Field Assessment of Hot Pixel Data: A Case Study of the Alcobrás Settlement Wildfire to Evaluate the Accuracy of AVHRR, GOES, and MODIS Sensors as Indicators of Biomass Burning Events
Leigh
Johnson, Setor de Estudos do Uso da Terra e Mudanças Globais, Parque Zoobotânico, Universidade Federal do Acre, leighjohnson@fulbrightweb.org
Jorge
Henrique
Garcia, Federação dos Trabalhadores na Agricultura do Estado do Acre, fetacre@uol.com.br
Diogo
Selhorst, Setor de Estudos do Uso da Terra e Mudanças Globais, Parque Zoobotânico, Universidade Federal do Acre, dselhorst@pop.com.br
(Presenting)
Elsa
Mendoza, Instituto de Pesquisa Ambiental da Amazônia & Setor de Estudos do Uso da Terra e Mudanças Globais, Parque Zoobotânico, Universidade Federal do Acre, elsa_mendoza@uol.com.br
Erika
Nascimento, ProArco/IBAMA, erika.nascimento@ibama.gov.br
Irving
Foster
Brown, Woods Hole Research Center & Setor de Estudos do Uso da Terra e Mudanças Globais, Parque Zoobotânico, Universidade Federal do Acre, fbrown@uol.com.br
Manuel
Cesario, Observatório da Amazônia Sul-Ocidental em Saúde Coletiva e Ambiente, Universidade Federal do Acre, manuel.cesario@uol.com.br
Fire is a revealing indicator of changing land use patterns in Amazonia. As densely settled colonization projects increase across the region, agricultural fires are ever more likely to slip out of control and become wildfires, producing intense, continuous burning events. This study analyses one such event, comparing the utility of AVHRR, GOES, and MODIS hot pixel data for monitoring the fire’s development over time and space. Between 16 August and 9 September 2003, a several thousand-hectare wildfire developed in the contiguous Alcobras and Zaqueu Machado colonization projects (~600 families in 15-ha lots, latitude -10.295 to -10.465 and longitude -67.715 to -67.885) in the Municipality of Capixaba, Acre State, Brazil. Community mapping activities have both confirmed and contradicted satellite data; they have also provided a property-level picture of the fire's evolution. AVHRR, GOES, and MODIS hot pixel data sets each yield different pictures of the fire's progression, due to the respective satellite's overpass time and frequency, spatial resolution, and the processing algorithm used. While AVHRR data from INPE show 45 points spanning a period of 13 days, with burning events peaking on August 27, MODIS data provided by the University of Maryland show 111 points spanning a period of 19 days, with a burning peak on August 27 and another crest on September 1. GOES data processed by INPE display 12 points spanning a period of 17 days, with burning events peaking on August 28. GOES data for the same area, processed by the University of Wisconsin, display 103 points spanning a period of 24 days, peaking on August 29. Preliminary results indicate that while AVHRR and MODIS-U Maryland data provide more accurate and nuanced spatial histories of the wildfire, GOES-U Wisconsin data present a superior temporal illustration of the fire's evolution. The study demonstrates the benefits of community-level validation of sensor accuracy and the importance of developing capacity for continued field monitoring.
Submetido por Leigh Johnson em 25-MAR-2004
Tema Científico do LBA: LC (Mudanças dos Usos da Terra e da Vegetação)