General Comments

The paper aims to use models to understand how the climate patterns of the Last Interglacial (LIG) influenced patterns of extreme sea levels related to storm surges around the world.  To do so, the authors use large-scale variables from LIG and pre-industrial global climate model simulations (meridional and zonal wind speed as well as sea level pressure) within a hydrodynamic model to simulate water levels along the world's coastlines. The authors find that in a warmer world (LIG), there are substantial seasonal and annual variations in sea level extremes associated with storm surge around the world's coastlines. 

While I find the premise of this study very intriguing, and potentially very informative, I have a number of concerns about the study and how much it can really tell us, being based off of only a few large-scale climate variables from one GCM (though I note that the lack of a full model ensemble is listed as an important caveat by the authors).  Portions of the results and discussion feel very vague, and are in need of more detailed analysis or description before final publication.

Specific Comments

In many ways, the introduction seems under-referenced.  For example, omitting large but highly relevant review papers like Knutson et al., 2019 seems problematic

"Since geological proxies do not have the stratigraphic and temporal resolution needed to address storms and cyclones directly, a proposed method to examine these phenomena in the past is climate modelling (Raible et al., 2021)." -- This seems like an overly broad, if not outright incorrect statement.  There's a lot of literature out there dealing with geological proxies and frequency of storms (see for example Rodysill et al., 2020 [https://www.nature.com/articles/s41598-020-75874-0], Bramante et al, 2020 [https://www.nature.com/articles/s41561-020-00656-2], or Wallace et al., 2019 [https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019PA003665])  Please revise the statement and be more attentive to citing appropriate literature.

"GCMs indicate significant changes in storm tracks under different climate change scenarios (Haarsma et al., 2013; Harvey et al., 2020),"-- Consider adding also Garner et al., 2021, for a recent study in the North Atlantic [https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2021EF002326].

2.5-km resolution along the coastline for a surge model still seems quite coarse.  For instances, other papers, like Lin et al., 2012 use an ADCIRC grid to model surges near NYC under different climate conditions that has a resolution of about 100 m near the coast.  Is there anything that can be done to improve this resolution?

"We note that results in the extra-tropical latitudes must be considered more reliable that in the tropics. This is because the spatial resolution of the climate forcing does not allow GTSM to simulate tropical cyclones with realistic frequency and magnitude (Roberts et al., 2020)." -- I find this statement (and the general methodological approach) somewhat confusing.  I agree that the resolution does not allow for either the GCM or GTSM to simulate tropical cyclones with any realistic frequency or magnitude.  However, the resolution of the GCM and use of only broad scale atmospheric variables in the hydrodynamic model doesn't seem as though it would allow reasonable simulation of frequency or intensity of individual extratropical cyclones either.  Also, we know that tropical cyclones and their impacts are hardly contained with in the tropical latitudes (e.g., Hurricane Sandy, 2012).  Therefore I am confused as to why the authors would think that the approach used is generally more reliable outside the tropics.  It seems to me that the methodology is, at best, able to produce broad scale findings of plausible maximum water levels, which may be used to guide more in-depth analyses in the future (e.g., downscaling individual storms in particular basins/time periods, etc.)

"Several other seasonal anomalies in zonal and meridional winds speed emerge that we do not mention here." -- Then why bother mentioning it?  If the results for these "other seasonal anomalies" were deemed inconsequential to the point of being irrelevant, don't bring it up at all.  If these "other seasonal anomalies" are worth noting, please explain why.

Please provide a citation or more background that justifies using EKE calculated from 10-m zonal and meridional winds as a proxy for storminess in climate models.

Section 3.3 is in great need of some quantitative analysis.  Even simple calculations of the correlation between variables, or a principle component analysis would be extremely useful.

The opening of the Discussion seems to dismiss differences between this work and similar studies of modeled future climates as being due to differences in models and climate benchmarks--so much so that a comparison is not relevant or reasonable.  Yet, the discussion goes on to say that because sea surface temperature patterns from LIG "somewhat resemble" those of projected futures, it is reasonable to use the results from this study to qualitatively draw conclusions about possible future climates.  This section seems to substantially contradict itself, and I find that quite concerning.

Figure 3:  It could be helpful to include some labels on maps, or else some other method of highlighting the areas the authors specifically note in their results.  Same for Figure 4 and 5 maps

Figure 4:  This figure is very hard to read, particular part A.  I recommend a bolder or darker color bar to make results more clear.

Technical Corrections

There are some typos throughout that the authors should work to address before final publication

General Comments

The purpose of this paper is to understand how global storm surges varied during the Last Interglacial period (LIG) and relate the variations in water levels to storm climate. The LIG period is of interest as it represents a potential future climate state of the earth and the overall relationship between changes in storminess and storm surges is an ongoing topic of research. The approach taken in this paper involves using modeled wind and pressure outputs from a GCM and then forcing a global hydrodynamic model to predict storm surges. The authors find various changes in storminess in a warmer climate during the LIG period note qualitative patterns between the two.

I find the topic and the overall scope of the work very interesting and potentially very informative; I would be able to recommend this article pending major revisions and would like to see this happen. I find the discussion, presentation and analysis of results vague and somewhat superficially analyzed in several places. In my opinion, more detailed analyses are required before final publication can be considered. Further, several technical details need to be clarified that as well are elaborated in my review. Of major concern for the reviewer is how one can reliably use relatively coarse resolution approximately 1-degree horizontal resolution GCM outputs to directly force a hydrodynamic model without quantifying/adjusting for potential biases in meteorological forcing inputs.   

