Spatial prediction of the onset of spruce budworm defoliation

Abstract

A logistic regression model for spatially explicit predictions of the likelihood of an onset of stand-level spruce budworm (Choristoneura fumiferana, Clemens) defoliation in a 15 000 km2 study area in northern British Columbia, Canada is developed. Predictions are derived from stand (volume and needle biomass) and topographic attributes (distance from nearest river, distance from previous year defoliation, and stand elevation) collected during the first 12 years of a current budworm outbreak. The likelihood of an onset of defoliation increased with an increase in stand volume and biomass of current needles and it decreased with an increase in the distance to the nearest river and to the nearest stand with a defoliation recorded for the year prior to the year of prediction. Stands located at higher elevations sustained less defoliation than stands located at lower elevations. A single model is assumed adequate for predictions throughout an entire outbreak cycle. Observed and predicted relative frequencies of locations with an onset of defoliation compared relatively well. Key words: Choristoneura fumiferana; Picea; logistic regression; nearest neighbour distance