No abstract

No abstract

Spectral-Temporal Analysis by Response Surface (STARS), which exploits both multispectral and multitemporal information using fitted response surfaces, was used to describe deforestation patterns in the Brazilian Amazon. The STARS was conducted upon a MODIS dataset formed by 21 selected cloud free images (eight days composition) acquired from August 2003 to August 2004. The Multi-Coefficient Image (MCI) resulted from the STARS was used as input attributes for the three tested classifiers: Instance Based Knearest neighbor (IBK), Decision Tree (DT) and Neural Network (NN). The IBK classifier presented the highest accuracy (κ=0.93) in detecting deforestation also indicating the deforestation period (early or late in the year). The results showed that the STARS is promising to describe spectral change patterns over time, allowing detection of the deforestation process which occurs in the Brazilian Amazon.

Abstract. The mapping of vegetation and Land Cover (LC) is important for research and for public policy planning but, in Brazil, although diverse maps exist there are few studies comparing them. The semiarid region of the Caatinga, in northeastern Brazil is an area long neglected by scientific research and its vegetation is diverse and relatively rich despite years of human occupation and very little preservation effort. In this study we make a comparison between the main maps made for the Caatinga from four different sources: IBGE (Brazilian Institute of Geography and Statistics), TCN (Third National Communication), ProBio (Project for Conservation and Sustainable Use of Biological Biodiversity) and MapBiomas. We also test these maps against well-known Land Cover maps from ESA and NASA: ESA’s GlobCover and Climate Change Initiative (CCI) Land Cover, and NASA’s MODIS MCD12Q1. This was done on a sample area where many of the Caatinga’s vegetation physiognomies can be found, using well-established Difference metrics and the new SPAtial EFficiency (SPAEF) algorithm as they present complementary viewpoints to test the correspondence of mapped classes as well as that of their spatial patterns. Our results show considerable disagreement between the maps tested and their class semantics, with IBGE’s and ProBio’s being the most similar among all national maps and MapBiomas’ the most closely related to global LC maps. The nature of the observed disagreement between these maps shows they diverge not only in the application of their classification systems, but also in their mapped spatial pattern, signaling the need for a better classification system and a better map of vegetation and land cover for the region.