Araza Arnan, de Bruin Sytze, Hein Lars, Herold Martin
Laboratory of Geo-information and Remote Sensing, Wageningen University and Research, Wageningen, The Netherlands.
Environmental Systems Analysis, Wageningen University and Research, Wageningen, The Netherlands.
Sci Rep. 2023 Aug 5;13(1):12704. doi: 10.1038/s41598-023-38935-8.
Countries have pledged to different national and international environmental agreements, most prominently the climate change mitigation targets of the Paris Agreement. Accounting for carbon stocks and flows (fluxes) is essential for countries that have recently adopted the United Nations System of Environmental-Economic Accounting - ecosystem accounting framework (UNSEEA) as a global statistical standard. In this paper, we analyze how spatial carbon fluxes can be used in support of the UNSEEA carbon accounts in five case countries with available in-situ data. Using global multi-date biomass map products and other remotely sensed data, we mapped the 2010-2018 carbon fluxes in Brazil, the Netherlands, the Philippines, Sweden and the USA using National Forest Inventory (NFI) and local biomass maps from airborne LiDAR as reference data. We identified areas that are unsupported by the reference data within environmental feature space (6-47% of vegetated country area); cross-validated an ensemble machine learning (RMSE=9-39 Mg C [Formula: see text] and [Formula: see text]=0.16-0.71) used to map carbon fluxes with prediction intervals; and assessed spatially correlated residuals (<5 km) before aggregating carbon fluxes from 1-ha pixels to UNSEEA forest classes. The resulting carbon accounting tables revealed the net carbon sequestration in natural broadleaved forests. Both in plantations and in other woody vegetation ecosystems, emissions exceeded sequestration. Overall, our estimates align with FAO-Forest Resource Assessment and national studies with the largest deviations in Brazil and USA. These two countries used highly clustered reference data, where clustering caused uncertainty given the need to extrapolate to under-sampled areas. We finally provide recommendations to mitigate the effect of under-sampling and to better account for the uncertainties once carbon stocks and flows need to be aggregated in relatively smaller countries. These actions are timely given the global initiatives that aim to upscale UNSEEA carbon accounting.
各国已承诺遵守不同的国家和国际环境协定，其中最突出的是《巴黎协定》的气候变化缓解目标。对于最近采用联合国环境经济核算体系——生态系统核算框架（UNSEEA）作为全球统计标准的国家而言，核算碳储量和流量（通量）至关重要。在本文中，我们分析了如何利用空间碳通量来支持五个拥有实地数据的案例国家的UNSEEA碳账户。利用全球多日期生物量地图产品和其他遥感数据，我们以国家森林资源清查（NFI）和机载激光雷达的局部生物量地图作为参考数据，绘制了巴西、荷兰、菲律宾、瑞典和美国2010 - 2018年的碳通量图。我们在环境特征空间内确定了参考数据未覆盖的区域（占植被覆盖国土面积的6 - 47%）；对用于绘制碳通量并带有预测区间的集成机器学习方法进行了交叉验证（均方根误差=9 - 39 Mg C [公式：见正文]，[公式：见正文]=0.1至0.71）；并在将1公顷像素的碳通量汇总到UNSEEA森林类别之前评估了空间相关残差（<5公里）。由此得出的碳核算表显示了天然阔叶林的净碳固存情况。在人工林和其他木本植被生态系统中，排放均超过了固存。总体而言，我们的估计与粮农组织森林资源评估以及各国研究结果相符，在巴西和美国偏差最大。这两个国家使用的参考数据高度集中，鉴于需要外推到采样不足的区域，这种集中导致了不确定性。我们最终提出了一些建议，以减轻采样不足的影响，并在相对较小的国家汇总碳储量和流量时更好地考虑不确定性。鉴于旨在扩大UNSEEA碳核算的全球倡议，这些行动很及时。
