Modeling red oak borer, Enaphalodes rufulus (Haldeman), damage using in situ and ancillary landscape data

Abstract

Red oak borer, Enaphalodes rufulus (Haldeman) (Coleoptera: Cerambycidae), has been implicated as a contributing factor to oak decline and mortality in forests of Arkansas, Missouri, and Oklahoma. A non-destructive rapid estimation procedure was used to determine red oak borer infestation histories of northern red oaks, Quercus rubra L., in a series of forest stands. Twenty-three biotic and abiotic variables in 364 vegetation-monitoring plots were analyzed for possible inclusion in a data distribution-independent machine-learning decision tree model to predict red oak borer hazard conditions on the Ozark National Forest. Decision tree models generated in this study of red oak borer damage were relatively successful in explaining patterns in the training data (71–81% overall accuracy), but relatively unsuccessful in predicting red oak borer hazard in unknown cases (42–49% overall accuracy based on cross-validation). Average clay content, distance to roads, and ridge-top topographic position were input variables that yielded the highest information content. Increased predictive accuracy likely depends on technology for optimizing the spatial aggregation scale of each input variable.