Spatial and temporal patterns of Landsat-based detection of tree mortality caused by a mountain pine beetle outbreak in Colorado, USA

• Landsat imagery was used to predict percent beetle-caused tree mortality.
• Cumulative tree mortality was 22% of the study area, equivalent to 228 million trees.
• Mortality patterns showed expansion early and intensification later in the outbreak.
• Plot-scale outbreak areas (30-m grid cells) experienced 60% mortality on average.
• Beetles moved through plot-scale areas on average in 3–4 years and killed 20%/year.

Mountain pine beetles (Dendroctonus ponderosae Hopkins) have killed billions of trees in the United States and Canada in the last decades, thereby causing major alterations to forest ecosystems. Therefore, monitoring the extent and patterns of these major disturbance events are important for subsequent forest management. At 30-m spatial resolution, Landsat imagery affords the opportunity for quantifying spatial and temporal patterns of bark beetle outbreaks at plot to stand resolution. We developed a continuous measure of bark beetle-caused tree mortality using multi-temporal Landsat data (1996–2011) in the Rocky Mountains of northcentral Colorado. We report the year of detection of tree mortality, which is 1 year after attack by beetles. Two approaches were used to predict percent red stage tree mortality within a 30-m grid cell, multiple linear regression models and generalized additive models (GAM) with a nonlinear spatial term. Both models explained >75% of the variance using three Landsat spectral explanatory variables. We used the linear model to predict tree mortality across the entire study area and time series because of its simplicity, its capability for extrapolation beyond the training area, and similar performance compared with the GAM. From 1996 to 2011, cumulative tree mortality was 22% of forested areas within the Landsat scene, equivalent to 228 million mature lodgepole pine trees (range: 174–332 million trees using the 95% confidence interval of field-measured crown areas). Early in the outbreak, tree mortality was associated with expansion, whereas later in the outbreak, mortality was due to intensification (increase in mortality within areas already having beetle activity). We used three metrics (cumulative tree mortality, duration of tree mortality, and average rate of tree mortality) to investigate the temporal and spatial patterns of bark beetle-caused mortality using different grid cell resolutions within the Landsat scene. At the Landsat spatial resolution (900 m2), we estimated that grid cells within the outbreak experienced means of 60% mortality, a duration of 3–4 years, and an average rate of mortality of 20%/year. With coarser spatial resolution, cumulative tree morality and rate of tree mortality decreased whereas duration of tree mortality increased. Our results improve the understanding of spatial and temporal patterns of tree mortality caused by bark beetle outbreaks, and provide specific information for forest managers and scientists about a severe mountain pine beetle infestation in northcentral Colorado.