随着气候变化对全球生态系统产生日益深远的影响，了解和预测生态系统碳循环的动态演变成为全球科学研究的重点。高寒生态系统是全球气候变化的“放大器”，其对温度敏感且冻土广布，易受到全球变暖的不利影响。本文旨在利用先进的陆面过程模型，揭示生态系统在历史时期和未来气候变化情景下碳循环的动态变化及其驱动因素，对于制定生态增汇措施和碳中和策略具有重要指导意义。

       本文以青海省为研究区，其地处青藏高原东北部，作为典型的内陆高寒生态区，生态环境脆弱，是反映整个高原响应气候变化的重要样本区域，但当前对该地区碳循环过程和驱动机制的理解依然匮乏。在高时空分辨率气候强迫数据的驱动下，利用陆面过程模型CLM构建了适应于高寒生态系统的碳动态模拟系统，在评估青海省碳动态模拟精度的前提下，模拟研究区2000-2018年碳通量(GPP/NPP/NEP/Ra/Rh)、碳储量(植被碳与土壤有机碳)及不同深度土壤有机碳的时空分布格局及变化趋势，采用随机森林模型和SHAP方法，量化不同驱动因子对区域碳通量与碳储量的影响。利用CMIP6气候情景数据和统计降尺度算法提高气候模式数据分辨率，选择GFDL-ESM4、IPSL-CM6A-LR、MPI-ESM1-2-HR、MRI-ESM2-0和UKESM1-0-LL共5个气候模式提供的输出结果，引入MBCnSD (Multivariate Bias Correction Statistical Downscaling)算法，逐步降尺度将多气模式情景数据的分辨率提升至0.1°，以此来驱动陆面过程模型，预测不同气候情景(SSP1-2.6、SSP2-4.5和SSP3-7.0)下2015-2100年碳通量、碳储量及不同深度土壤有机碳时空格局与变化特征。在此基础上，分析未来气候、土壤水热条件等因素对青海省碳通量、碳储量和多层土壤有机碳变化的驱动机制，探究气候变化背景下生态系统碳固持及碳释放之间的动态平衡关系及影响因素。

       研究结果表明：

       (1)在2000至2018年期间，青海省生态系统表现出复杂的碳通量与碳储量时空变化趋势，揭示出生态系统碳固持地域性差异。GPP表现出自东南向西北的递减分布格局，显著增加趋势的区域占总面积的14.70%，年际变动较为平稳。NPP的多年平均值约为0.270 kg C/m²/a，且近20年间持续增加。青海省NEP的平均值为0.142 kg C/m²/a，年增长率为4.57%，呈增加和显著增加趋势的区域共占研究区的35.18%，体现了青海省作为碳汇区的功能。植被碳储量和土壤有机碳储量分别为0.22 PgC和9.12 PgC，碳密度整体呈现上升趋势，植被碳密度的年增长率(1.11%)超过了土壤有机碳密度的增长率(0.39%)。同时，不同植被类型、冻土类型下碳储量也存在差异，森林总碳密度最高为21.74 kg C/m²，比高寒草甸和高寒草原分别高1.20倍和2.26倍。多年冻土区的平均碳储量为14.73 kg C/m²，土壤有机碳密度是其植被碳密度的44.75倍，高于季节性冻土区的碳储量。三个深度的土壤有机碳密度均呈现出自青海省东南部向西北部递减的趋势，超过90%的土壤有机碳集中在0-30cm的土壤层中，随着土层深度的增加，土壤有机碳密度显著下降。

       (2)本文深入揭示了影响青海省碳通量与碳储量的关键驱动力，凸显出不同气候和生态因素在生态系统碳平衡中的作用。太阳辐射是影响青海省所有碳通量的最主要驱动因子，其对自养呼吸(RA)和生态系统呼吸(RECO)的贡献率超过52%，对GPP、NPP等其他碳通量的贡献超过48%。NDVI主要影响植被异养呼吸(RH)，而降水则对净初级生产力(NPP)和碳汇的贡献率稍高，气温则是影响生态系统异养呼吸和总体呼吸过程的关键。此外，地形地貌和土壤水热状态对各碳通量的影响相对较小。对于青海省碳储量的驱动分析表明，太阳辐射、NDVI和温度是主导驱动因子，其中NDVI和温度对植被碳有显著正向影响。对于植被碳密度，贡献率排序分别为太阳辐射(52%)，NDVI(13%)和温度(12%)；对于土壤有机碳密度，贡献率排序为NDVI(45%)、太阳辐射(24%)和温度(9%)。

