2000、2010与2020年中国水土保持梯田30米分辨率数据集

梯田作为中国分布最广、投入最多的水土保持措施之一，目前缺乏包含时空分布与多样分类类型的综合数据库，这一缺失严重制约了中国水土流失评估与水土保持规划的准确性。为填补这一空白，本研究开发了分阶段制图框架以划分多种梯田措施，并构建了名为“水土保持梯田措施数据集”（SWCTMD）的新数据产品。该数据集覆盖2000年至2020年，基于时间序列Landsat影像与数字高程模型数据融合生成，整合了水土保持措施因子及四种梯田类型：水平梯田、坡式梯田、隔坡梯田与反坡梯田。验证结果显示，SWCTMD对梯田识别的总体精度（OA）平均达91.7%，F1分数为83.3%；对不同梯田类型分类的OA达89.4%，F1分数为78.9%，证明了其在梯田制图方面的高精度。SWCTMD可用于提升水土流失模拟精度，支持水土流失长期趋势分析，同时为地球系统模拟提供重要数据基础，助力土地资源管理、粮食安全、生物多样性及水循环等领域的研究。

As one of the most widely distributed and most heavily invested soil and water conservation measures in China, terraced fields currently lack a comprehensive database that documents their spatiotemporal distribution and diverse classification types. This gap severely restricts the accuracy of soil erosion assessment and soil and water conservation planning in China. To fill this gap, this study developed a phased mapping framework for classifying various terraced field measures, and constructed a new data product named "Soil and Water Conservation Terraced Field Measures Dataset" (SWCTMD). Covering the period from 2000 to 2020, this dataset was generated by fusing time-series Landsat imagery and Digital Elevation Model (DEM) data, and integrates the soil and water conservation measure factor and four types of terraced fields: level terraces, sloping terraces, slope-interval terraces, and reverse-slope terraces. Validation results show that the average Overall Accuracy (OA) of SWCTMD for terraced field identification reaches 91.7%, with an F1-score of 83.3%; the OA for classification of different terraced field types reaches 89.4%, with an F1-score of 78.9%, which demonstrates its high accuracy in terraced field mapping. SWCTMD can be used to improve the accuracy of soil erosion simulation, support long-term trend analysis of soil erosion, provide an important data foundation for Earth system simulation, and facilitate research in fields such as land resource management, food security, biodiversity, and water cycle.