Impact of the temperature-cloud phase relationship on the simulated Arctic warming during the last interglacial
Abstract. The Arctic during the last interglacial period (LIG) was considered warmer than today. While a recent proxy-based study suggests the disappearance of summer sea ice in the Arctic at the LIG, many climate models fail to capture this feature. It is thus essential to investigate sources of uncertainty in numerical models. The current study examines the impact of the temperature-cloud phase relationship. Sensitivity studies are conducted for the first time to explore the potential importance of this relationship in simulating the LIG climate. Two different cloud parameter sets are used for an atmosphere-ocean general circulation model with and without the dynamic vegetation feedback. The model with cloud parametrization permitting liquid water at a lower temperature and a larger fraction of supercooled liquid water at the same temperature simulates a warmer preindustrial (PI) climate, larger annual mean Arctic warming at the LIG, and substantially reduced sea ice cover during summer at the LIG. It is demonstrated that the low-level clouds play a crucial role in controlling the Arctic response via the greenhouse effect. The result indicates the importance of the temperature-cloud phase relationship in simulating the Arctic climate at the LIG. It also highlights the importance of accurately simulating modern sea ice thickness and representing the processes that affect the fraction of supercooled liquid water in clouds.
This study presents investigation into the role of cloud phase parameterization in simulating the Arctic climate during LIG. The topic is highly relevant, given the persistent model-data discrepancy regarding the magnitude of Arctic warming and summer sea ice extent during this period. The authors convincingly demonstrate that a parameter set allowing for more supercooled liquid water can lead to a significantly warmer Arctic and a closer match to proxy-based summer sea ice reconstructions. While the findings are intriguing and the experimental design is robust in its sensitivity framework, several key aspects of the argumentation require clarification and a more nuanced discussion to strengthen the paper's conclusions.
1 Impact of Model Biases and the Validity of Proxy Comparison
It seems there are significant model biases in the PI control simulations, notably a warm bias over North America and an unrealistic sea ice thickness distribution near Canada and Greenland. If the model's description of the modern Arctic climate is flawed, the feedback processes it simulates (e.g., sea ice albedo, sea ice clouds) may also be inaccurate. The fact that the PI simulations using parameter set A (Figure 6a) appear to be closer to observed sea ice concentrations. The authors should clearly explain how the modern biases may affect their interpretation of the LIG results and the generalizability of their findings. In addition, the authors admit they cannot claim
which model versions is more realistic. Then is it necessary to compare the LIGvC and LIGvA with the proxy reconstruction?
It is necessary to discuss whether the improved agreement with the LIG proxy parameters stems from more realistic physical processes or from accidental compensation of model errors.
2 Initial Condition Sensitivity
The central finding—that a warmer, thinner sea-ice initial state (PIvC) leads to an ice-free LIG Arctic—raises the critical question of whether the cloud-phase process would be sufficiently powerful to generate a similar outcome if the model were initialized from a more realistic, colder glacial state. Is the result demonstrating a unique, enhanced physical response to LIG forcing, or is it primarily a consequence of sensitivity to the initial state? The manuscript does not sufficiently disentangle these two effects.
3 Overemphasis on Annual Mean Temperature
The analysis of surface air temperature changes focuses heavily on annual means (e.g., Fig. 2). However, the crucial processes for sea ice evolution—melting in summer and freezing in autumn—are inherently seasonal. The key mechanism proposed, the liquid cloud longwave feedback, is explicitly shown to be a autumn/winter phenomenon. Therefore, the annual mean temperature is a less informative metric for the core thesis than a detailed seasonal analysis.
The authors should place greater emphasis on the seasonal progression of temperature and energy budget changes when discussing the causes of sea ice loss. Seasonal (e.g., JJA, SON) temperature and radiation anomaly maps could be prioritized in the main text to directly link the forcing to the response.
4 Oversimplified Explanation for Polar Asymmetry
The explanation for the asymmetric temperature response between the Arctic (warming) and Antarctic (cooling) to the cloud parameter change is attributed solely to the difference in summer climatological temperatures. This explanation is plausible but likely incomplete. The discussion of this asymmetry should be expanded to acknowledge the potential role of these large-scale dynamical factors, even if a full analysis is beyond the paper's scope. atmospheric circulation, moisture transport,
5 The attribution of autumn/winter warming
The attribution of autumn/winter warming primarily to the cloud parameterization requires refinement. A compelling alternative explanation is the summer remnant effect whereby the larger summer sea-ice loss in LIGvC creates a warmer, moister autumn lower atmosphere, passively leading to more clouds and warming. While the AGCM experiments provide some evidence for an active role of the cloud phase, the coupled model result is likely a combination of this passive response and the active cloud-phase amplification. The authors should clearly distinguish between these two effects and present the autumn/winter warming as a result of a positive feedback loop where the cloud parameterization acts as a powerful amplifier of the conditions created by prior sea-ice loss.