Development of a hazard map for oak wilt disease in Japan

摘要

We constructed a hazard map for oak wilt disease in Yamagata and Fukushima Prefectures, Japan, based on a logistic regression model. Within the framework of a regression model, we incorporated data from previously damaged areas on vegetation type, geography and meteorology, and pest immigration. We evaluated the predictive power of the model using defoliation records within the same prefecture taken from the subsequent year. The results obtained for Yamagata Prefecture were highly predictive and we conclude that practical hazard maps can be constructed based on damage records from the previous year combined with other environmental data. The predictive value of the model was improved dramatically by incorporating an immigration effect. The model used to predict damage in Yamagata Prefecture was correct for only 46.7% of the actual damage in Fukushima Prefecture. Several difficulties were identified regarding the application of the model in other prefectures. Based on our results, we suggest that defoliation data should be collected quickly in the target area when constructing a localized regression model that includes immigration. Collection and organization of defoliation data could help to quickly draw up hazard maps before oak wilt disease begins to occur in a given year.