Classification of crops in different European regions based on TerraSAR-X data

Abstract:
The presented study deals with classifications of agricultural land use based on stacks of TerraSAR-X images taken during the vegetation season and the Maximum Likelihood method. The investigations took place in three different regions that are located in the West and North of Germany as well as the Southeast of Poland. The highest producer’s accuracies could be achieved for the classes oilseed rape (81% to 93%), potato (77% to 64% ) and the group of grain crops (87% to 95%). A higher discrepancy of accuracies between the three study areas can be observed for maize (22%, 48%, and 82%) grasslands (39%, 71%, and 89%) and single grain classes e.g. wheat (43%, 65%„ and 70%). Furthermore, strong variations of accuracies were observed for different test fields of the same class. This is probably due to either differences in the farmers’ land management or a variation of local conditions (e.g. hydrology, inclination, soils). The presented results underline the capability of satellite radar for multitemporal classification of agricultural areas.