Forest resilience to fire in the western US: a test case for using satellite metrics to assess spatial and temporal patterns of forest recovery

Abstract:

Fire and other climate driven disturbances are shaping the future of our forests. Given the significance of forests and what their loss represents to humankind, it is crucial that we manage forests for their continued existence under contemporary climate conditions. This requires the immediate investigation of forest resilience, the capacity of a forest to return to its original state following a perturbation. The available research on forest resilience to fire is limited to small study regions, focusing on field observations that are limited in their ability to characterize large swaths of forests. It is critical to accurately measure postfire vegetation recovery when assessing forest resilience. Here we demonstrate a new methodology for broad scale assessment of forest resilience by combining non-spatial and spatial statistical models to characterize the recovery of a 11,400 acre forest in northwestern Montana, 15 years postfire. We used a generalized additive model with 33 years of satellite derived net primary production data to determine percent recovery across the landscape at 30m resolution. We used a spatial autonormal model to test for the presence of autocorrelation in percent recovery and the explanatory power of covariates. As expected, higher burn severity resulted in reduced recovery. For a 1 unit increase in burn severity, the percent recovery is predicted to decrease by between 0.10-0.11 percentage points (95% CI). The topographic position index (TPI), continuous heat-insolation load index (CHILI) and elevation were also significant predictors of postfire recovery. Following burn severity, elevation was the greatest predictor of recovery; however, we were surprised to find low elevation sites were more likely to experience higher rates of recovery. For a 1 meter decrease in elevation, the percent recovery is predicted to increase by between 0.15-0.16 percentage points (95% CI). TPI and CHILI, respectively were the next most important predictors. Combining these results, percent recovery was predicted to be greatest in forested areas with a lower burn severity, at lower elevations, in convergent landscapes and in areas with lesser heat loads. This methodology will be expanded to fires across the western US to help identify local and regional patterns of resilience to inform management practices.