A plausible simulation of the global energy balance is a first-order requirement for a credible climate model. Here I investigate the representation of the global energy balance in 40 state-of-the-art global climate models participating in the Coupled Model Intercomparison Project phase 6 (CMIP6). In the CMIP6 multi-model mean, the magnitudes of the energy balance components are often in better agreement with recent reference estimates compared to earlier model generations on a global mean basis. However, the inter-model spread in the representation of many of the components remains substantial, often on the order of 10–20 Wm⁻² globally, except for aspects of the shortwave clear-sky budgets, which are now more consistently simulated by the CMIP6 models. The substantial inter-model spread in the simulated global mean latent heat fluxes in the CMIP6 models, exceeding 20% (18 Wm⁻²), further implies also large discrepancies in their representation of the global water balance. From a historic perspective of model development over the past decades, the largest adjustments in the magnitudes of the simulated present-day global mean energy balance components occurred in the shortwave atmospheric clear-sky absorption and the surface downward longwave radiation. Both components were gradually adjusted upwards over several model generations, on the order of 10 Wm⁻², to reach 73 and 344 Wm⁻², respectively in the CMIP6 multi-model means. Thereby, CMIP6 has become the first model generation that largely remediates long-standing model deficiencies related to an overestimation in surface downward shortwave and compensational underestimation in downward longwave radiation in its multi-model mean.

对全球能源平衡进行合理的模拟是建立可靠的气候模型的首要条件。在这里，我研究了40种最先进的全球气候模型中全球能量平衡的表示形式，这些模型都参加了耦合模型比较项目6（CMIP6）。在CMIP6多模型均值中，与在全球均值基础上较早的模型相比，能量平衡分量的大小通常与最近的参考估计更好地吻合。但是，模型间的差异在许多组件的表示中仍然很重要，通常在10-20 Wm ^{-2的数量级}在全球范围内，短波晴空预算的各个方面除外，而CMIP6模型现在可以更一致地对其进行模拟。在CMIP6模型中，模拟的全球平均潜热通量中的大量模型间扩散超过20％（18 Wm ^-2），这进一步暗示了它们在表示全球水平衡方面的巨大差异。从过去几十年的模型开发的历史角度来看，当前模拟的全球平均能量平衡分量的幅度最大的调整发生在短波大气晴空吸收和地表向下长波辐射中。在几代模型中，两个组件都逐渐向上调整，幅度为10 Wm ^-2，分别达到73和344 Wm ^-2，分别以CMIP6多模型表示。因此，CMIP6已成为第一代模型，该模型以其多模型平均值极大地弥补了与长期以来对表面下行短波的高估以及对向下长波辐射的补偿低估有关的长期模型缺陷。