Science et changements planétaires / Sécheresse


Use of remote sensing to analyse and spatially extrapolate soil landscapes in Saharian regions Volume 11, issue 3, Septembre 2000


See all figures

Institut d'hydraulique et d'agronomie sahariennes, Centre universitaire de Ouargla, route de Ghardaïa, BP 163, Ouargla 30000, Algérie.
  • Page(s) : 179-88
  • Published in: 2000

In arid regions, soil surface states (SSS) are closely related to soil dynamics. This paper ains at studying the contribution of remote sensing in these regions (case study of the Ouargla depression in the Algerian Sahara) for the spatial extrapolation of results obtained from a reference area (subimage: 1/16th of the whole Spot image) for which field data are known throughout the region covered by the Spot image. The subscene hierarchical ascending classification (HAC) using the field data identified 13 classes of SSS (4 salines, 2 gypsics, 3 regs, 3 sandy and 1 rock) and one class of vegetation and others. Extrapolation of these results to the whole image using maximum likelihood classification (ML) showed that the 13 SS defined in the subimage are also represented in the whole image. The extrapolation for the whole image is validated by comparing the two classifications applied to subscene: 13 SSS resulting from the HAC and ML. 82.6% of the pixels are classified in the same SSS class before and after extrapolation. Based on these satisfactory extrapolation results, we established a map of regional recognition of soil landscapes in 6 units.