CoastBench: A global training dataset for coastal classification using satellite imagery and elevation data

CoastBench is a global training dataset for coastal classification using satellite imagery and elevation data. It contains approximately 1,800 expert-labeled samples from coastlines around the world. Each sample includes labels for four coastal attributes across the cross-shore coastal profile: sediment type, coastal type, presence/absence of built environment, and presence/absence of human-made coastal defenses. Labels were assigned via a custom web interface using high-resolution satellite imagery. All samples are spatially anchored to the Global Coastal Transect System (GCTS) grid. CoastBench is provided in STAC-compliant format under a CC-BY-4.0 license. We welcome additional contributions through the open web application. This training dataset serves as the foundation for the global coastal typology produced in:Calkoen, F. R., Luijendijk, A. P., Hanson, S., Nicholls, R. J., Moreno-Rodenas, A., De Heer, H., and Baart, F.: Mapping the world's coast: a global 100-m coastal typology derived from satellite data using deep learning, Earth Syst. Sci. Data Discuss. [preprint], https://doi.org/10.5194/essd-2025-388, in review, 2025. More information about the transect grid can be found in:Calkoen, F.R., et al. (2025), Enabling coastal analytics at planetary scale, Environmental Modelling & Software, 183: 106257.https://doi.org/10.1016/j.envsoft.2025.106257 The derived global coastal typology dataset is available for download at:https://doi.org/10.5281/zenodo.15599096