Multitemporal fraction images derived from Terra MODIS data for analysing land cover change over the Amazon region

摘要: L. O. ANDERSON*, Y. E. SHIMABUKURO and ARAIInstituto Nacional de Pesquisas Espaciais, Divisao Sensoriamento Remoto, Av. dosAstronautas, 1758, CEP 12227-010, Sao Jose dos Campos, SP, BrazilThe Moderate Resolution Imaging Spectroradiometer (MODIS) instrtimentonboard Earth Observing System (EOS) Terra plataform has been designed toprovide improved information for monitoring land, ocean, atmosphereconditions. The design combined characteristics oi' the Advanced Very HighResolution Radiometer (AVHRR) Landsat Thematic Mapper (TM),adding spectral channels in middle thermal infrared wavelength andproviding data 250 m. 500 m 1 km spatial resolutions. Spectral forattnospheric cloud characterization have included to permit both theremoval of atmospheric effects on surface observations provision ofatmospheric measurements (Justice et al. 1998). This work utilized land productM0D13Q1, which is a vegetation index product with 250m resolution isa composite 16 days observation over Mato Grosso State (figure 1). Thisproduct contains NDVI (Normalized Difference VegetationIndex), EVI (Enhanced Vegetation Index), red reflectance band {250m spatialresolution, 620-670 nm bandwidth), near (250 841-876 blue (500 resolution,rearranged 250m, 459-479nm medium band(500m resolution, rearranged 2105-2155nm qualityassurance EVI. view zenith angle, sun relativeazimuth angle average parameters (Lozar Balbach 2002). composites ofNDVI were performed using Constrained View Maximum ValueComposite (CV-MVC) algorithm.The pixel moderate sensor images, due its resolution,generally includes more than one type terrain cover. When these sensors observethe Earth, measured radiance integration all objectsthat are contained within pixel, implying existence so-called mixtureproblem (Aguiar el 1999). linear mixing model used analyse themixture signatures vegetation, soil, shade each either high(Airborne Visible/Infrared Spectrometer (AVIRIS). TM, etc.) coarse(AVHRR Systeme Pour l'Observation la Terre (SPOT-Vegetation))resolution data. available methods estimate proportion componentinside by minimizing sum squares errors. proportionvalues must be non-negative, they also equal (Shimabukuro Smith1991). In this work, we accept that it possible find pure pixels cover typein data, purest soil,and endmembers can seen figure 2. For coming investigate*Corresponding author. Email: liana@ltid.inpe.br