< Back to previous page

Project

Nonlinear spectral satellite image un-mixing for sustainable ecosystem management.

Spectral sensors onboard satellites have become the diagnostic tool of excellence where it concerns the monitoring of vegetation dynamics at local, regional and global scales. To date, the full potential of satellite image analysis for terrestrial ecosystem monitoring is still not fully employed. One reason is the composite nature of the pixels, i.e. the square elements of the 2D ground space. Satellite-measured surface reflectance values are mixed results of signals reflected from different ground components, e;g. vegetation, bare soil and shadow. The reflectance of a canopy is as such strongly influenced by the soil reflectance especially when plant density is low, or the canopy has not reached full closure. Consequently, the monitoring of vegetation characteristics is drastically hindered. To address this issue, the current project presents an alternative spectral unmixing technique. Specific focus is put on the effect of multiple scattering resulting in a nonlinear mixing process. The proposed technique allows to unmix the heterogeneous image pixels into the signal that solely originates fril tge vegetatuib component and the signal originating from other components (i;e., soil, shadow). Such a technique for extracting the pure spectral information of the vegetation from satellite image pixels would allow using temporal spectral information for monitoring vegetation dynamics without the interference of undesired background 'noise' from soils, shadows or other components inherent in remote sensing data. This would be a breakthrough as a more effective implementation of satellite images in ecosystem research is then feasible, for instance using spectral indices linked to biophysical features.
Date:1 Jan 2010 →  31 Dec 2010
Keywords:Terrestrial ecosystems, Satellite, Optical remote sensing, Vegetation, Spectral mixture analysis, Soil reflectance modeling, Nonlinearity
Disciplines:Scientific computing, Bioinformatics and computational biology, Public health care, Public health services