Land use and land cover fraction estimation for Sentinel-2 RGB images: A new LULC mapping task Rodriguez Ortega, José Tabik, Siham Benhammou, Yassir Khaldi, Rohaifa Alcaraz-Segura, Domingo Land use and land cover mapping Sentinel-2 RGB images Deep learning Convolutional neural networks Multi-input model Governmental institutions provide regional Land Use Land Cover (LULC) maps,1 but their complex formats, varied resolutions and diverse annotations limit usability. Traditionally, LULC mapping is framed as multiclass classification (assigning one dominant label per image) or multi-label classification (identifying coexisting classes). Alternative approaches remain unexplored due to challenges in leveraging existing LULC products for new tasks. This work presents a novel reformulation — LULC fraction estimation per public Sentinel-2 RGB image — that predicts both the presence and the fractional abundance of coexisting LULC categories within each image. Our contributions include: (1) Land-1.0,2 the first open source dataset with LULC fractions, climatological, and topographic data for 21,489 tiles; (2) a systematic method to build such datasets from existing LULC products; and (3) three deep learning solutions, where multitask models outperform single-task approaches. Future remote sensing foundation models could further improve the results by expanding the representation beyond supervised CNNs. This scalable and cost-effective method will help practitioners in environmental science and many other fields establish better monitoring of natural resources and biodiversity conservation using affordable RGB imagery and environmental data without the need to obtain and process expensive and complex hyperspectral or multispectral data. 2025-07-22T10:58:14Z 2025-07-22T10:58:14Z 2025-06-27 journal article Rodríguez-Ortega, J., Tabik, S., Benhammou, Y., Khaldi, R., & Alcaraz-Segura, D. (2025). Land use and land cover fraction estimation for Sentinel-2 RGB images: A new LULC mapping task. Remote Sensing Applications: Society and Environment, 101626. https://hdl.handle.net/10481/105557 https://doi.org/10.1016/j.rsase.2025.101626 eng http://creativecommons.org/licenses/by-nc-nd/3.0/ open access Creative Commons Attribution-NonCommercial-NoDerivs 3.0 License Elsevier