Land cover classification at a regional scale in Iberia: separability in a multi-temporal and multi-spectral data set of satellite images

Earth observation at regional scales, such as of the Iberian Peninsula or Mediterranean Basin, is an important tool to understand the relationships between climate and surface properties. Among the different layers of information that can be derived from satellite imagery, land cover maps are important by themselves and as an aid to infer other variables. Land cover legends at regional scales require finer categories than those used at a global scale, which implies processing multi-spectral imagery acquired by Earth observing systems with daily acquisition rates. In this article we discuss several alternatives to analyse satellite image datasets that are both multi-temporal and multi-spectral, with spatial resolution of 1 km. In order to facilitate the interpretation of our results, we restrict our analysis to pixels that correspond to cells with a uniform and known cover on the ground, as described by a detailed vegetation map, in Catalonia (NE Spain). Our results indicate that canonical redundancy analysis is efficient at reducing the multi-spectral and multi-temporal space while keeping high statistical separability among habitat types. The small fraction of uniform pixels (y2%) suggests that, at least for the Mediterranean Region, data fusion techniques would be convenient to increase spatial resolution in the dataset, and that instruments keeping daily acquisition rates but with higher spatial resolution (y1 ha) should be considered.