Monitoramento de queimadas no sudoeste do Pará, a partir de séries temporais do sensor modis

Fires associated with the expansion of cattle ranching and agriculture has become a problem in the Amazon biome, causing severe environmental damages. Remote sensing images have been widely used in the fire monitoring on the extensive Amazon forest, but an accurate automated detection still need improvements. This research aims to evaluate MODIS time series spectral indices for mapping burned areas in the municipality of Novo Progresso, Para, and determine their occurrences in the different types of land use/land cover during the period 2000-2014. In burned area detection, the following data were compared: near-infrared band and spectral indices (NBR, NDVI and NBRT), considering daily images and 8-day composite products. MODIS time series were filtered and standardized temporally to eliminate false fire events. Threshold-value determination for the fire occurrences was obtained from the comparison of MODIS series with visual image classification of Landsat TM and ETM+ data using the Kappa coefficient. The best result considered the following factors: near-infrared band, daily data, and mean standardization, obtaining the Kappa coefficient value of 0.72 and Overall accuracy of 99%. The deforested areas are responsible for more than 70% of fire events. Private properties showed a higher percentage of the burned area, while private and Indigenous Lands Environmental Reserves had the lowest rates. The result of the proposed method was better than the burned area product (MCD45), but still shows cloud cover interference that should be improved in future work.