Relationships Between Fire and Land Use Change in the Brazilian Amazon Based on Satellite Data

Wednesday, 17 December 2014
Thierry Fanin and Guido van der Werf, VU University, Amsterdam, Netherlands
Fires are used as a tool in the process of deforestation. The relationship between fire and deforestation varies temporally and spatially according to the type of deforestation and climatic conditions. This study evaluates spatiotemporal variability between fire and deforestation over the 2002–2012 period in the Brazilian Legal Amazon (BLA). We based our study on four datasets: deforestation estimates from PRODES (Amazon Deforestation Monitoring Project) and forest cover loss from the Global Forest Change (GFC) project based on Landsat data, and burned area and land cover based on Moderate Resolution Imaging Spectroradiometer (MODIS) data.

While GFC and PRODES supported similar findings on spatial and temporal dynamics, the Landsat-scale comparison also highlighted a number of differences. Both datasets show a decrease after 2004 in forest loss or deforestation extent mainly from decreasing clearing rates in evergreen broadleaf forest, mostly in the states of Mato Grosso and Rondonia. However, the drop is larger and more gradual in PRODES than in GFC, with the former having less than half the forest loss of the latter. GFC indicates anomalous high forest loss in the years 2007 and 2010 not seen in PRODES. Rescaling these forest dynamics datasets to 500-meter resolution, allowed for a comparison against the MODIS datasets. The burned area data indicates that the mismatch between PRODES and GFC is largely related to increased fire occurrence during these dry years, mainly in Para. In addition it indicates that the time interval between deforestation and fire differs according to land cover, which is important when estimating the atmospheric impact of forest loss. We found that evergreen broadleaf forests are burned shortly after deforestation due to slash and burn techniques, while croplands have longer intervals depending on the crop variety. As a final step, we used these insights to better quantify carbon emissions from this region.