Combining Land Cover Data with MODIS Active Fire Detections to Establish Fire Type and Estimate Burned Area
Douglas
Morton, University of Maryland, morton@geog.umd.edu
(Presenting)
Jeffrey
Thomas
Morisette, NASA GSFC, jeff.morisette@nasa.gov
Ivan
Csiszar, University of Maryland, icsiszar@hermes.geog.umd.edu
Wilfrid
Schroeder, IBAMA, wilfrid.schroeder@ibama.gov.br
Ruth
DeFries, University of Maryland, rdefries@geog.umd.edu
João
Antonio Raposo
Pereira, IBAMA, jraposo@ibama.gov.br
Christopher
O.
Justice, University of Maryland, justice@hermes.geog.umd.edu
Accurate, spatially explicit measurements of fire type and burned area are necessary to estimate the magnitude of smoke emissions, carbon losses, ecosystem impacts, and climate feedbacks from fire in Amazonia. The fire type for active fire detections from moderate or coarse resolution sensors can be difficult to evaluate with high resolution imagery due to limited spatial and temporal coverage from high resolution sensors. MODIS data products at 250 m – 1 km resolution provide several possibilities for integrating regional land cover products and active fire detections to establish fire type and estimate burned area. We compare the MODIS active fire time series for 2000-2003 to the MODIS Vegetation Continuous Fields (VCF) percent tree cover and cloud-filtered 16-day NDVI time series at 250 m resolution to classify fires as deforestation or agricultural maintenance fires. For each analysis, deforestation fires are compared to field data and basin-wide deforestation maps from the INPE PRODES program to evaluate errors of omission and commission. The time series approach permits classification of land use following fire as pasture or agriculture based on the phenological signature of the NDVI time series from the subsequent wet season. We also explore the possibility of identifying understory fires using the phenological response from known understory fires to train the decision tree classification algorithm. Our results suggest that integration of MODIS land cover and active fire products can improve our understanding of the spatial and temporal patterns of maintenance, deforestation, and understory fires in Amazonia.
Submetido por Douglas Morton em 17-MAR-2004
Tema Científico do LBA: LC (Mudanças dos Usos da Terra e da Vegetação)