Time series of vegetation índices (NDVI and EVI) from MODIS for detecting deforestation in the Cerrado biome

The Cerrado is a biome characterized by a wide variety of vegetation formations, strong seasonality and high anthropogenic pressure. This study analyzed the use of time series (2000-2013) of MODIS EVI and MODIS NDVI to detect deforestation in the Cerrado. The study areas corresponded to the municipalities of Jataí/GO, Luís Eduardo Magalhães/BA, Mateiros/TO and São Miguel do Araguaia/GO. The time series were smoothed by the double logistic filter, available in the TIMESAT program. Kruskal-Wallis statistics was used to determine whether the representative temporal signatures of forestlands from Jataí and São Miguel do Araguaia and shrublands from all municipalities were statistically equal. Then the deforestation thresholds were identified for each vegetation formation (EVI and NDVI values below of which are considered deforestation). Results indicated that it is not possible to define a single threshold for each type of vegetation, however, it is possible to detect deforestation in forestlands and shrublands. The performance of NDVI was higher than that from EVI since the decrease in the NDVI values during the events of deforestation was higher.

Keywords:
Remote Sensing; Temporal Signatures; Tropical Savanna; Satellite Images