Intégration des données spectrales et géomorphométriques pour la caractérisation de la dégradation des sols et l'identification des zones desusceptibilité à l'érosion hydrique

Abstract

In the field of spatial observations of arid and semiarid ecosystems, the interest of remote sensing has long been recognised. The present work investigates the use of remote sensing techniques and digital elevation model analysis to characterise land degradation processes. The main objective is to evaluate the potential of spectral data and geomorphometric variables to discriminate different levels of soil degradation, and to assess ecosystems fragility and their susceptibility to degradation and desertification phenomena. The methodology adopted uses an integrated approach that combines spectral measurements, provided by remote sensing images, and geomorphometric variables, derived from a digital elevation model, in an attempt to define a set of indicators (spectral and topographic) of geomorphic processes and land degradation. Remote sensing techniques are based on two approaches: spectral mixture analysis that deals with heterogeneity at the sub-pixel level, and a set of indices describing the spectrum shape, which are sensitive to soil surface conditions. Integration procedures were involved in two ways. The first is in the correction of terrain-induced image distortions, which provide images free from relief displacement effects (ortho-rectification), and in the removal of topographically induced effects on TM images through a combined atmospheric and topographic correction. The second is in a parametric integration of spectral data and geomorphometric attributes to assess land susceptibility to degradation and desertification processes. Two types of data were collected for this research: satellite optical imagery and ground-based spectro-radiometric measurements. While indices describing soil colour (corresponding to colour parameters Intensity, Hue and Saturation) were used to discriminate different levels of soil degradation based on both ground and satellite data, the spectral mixture analysis was performed on the image to derive relative abundance of scene components. The results show that the spectral indices have enough potential to discriminate different levels of degradation, particularly when bands from the short-wave infrared domain are included (TM5 and TM7). They demonstrate results similar to those generated by spectral unmixing for the assessment of land degradation features in general, and soil erosion in particular. Concerning terrain analysis, this study points up the interest of the integrated use of local topographic attributes and combined topographic indices to characterise the hydrologic behaviour of terrain units, and to understand its effects on landscape evolution. These topographic variables quantify the contextual nature of points and characterise the spatial variability of processes occurring in the landscape. They are the major factors controlling the direction and the intensity of hillslope and hydrologic processes. The latter are responsible for the landscape evolution and its exposure to degradation risks by water erosion. Compared to the method based on curvatures analysis, our approach allows a better identification of homogeneous response units to hydrologic processes. These units are in agreement with flow directions and with the principles governing water and substance motion on the hillslope. Finally, we carried out an integrated analysis of geo-ecological parameters describing the studied ecosystem. It consists in defining hydrologic response units through an integration of spectral information and geomorphometric attributes. This allows us to determine the ecosystem fragility and to evaluate its susceptibility to land degradation and desertification processes.