The Relationship of Multispectral Satellite Imagery to Immediate Fire Effects

Abstract

The Forest Service Remote Sensing Applications Center (RSAC) and the U.S. Geological Survey Earth Resources Observation and Science (EROS) Data Center produce Burned Area Refl ectance Classifi cation (BARC) maps for use by Burned Area Emergency Response (BAER) teams in rapid response to wildfi res. BAER teams desire maps indicative of fi re effects on soils, but green and nonphotosynthetic vegetation and other materials also affect the spectral properties of post-fi re imagery. Our objective was to assess how well satellite image-derived burn severity indices relate to a suite of immediate post-fi re effects measured on the ground. We measured or calculated fi re effects variables at 418 plots, nested in 50 fi eld sites, located across the full range of burn severities observed at the 2003 Black Mountain, Cooney Ridge, Robert, and Wedge Canyon wildfi res in western Montana, the 2003 Old and Simi wildfi res in southern California, and the 2004 Porcupine and Chicken wildfi res in interior Alaska. We generated the Normalized Burn Ratio (NBR), differenced Normalized Burn Ratio (dNBR), Relative dNBR (RdNBR), Normalized Difference Vegetation Index (NDVI), and differenced NDVI (dNDVI) burn severity indices from Landsat 5 Thematic Mapper (TM) imagery across these eight wildfi res. The NBR correlated best with the fi re effects measures but insignifi cantly, meaning other indices could act as suitable substitutes. The overstory (trees in Montana and Alaska, shrubs in California) measures appear to correlate best to the image variables, followed by understory and surface cover measures. Exposed mineral soil and soil water repellency were poorly correlated with the image variables, while green vegetation was most highly correlated. The BARC maps are more indicative of post-fi re vegetation condition than soil condition. We conclude that the NBR and dNBR, from which BARC maps of large wildfi res in the United States are currently derived, are sound choices for rapid assessment of immediate post-fi re burn severity across the three ecosystems sampled. Our future research will focus on spectral mixture analysis (SMA) because it acknowledges that pixel refl ectance is fundamentally a mixture of charred, dead, green and nonphotosynthetic vegetation, soil, rock and ash materials that are highly variable at fine scales.