2022-2024 Dataset of 10-meter Resolution Rice Planting Distribution in the Middle and Lower Reaches of the Yangtze River, China

The Yangtze River Basin is a major rice-producing region in China where accurate and high-resolution rice mapping is essential for food security and sustainable land management. we propose Progressive Deep Learning Crop Mapping (PDLCM), a novel framework that integrates Sentinel-1/2 time-series imagery, spatiotemporal feature fusion, and iterative sample enhancement using both positive and hard negative instances. Based on this approach, we generated the first deep learning–based, 10 m resolution rice distribution maps for the entire Yangtze River Basin (~1 million km²) over three consecutive years (2022–2024). The classified data achieved overall accuracies of 96.81% and an F1 score of 0.88.This study delivers a scalable, transferable solution for operational crop mapping and offers new insights into the spatiotemporal behavior of deep learning models in complex agricultural regions. Classification results: 0: non-rice, 1: rice.The relevant code can be obtained through the following link:https://pan.baidu.com/s/17_CLtikOzTA2De57apNjAw?pwd=H8Yv

长江流域是中国主要水稻产区，精准且高分辨率的水稻制图对保障粮食安全与可持续土地管理至关重要。本研究提出渐进式深度学习作物制图（Progressive Deep Learning Crop Mapping, PDLCM）这一新颖框架，该框架整合了哨兵-1/2号（Sentinel-1/2）时序影像、时空特征融合技术，以及基于正样本与难例样本的迭代样本增强策略。基于该方法，我们生成了首套基于深度学习、分辨率为10米的长江流域（总面积约100万平方千米）连续三年（2022—2024年）水稻分布制图数据。该分类数据的总体精度达96.81%，F1值为0.88。本研究为业务化作物制图提供了一套可扩展、可迁移的解决方案，并为复杂农业区域内深度学习模型的时空行为特征提供了全新认知。分类结果说明：0代表非水稻区域，1代表水稻区域。相关代码可通过以下链接获取：https://pan.baidu.com/s/17_CLtikOzTA2De57apNjAw?pwd=H8Yv