Regional land cover and agricultural area statistics and mapping in The Departement Ardeche, France, by use of Thematic Mapper data

Abstract

Based on the uni-temporal analysis of Thematic Mapper data, a crop and land use inventory experiment over the French Departement Ardeche is currently being carried out at the Joint Research Centre (JRC), Ispra. The study area is classified almost completely as a "less-favoured area" within the corresponding sectorial policy of the Commission of the European Communities (CEC) Agriculture General Directorate, which imposes restrictive conditions on the use of remotely-sensed data (very small average field size – 0–6–1 ha – and dominance of mountainous and rugged terrain types). Additionally, the climatic and edaphic heterogeneity of the study region required an adequate spatial subdivision into homogeneous agro-physical units in order to cope with local variations in plant or crop phenology. Ground-collected information could be successfully registered to the satellite data. Further analysis however, revealed problems being related to the sampling scheme definition which was designed with respect to sampling theory but did not cope, at the same time, with physical and structural heterogeneities of the study area. Due to these difficulties, an interpretative approach was adopted where, for each zone, a set of important spectral classes was defined by image sampling and cluster analysis. After thematic identification of these clusters, a minimum distance classifier was used to map the information classes to the image domain. Accuracy assessments of the resulting classification were hampered by the insufficiently precise description of some ground observation classes, which also limits the potential application of regression estimate techniques.