Hu et al. (2026) Hcropland30: A hybrid 30-m global cropland map generated by leveraging global land cover products and Landsat data via deep learning

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-07-02
• Authors: Qiong Hu, Zhiwen Cai, Liangzhi You, Steffen Fritz, Xi Zhang, He Yin, Haodong Wei, Jingya Yang, Zexuan Li, Qiangyi Yu, H B Wu, Baodong Xu, Wenbin Wu
• DOI: 10.1016/j.rse.2026.115552

Research Groups

• Key Laboratory for Geographical Process Analysis & Simulation of Hubei Province/College of Urban and Environmental Sciences/ Academy of Frontier Interdisciplinary Research, Central China Normal University
• Macro Agriculture Research Institute, College of Plant Science and Technology, Huazhong Agricultural University
• International Food Policy Research Institute
• Novel Data Ecosystems for Sustainability Research Group, International Institute for Applied Systems Analysis (IIASA)
• Department of Geography, Kent State University
• State Key Laboratory of Efficient Utilization of Arid and Semi-arid Arable Land in Northern China, the Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences
• College of Resources and Environment, Huazhong Agricultural University

Short Summary

This study developed a hybrid 30 m global cropland map (Hcropland30) using deep learning to advance cropland mapping at a global scale. The Hcropland30 map achieved an overall accuracy of 89.8% and a cropland F1-score of 0.85.

Objective

• To generate a high-accuracy global cropland dataset for 2020.
• To introduce a robust methodology for diverse global mapping applications by leveraging deep learning and varied prior knowledge.

Study Configuration

• Spatial Scale: Global, with a spatial resolution of 30 m.
• Temporal Scale: The study focused on the year 2020, but the proposed framework can be extended to create multi-temporal maps.

Methodology and Data

• Models used: Hierarchical sampling strategy, ensemble learning technique, U-Net model for predicting global cropland.
• Data sources: Landsat-derived seasonal NDVI composites, six LULC datasets (including existing 10–30 m LULC products).

Main Results

• The Hcropland30 map achieved an overall accuracy of 89.8% and a cropland F1-score of 0.85.
• Considerable spatial discrepancies among existing 10–30 m LULC products were revealed, with areas exhibiting ‘high’ to ‘very high’ inconsistency significantly outnumbering those of ‘medium’ to ‘low’ inconsistency.

Contributions

• The study introduced a robust methodology for diverse global mapping applications by leveraging deep learning and varied prior knowledge.
• The Hcropland30 map provided a high-accuracy global cropland dataset, which can be used as a reference for future studies.

Funding

• This research was funded by the National Natural Science Foundation of China (Grant No. 42171387).
• The study also received funding from the International Institute for Applied Systems Analysis (IIASA) and the State Key Laboratory of Efficient Utilization of Arid and Semi-arid Arable Land in Northern China.

Citation

@article{Hu2026Hcropland30,
  author = {Hu, Qiong and Cai, Zhiwen and You, Liangzhi and Fritz, Steffen and Zhang, Xi and Yin, He and Wei, Haodong and Yang, Jingya and Li, Zexuan and Yu, Qiangyi and Wu, H B and Xu, Baodong and Wu, Wenbin},
  title = {Hcropland30: A hybrid 30-m global cropland map generated by leveraging global land cover products and Landsat data via deep learning},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115552},
  url = {https://doi.org/10.1016/j.rse.2026.115552}
}