For the past 10 years the Cooperative Institute for Meteorological Satellite Studies (CIMSS) at the University of Wisconsin - Madison has used the GOES series of satellites to monitor fires and smoke in the Western Hemisphere. To date most of our investigations have concentrated on using multi-spectral GOES-8 imagery (visible, 3.9 10.7, and 12 microns) to identify and catalogue fire activity in South America associated with deforestation, grassland management, and agricultural applications.

This 4-panel composite shows how fires are observed in South America in multi-spectral GOES imagery. Each panel represents a region 780 by 800 km centered in central Brazil at 1:45 P.M. local time on 24 August 1995. The milky regions in the GOES-8 "visible" panel show smoke plumes emanating from fires burning along the boundary between the selva (forest, on the left) and the cerrado (grassland, on the right). The GOES-8 3.9 and 10.7 micron bands are located in the infrared window region of the electromagnetic spectrum. In cloud-free conditions these bands provide a "window" through the atmosphere that enable us to better observe the earth's surface. In the 4-panel composite the 11 (or 10.7) and 4 (or 3.9) micron panels have been color enhanced to highlight the variations in GOES-8 observed brightness temperatures along the boundary between the selva and cerrado. The color scale is located below the 4 micron panel. Both the 4 and 11 micron panels show the contrast between the cooler tropical forest in the west and the warmer grassland in the east. The 4 micron panel also shows a number of hotter regions corresponding to fires along the forest/grassland boundary and along a road in the west. The 4-11 micron panel shows the differences between the 3.9 and 10.7 micron bands. Typically the difference is about 5 Kelvin due to differences in atmospheric transmittance, surface emissivity, and solar contamination (during the day) in the 3.9 micron band. The differences become larger when part of the pixel contains fire activity. The 3.9 micron band is much more sensitive to sub-pixel hot spots. The hotter portion of the pixel will contribute relatively more radiance in the shorter wavelengths than in the longer wavelengths (Prins and Menzel, 1992; 1994)

In order to monitor fire activity in GOES imagery for large areas the fire identification process was automated. CIMSS developed the Automated Biomass Burning Algorithm (ABBA) to characterize sub-pixel burning throughout South America. The GOES ABBA is a contextual multi-spectral thresholding algorithm which utilizes dynamic local thresholds derived from the GOES satellite imagery and ancillary databases to locate fire pixels and provides estimates of the sub-pixel area and mean temperature of the fire(s) (Prins and Menzel, 1992; 1994). The GOES-8 ABBA fire product includes: fire location (lat./lon.), estimates of fire size and temperature, 3.9 and 10.7 micron observed brightness temperatures, background brightness temperatures, albedo statistics, ecosystem type, and a flag for non-processed fire pixels to indicate the reason for not processing.