Mapeamento da cobertura e uso da terra: uma abordagem utilizando dados de sensoriamento remoto óptico multitemporais e provenientes de múltiplas plataformas

Abstract

Land cover and land use maps are instrumental for the territorial governance, particularly when capable of representing the occupation dynamics. Within this context, this study was focused on understanding and quantifying the changes in the Ribeirão João Leite watershed (including the seven municipalities under its direct infl uence), Goiás State, Brazil, along four distinct periods (1984, 1992, 2000 e 2011), based on the supervised classifi cation (Bhattacharya algorithm) of Landsat 5 and 7 imagery, in addition to the use of seasonal information and masks (obtained via linear spectral mixture model and application of the NDBI index - normalized difference built-up land index) aiming at improving the discrimination of agriculture and urban areas, as well as reducing errors induced by shaded relief. For the year 2011, class identifi cation and separability were also supported by fi eld data and the interpretation of higher resolution (2m) satellite image (THEOS / Thailand Earth Observation) and aerial photographs acquired by an Unmanned Airborne Vehicle (UAV / Sense Fly – Swinglet CAM), resulting in a total accuracy between 80 and 90%. In 27 years, period comprised in this study, the remnant vegetation in the watershed and adjacent areas decreased about 35%, while substantial expansions were observed in the agricultural areas (~47%) and urban settings (~122%). Our results also corroborate the synergistic utilization of remote sensing data with multiple resolutions and obtained via distinct platforms. An example in this direction is the combined use of Landsat like and higher resolution (> 5m) data, in association with centimeter airborne mosaics.