       (3)对未来情景的预测分析表明，青海省的碳通量与碳储量将面临不同程度的变化，从而对生态系统的碳汇功能产生潜在影响。在SSP126情景下，GPP、NPP和NEP的空间格局显示出从2030年到2090年温和增长的趋势。在SSP370情景中，GPP、NPP和RH显示了2060年和2090年青海省局部地区的显著增加，与SSP126相比，在东北部和西南部区域的碳通量激增。NEP的空间分布也显示了增长趋势，青海省NEP高值区逐渐向西南部扩张，且NEP小于0的区域减少，反映出生态系统碳汇能力的潜在增强。SSP585的高排放情景下，表现出较大的变化，碳通量的增长尤其显著。在青海省碳储量的时空动态上，SSP126情景中，植被碳密度的平均增长率为1.9%，SSP370情景为3.1%，而在SSP585情境下，植被碳密度的增长率为6.4%。同样地，土壤有机碳在SSP126情景显示了0.6%的增速，在SSP370和SSP585情景下增速分别为1.5%和1.98%。

    (4)气候因子特别是温度和降水的变化，对青海省未来碳动态起着决定性的作用。气温(TEMP)和降水(PRECT)是影响碳通量的主要驱动力，对碳通量均整体表现为正向影响。对于GPP和NPP，从SSP126到SSP585情景，温度的影响变得更加显著，而降水对GPP的相对贡献略有下降。NEP作为生态系统碳收支的表达，在SSP126情景中，降水是最大的驱动力，在SSP370和SSP585情景中，气温贡献率显著上升，温度和降水对NEP表现为共同影响。Rh对温度的响应在SSP126情景中相对平稳，然而在SSP370和SSP585情景中，其对气温升高的敏感性有所增强。对于植被碳的驱动分析，气温在SSP126情景下的贡献率为24.66%，随着情景移向SSP370升至24.98%，然后在SSP585情景中进一步增加至25.98%。降水在SSP126情景下占主导地位，贡献率为31.34%，但在SSP370中微幅下降至25.18%，而在SSP585中上升为27.48%。降水在SSP126情景下为影响土壤有机碳储量最主要的因素，贡献率高达36.29%，随着气候情景的演变，在SSP370情景下降至32.93%，并在SSP585情景中进一步降至31.61%。气温的相对贡献率从SSP126情景的33.32%逐步增加，并在SSP585情景中增至38.65%。

     (5)在不同深度的土壤中，青海省土壤有机碳储量对于多样化气候情景的响应揭示了生态系统应对未来气候变化的可能途径和趋势。在0-30cm深度的表层土壤中，青海省西南部的土壤有机碳储量相较于2015年有显著增加，而东北部的增加较小。SSP585情景预测的土壤有机碳储量增加区域更广，特别在南部和东北部地区。在30-60cm的土层深度，SSP126情景表现出土壤有机碳储量变动较为显著，11.37%的区域土壤有机碳储量下降0.3-1.5 kg C/m²，而SSP585情景中土壤有机碳储量增加，尤其在青海省南部地区显著。至于60-100cm的深层土壤，各情景下变化均较为轻微，但SSP585情景中土壤有机碳减少的区域面积最小，集中在青海省南部边缘地区。在气候变化对有机碳密度响应的分析中，SSP126情景下降水(PRECT)对0-30cm土壤层有机碳密度的贡献率最高，随着情景向SSP370和SSP585转变，气温对土壤有机碳变动的影响显著增加，成为影响土壤有机碳储量的最关键因素。在30-60cm土层，降水的贡献率自SSP126情景至SSP585情景呈下降趋势，而气温贡献率则相反。气温和降水对土壤有机碳的影响在所有情景中均呈现正向效应，尤其在表层土壤中。三种气候情景中，土壤水分和土壤温度都对土壤有机碳产生正向影响，在0-30cm深度上土壤水分始终是土壤有机碳变化最主要的驱动因子。

       With the increasingly profound impact of climate change on global ecosystems, understanding and predicting the dynamic evolution of ecosystem carbon cycles has become a focal point of global scientific research. Alpine ecosystems act as an "amplifier" of global climate change, and they are highly sensitive to temperature and characterized by widespread permafrost, thus vulnerable to the adverse effects of global warming. This paper aimed to use an advanced land surface process model to reveal the dynamics and driving factors of ecosystem carbon cycles during historical periods and under future climate change scenarios. It was of directive significance for guiding regional carbon emissions and climate change policies.

