始新世大洋环流模拟的模式依赖性

MODEL DEPENDENCY OF SIMULATED EOCENE OCEAN CIRCULATION IN THE DEEP⁃TIME MODEL INTERCOMPARISON PROJECT(DEEPMIP)

随着模式的复杂化和多元化发展,国际上不仅组织了现代气候模拟的模式比较计划,也开展了古气候模拟的比较计划。在模式比较计划中,如何有效处理模式间的差异和剔除模式依赖性已成为重要问题。本项研究基于国际深时多模式比较计划(DeepMIP)中的始新世大洋环流模拟结果,选取pCO_(2)倍增试验、模式参数化、古地理边界条件这3个案例,深入分析模式依赖性如何影响古气候模拟结果和模式之间差异性,归纳和总结古气候模拟研究中模式依赖性产生的可能原因。结果显示,部分模式的关键变量对驱动因子响应过于敏感、模式参数化方案过于简单、模式中对古地理边界条件表达的欠准确性均可导致模式依赖性的产生以及模式之间显著差异。这表明在分析数值模式结果时,需要参考多模式结果,并特别关注模拟结果变化的各个环节,从而减少模式依赖性的干扰,增加模拟结果的准确性,以便于更加准确认识气候环境演变过程。

Inter⁃model comparison projects have not been only performed for modern and future climate simulations but also for palaeoclimate simulations.Climate model comparisons are performed to diagnose inter⁃model discrepancies,so that a weighting of model⁃dependencies can be avoided.In the present study,we selected three model⁃dependency cases—Eocene ocean circulation responses to a pCO_(2) tripling,to various vertical diffusivity parameterization and to palaeo⁃geography boundary condition setups—from the Deep⁃Time Model Intercomparison Project(DeepMIP).This study aims to evaluate where inter⁃model dependencies could arise and how those contribute to simulated ocean circulation results.We show that:(1)Divergent responses of ocean circulation to a pCO_(2) tripling are found in DeepMIP models.Unlike other models,the Community Earth System Model(CESM)model suggests a basin⁃scale shift from North Atlantic to North Pacific in response to a pCO_(2) tripling,which most likely arises from its extreme evaporation sensitivity.The lack of geological evidence of the Pacific meridional overturning circulation(PMOC)and extreme evaporation during the Eocene implies that CESM would overresponse to an increase in pCO_(2).(2)Highly simplified vertical diffusion parameterization reduces model freedom and tends to underestimate changes in ocean circulation.(3)Large uncertainties are introduced by the paleogeographic boundary conditions,which is related to the reconstruction and adaption of geography into the model⁃grid frame.Our analyses reveal that model discrepancies can originate from overestimated model sensitivities to a pCO_(2) tripling,highly simplified parameterizations and a less accurate representation of the paleogeographic boundary conditions of the individual models.Our results suggest that it is crucial to carefully diagnose the influences of pCO_(2),parameterization and palaeogeography on the simulation results.Thus,model⁃dependencies can be avoided and a fundamental understanding of palaeoclimate/palaeococeanographic simulations can be obtained.