Remotely sensed indices of burn severity are commonly used by natural resource managers for wildfire recovery planning and by researchers studying fire behavior and ecology across communities and over time. NASA’s Landsat satellites have long been used for mapping burn severity at 30-meter resolution, however the recently launched pair of Sentinel 2 satellites have improved sensors providing both 10- and 20-meter resolution in the wavelengths used for estimating burn severity. Because burn severity can vary across a fire at fine spatial scales, Sentinel’s increased resolution provides an opportunity to quantify the heterogeneity of burn severity at sub 30-meter scales in order to contextualize past work and inform future research on the appropriate scale to measure the ecological effects of fire. We utilized Google Earth Engine to assess both the scale of burn severity heterogeneity and the factors that influence it across fires occurring in different ecoregions of the western US between 2016 and 2018.
Results/Conclusions
We found that burn severity was significantly more heterogeneous within pixels classified as moderate severity under a categorical ranking system, suggesting that with higher resolution imagery it is possible to partially resolve mixed severity into either low or high severity fire. This helps expose small unburned or surviving pockets of vegetation, known as refugia, which play a critical role in post-fire recovery. While the factors influencing burn severity heterogeneity varied in strength across ecoregions, important predictors included vegetation heterogeneity, species composition, and topographic complexity, when top-down effects from daily fire weather were relatively weak. This highlights the important influence of bottom-up factors of landscape variability on burn severity under moderate fire weather, while these factors tended to break down under more extreme weather. We also tested the sensitivity of landscape metrics quantifying the patch size and arrangement of stand replacing fire to imagery resolution. We found that estimates of patch core area and distance to edge both decreased at higher resolutions due to greater fragmentation of high severity patches as small refugia were exposed. These results point to the potentially important contribution of new sources of remote sensing data when analyzing the ecological effects of wildfire.