Análise metodológica do tratamento de dados sar r99b para discriminar incremento de desflorestamento no sudoeste da amazônia brasileira

AUTOR(ES)
DATA DE PUBLICAÇÃO

2008

RESUMO

- The estimate and mapping of deforestation in Brazilian Amazon has been carried out yarly, with the objective to monitor the forest resources, and as a mean for giving support to governmental control actions. This systematic monitoring is executed by the PRODES Digital project, developed by the Brazilian National Institute for Space Research (INPE), using Landsat TM images, as primary data source. The use of optical remote sensing data in large tropical forest areas has an important limitation due to cloud cover. Synthetic aperture radar (SAR) can be a valuable alternative in Amazon regions where cloud cover is critical, because the data acquisition is independent on atmospheric conditions. The data acquired by polarimetric SAR sensors contain more information on transmitted signal interaction with the target than that from monopolarized conventional data. This increase in information can represent an advance in land cover class discrimination and change detection in tropical forest regions. The main objective of this work is to evaluate the potential of L band PolSAR data to discriminate deforestation increment. Exploratory analysis and digital classification (MAXVER-ICM assuming normal distribution and with multiple sources of statistical evidences, ISOSEG, H/, Wishart k-médias H/ e Wishart k-médias H//A) have been performed with SAR and PolSAR data of a test site located in Acre State. PolSAR and multipolarized SAR R99B data have good potential to discriminate deforestation increment. The multipolarized and polarimetric SAR R99B presented good potential to discriminate deforestation increment areas automatically. MAXVER-ICM classifier showed better performance in the classification of SAR and PolSAR data. Individual polarizations (HH, HV or VV) presented limitation to separate deforestation increment and forest classes. When two channels are considered, the increase in information was significant. The information contained in HH + HV was enough to separate to discriminate deforestation increment with good accuracy (kappa = 0.68). VV channel amplitude and phase difference data did not add information to discriminate deforestation increment. In general, deforestation increments neglected in classifications consisted in burned areas, possibly in initial stage of deforestation. Some deforested areas showing signs of recent regeneration were also neglected.

ASSUNTO(S)

amazon (region) amazônia sar r99b polarimetria desflorestamento deforestation polarimetry

ACESSO AO ARTIGO

Documentos Relacionados