Ethiopian crop type 2020 (EthCT2020) dataset

The observation of crop types is paramount for a myriad of agricultural and environmental remote sensing applications, including land use mapping, crop growth monitoring, and yield forecasting. Accurate georeferenced and quality-controlled crop type data is integral to the calibration of machine learning algorithms. However, publicly available field data remains scarce, particularly within the dynamic smallholder farming systems of sub-Saharan Africa.

To address this gap, the EthCT2020 dataset was established for the 2020/21 cropping season in Ethiopia. This dataset comprises 2,793 harmonized and quality-controlled in-situ samples of annual crops, representing seven crop groups and 22 classes across the diverse agricultural landscape of Ethiopia, with a specific emphasis on wheat-based farming systems. The data collection process involved a nationwide campaign implemented through a stratified sampling approach grounded in the wheat crop calendar, resulting in the acquisition of 1,263 samples from this campaign (GDCC; Source 1). Additionally, the dataset was enhanced by incorporating 1,530 wheat samples sourced from specialized databases dedicated to wheat disease surveillance, specifically the Wheat Rust Toolbox (WRTB; Source 2), which contributed 734 samples, and from an internal farm household survey database (FHSD; Source 3), which added 796 samples. The field data were meticulously labeled in accordance with the guidelines established by the Joint Experiment for Crop Assessment and Monitoring (JECAM) and the FAO Indicative Crop Classification for the definition and collection of agricultural land data.

Extensive processing of the EthCT2020 dataset was conducted to ensure its quality and accuracy, which included data harmonization and mixed pixel assessment through visual interpretation of satellite imagery, coupled with quality control measures based on Sentinel-2 NDVI homogeneity analysis. The EthCT2020 dataset is distinctive in its diversity and spatial precision, and it represents a valuable resource for further development of innovative remote sensing technologies and predictive models within the Ethiopian agricultural context. Additionally, it serves as a foundational input for climate models and forecasting systems for crop diseases and pests.