Accueil du site
Master
Brésil
2015
Uso de séries temporais NDVI do MODIS para mapeamento da cobertura vegetal natural e exótica em uma região de transição cerrado-caatinga
Titre : Uso de séries temporais NDVI do MODIS para mapeamento da cobertura vegetal natural e exótica em uma região de transição cerrado-caatinga
Auteur : Abade, Natanael Antunes
Université de soutenance : Universidade de Brasília (UNB)
Grade : Mestrado em Geografia 2015
Résumé
The orbital monitoring of Brazilian flora began in the 1970s with Radam project, when were used data of the Side-Looking Airborne Radar (SLAR). Other programs were conducted in order to know and monitor the conservation status of Brazil’s biomes. The Programa de Cálculo de Desflorestamento da Amazônia (PRODES), the Detecção em Tempo Real (DETER), the Projeto de Conservação e Utilização Sustentável da Diversidade Biológica Brasileira (PROBIO) and Projeto de Monitoramento do Desmatamento nos Biomas Brasileiros por Satélite (PMDBBS) were some such initiatives. In addition to government actions, research is being conducted by institutions of national and international academic and scientific nature. However, most efforts were directed to the Amazon ecosystem or used techniques supported by monotemporal images, resolution space-times is not compatible with the current dynamics of change of the landscape. Given this reality, we developed this work aiming to study a transition region of semi-arid ecosystems (Cerrado-Caatinga), located in the north of Minas Gerais and Bahia southwest, Brazil. This place has rich biodiversity, consisting of a complex flora and that has not been well studied. Its landscape begins to undergo changes due to significant growth of human activities related to agriculture, leveraged by the realization of major irrigation projects created between the 1970s and 1990s. Thus, we used the following method : use of NDVI-MODIS time-series from 2011 to 2013, deriving from the MOD09Q1 product ; use of median and Savitzky-Golay filters ; selection of temporal-signature references of land use and vegetation cover and use of classifiers similarity and distance Spectral Angle Mapper (SAM), Spectral Correlation Mapper (SCM) and Euclidian Distance Measure (ED).The results were satisfactory especially when using the classifier ED on time series NDVI-MODIS, reaching levels of agreement, Global Accuracy and Kappa, more than 82% and 0.75. However, more studies should be made to identify the shrub and herbaceous vegetation types of the Cerrado biome and the crops grown on farms of small extent, present in large numbers on the left of São Francisco river
Page publiée le 27 juillet 2017