We investigated the effects of forest management on the carbon (C) dynamics in Romanian forest soils, using two model simulations: CBM-CFS3 and Yasso15. Default parametrization of the models and harmonized litterfall simulated by CBM provided satisfactory results when compared to observed data from National Forest Inventory (NFI). We explored a stratification approach to investigate the improvement of soil C prediction. For stratification on forest types only, the NRMSE (i.e., normalized RMSE of simulated vs. NFI) was approximately 26%, for both models; the NRMSE values reduced to 13% when stratification was done based on climate only. Assuming the continuation of the current forest management practices for a period of 50 years, both models simulated a very small C sink during simulation period (0.05 MgC ha⁻¹ yr⁻¹). Yet, a change towards extensive forest management practices would yield a constant, minor accumulation of soil C, while more intensive practices would yield a constant, minor loss of soil C. For the maximum wood supply scenario (entire volume increment is removed by silvicultural interventions during the simulated period) Yasso15 resulted in larger emissions (−0.3 MgC ha⁻¹ yr⁻¹) than CBM (−0.1 MgC ha⁻¹ yr⁻¹). Under ‘no interventions’ scenario, both models simulated a stable accumulation of C which was, nevertheless, larger in Yasso15 (0.35 MgC ha⁻¹ yr⁻¹) compared to CBM-CSF (0.18 MgC ha⁻¹ yr⁻¹). The simulation of C stock change showed a strong “start-up” effect during the first decade of the simulation, for both models, explained by the difference in litterfall applied to each scenario compared to the spinoff scenario. Stratification at regional scale based on climate and forest types, represented a reasonable spatial stratification, that improved the prediction of soil C stock and stock change.

中文翻译：

---

根据两个模型模拟，造林干预推动了罗马尼亚森林土壤有机碳的变化

我们使用两个模型模拟：CBM-CFS3 和 Yasso15，研究了森林管理对罗马尼亚森林土壤中碳 (C) 动态的影响。与来自国家森林清单 (NFI) 的观测数据相比，模型的默认参数化和 CBM 模拟的协调凋落物提供了令人满意的结果。我们探索了一种分层方法来研究土壤碳预测的改进。仅就森林类型分层而言，两种模型的 NRMSE（即模拟与 NFI 的归一化 RMSE）约为 26%；当仅基于气候进行分层时，NRMSE 值降低到 13%。假设当前的森林管理实践持续 50 年，两个模型在模拟期间都模拟了一个非常小的碳汇 (0.05 MgC ha ^-1年^-1 )。然而，向粗放式森林管理做法的改变会产生持续的、少量的土壤碳积累，而更密集的做法会产生持续的、较小的土壤碳损失。 对于最大木材供应情况（整个体积增量被造林干预去除在模拟期间）Yasso15导致较大的排放量（-0.3的MgCl公顷^-1年^-1）比CBM（-0.1的MgCl公顷^-1年^-1）。在“无干预”情景下，两种模型都模拟了 C 的稳定积累，尽管如此，与 CBM-CSF（0.18 MgC ha ^-1 yr ^-1）相比，Yasso15（0.35 MgC ha ^-1 yr ^-1）。在模拟的第一个十年期间，对 C 库变化的模拟显示出强烈的“启动”效应，对于两个模型，这可以通过应用于每个场景的凋落物与衍生场景的差异来解释。基于气候和森林类型的区域尺度分层，代表了合理的空间分层，提高了对土壤碳库和库变化的预测。