Wheat, as one of the world's major food crops, has a significant impact on the formation of agricultural trade patterns. China is the world's largest producer and consumer of wheat, with a vast planting area and diverse planting systems. However, current wheat mapping research based on remote sensing technology often relies on unified phenological characteristic variables and fails to fully consider the significant differences in wheat growth cycles among different agricultural ecological regions in China. In addition, the lack of large-scale training samples severely restricts the accuracy and spatiotemporal generalization ability of the model. The existing research in China mainly focuses on winter wheat monitoring and mapping, while the field of spring wheat, especially in the major spring wheat producing areas in North China, is still in a research gap, resulting in a serious shortage of targeted remote sensing products. These limitations hinder the development of high-precision and spatially comprehensive wheat distribution atlases, weakening the integrity of agricultural monitoring and food security assessment. To address the aforementioned issues, this dataset proposes a cross regional training sample generation method that integrates time-series remote sensing data with crop distribution products. At the same time, a provincial-level differentiated feature selection strategy is introduced to enhance the regional adaptability and classification performance of the model. Based on the above method, a 10 meter resolution wheat distribution dataset (CN-Wheat10) covering 2018-2024 was constructed, including spring and winter wheat harvest area maps of 15 provinces in China and detailed winter wheat planting area maps of 10 provinces. CN-Wheat10 not only provides spatial distribution information of wheat harvesting area in winter and spring seasons, but also covers the winter wheat planting areas in major production areas. Compared to existing products mainly based on winter wheat, this dataset has expanded in spatial coverage and crop types, providing more comprehensive data support for agricultural monitoring and management in China.
This dataset uses a cross regional training sample generation method to integrate time-series remote sensing data with crop distribution products. At the same time, a provincial-level differentiated feature selection strategy is introduced to enhance the regional adaptability and classification performance of the model.
The validation of a large-scale reference dataset constructed through field investigation and high-resolution image visual interpretation shows that the mapping accuracy of CN-Wheat10 for winter wheat is above 0.93, and for spring wheat it is above 0.91. When compared with the wheat area statistics in the China Statistical Yearbook, the determination coefficient (R ²) at the provincial level exceeds 0.94, while at the municipal level it remains above 0.88. In terms of spatial distribution, Chinese wheat shows a pattern of clustering in the east and dispersion in the west, with winter wheat as the main body and spring wheat as a supplement.
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.FIGSHARE.DB7051.2025
10.6084/m9.figshare.28852220
