Soil organic carbon (SOC) is the largest carbon pool in the terrestrial carbon cycle, which is closely linked to climate change and global warming feedbacks. Based on proxy observation data and outputs from the fifth phase of the Coupled Model Inter-comparison Project (CMIP5), we analyzed quantitatively the spatial distribution and temporal change for SOC using boosted regression trees (BRT) over the Tibetan Plateau (TP). The ensemble proxy observation SOC stock data indicated a decreasing spatial distribution from southeast to northwest over the TP. We used data from ten CMIP5 earth system models (ESMs) for SOC, which exhibited differences. However, the CMIP5 multi-model ensemble (MME) result presented a similar spatial distribution pattern to the ensemble proxy observation SOC, while SOC storage and turnover times from the CMIP5 MME model were less than the ensemble proxy observation. The BRT results indicated that air temperature (Ta) accounted for 44.81% of the relative contribution and was the most influential variable on MME SOC. This was followed by net primary production (NPP), with a 19.09% relative contribution. The relative influence of the top 10 cm soil temperature, total soil water, and precipitation on MME SOC was 13.55%, 12.68%, and 9.87%, respectively. Using the BRT method to determine the spatial distribution of relative contribution of SOC for these five variables demonstrated that Ta was mainly higher over the central and northwestern regions of the TP, and NPP was higher over the western central regions and along the plateau’s eastern edge. The statistical frequency of maximum relative contribution to SOC for the five variables indicated that the relative contribution of NPP covered the largest area, with 32% of the total grid numbers, followed by Ta, with 25%.

土壤有机碳（SOC）是陆地碳循环中最大的碳库，与气候变化和全球变暖反馈密切相关。基于耦合模型比对项目 (CMIP5) 第五阶段的代理观测数据和输出，我们使用增强回归树 (BRT) 在青藏高原 (TP) 上定量分析了 SOC 的空间分布和时间变化。集合代理观测 SOC 储量数据表明青藏高原上从东南向西北递减的空间分布。我们使用来自十个 CMIP5 地球系统模型 (ESM) 的 SOC 数据，这些数据表现出差异。然而，CMIP5 多模式集合 (MME) 结果呈现出与集合代理观测 SOC 相似的空间分布模式，而来自 CMIP5 MME 模型的 SOC 存储和周转时间少于集合代理观察。BRT结果表明，气温（Ta）占相对贡献的44.81%，是对MME SOC影响最大的变量。其次是净初级生产（NPP），相对贡献率为 19.09%。表层10 cm土壤温度、土壤总水分和降水对MME SOC的相对影响分别为13.55%、12.68%和9.87%。使用 BRT 方法确定这五个变量的 SOC 相对贡献的空间分布表明，Ta 主要在青藏高原中部和西北部地区较高，NPP 在中部西部地区和高原东部边缘较高。