1985-2024年中国首个30m分辨率年度树木覆盖数据集

中国年度林木覆盖数据集（CATCD）是国内首套长时序、高空间分辨率的年度林木覆盖数据集，通过集成Landsat影像时间序列与基于随机森林的集成学习方法生成。CATCD的可靠性与准确性已通过多源参考数据验证（相关性：0.70-0.96，均方根误差：5.6%-25.2%）。当前版本（v 0.0.1）包含1985年至2024年中国30米分辨率年度林木覆盖数据，每年数据存储为包含"TreeCover"和"Uncertainty"波段的GeoTIFF文件。其中TreeCover数据范围为0-100，代表像元内林木覆盖百分比；Uncertainty为通过5折交叉验证得出的林木覆盖不确定性估计值，以五个模型预测结果的标准差表示。该数据集除提供中国年度林木覆盖信息外，用户还可通过设定林木覆盖阈值提取森林动态变化范围。

The Chinese Annual Tree Cover Dataset (CATCD) is the first long-time-series, high-spatial-resolution annual tree cover dataset in China. It is generated by integrating Landsat image time series and random forest-based ensemble learning methods. The reliability and accuracy of CATCD have been validated using multi-source reference data, with a correlation coefficient ranging from 0.70 to 0.96 and a root mean square error (RMSE) of 5.6% to 25.2%. The current version (v 0.0.1) contains annual tree cover data of China at 30-meter resolution from 1985 to 2024. Each year’s data is stored as a GeoTIFF file with two bands: "TreeCover" and "Uncertainty". The "TreeCover" band ranges from 0 to 100, representing the tree cover percentage within each pixel; the "Uncertainty" band is an estimate of tree cover uncertainty derived from 5-fold cross-validation, expressed as the standard deviation of predictions from five models. In addition to providing annual tree cover information for China, users can also extract forest dynamic change ranges by setting tree cover thresholds.

搜集汇总

背景与挑战

背景概述

该数据集是中国首个长期、高空间分辨率的年度树木覆盖数据集，覆盖1985年至2024年，分辨率为30米。数据通过Landsat影像时间序列和随机森林集成学习技术生成，并经过多源数据验证，具有较高的可靠性。