Grassland identification using multi-temporal RapidEye image series

Grasslands cover large areas of the earth´s surface and have been extensively converted to other uses such as cultivation and urbanization. The monitoring of grasslands is needed for any land use planning and environmental management. Remote Sensing techniques are suitable to provide detailed spatial information on grassland to support this process. The RapidEye satellite constellation represents a unique potential of multi-temporal acquisition of high resolution image data, therefore, offering a reliable data source for detailed multi-temporal analysis. In the presented study a semi-automatic land-cover classification approach with emphasis on the identification of grassland was developed. The methodology is based on the analysis of multi-temporal RapidEye images using the supervised decision tree (DT) classifier C5 in combination with prepended image segmentation. The results presented correspond to an area of 2500 km² in the State of Brandenburg / Germany. The classification accuracy was assessed by using randomly distributed independent reference points and the confusion matrix to derive users´ and producers´ accuracies. The grassland classification of the test area reached an overall accuracy of about 90%.