1990-2024年中国30米高分辨率水稻分布数据集

水稻是全球超半数人口的主食并贡献了全球人为甲烷排放量的10%以上。亚洲作为水稻主产区其产量占全球60%以上对该区域水稻分布的精准制图对于监测粮食安全和温室气体排放至关重要。然而由于云层对光学遥感影像的影响目前仍缺乏覆盖整个亚洲区域的长期、高分辨率水稻分布数据集。本研究开发了全球作物数据集-水稻GCD-Rice数据集以绘制1990年至2023年间16个亚洲国家三个水稻季的种植分布图。利用Landsat和Sentinel-1数据集结合物候学方法和随机森林模型这些地图使用258,547个田间样本进行了验证。结果显示平均用户精度为89.88%生产者精度为88.52%总体精度为88.85%。此外与统计面积相比整体平均R²值为0.768斜率为0.874RMSE为0.346。Rice is a staple food for over half the global population and contributes to more than 10% of global anthropogenic methane emissions. Precise mapping of rice distribution in Asia, the primary region for rice cultivation responsible for over 60% of global production, is crucial for monitoring food security and greenhouse gas emissions. However, due to cloud cover impacts on optical remote sensing imagery, there is still a lack of long-term, high-resolution rice distribution datasets for the entire Asian region. This study develops the Global Crop Dataset-Rice (GCD-Rice) dataset to map rice cultivation across three seasons in 16 Asian countries from 1990 to 2023. Using Landsat and Sentinel-1 datasets, along with a phenological approach and a random forest model, the maps were validated with 258,547 field samples. Results show an average user accuracy of 89.88%, a producer accuracy of 88.52%, and an overall accuracy of 88.85%. Furthermore, comparing with statistical area reveals an overall average R² value of 0.768, a slope of 0.874, and an RMSE of 0.346.主要发现1. 发布了填补空白的核心数据集GCD-Rice该数据集首次提供了覆盖整个亚洲主要水稻生产国16国、跨越34年1990-2023年、包含三个种植季的连续高分辨率水稻分布信息。2. 采用了创新的多源遥感融合方法为了克服光学影像在雨季云层覆盖严重的问题研究同时使用了光学影像Landsat和雷达影像Sentinel-1。雷达影像不受云雾影响保证了数据的连续性和可靠性。结合物候学方法利用水稻独特生长周期特征和随机森林机器学习模型实现了高精度的自动识别。3. 数据集精度经过大规模实地验证结果优异样本量巨大使用了超过25.8万个实地样本进行验证保证了评估的稳健性。精度指标高分类精度总体精度接近89%用户精度和生产者精度均在89%和88%左右说明数据集在识别“是/否”为水稻田方面非常可靠。面积一致性与官方统计面积对比R²值达到0.768斜率0.874显示出数据集估算的面积与真实统计数据高度相关且偏差较小。4. 具有重大的应用价值粮食安全为监测亚洲水稻生产动态、评估产量波动、预警粮食危机提供了前所未有的精细时空数据。气候变化为准确估算水稻田甲烷排放这一关键温室气体源支持国家温室气体清单编制和减排政策制定提供了至关重要的空间本底数据。农业管理有助于分析种植制度变化、灌溉用水需求以及农业政策的影响。原文请在文末点击阅读原文主要图表Fig.1 Geographical distribution of the study area, including 16 countries.Fig.2 Overall Flowchart of this study to identify Asian rice.Fig.3 The percentage of effective optical observations within the study area, presented for the periods (a) 1990–2015, (b) 2003–2011 and (c) 2016–2023.Fig.4 Provincial or municipal comparisons of the identified acreage and statistical acreage of rice in each country, includingR2, RMAE, and Slope.Fig.5 Planting frequency of rice in Asia from 1990 to 2023. Panels (a–d) show a zoomed-in view of the local details in different regions.Fig.6 Planting frequency of single-season rice in Asia from 1990 to 2023. Panels (a–d) show a zoomed-in view of the local details in different regions.Fig.7 Planting frequency of double-season rice in Asia from 1990 to 2023. Panels (a–d) show a zoomed-in view of the local details in different regions.Fig.8 Planting frequency of triple-season rice in Asia from 1990 to 2023. Panels (a–d) show a zoomed-in view of the local details in different regions.Fig.9 Comparative analysis of rice planting distribution with MSEAsia and NESEA-Rice10 datasets (linear regression: R2, RMAE, slope).Fig.10 Comparison of this study with two existing products across four regions (MMR, THA, IDN, and JPN), in 2020, 2022, 2019, and 2021, respectively.【下载→方式一推荐主页 *个人* 简介城数派-CSDN博客方式二数据下载方式汇总-CSDN博客