EnMAP Cloud and Cloud Shadow Benchmarking Dataset Accompanying the EnICCS Package

This dataset contains hand-drawn reference masks for cloud and cloud shadow detection in EnMAP hyperspectral imagery. It accompanies the EnICCS package and research paper. The dataset includes 5 pairs of binary masks (cloud and shadow) for representative EnMAP scenes acquired between 2023 and 2024 over the broader region surrounding Siaya County, western Kenya. Scenes were selected to represent varying seasonality, surface types (smallholder farms, forests, towns, etc.), sensor geometries and cloud characteristics. Masks were hand-drawn in QGIS based on EnMAP VNIR composites without access to operational cloud and cloud shadow masks. The masks are provided as GeoTIFF files with clear naming conventions that include datatake and tile IDs, enabling users to retrieve the corresponding EnMAP imagery. This dataset is intended for benchmarking cloud masking algorithms on EnMAP data, taking into account its limitations and associated considerations. EnICCS uses EnMAP L2A data for cloud and cloud shadow masking. Please refer to the paper, SI, and GitHub repository for more information.  Full documentation, including file formats, naming conventions, and evaluation methodology, is provided in the README file. Data Notice: EnMAP data are licensed products of DLR (© 2023, 2024, all rights reserved) and are freely available (see www.enmap.org) The CCS masks were hand-drawn on EnMAP scenes and as such contain modified EnMAP data ©DLR [2023, 2024].