Soil is the largest organic carbon pool in the terrestrial biosphere, and a small variation of soil organic carbon (SOC) can substantially affect the global carbon cycle and climate. Climate change is a major driver affecting the dynamics of SOC; however, our understanding about the responses of SOC in alpine ecosystems to climate change is quite limited. In particular, the differences of SOC dynamics at different depths were rarely reported. In this study, we investigated the impacts of precipitation variations and warming on SOC dynamics at both top and deep soils in the Qilian Mountains in Northwestern China using a machine learning approach and climate sensitivity experiments. Our simulation revealed the temporal inconsistency between topsoil SOC (in the top 20 cm, denoted as SOC20) and deeper soil SOC dynamics—SOC20 showed a decreasing trend since 2009 which is earlier than that (2012) in the top 100 cm soil (SOC100). We also found that SOC100 may be more sensitive to warming due to the strengthened microbial decomposition rate and additional carbon source through deepened active layer. On the contrary, SOC20 presented more intense responses to precipitation than SOC100, which was mainly attributed to the different responses of upland and lowland SOC to precipitation variations. Our projection indicated that SOC20 may not substantially change under future climate trajectories because the enriched SOC induced by increased precipitation may offset the carbon loss via warming. However, SOC100 was projected to decrease significantly due to the enhanced carbon emissions via warming induced strengthened decomposition rate, additional carbon source from the deepened active layer, and exposed soil carbon to the atmosphere caused by ground subsidence and disrupted soil horizons resulting from thawed frozen soil. In brief, this study deepened our understanding of the mechanism of climate effect on SOC dynamics and can be helpful for regional soil ecological security assessment and risk projection.

土壤是陆地生物圈中最大的有机碳库，土壤有机碳（SOC）的微小变化会严重影响全球碳循环和气候。气候变化是影响SOC动态的主要驱动力。然而，我们对高山生态系统中SOC对气候变化的响应的了解非常有限。尤其是，很少报道不同深度的SOC动力学差异。在这项研究中，我们使用机器学习方法和气候敏感性实验研究了降水变化和变暖对西北祁连山表层和深层土壤SOC动态的影响。我们的模拟显示，表层土壤SOC之间的时间不一致（在顶部20厘米，（表示为SOC20）和更深的土壤SOC动力学-自2009年以来，SOC20呈下降趋势，而在100 cm高土壤（SOC100）中则早于（2012）。我们还发现，由于微生物分解速率增强以及通过加深的活性层增加了碳源，SOC100可能对变暖更为敏感。相反，SOC20对降水的响应比SOC100更为强烈，这主要归因于高海拔和低海拔SOC对降水变化的不同响应。我们的预测表明，在未来的气候变化趋势下，SOC20可能不会发生实质性变化，因为降水增加引起的富集SOC可能抵消了变暖带来的碳损失。然而，预计SOC100会显着下降，这是由于变暖导致的分解速率增加，活性层加深而产生的额外碳源以及地面沉降和融化的冻土导致的土壤层位破坏而使土壤中的碳暴露于大气中，从而使碳排放量显着减少。简而言之，本研究加深了我们对气候影响SOC动态机制的理解，并有助于区域土壤生态安全评估和风险预测。