VISIBLE—NIR HYPERSPECTRAL IMAGERY FORDISCRIMINATING SOIL TYPES IN THE LA PEYNEWATERSHED (FRANCE)

摘要

Remotely sensed data might be an important tool for acquiring geographical information about soil attributes especially in areas where the soil surface is permanently or tem-porarily exposed, as in the extensive vineyard region of the Languedoc Roussillon region of France. By providing detailed spectral signature for every pixel, imaging spectrometry can potentially be used to identify the nature and abundance of some soil surface components. This chapter reports results obtained in the analysis of soil spectral data of the Low Languedoc Plains natural region of southern France acquired by an image spec-trometer (Hyperspectral Mapper – HyMap) after laboratory calibration to make quan-titative estimations of CaCO_3 and clay content of surface soil horizons. Two different approaches for extracting information from soil reflectance spectra were used: (a) band depth analysis for estimating calcium carbonate and clay abundances and (b) a redness index to identify soil colour variations. Validation data are available on surface soil samples as visible–NIR and X-ray spectra, and chemical characterisations. For areas with soils developed from sedimentary materials of variable calcium car-bonate and particle-size composition, we have demonstrated the possibility of obtaining quantitative information for CaCO_3 content from the intensity of absorption of the feature centred at 2341 nm. CaCO_3 estimates were highly satisfactory and at least as good as field effervescence test usually used in fieldwork. Clay particle-size content classes can also be estimated using the absorption depth of the clay minerals absorption feature at 2206 nm. Redness index along with clay and CaCO_3 estimates allows characterisation and expla-nation of the intra-unit variability of a pre-existing 1:100,000 soil map, and could be further used as new covariates in order to refine limits or segment existing units.