Conclusions

Multiple disturbances, such as wildfire-insect outbreak interactions, are not well understood at provincial scales. Lack of large-scale datasets and easily available analytical techniques to address multiple disturbances have contributed to a limited number of quantitative studies. These studies have had confounding results. Here, we have described a probabilistic model framework that was used in a case study of spatiotemporal interactions between wildfire and insect outbreaks in the Pacific Northwest. To create the model, we used historic spatial data (1980–2004) on fire occurrence and insect outbreaks on forested lands of Oregon and Washington along with average monthly temperature and PDSI.

Results of our analyses seem to imply an increasing risk of new bark beetle infestations with increasing size of nearby fires (within 1 km) up to fires of ~750 ha. The latter was indicated by the estimated odds relative to locations with no fire damage. Also, there appears to be a decrease in risk of a fire getting large at locations with 3-6-year-old spruce budworm damage of size greater than 100 ha.

The methods we employed provide an objective tool for modeling complex processes with time lags in the predictors and estimating associated probability maps for competing risks (multiple disturbances). The framework is especially amenable to quantifying uncertainties from multiple sources (stochastic disturbance, sampling and measurement errors, approximate models, and missing predictor variables), which are essential for model appraisal purposes. The analyses can be expanded to other areas using the rich, spatially explicit, historical data set available for wildfire occurrence across all of the United States (NIFMID). It also has application for forest insects and diseases through Forest Health Protection annual aerial survey data. Although we only examined three insect species (mountain pine beetle, western pine beetle, and western spruce budworm), other stressors could be considered, including nonnative invasive species and stress level due to pollution. Risk assessment maps can be produced by multiplying the estimated probabilities by a loss function.