Prediction model of the occurrence probability of bark-stripping by sika deer (Cervus nippon) in plantation forests in Kumamoto prefecture

We analyzed the characteristics of 283 sites where sika deer aused bark-stripping damage in plantation forests in Kumamoto Prefecture, between 2000 and 2001. We developed four prediction models of the occurrence probability of bark-stripping damage using logistic regression. As independent variables, we used 16 factors, including topographic (elevation and slope angle), geographic (distance from the nearest road, etc.), vegetation (the number of cells used for forestry, agriculture, etc.), and stand (planted species and stand age) factors. The model with the lowest Akaike Information Criterion (AIC) and the greatest accuracy of classification showed that bark-stripping damage was strongly related to elevation, slope angle, and distance from the nearest road. The accuracy of the model was improved when the factors "planted species" and "stand age" were added. Finally, we drew a map showing the probability of bark-stripping damage. Such information should prove very useful in forest management planning.