CCSM3 simulations

Soil-temperatures simulated by the fully coupled Community Climate System Model version 3.0 (CCSM3) are evaluated using three gridded Russian soil-temperature climatologies (1951-1980, 1961-1990, and 1971-2000) to assess the performance of permafrost and/or soil simulations. CCSM3 captures the annual phase of the soiltemperature cycle well, but not the amplitude. It provides slightly too high (low) soiltemperatures in winter (summer) with a better performance in summer than winter. In winter, soil-temperature biases reach up to 6 K. Simulated near-surface air temperatures agree well with the near-surface air temperatures from reanalysis data. Discrepancies in CCSM3-simulated near-surface air temperatures significantly correlate with discrepancies in CCSM3-simulated soiltemperatures, i.e. contribute to discrepancy in soil-temperature simulation. Evaluation of cloud-fraction by means of the International Satellite Cloud Climatology project data reveals that errors in simulated cloud fraction explain some of the soil-temperature discrepancies in summer. Evaluation by means of the Global Precipitation Climatology Centre data identifies inaccurately-simulated precipitation as a contributor to underestimating summer soil-temperatures. Comparison to snow-depth observations shows that overestimating snow-depth leads to winter soil-temperature overestimation. Sensitivity studies reveal that uncertainty in mineral-soil composition notably contributes to discrepancies between CCSM3-simulated and observed soil-temperature climatology while differences between the assumed vegetation in CCSM3 and the actual vegetation in nature marginally contribute to the discrepancies in soil-temperature. Out of the 6 K bias in CCSM3 soil-temperature simulation, about 2.5 K of the bias may result from the incorrect simulation of the observed forcing and about 2 K of the bias may be explained by uncertainties due network density in winter. This means that about 1.5 K winter-bias may result from measurement errors and/or model deficiencies. Overall, the performance of a permafrost/soil model fully coupled with a climate model depends partly on the permafrost/soil model itself, the accuracy of the forcing data and design of observational network.本研究采用三套网格化（gridded）俄罗斯土壤温度气候学数据集（时间跨度分别为1951-1980年、1961-1990年及1971-2000年），对完全耦合的社区气候系统模型第3.0版（Community Climate System Model version 3.0, CCSM3）模拟的土壤温度进行评估，以检验该模型在多年冻土（permafrost）与土壤模拟方面的性能。CCSM3能够较好地捕捉土壤温度循环的年相位特征，但无法准确再现其振幅。该模型在冬季模拟的土壤温度略偏高，夏季则略偏低，且夏季模拟效果优于冬季。冬季土壤温度偏差最高可达6开尔文（K）。模拟的近地表气温与再分析（reanalysis）数据中的近地表气温吻合度较高。CCSM3模拟的近地表气温偏差与模拟的土壤温度偏差显著相关，即前者会加剧土壤温度模拟的偏差。基于国际卫星云气候学计划（International Satellite Cloud Climatology Project, ISCCP）数据开展的云量评估显示，模拟云量的误差可解释夏季部分土壤温度偏差。通过全球降水气候学中心（Global Precipitation Climatology Centre, GPCC）数据进行的评估则表明，降水模拟失准是导致夏季土壤温度被低估的原因之一。与雪深观测数据的对比显示，雪深模拟偏高会导致冬季土壤温度被高估。敏感性试验结果表明，矿质土壤组分的不确定性对CCSM3模拟结果与观测土壤温度气候学之间的偏差具有显著贡献，而模型预设植被与自然实际植被之间的差异对土壤温度偏差的贡献则相对微弱。在CCSM3模拟的6开尔文土壤温度总偏差中，约2.5开尔文可归因于观测强迫场的模拟失准，约2开尔文则可由冬季观测站网密度不足带来的不确定性解释。这意味着剩余约1.5开尔文的冬季偏差，可能源于观测误差和/或模型自身缺陷。总体而言，与气候模型完全耦合的多年冻土/土壤模型的性能，部分取决于该多年冻土/土壤模型本身、强迫场数据的精度以及观测站网的设计方案。