Monitoring of agricultural investment areas in Ethiopia based on remote sensing time-series data

Abstract

Ethiopia is known to be currently suffering from considerable food deficits. Therefore, Ethiopia has been promoting large-scale agricultural investment (LSAI) to transform the agricultural sector. In this context, GIZ and DLR entered a cooperation to develop a remote sensing approach for monitoring the actual state of land use of the LSAI areas. The approach was developed based on Sentinel-1 and Sentinel-2 data using Google Earth Engine. First step is a land use/land cover classification to identify cropland. Second, different input parameters and combinations thereof were evaluated for crop type classification. The accuracies of 33 supervised random forest models were assessed for crop classification for three study areas and two years. The finally selected method is applied for crop type mapping and used to investigate class-specific accuracies. The results reveal that large parts of the LSAI areas are currently not used for crop production and indicate great potential for further agricultural development in two of the three study areas. The developed methods and classification results are provided to GIZ and regional experts to support agricultural investment monitoring in Ethiopia.