Sci Data. 2026 May 11. doi: 10.1038/s41597-026-07370-5. Online ahead of print.
ABSTRACT
Accurate, continuous and high-resolution mapping of multiple crop types, including both grain and cash crops, is vital for supporting sustainable agricultural development. While substantial progress has been made in mapping major grain crops, China still lacks a long-term, high-resolution dataset that simultaneously captures winter wheat, maize, rice and sugarcane. In this study, we used machine learning method to produce the China Crop Dataset (CCD), a 30 m resolution multi-crop dataset spanning 2001-2024, by integrating Landsat imagery and fused product with high spatial resolution. Validation based on field surveys and visually interpreted Google Earth samples confirmed the high accuracy of the CCD, with producer’s accuracy, user’s accuracy and overall accuracy reaching 88.45%, 87.2% and 91.01%, respectively. Furthermore, the CCD exhibited strong spatial consistency with existing datasets. The classified areas of winter wheat, maize, single-season rice, double-season rice and sugarcane showed good agreement with statistical area, with correlation coefficients (R2) exceeding 0.6 in most provinces. This dataset provides a robust and long-term resource for supporting agricultural planning, and facilitating research on land use and food security in China.
PMID:42115702 | DOI:10.1038/s41597-026-07370-5