       This paper focused on Qinghai Province, located in the northeastern Tibetan Plateau, a typical fragile alpine ecosystem indicative of the broader plateau's response to climate change. The region's ecological environment was delicate, making it an essential sample area reflecting the entire plateau's response to climate change. However, the current understanding of the carbon cycle processes and driving mechanisms in this region remains sparse. Driven by high spatial and temporal resolution climate forcing data, the Community Land Model (CLM) was employed to build a carbon dynamics simulation system tailored for cold alpine ecosystems. Based on the evaluation of the simulation accuracy of Qinghai Province's carbon dynamics, the study simulated the temporal and spatial patterns and trends of carbon fluxes (GPP, NPP, NEP, Ra, Rh), carbon storages (vegetation and soil organic carbon), and soil organic carbon at different depths from 2000 to 2018. The Random Forest model and SHAP method were applied to quantify the impact of various drivers on regional carbon flux and storage. Utilizing CMIP6 climate scenario data and statistical downscaling algorithms, the resolution of model data was enhanced. Outputs from five climate models-GFDL-ESM4, IPSL-CM6A-LR, MPI-ESM1-2-HR, MRI-ESM2-0, and UKESM1-0-LL—were selected, and the Multivariate Bias Correction Statistical Downscaling (MBCnSD) algorithm was introduced to stepwise increase the resolution of multimodal scenario data to 0.1°. This drove the land surface process model, predicting spatial-temporal patterns and change features of carbon fluxes, storages, and soil organic carbon at different depths under various climate scenarios (SSP1-2.6, SSP2-4.5, and SSP3-7.0) for 2015-2100. On this basis, the study analyzed the mechanisms driving changes in carbon flux, storages, and multi-layer soil organic carbon under future climatic and soil hydrothermal conditions in Qinghai Province. It explored the dynamic balance between carbon sequestration and release within ecosystems against the backdrop of climate change and the influencing factors, providing theoretical and technical support for carbon management and ecological construction in Qinghai Province.

       The results of the study indicated:

       (1) From 2000 to 2018, the ecosystems of Qinghai Province exhibited complex temporal and spatial trends in carbon fluxes and storages, revealing regional differences in the capacity for carbon sequestration and emission. Gross Primary Productivity (GPP) displayed a decreasing distribution pattern from southeast to northwest, with the areas showing a significant increasing trend accounting for 14.70% of the total area,  while the interannual variability remained relatively stable. The multi-year average value of Net Primary Productivity (NPP) was approximately 0.270 kg C/m²/year, and it continued to increase over the past two decades. The average Net Ecosystem Productivity (NEP) of Qinghai Province was 0.142 kg C/m²/year, with an annual growth rate of 4.57%. Regions with increasing and significantly increasing trends constituted 35.18% of the study area, reflecting Qinghai Province's function as a carbon sink. The carbon storages of vegetation and soil organic carbon were 0.22 PgC and 9.12 PgC, respectively, showing an overall upward trend in carbon density. The annual growth rate of vegetation carbon density (1.11%) exceeded that of soil organic carbon density (0.39%). Additionally, carbon storage varied among different vegetation types and permafrost types, with the highest total carbon density in forests at 21.74 kg C/m², which was 1.20 times and 2.26 times higher than that of alpine meadows and alpine grasslands, respectively. The average carbon storage in permafrost areas was 14.73 kg C/m², with soil organic carbon density being 44.75 times that of vegetation carbon density, exceeding that of the seasonally frozen soil areas. The soil organic carbon densities at three depths all showed a decreasing trend from the southeast to the northwest of Qinghai Province, with over 90% of the soil organic carbon concentrated within a depth of 0-30 cm, and the soil organic carbon density sharply declined with increasing soil depth.

       (2) The study delved into the key driving forces affecting carbon fluxes and storage in Qinghai Province, highlighting the role of various climatic and ecological factors in the carbon balance of the ecosystem. Solar radiation emerged as the most significant driving factor for all carbon fluxes in Qinghai Province, contributing more than 52% to autotrophic respiration (RA) and ecosystem respiration (RECO), and over 48% to other carbon fluxes such as GPP and NPP. The Normalized Difference Vegetation Index (NDVI) primarily influenced vegetation respiration, while precipitation had a slightly higher contribution to Net Primary Productivity (NPP) and carbon sequestration. Temperature was crucial in affecting heterotrophic respiration and the overall respiration processes of the ecosystem. Moreover, topography and soil thermal conditions had relatively smaller impacts on the carbon fluxes. Analyses of drivers for carbon storage in Qinghai Province indicated that solar radiation, NDVI, and temperature were the dominant driving factors, with NDVI and temperature having a significant positive impact on vegetation carbon. For vegetation carbon density, the contribution rates were in the order of solar radiation (52%), NDVI (13%), and temperature (12%). For soil organic carbon density, the contribution rates ranked as NDVI (45%), solar radiation (24%), and temperature (9%).

