Passive ground-based analyses for interpreting satellite fire data - Applications to AVHRR and MODIS active fire detections in Amazonia
Manoel
Cardoso, Complex Systems Research Center, University of New Hampshire, Durham, NH 03824 United States, manoel.cardoso@unh.edu
(Presenting)
George
Hurtt, Complex Systems Research Center, University of New Hampshire, Durham, NH 03824 United States, george.hurtt@unh.edu
Berrien
Moore III, Complex Systems Research Center, University of New Hampshire, Durham, NH 03824 United States, b.moore@unh.edu
Carlos
Afonso
Nobre, Centro de Previsao de Tempo e Estudos Climaticos, Rod. Pres. Dutra, Km 40, Cachoeira Paulista, SP 12630-000 Brazil, nobre@cptec.inpe.br
Heather
Bain, Complex Systems Research Center, University of New Hampshire, Durham, NH 03824 United States, h_bain03@yahoo.com
Because of their broad spatial and temporal coverage, satellites are very important for providing information on fire activity in Amazonia. A key to the application of these tools for environmental studies is the accurate interpretation of the data they provide. Examples of factors that should be considered include temporal sampling, cloud coverage, fire intensity below detection, and confounding reflective surfaces. To enhance the interpretation of satellite data for this region, we collected ground-based information on fire activity and statistically related them to corresponding satellite-based data. Ground-based data were collected in Para in 2001 and in Mato Grosso in 2002 using a simple and passive method. Corresponding fire data from AVHRR and MODIS were then obtained and related to the ground-based data using error matrixes. Results from these analyses indicate that the total accuracy for both fire products was very high and dominated by accurate non-fire detection. Fire-detection accuracy was lower, and errors of commission were less than errors of omission. Satellite fire products differed in the frequency of omission and commission errors for fires. Omission errors were lower for AVHRR, and commission errors were lower for MODIS. Preliminary attribution studies suggest that sampling time, fire size and land cover are important complicating factors for active fire detection in the region. Results from this study show that passive ground-based analyses can substantially contribute to the interpretation of satellite fire data.
Submetido por Manoel Ferreira Cardoso em 18-MAR-2004
Tema Científico do LBA: LC (Mudanças dos Usos da Terra e da Vegetação)