Finding the Limits of GOES-R ABI Fire Detection

“How big of a fire can GOES see?” was the first question asked when biomass burning research using the Geostationary Operational Environmental Satellites (GOES) began in the 1990s. And it remains one of the most common questions today. The answer is not straightforward. The radiative output of the fire compared to the local environment, when considering the capabilities of the detector and any downstream data modification, determines the detectability of a fire. A small fire that is seen clearly at night may be indiscernable during the day because the reflected solar flux exceeds that added by the fire. A large, smoldering fire under a forest canopy will likely be hard to detect. A smaller but hotter fire with the same, if not less, radiant flux, such as a burning structure surrounded by grass, could show up at the same time. A lack of reliable data about fire start and end times, extents, fuels burnt, etc has made parameterizing the bounds of fire detection difficult. The current GOES Advanced Baseline Imager (ABI) sees more fires than ever, and the lower limit for detection appears to be lower than anticipated. By combining various types of fire information, such as the records maintained by prescribed fire practitioners, the carefully investigated, officially reported information about the starts of major events, and even media reports on structure fires, we can begin to put bounds on the properties of the fires that we can see. The data from prescribed fires in particular can be extremely useful, as it may contain not just start and end times and a centroid, but a polygon of the burn area, an assessment of the types of fuels and their moisture content, and the weather conditions at the time. Coordinating with the prescribed fire community yields symbiotic benefits, as the automated detection data produced by the current operational Fire Detection and Characterization (FDC) algorithm and the forthcoming Next Generation Fire System (NGFS) can aid them in cataloging the controlled burns that are not reported directly to the regional prescribed fire councils and state and local agencies. The first results of an examination of an expanded validation dataset will be presented, including an examination of the conditions that can confound fire detection.