Close Window

Use of the Mixture Model in the attainment of the training data futures to be employed in the integration method of Multi Resolutions Remote Sensing data, for the representation of the land covering

Marcelo Lopes Latorre, Comando-geral de Tecnologia Aeroespacial (CTA)/INPE, latav@ltid.inpe.br (Presenting)
Yosio Edemir Shimabukuro, Instituto Nacional de Pesquisas Espaciais (INPE, yosio@ltid.inpe.br
Matthew C Hansen, South Dakota State University, Matthew.Hansen@sdstate.edu
Ruth DFries, Maryland University, rdefries@geog.umd.edu

This work had objective to establish an analysis of the training data use viability, gotten through the mixture technique, in the elaborated model to monitoring the vegetation dynamic cover in Amazonia. This process is based on the techniques developed and applied for the team of the Maryland University , where to larger emphasis is in the integration of different sensory spatial date with different resolutions (Terra/MODIS, Landsat/TM, ETM+ and CBERS 2/CCD). The study area, it was chosen one region in Mato Grosso State, will be presenting great representativity of the Brazilian Amazonia in terms of land to cover characteristics. The process elaboration occurred, jointly, with the application of several tests, using different training dates (ETM+, TM and CBERS/CCD scenes) obtained through the same technique, individually or in group. Through its preliminary results and its comparison with the dates obtained from PRODES Project (to year of 2002), the feasibility of the mixture application was verified. It is expected that with this analisys, the methodology of the monitoring of vegetation cover dynamics contribute, especially, to projects already developed for the Brazilian Amazonia, as it is the case of PRODES project.

Science Theme:  LC (Land Use and Land Cover Change)

Presentation Type:  Poster

Abstract ID: 35

Close Window