Recent ComparisonValidation Studies of the Wildfire ABBA WFABBA

Recent Comparison/Validation Studies of the Wildfire ABBA (WF_ABBA) in North and South America Joleen M. Feltz*, Michel Moreau^, Elaine M. Prins+, Kirsten Mc. Claid-Cook#, Irving Foster Brown## *Cooperative Institute for Meteorological Satellite Studies, University of Wisconsin, Madison, WI ^Environment Canada/Meteorological Services/Quebec Region, Quebec, Canada +NOAA/NESDIS/ORA #State ##Woods Advanced Satellite Products Team University of New York, Albany, New York Hole Research Center-WHRC, Federal University of Acre-UFAC, Federal University Fluminense-UFF

8 -Year Comparison of 3 -Hourly GOES ABBA and AVHRR INPE Fire Products in South America: 1995 - 2002 Arc of Deforestation In every year the GOES-8 ABBA diurnal fire product identified more fire pixels in the “arc of deforestation” than the single time period NOAA (pm overpass) AVHRR INPE fire product. This study demonstrates the complementary nature of multi-sensor fire detection programs.

Validation/Comparison Studies in Western Brazil LBA Activity

The burning patterns in Acre are evident in a composite image of GOES-8 WF_ABBA • The linear features represent burning along roadways. • More intense burning is also evident along the border between Brazil and Bolivia.

Ground Truth Information for Fires in Acre During September/October 2002 Information provided by Kirsten Mc. Claid-Cook • Start dates and most start times listed for 88 fires in the Acre regi • Active fire or fire scar • Latitude/Longitude • Size in hectares • Biomass and weather information

Distribution of Biomass Type Determined by Field Observation field observations (88 observations) WF_ABBA observed fire pixels (76 observations)

Comparison of GOES WF_ABBA with. Ground Truth in Acre, Brazil GOES WF_ABBA Fire Categories 0 – Processed Fire Pixel 1 – Saturated Fire Pixel 2 – Cloudy Fire Pixel Possible Categories: 3 – High 4 – Medium 5 – Low

Number of observed fire pixels versus the percentage of region determined to be cloudy at 1745 UTC

Number of observed fire pixels versus the percentage of region determined to be cloudy at 2045 UTC

Quebec, Canada

2002 Quebec Validation Study Dates: 20 June through 31 August a. b.

2002 Validation Study in Quebec Red markers indicate fires first detected by the WF_ABBA. Filled red circles are fire pixels that were only detected by the WF_ABBA Courtesy of Michel Moreau Environment Canada, Meteorological Ser vices, Quebec Region In one case, the WF_ABBA detected a fire 17 days in advance of the first fire agency report. This fire eventually burned more than 55, 000 hectares. This fire was located in Northern Quebec where there is no need for systematic daily detection by SOPFEU

2002 Quebec Fire Season Results Fire Pixel Flag Positive Fire Detections Possible Fire Detections False Detections Processed 2775 (3030) 20 (20) 52 (133) Saturated 1979 (2001) 0 (0) 0 (6) Cloud Covered 1978 (2120) 2 (3) 3 (11) 598 (689) 1 (1) 5 (170) 61 (88) 9 (12) 18 (197) 448 (539) 75 (113) 168 (1660) High Possibility Medium Possibility Low Possibility Un. Filtered 78% confirmed 2% possible 20% false Filtered 96% confirmed 1% possible 3% false

Western United States: Wildfires in June 2002 1. Smaller fires are more difficult to identify – large view angle. 2. Mountainous topography can hide fire signal 3. Diurnal heating will cause barren desert and rocky landscape to radiate close to the 3. 9 micron saturation temperature.

Comparison of GOES WF_ABBA with Nature Conservancy Ground Truth in Minnesota GOES WF_ABBA Fire Categories 0 – Processed Fire Pixel 1 – Saturated Fire Pixel 2 – Cloudy Fire Pixel Possible Categories: 3 – High 4 – Medium 5 – Low

Conclusions • Initial validation/comparison studies throughout the Western Hemisphere have shown that a consistent geostationary satellite fire monitoring algorithm can be applied in different biomes, under varying view angles and observing conditions. • Algorithm is useful for varying user applications: initial detect, mapping of distribution of fires, etc. • Temporal filtering significantly reduces false alarms, but also eliminates some short-lived fires. • The diurnal information available from the GOES WF_ABBA can complement fire products from higher resolution systems (MODIS).