Short-Term Observations of the Temporal Development of Active Fires From Consecutive Same-Day ETM+ and ASTER Imagery in the Amazon: Implications for Active Fire Product Validation

In this study, we used same-day 30-m spatial resolution Landsat-7/ETM+ and Terra/ASTER data to study the short-term development of active fires in the Brazilian Amazon between the overpasses of the two satellites at approximately 10:00 and 10:30 local times, respectively. We analyzed the spatial progression of fire fronts and the temporal changes in the extent of burning at the scales of the pixel sizes of MODIS (1 km) and GOES Imager (4 km). The progression of fire fronts varied between individual fires, but in most cases remained within the scale of a few 30-m pixels, while the total extent of burning detected typically increased during the 30 min between the ETM+ and ASTER observations. This is in accordance with the typical mid-morning upslope part of the diurnal cycle of fire activity observed previously by coarse resolution sensors. We also assessed the potential changes in derived validation results of active fire products from medium and coarse resolution sensors. We derived fire detection probabilities from Terra/MODIS as a function of the total number of 30-m fire pixels from ASTER (representing simultaneous reference data and, hence, the ldquotruthrdquo) and separately from ETM+ (representing nonsimultaneous reference data). We found spuriously increased detection probabilities using ETM+ resulting from the observed increase of fire activity between the ETM+ and ASTER acquisitions. While this effect can potentially be corrected for by a statistical adjustment, the overall recommendation is that a temporally unbiased sample of nonsimultaneous reference data should be used for validation.