Great Lisbon Metropolitan Area land use/cover characterization through multi-temporal and multi-resolution VIS components analysis

Abstract

Urban environments are heterogeneous by nature. Hence, to allow quantitative studies it is necessary to simplify them in combinations of basic land use/cover materials. The Ridd’s VIS model (1995) is a conceptual representation that allows simplifying urban environments through the combination of three basic components: vegetation (V), impervious surface (I), and soil (S). More recently, Lu and Weng (Lu & Weng 2004) successfully tested a new combination (vegetation, impervious surface and shadow), more adapted to urban reality. The majority of urban uses can be interpreted as a combination of those three basic components. The VIS analysis allows to disclose that the mainstream of urban features has its own VIS signature, which is difficult to detect through pixel-by-pixel based classifiers. This work examines the land use/cover characteristics of the Great Lisbon Metropolitan Area (GAML) using sub-pixel classification techniques, mainly linear spectral unmixing (LSU), developing a conceptual model to characterize the occupation standards. The LSU ability to measure the physical composition of urban morphology is also explored and tested. In this work we use Landsat 5 TM multispectral images (1987 and 1997), Landsat 7 ETM+ panchromatic and multispectral images (2000) and SPOT 5 HRVIR panchromatic (2.5 m supermode pixel size) and multispectral images (2004), evaluating at the same time the land use/cover signatures evolution and the effect of spatial resolution differences on the same signatures measurements.