A correction for serial nonindependence in mountain pine beetle aerial survey data to reduce overestimation of cumulative damage

Abstract

At the beginning of this century, an unprecedented mountain pine beetle ( Dendroctonus ponderosae Hopkins) outbreak expanded across western North America. Methods are needed to quantify its impact on forests. Although aerial survey data of infestation are readily available, their translation to tree mortality estimates is challenging because the polygons sketched by aerial surveyors to delimit infested regions often overlap from one year to the next, which produces a lack of independence in yearly estimates of infested landscape. This problem manifests when annual proportions of infested land are summed over years because the sum often exceeds one. Cumulative proportions of infested land larger than one implies overestimation of tree mortality if initial densities of available host trees are multiplied by annual proportions of infested landscape. To address this problem, we developed a probabilistic method for correcting for nonindependence in proportions of infested land recorded using overlapping polygons. We demonstrate our approach in Jasper National Park where the probabilistic correction reduces overestimation bias in regions that were infested multiple years when predictions are compared to validation data. The approach and the spatial raster data sets we generated will be useful to forest managers seeking to quantify the impact of mountain pine beetles.