Uso de técnicas de análise multivariada para a predição de desmatamento na Amazônia : o modelo AMAZON-PD

AUTOR(ES)
DATA DE PUBLICAÇÃO

2007

RESUMO

Owner of the world largest rainforest, Brazil put forth its best effort in many different segments of the society looking forward to mitigating Amazon deforestation effects. This thesis proposes a prediction study of Amazon deforestation throughout multivariate analysis techniques, more specifically a logistic regression model. Given regional characteristics, usage of remote sensing data of medium resolution, organized and modeled in geographical information systems, constitutes an increasingly adopted praxis. In this case, the study was intended to raise space-temporal relationships that exist among biophysical variables identified in a single Landsat scene, located in São Félix do Xingu, Southern of Pará State, Brazil, observed along with six different dates from 1985 to 2004. Therefore, the Multivariate Analysis Applied to Zoning for Deforestation Prediction model (AMAZON-PD) was conceived, tested, and validated. In order to achieve such goal, during the pre-processing stage, the data were divided in four categories forest, non-forest, hydrography, and deforestation taking into account the methodological procedures developed to PRODES DIGITAL by the Brazilian National Institute for Space Research (INPE). Right after selection and formation of the variables to be used in the model, the following step was to generate the logistic regression coefficients in a year basis. This procedure played a key role in the understanding of the independent variables effects on the response-variable (environmental impact = deforestation), first in a static fashion and later on in a dynamic way. The attained results, which accuracy were verified upon classification tables, allowed to infer the future location of deforestation sites.

ASSUNTO(S)

regressão logística logistic regression deforestation geoprocessamento desmatamento sistemas de informação geográfica remote sensing ciencias exatas e da terra land cover change models geographical information systems sensoriamento remoto análise multivariada geoprocessing multivariate analysis modelos de mudanças na cobertura do solo.

ACESSO AO ARTIGO

Documentos Relacionados