The dataset of main grain land changes in China over 1985–2020

We have collected spatial distribution datasets of rice, wheat, and maize in China within a resolution range of 30m-1km, spanning from 1985 to 2020. Based on these existing datasets, we have utilized data fusion techniques to create a nationwide fundamental sample dataset of main grain land (MGL). Considering 7 types comprehensively, including single wheat, single rice, single maize, wheat &amp; maize, wheat &amp; rice, double rice, and non-MGL, we have automatically selected 100,000 sample points nationwide from the fundamental sample dataset using stratified sampling. Meanwhile, utilizing Landsat imagery from the GEE platform, we have constructed a remote sensing imagery feature set for each year. Available spectral features are synthesized every three months, including red, green, blue, and near-infrared bands, as well as NDVI, NDBI, and NDWI. By combining four spectral features from each year, we have formed a phenological feature dataset. Finally, using the random forest algorithm, we individually mapped the MGL for each year. In addition, we assigned MGL and non-MGL values of 1 and 0 to all cropland, respectively. The planting intensity, gain time and loss time of cropland and MGL was calculated.<br>本研究收集了1985年至2020年间、分辨率覆盖30米至1千米的中国水稻、小麦、玉米空间分布数据集。基于上述现有数据集，本研究采用数据融合技术构建了全国尺度的主粮用地（Main Grain Land, MGL）基础样本数据集。综合覆盖单作小麦、单作水稻、单作玉米、小麦-玉米轮作、小麦-水稻轮作、双季稻以及非主粮用地共7种地类，本研究采用分层抽样法从基础样本数据集内自动选取全国范围内的10万个样本点。与此同时，本研究借助谷歌地球引擎（Google Earth Engine, GEE）平台的陆地卫星（Landsat）影像，构建了逐年遥感影像特征集。每3个月合成一次可用光谱特征，涵盖红、绿、蓝、近红外波段，以及归一化植被指数（Normalized Difference Vegetation Index, NDVI）、归一化建筑指数（Normalized Difference Built-up Index, NDBI）与归一化水体指数（Normalized Difference Water Index, NDWI）。结合逐年的4组光谱特征，本研究构建了物候特征数据集。最后，本研究采用随机森林（Random Forest）算法，逐年完成主粮用地的空间制图。此外，本研究为所有耕地赋予分类值：主粮用地赋值为1，非主粮用地赋值为0。本研究同时计算了耕地与主粮用地的种植强度、种植始期与种植终期。

搜集汇总

背景与挑战

背景概述

该数据集提供了1985-2020年中国主要粮食作物（水稻、小麦、玉米）的空间分布数据，分辨率为30m-1km，包含7种作物类型和10万个样本点。通过Landsat影像和随机森林算法，构建了逐年粮食用地分布图，为农业土地管理和可持续农业发展研究提供了重要数据支持。