REMVEG: Remote assessment of vegetation status by hyperspectral imagery

Broad-band, high-spatial resolution satellite Remote Sensing (i.e., Landsat-TM, Spot-HIV) has proved to be an essential tool for landcover change mapping, but insufficient to (i) resolve subtle categories like the ones often needed for ecological research, and (ii) measure biophysical magnitudes. Narrow-band hyper-spectral airborne Remote Sensing provides data that can be used for these purposes, but also generates specific processing needs. Classical multivariate clustering techniques and simple vegetation indexes, traditionally applied to broad-band satellite imagery, are insufficient to retrieve information from narrow-band hyper spectral imagery. The processing of this imagery needs (i) to be driven from dedicated field data, in particular from field spectroscopy and (ii) to abandon the pixel as the processing unit and to substitute it by patches, as produced by segmentation algorithms (Lobo 1997).
\nWithin the frame of our general goals in the project, we have focused on the evaluation of DAIS imagery to (i) characterize different vegetation covers in terms of their hyperspectral signatures and to (ii) use the DAIS signatures
\nto assess vegetation condition in a parallel way as this has been previously done with hand-held field spectroradiometry. We describe in this intermediate report our first steps and results processing DAIS hyperspectral imagery acquired over a humid Mediterranean forest in Les Gavarres (NE Spain, 41 ° 54' N, 2° 58' E).