Open-SEA-Rice-10:东南亚水稻收获面积与种植强度的开放获取高分辨率地图

东南亚贡献了全球20%的稻米产量和29%的全球稻米甲烷排放，凸显了准确收割面积数据以支持粮食安全和温室气体核算的必要性。然而，由于种植方式多样，现有水稻地图常常缺乏关于作物强度、空间分辨率和准确性的信息。本研究提供了一个10米分辨率的开放访问水稻种植强度数据集，使得东南亚地区种植和收获面积的精确估算成为可能。采用本地无监督分类（LUCK-PALM）结合Sentinel-1A和Sentinel-2A/B数据生成图谱（2020–2021）。像素层面的验证（n = 58,885）显示，在识别稻田方面，整体准确率为0.98，河童系数为0.870，F1分数为0.879。该综合数据集已公开存储，可用于提升粮食和水安全策略，并优化甲烷排放估算。

Southeast Asia accounts for 20% of global rice production and 29% of global methane emissions from rice cultivation, highlighting the imperative of accurate harvested area data to support food security and greenhouse gas accounting. However, due to diverse cultivation practices, existing rice maps often lack information on crop intensity, spatial resolution, and accuracy. This study presents a 10-meter resolution open-access dataset of rice cultivation intensity, enabling accurate estimation of planted and harvested areas across Southeast Asia. The dataset was generated using local unsupervised classification (LUCK-PALM) combined with Sentinel-1A and Sentinel-2A/B data spanning 2020–2021. Pixel-level validation (n = 58,885) demonstrated an overall accuracy of 0.98, a Kappa coefficient of 0.870, and an F1-score of 0.879 for rice paddy identification. This comprehensive dataset is publicly available, and can be used to enhance food and water security strategies and optimize methane emission estimates.

搜集汇总

背景与挑战

背景概述

该数据集为东南亚地区提供了10米高分辨率的水稻收获面积与种植强度开放地图，基于2020-2021年Sentinel卫星数据，采用本地无监督分类方法（LUCK-PALM）生成，验证显示水稻田识别总体精度达0.98。它旨在支持粮食安全策略优化和甲烷排放估算，具有高空间精度和公开可访问性。