National Scale Above-Ground Biomass in Nepal's Forests at 10m Resolution (Circa 2020)

AGB mapping in Nepal's forests circa 2020 at 10m resolution using Sentinel-1, Sentinel-2, and field data in the Google Earth Engine. The CART model achieved 92% accuracy (R² = 0.92), with RMSE of 30.42 Mg ha⁻¹ and MAE of 18.84 Mg ha⁻¹. 2,834 field-measured AGB samples were used to train and validate the machine learning model. Nepal's total tree cover AGB is estimated at 132 Pg, with a mean value of 150.32 Mg ha⁻¹ and maximum of 452.05 Mg ha⁻¹. High uncertainty in mountainous areas; low uncertainty in dense forests, aiding conservation strategies. The study supports REDD+ initiatives and sustainable forest management with open-access AGB data.基于哨兵一号（Sentinel-1）、哨兵二号（Sentinel-2）遥感数据与实地调研数据，依托谷歌地球引擎（Google Earth Engine）平台，生成2020年前后尼泊尔森林10米分辨率地上生物量（Aboveground Biomass, AGB）制图成果。 本研究所用的分类与回归树（Classification and Regression Tree, CART）模型精度达92%，决定系数（R²）为0.92，均方根误差（Root Mean Square Error, RMSE）为30.42 Mg·ha⁻¹，平均绝对误差（Mean Absolute Error, MAE）为18.84 Mg·ha⁻¹。 本次研究共使用2834份实地测得的地上生物量样本，用于机器学习模型的训练与验证。 经估算，尼泊尔全境树木覆盖地上生物量总量为132 Pg，单位面积平均生物量为150.32 Mg·ha⁻¹，最高值达452.05 Mg·ha⁻¹。 该制图成果在山区存在较高不确定性，而在茂密森林中不确定性较低，可为森林保护策略制定提供支撑。 本研究开放获取的地上生物量数据，可为减少毁林和森林退化所致排放量（Reducing Emissions from Deforestation and Forest Degradation+, REDD+）行动及可持续森林管理工作提供支持。