Mapping Percent Tree Mortality Due to Mountain Pine Beetle Damage

The mountain pine beetle (Dendroctonus ponderosae Hopkins) is a pervasive and particularly destructive species of insect that has killed vast areas of conifers in western North America. To map large areas of infestation satellite imagery is often used because of its ability to cover large areas, but the spatial resolution often precludes fine-scale analysis. Furthermore, maps of insect-caused tree mortality have been attempted on the basis of presence-absence. We present here a two-step method to estimate and map tree mortality as a percentage within a Landsat-sized pixel. The first step delineates presence-absence, and the second estimates percent tree mortality for those pixels with tree mortality. We tested a total of 25 two-step models. The binary presence-absence first step of the models produced overall accuracies between 89.9 and 98.0% and Cohen’s kappa values between 0.69 and 0.94. The full two-step models that predicted percent tree mortality produced estimated differences between modeled and observed percent tree mortality that ranged from 0.1 to 2.4% with associated root mean square errors between 10 and 18%. This method has several advantages over most current methodologies, including the use of a single image, nondependence on wetness measures, magnitude of error that is not density-dependent, and no restriction to trees with red-shaded faded crowns.