Evaluating the potential of Landsat TM/ETM+ imagery for assessing fire severity in Alaskan black spruce forests

Satellite remotely sensed data of fire disturbance offers important information; however, current methods to study fire severity may need modifications for boreal regions. We assessed the potential of the differenced Normalized Burn Ratio (dNBR) and other spectroscopic indices and image transforms derived from Landsat TM/ETM+ data for mapping fire severity in Alaskan black spruce forests (Picea mariana) using ground measures of severity from 55 plots located in two fire events. The analysis yielded low correlations between the satellite and field measures of severity, with the highest correlation (R2adjusted = 0.52, P 0.0001) between the dNBR and the composite burn index being lower than those found in similar studies in forests in the conterminous USA. Correlations improved using a ratio of two Landsat shortwave infrared bands (Band 7/Band 5). Overall, the satellite fire severity indices and transformations were more highly correlated with measures of canopy-layer fire severity than ground-layer fire severity. High levels of fire severity present in the fire events, deep organic soils, varied topography of the boreal region, and variations in solar elevation angle may account for the low correlations, and illustrate the challenges faced in developing approaches to map fire and burn severity in high northern latitude regions.