Soil organic matter decomposition and its interactions with climate depend on whether the organic matter is associated with soil minerals. However, data limitations have hindered global-scale analyses of mineral-associated and particulate soil organic carbon pools and their benchmarking in Earth system models used to estimate carbon cycle–climate feedbacks. Here we analyse observationally derived global estimates of soil carbon pools to quantify their relative proportions and compute their climatological temperature sensitivities as the decline in carbon with increasing temperature. We find that the climatological temperature sensitivity of particulate carbon is on average 28% higher than that of mineral-associated carbon, and up to 53% higher in cool climates. Moreover, the distribution of carbon between these underlying soil carbon pools drives the emergent climatological temperature sensitivity of bulk soil carbon stocks. However, global models vary widely in their predictions of soil carbon pool distributions. We show that the global proportion of model pools that are conceptually similar to mineral-protected carbon ranges from 16 to 85% across Earth system models from the Coupled Model Intercomparison Project Phase 6 and offline land models, with implications for bulk soil carbon ages and ecosystem responsiveness. To improve projections of carbon cycle–climate feedbacks, it is imperative to assess underlying soil carbon pools to accurately predict the distribution and vulnerability of soil carbon.

土壤有机质分解及其与气候的相互作用取决于有机质是否与土壤矿物质相关。然而，数据的限制阻碍了对矿物相关和颗粒土壤有机碳库的全球范围分析以及用于估计碳循环-气候反馈的地球系统模型中的基准测试。在这里，我们分析了观测得出的土壤碳库的全球估计值，以量化其相对比例，并计算其气候温度敏感性，即碳随温度升高而减少。我们发现颗粒碳的气候温度敏感性平均比矿物伴生碳高 28%，在凉爽气候下最高可达 53%。此外，这些底层土壤碳库之间的碳分布驱动了大量土壤碳库的新兴气候温度敏感性。然而，全球模型对土壤碳库分布的预测差异很大。我们表明，在耦合模型比对项目第六阶段的地球系统模型和离线土地模型中，概念上与矿物保护碳类似的模型池的全球比例范围为 16% 至 85%，这对整体土壤碳年龄和生态系统具有影响反应能力。为了改进碳循环-气候反馈的预测，必须评估潜在的土壤碳库，以准确预测土壤碳的分布和脆弱性。