Validation of representativeness with relief parameters based on the comparison of two landuse classifications

This study compares the results of a conventional landuse mapping to a Maximum Likelihood classification of Landsat-TM data for a study area of 9 km2 in the foothill section of the southern Upper Rhine Valley (South West Germany). For this purpose the 40 classes of the satellite-derived classification and the 7 classes of the mapped classification are summarized to the classes vineyard, pasture, forest, cropland, and urban area. In spite of the different scales and methodologies, the correspondence between the two classifications is high for the classes forest and pasture. The differences occurring in the urban class as well as in the classes cropland and vineyard are analyzed and explained. As the satellite-derived classification covers not only the study area, but an area of 36 000 km2, we have examined, using a DEM as an additional data layer, whether the geoecological conditions in the study area can be regarded as representative for the entire foothill section. GIs overlay operations are applied to the data layers landuse, elevation, aspect, and slope both for the study area and the Markgräfler Hügel-land area. This approach provides a new means of quantifying the term “representativeness” of a study area. It is shown that the correlations between landuse distribution and geomorphographical parameters correspond well in both scales. With regard to these parameters the study area can therefore be regarded as representative for the foothill section.