Soil carbon sequestration is seen as an effective means to draw down atmospheric CO₂, but at the same time warming may accelerate the loss of extant soil carbon, so an accurate estimation of soil carbon stocks and their vulnerability to climate change is required. Here we demonstrate how separating soil carbon into particulate and mineral-associated organic matter (POM and MAOM, respectively) aids in the understanding of its vulnerability to climate change and identification of carbon sequestration strategies. By coupling European-wide databases with soil organic matter physical fractionation, we assessed the current geographical distribution of mineral topsoil carbon in POM and MAOM by land cover using a machine-learning approach. Further, using observed climate relationships, we projected the vulnerability of carbon in POM and MAOM to future climate change. Arable and coniferous forest soils contain the largest and most vulnerable carbon stocks when cumulated at the European scale. Although we show a lower carbon loss from mineral topsoils with climate change (2.5 ± 1.2 PgC by 2080) than those in some previous predictions, we urge the implementation of coniferous forest management practices that increase plant inputs to soils to offset POM losses, and the adoption of best management practices to avert the loss of and to build up both POM and MAOM in arable soils.

土壤固碳被视为减少大气CO _{2的有效手段}，但与此同时，变暖可能会加速现有土壤碳的流失，因此需要准确估计土壤碳储量及其对气候变化的脆弱性。在这里，我们展示了如何将土壤碳分离为颗粒物和矿物相关有机物（分别为 POM 和 MAOM）有助于了解其对气候变化的脆弱性和确定碳封存策略。通过将欧洲范围内的数据库与土壤有机质物理分馏相结合，我们使用机器学习方法通​​过土地覆盖评估了 POM 和 MAOM 中矿物表土碳的当前地理分布。此外，利用观察到的气候关系，我们预测了 POM 和 MAOM 中碳对未来气候变化的脆弱性。在欧洲范围内累积时，可耕地和针叶林土壤包含最大和最脆弱的碳储量。尽管我们显示出与气候变化有关的矿物表土的碳损失（到 2080 年为 2.5 ± 1.2 PgC）低于之前的一些预测，但我们敦促实施针叶林管理实践，增加植物对土壤的投入以抵消 POM 损失，以及采用最佳管理实践来避免可耕土壤中 POM 和 MAOM 的损失和积累。