东北地区2000-2023年30米年尺度水稻田数据集

本研究运用以下方法，通过谷歌地球引擎（GEE）平台，以随机森林算法重建了21世纪以来中国东北地区30米分辨率的逐年水稻分布图：（1）采用分层筛选原理选取地面样本；（2）对Landsat 5/7/8时间序列数据进行线性插值与Whittaker平滑处理；（3）构建增强的光谱特征集。总体而言，所得水稻分布图总体精度（OA）较高，范围在0.93至0.99之间，且卫星估算结果与大部分城市统计数据显示出良好的一致性（R²≥0.7，p<0.01）。此外，与同类作物制图数据集相比，本研究结果具有更高的准确性。这是中国首次以30米分辨率重建如此详细且长时间序列的水稻格局。In this study, we reconstructed annual rice distribution maps at 30-meter resolution for Northeast China since the 21st century using the random forest algorithm via the Google Earth Engine (GEE) platform, following the steps below: (1) Ground samples were selected using stratified sampling principles; (2) Linear interpolation and Whittaker smoothing were performed on the Landsat 5/7/8 time-series dataset; (3) An enhanced spectral feature set was constructed. Overall, the obtained rice distribution maps achieved high Overall Accuracy (OA), ranging from 0.93 to 0.99. Additionally, the satellite-derived estimates showed good consistency with most urban statistical data (R² ≥ 0.7, p < 0.01). Furthermore, the results of this study exhibit higher accuracy compared to similar crop mapping datasets. This is the first reconstruction of such detailed, long-time-series rice distribution patterns at 30-meter resolution across China.

搜集汇总

背景与挑战

背景概述

该数据集提供了中国东北地区2000年至2023年每年30米分辨率的水稻田分布图，基于Google Earth Engine平台和Landsat卫星影像，采用随机森林算法重建，整体精度高（OA 0.93-0.99），并与统计数据显示良好一致性。这是中国首次在如此高分辨率和长期时间尺度上详细重建水稻模式的数据集，具有较高的科学和应用价值。