Test of a mountain pine beetle winter mortality model

Winter mortality of mountain pine beetle larvae caused by unexpectedly low winter temperatures is an important determinant of mountain pine beetle population dynamics. We tested a widely used mountain pine beetle winter mortality model developed by researchers at the Canadian Forest Service and the United States Forest Service. lt predicts larval survival probabilities given temperature time series over the winter season. To validate the model, we compared its predictions to observed lower lethal temperature thresholds and cold-associated mortality. Model predictions were biologically reasonable--especially when the model was driven using observed under-bark temperatures. However, when the model was driven using predicted under-bark temperatures, its predictions were slightly biased due in part to inaccurate translation of air temperature data to temperatures under the bark where larvae develop. Relative mortality predicted across the study area in Banff National Park was not well predicted by the model. We speculate that the poor predictive performance in this mountainous study region is likely because the topography presents a difficult prediction challenge for the model. We hope that our results and data will inform users of this model of its constraints and how to optimize the accuracy of model predictions.

Methods
At our detailed study site, we collected hourly time series of under-bark temperature in mountain pine beetle-infested trees on the north and south sides. In addition, we measured larval supercooling temperatures to quantify their level of cold hardiness, and assessed larval winter survival by pre and post-winter population surveys within each of our study trees. Observations were compared to model predictions when the winter mortality model was driven using under-bark temperatures.

At our spatially distributed sites, we assessed population success by quantifying the number of parents, and surviving progeny in the spring (after winter). Survival estimates based on these data were compared to model predictions of winter survival probability made using the mountain pine beetle winter mortality model as implemented in the publicly available BioSIM software.