COS 161-6 - From charred trees to satellite sensors: Translating remotely sensed indices into on-the-ground measures of burn severity in Rocky Mountain forests

Thursday, August 10, 2017: 3:20 PM
B116, Oregon Convention Center
Brian J. Harvey, School of Environmental and Forest Sciences, University of Washington, Seattle, WA and Robert A. Andrus, Department of Geography, University of Colorado, Boulder, CO

Fires are key disturbances that govern the structure and function of forest ecosystems worldwide. Detecting fire effects (i.e., “burn severity”, or the degree of ecological change caused by fire), therefore, is critical to understanding how forests are changing across space and time. Satellites provide valuable wall-to-wall information on nearly all forest fires occurring over the last several decades, but quantitative relationships between satellite and field measures of burn severity have yet to be rigorously tested in many regions. In this study, we tested a widely used satellite burn severity index [Relative differenced Normalized Burn Ratio (RdNBR)] against quantitative field measures of burn severity collected from nearly 600 plots distributed across biophysical gradients (e.g., latitude, topography, pre-fire forest structure) in the US Rocky Mountains. Specifically, we asked: 1) what is the direct relationship between RdNBR and field measures of burn severity (e.g., tree mortality, charring on trees, surface charring)? and 2) Are relationships between RdNBR and field measures of burn severity affected by factors such as latitude, topographic setting (e.g., slope and aspect), forest structure (e.g., basal area), or recent pre-fire disturbances (e.g., bark beetle outbreaks)?


Overall, we found that RdNBR corresponded reliably to field measures of burn severity throughout the US Rocky Mountains; root-mean square error across models ranged from 20 to 23%. However, a few key exceptions were detected. Relationships between field measures and RdNBR were better for tree-canopy (R2 ranging from 0.71 to 0.73) than for forest-floor (R2 ranging from 0.54 to 0.63) measures of burn severity. Models for different field measures of burn severity also varied in their relationship with RdNBR. That is, the same value of RdNBR corresponded to uniquely different fire effects (e.g., tree mortality vs. char height) on the ground. The relationship between field and satellite measures of burn severity was mostly unaffected by non-fire-related biophysical variables (e.g., latitude or topography). One key exception was that the relationship between field measures of fire-caused tree mortality and RdNBR weakened (i.e., shallower slope) where there was greater pre-fire tree basal area and where there was greater pre-fire beetle outbreak severity. Therefore, caution should be taken when interpreting RdNBR across burned areas with wide variability in pre-fire tree biomass or where recent prior disturbances (e.g., bark beetle outbreaks) have altered pre-fire forest structure.