An evaluation of remotely sensed indices for quantifying burn severity in arid ecoregions

It is sometimes assumed the sparse and low statured vegetation in arid systems would limit the effectiveness of two remote-sensing derived indices of burn severity: the difference Normalised Burn Ratio (dNBR) and relativised difference Normalised Burn Ratio (RdNBR). Wecompared the relationship that dNBR, RdNBR and a ground-based index of burn severity (the Composite Burn Index, CBI) had with woody cover and woody density 1 year after burning in five fires that occurred in the Mojave Desert during 2005. Data were collected within 437 plots spanning geographic and elevation gradients representative of vegetation associations in low- (,1200 m), mid- (1200 to 1700 m) and high-elevation (.1700 m) zones. Statistically, dNBR and RdNBR were both effective measures of severity in all three elevation zones; woody cover and density had steep exponential declines as the values of each remote-sensing index increased. We found though that dNBR was more ecologically interpretable than RdNBR and will likely be of most relevance in the Mojave Desert. It will be necessary though to test these, as well as other remote-sensing burn-severity indices, across more desert regions before inferences can be made of the generality of the patterns we observed.