. crenata forest absence’ and ‘F. crenata individual presence’ covered 5.6 % of the study area. These areas consist of relatively higher proportions of deciduous Quercus crispula-Castanea crenata forest and Q. serrata-Cerasus verecunda forest at low elevations, and Abies mariesii forest and Betula ermanii forest at high elevations. Two important threshold warmth index (WI) values for F. crenata individuals were found to be 38.5 and 93.2. This was wider than the WI range of 46.2 and 87.8, which were found in the F. crenata dominant forest model. These findings revealed that PRDB is a useful source for modeling plant species distributions and for impact assessment studies of climate change.