Paranã river basin has one of the major fragments of Decidual Seasonal Forest in Brazil. This vegetation is widely fragmented due to the selective wood exploitation and the growth of pasture areas, what justifies the development of studies in order to understand its dynamics and preserve its diversity. Thus, the present study aimed at defining a method for regional identification of the Deciduous Forest in the Paranã river basin. The deciduous forest has a typical phenological cycle in comparison with other savanna physiognomies. Due these characteristics, a temporal series of normalized difference vegetation index (NDVI) images of the MODIS sensor was used for its detection. The adopted methodology may be subdivided into the following steps: (a) elaboration of the 3D cube of NDVI images, where the z profile corresponding to temporal signature or NDVI spectrum, (b) noise elimination using the Minimum Noise Fraction (MNF) transformation, and (c) NDVI temporal variability examination of deciduous forest vegetation, with the establishment of the best NDVI band applied in the vegetation index differencing method. The Deciduous Forest presents a typical NDVI spectral behaviour, with higher values in the raining season and lower values in the dry season, what makes this kind of vegetation different from others. The employment of a changing detection algorithm between two images: one for the dry season and the other for the raining season enhances the localization of the Decidual Seasonal Forest. So, the methodology has proved to be effective for regional delimitation of Deciduous Forests considering the MODIS sensor. Considering the changing detection method, Deciduous Forest region is characterized by presenting NDVI alteration values.
Decidual Seasonal Forest; MODIS; NDVI; changing detection; multitemporal analysis
