Title:
Abstract:
At present, the estimation of parameters used in global biomass fire emissions models are based on low and moderate spatial resolution satellite data, and even though there is an important advanced since the rising of the first orbital platforms, it is still necessary to reduce the uncertainty of these estimates.This improvement depends basically, on the capacity of the scientific community to develop techniques and procedures that would allow a more accurate estimate of burned areas. Thus, the main objective of this study was to evaluate three MODIS burned area products and one MODIS active fire products developed from different automatic change detection algorithms, based on spectral mixing model, empirical derived thresholds and classification techniques using satellite data from MODIS/Aqua and Terra, with different spatial resolutions (250, 500 and 1000 m). Automated change detection algorithms was used to estimate the amount of burned areas in the Brazilian Amazonia and Cerrado biomes for the year 2005, as a subsidiary parameter to biomass burning global emissions greenhouse gases models. Burned area products quantification algorithms results showed expressive discrepancies in biomass burning estimates. In Amazonia biome the total area burned labeled by the MODIS burned area products is 70.500 km2 (Burned Scars Mapping), 20.900 km2 (Change Detection Algorithm - MCD45) and 64.100 km2 (Burned Area Detection Algorithm), while MODIS Active Fire Product (MOD14) is 149.200 km2. The difference between lowest and highest values estimates is 128.300 km2 (85%). In Cerrado biome the total area burned labeled by the MODIS fire products is 115.700 km2, 77.400 km2, 26.000 km2 and 80.400 km2, respectively. The difference between lowest and highest values estimates is 89.700 km2 (77%). These results confirm that uncertainties in burned area estimates, based on low and moderate spatial resolution satellite data, contribute significantly to the uncertainties in greenhouse gases emissions estimation due to biomass burning.To quantitatively evaluate the results a validation of the MODIS fire products estimates was applied by comparing the results with a burned area visual interpretation of TM/Landsat images (reference classification), using a error matrix and a linear correlation analysis in two sites located at Acre and Rondônia States. The MODIS burned area product Burned Scars Mapping shows the highest accuracy between the observed values and the reference classification (81 e 93%); and by the linear correlation coefficient (0,57 and 0.77), respectively, for both validation sites.
Related Publications:
Lima, A.; Shimabukuro, Y. E.; Adami, M.; Freitas, R. M.; Aragao, L. E.; Formaggio, A. R.; Rivera-Lombardi, R. J. Burnt scars mapping in Brazilian Amazon using linear spectral mixing model in MODIS images. In: SIMPÓSIO BRASILEIRO DE SENSORIAMENTO REMOTO, 14. (SBSR), 2009, Natal. Anais do XIV Simpósio Brasileiro de Sensoriamento Remoto, 2009.
Piromal, R. A. S.; Rivera-Lombardi, R. J.; Shimabukuro, Y. E.; Formaggio, A. R.; Krug, T. Burned area detection in the Brazilian Amazon using MODIS data products (MOD14). Acta Amazonica, v. 38, n. 1, p. 77-84, 2008.
Setzer, A. W.; Morelli, F.; Rivera-Lombardi, R. J. Burned area fortnightly estimates. In: B.F.T.Rudorff; Y.E.Shimabukuro; J.C.Ceballos. (Eds.). Modis sensor and environment application in Brazil. S. J. Campos: Editora Bookimage, 2007, Chapter 28, p. 403-417. Book chapter.
Thelma K.; Rivera-Lombardi, R. J.; Dos Santos, J. R. Burned area, Recurrence of Fires and Permanence of Burnt scars in Selected Areas of the Brazilian Cerrado Using TM-Landsat Imagery. In: XXth International Society for Photogrammetry and Remote Sensing Congress, 2004, Istanbul. . XXth ISPRS Congress: Geo-Imagery Bridging Continents. Istanbul: ISPRS, 2004. Instanbul: M.Orhan Altan, Congress Director, 2004. v. XXXV. p. 243-246.