       (3) Predictive analyses of future scenarios indicated that the carbon flux and storage in Qinghai Province would experience varying degrees of change, potentially impacting the ecosystem's carbon sink function. Under the SSP126 scenario, the spatial patterns of GPP, NPP, and NEP showed a mild increasing trend from 2030 to 2090. In the SSP370 scenario, significant increases in GPP, NPP, and RH were observed by 2060 and 2090 in parts of Qinghai Province, with a surge in carbon flux in the northeastern and southwestern regions compared to the SSP126 scenario. The spatial distribution of NEP also indicated a growing trend, with the high-value areas of NEP expanding southwestward and the areas where NEP was less than zero diminishing, suggesting a potential enhancement in the ecosystem's carbon sink capacity. Under the high-emission SSP585 scenario, the most dramatic changes were displayed, with particularly notable increases in carbon flux. In terms of the temporal and spatial dynamics of carbon storage in Qinghai Province, under the SSP126 scenario, the average growth rate of vegetation carbon density was 1.9%, 3.1% for the SSP370 scenario, and 6.4% under the SSP585 scenario. Similarly, soil organic carbon showed an increase of 0.6% in the SSP126 scenario, with the growth rate accelerating to 1.5% and 1.98% under the SSP370 and SSP585 scenarios, respectively.

       (4) Climate factors, especially changes in temperature and precipitation, played a decisive role in the future carbon dynamics of Qinghai Province. Temperature and precipitation rate were the main driving forces affecting carbon flux, showing an overall positive effect. For GPP and NPP, the influence of temperature became more pronounced from the SSP126 to SSP585 scenarios, while the relative contribution of precipitation to GPP slightly declined. NEP, as an expression of the ecosystem's carbon budget, revealed that precipitation rate was the dominant driving force in the SSP126 scenario, but the contribution of temperature significantly increased in the SSP370 and SSP585 scenarios, with both temperature and precipitation jointly influencing NEP. The response of Rh to temperature remained relatively stable in the SSP126 scenario; however, its sensitivity to rising temperatures was enhanced in the SSP370 and SSP585 scenarios. In the driving analysis for vegetation carbon, temperature had a contribution rate of 24.66% in the SSP126 scenario, which increased to 24.98% in the SSP370 scenario before further rising to 25.98% in the SSP585 scenario. The precipitation rate dominated in the SSP126 scenario with a contribution rate of 31.34%, slightly decreased to 25.18% in the SSP370 scenario and rose to 27.48% in the SSP585 scenario. Precipitation rate was the most significant factor affecting soil organic carbon storage in the SSP126 scenario, contributing 36.29%, which decreased to 32.93% in the SSP370 scenario and further to 31.61% in the SSP585 scenario. The relative contribution of temperature gradually increased from 33.32% in the SSP126 scenario and reached 38.65% in the SSP585 scenario.

       (5) The study revealed that soil organic carbon storage at various depths in Qinghai Province responded to diversified climate scenarios, demonstrating potential pathways and trends for the ecosystem to cope with future climatic changes. In the surface soil layer of 0-30cm depth, there was a significant increase in the soil organic carbon storage in the southwestern part compared to the year 2015, whereas the northeastern part experienced a smaller increase. Under the SSP585 scenario, an expansion in the areas of increased soil organic carbon storage was predicted, particularly in the southern and northeastern regions. For the soil layer at 30-60cm depth, the SSP126 scenario exhibited more noticeable variations in soil organic carbon storages, with a 0.3-1.5 kg C/m² decrease in 11.37% of the area, while the SSP585 scenario anticipated an increase in soil organic carbon storages, notably in the southern regions of Qinghai Province. As for the deep soil layer of 60-100cm, all scenarios showed minor changes, but the SSP585 scenario presented the smallest area of reduced soil organic carbon, concentrated along the southern fringe of Qinghai Province. In the analysis of the response of organic carbon density to climatic changes, the precipitation rate had the highest impact on the density of soil organic carbon in the 0-30cm layer under the SSP126 scenario. However, as the climate scenarios shifted toward SSP370 and SSP585, the influence of temperature on variations in soil organic carbon significantly increased, becoming the most crucial factor affecting soil organic carbon storage. In the 30-60cm layer, the contribution of precipitation rate decreased from 46.5% in the SSP126 scenario, whereas the contribution of temperature even rose to 37.53% in the SSP585 scenario. Both temperature and precipitation rate exhibited positive effects on soil organic carbon across all scenarios, especially in the topsoil layer. Throughout the three climate scenarios, soil moisture and temperature had a positive influence on soil organic carbon, with soil moisture continually being the primary driver of changes in organic carbon at a depth of 0-30cm.