干旱区生态环境知识资源中心

In this experiment we use reflectance spectra of soil and plants typical of arid Mediterranean environments in both types of situations: sandy surface and non-sandy surfaces. Spectra were recorded on the ground during a field campaign in April 1995 in southern Tunisia with an GER-IRIS field spectrometer covering the 400-2500 nm spectral range. From these spectra SPOT4-VEGETATION Signal is simulated to reproduce desertification patterns of the area, particularly regarding presence or absence of sand at the surface.

The best band combinations for soil surface discrimination have been tested, the contrast between MIR and VIS band is very efficient in discriminating the sandy (quartzic) soil surfaces, from carbonate or sulphate soils (normalised difference). RED/BLUE bands ratio is also very efficient because of the redder colour of the iron oxide coated sand grains.

Two Landsat Thematic Mapper images over the area are used to simulate VEGETATION imagery at two dates. Unmixing techniques have been used to decompose the the TM images and rebuild them with the proper spectral characteristics, this allows particularly a better spectral simulation of the blue band. Spatial resolution is simulated by simple averaging which is a reasonable approximation in this context of flat soil and scarce prostrated vegetation. Selected band combinations are then applied to the simulated imagery to Stress differences between the two dates.

By comparing with ground information on sand extent, this experiment shows that the spectral characteristics of VEGETATION allow to discriminate sandy surfaces very efficiently. Because of its high revisit rate this instrument will thus be a unique tool for real time monitoring of sand movement for desertification control and sand encroachment prevention, which could not be achieved by previous RS programs.