Specific Comments

Overall, the introduction could use more precise language and focus.

“Specifically, for the LIG, Kaspar et al. (2007) found a strengthening of the winter mid-latitude storm tracks, along with a northward shift and an extension to the east.”  There are several sentences that state very broad and vague changes to “future storms”. Sometimes the word ‘storm’ is qualified as ‘tropical’ or ‘extratropical’, or even ‘winter mid-latitude storm tracks’ and sometimes it is not. Please be consistent and clarify.

“Ensembles of climate models project a future poleward shift of boreal extra-tropical cyclones, and a decrease in their occurrence (Chang et al., 2012). For the boreal mid-latitudes, the most recent generation of global climate models associate future global warming with a southern shift of winter storm tracks, and weakening of summer storm tracks (Harvey et al., 2020).” To me, these last two sentences appear to be in direct contradiction from one GCM model generation to the next, which raises some serious concerns in the application of GCMs if interpreted literally. Perhaps the authors mean to express that there is great uncertainty in what can happen in terms of future changes to storm tracks? More explanation here is necessary as to why these changes are occurring in modeled simulations from a dynamical perspective.

“To achieve this, we employ meridional and zonal wind speed and sea level pressure from simulations of LIG and PI climate with a global climate model to force a global 90 hydrodynamic model to simulate the extreme water levels along coastlines resulting from storm surges.” This statement raises the important question: how are the meteorological inputs deemed reasonable to force a hydrodynamic model? Normally historical periods are used for a statistical validation of inputs and compared with reanalysis datasets. It is well known that relatively coarse horizontal resolutions associated with the GCMs produce largely biased surface winds at weather time scales that are important for storm surge prediction.

“We note that results in the extra-tropical latitudes must be considered more reliable that in the tropics. This is because the spatial resolution of the climate forcing does not allow GTSM to simulate tropical cyclones with realistic frequency and magnitude (Roberts et al., 2020).” Please clarify what you mean by “reliable”. Both extratropical and tropical cyclones are poorly represented in GCMs hence the large body of research on both statistical and dynamical downscaling to improve the representation of these storms. This brings me back to the previous point about how you determine that the meteorological inputs are physically representative/accurate for storm surge prediction.

Section 3 needs improvement and more quantitative analysis. Perhaps a clustering algorithm could be used to relate the changes in storm surge return periods (globally) and atmospheric variables to better understand their relationship(s). Correlation maps between variables and predictions could also demonstrate the relationship between them and strengthen the discussion. Perhaps also separating out the tropical “warm-season” from the “cool-season” would be helpful to further refine the results since the seasonal separation is opposite for each hemisphere (e.g., summer is JJA in the northern hemisphere while winter in the southern).

I find the visualization of points colored by magnitude in Figures 3, 4 and 5 difficult to analyze. Perhaps these figures could be divided into regions using subplots for select areas of interest that are discussed in the text instead of showing the entire global picture? I also find the colormaps somewhat non-intuitive in Figures 4 and 5 (i.e., blue and red should be inverted).

Section 4 is also confusing in that the authors state that a comparison with previous modeled results would not be valid but then later it is stated that the surface temperature patterns are similar between LIG simulations and modeled warmer future climates, so it’d be “meaningful” to perform intercomparisons.

“Nevertheless, the validity of a comparison with those studies is limited by differences in the climate models and reference climate benchmarks employed, such that it is not possible to separate the effect of differences in climate forcing and of different models” This statement is then later followed by a “On the other hand, spatial patterns of warming during JJA in the Northern Hemisphere are similar across simulations of the LIG and of warmer futures, as evident from the fact that boreal summer sea surface temperature patterns in the LIG simulation somewhat resemble those of the projected futures (Fig. S8). A qualitative 230 comparison of results for the summer of the Northern Hemisphere is therefore meaningful.”  

Of concern by the reviewer is this statement in Section 4.2: “While regional studies have attempted to correct for such biases (Marsooli et al., 2019), global studies have not.” Do the authors have confidence that this statement is correct?

 “However, if we assume biases to equally affect the PI and the LIG simulations, the anomalies between results based on the two simulations, which we report here, should not be impacted by the presence of biases.” That is a significant and important assumption that the biases are equal throughout the PI and LIG simulations. Given the significant changes to the atmosphere and general circulation between the PI and LIG simulations, I would not expect biases to be equivalent at all. The authors have also presented no information to back up this assumption.  I would recommend that this is further investigated before publication can be recommended.

Technical Corrections

There are some grammar and spelling errors throughout that the authors should work to address before my other comments mentioned are addressed. Please be consistent with how the word “storms” are referred to whether that means extra-tropical or tropical, as discussed earlier on.

This manuscript presents simulations of the storm surges at the LIG through combining the climate modeling using CESM1.2 and hydrodynamic modeling using GTSM. The authors’ results show spatial heterogeneity in the seasonal LIG sea level extremes, which they attribute to changes in the atmospheric circulation. The implication of the findings on the interpretation of sea-level proxies is also discussed.

The manuscript is very descriptive and does not have many in-depth analyses into the physical processes. I suggest a major revision before considering for publication. Please see details below.

1. How was the CESM LIG simulation initialized? How long was the simulation? What criteria was used to determine the simulations are in equilibrium (e.g., Line 98)?
2. How well does the authors’ modeling approach reproduce the climate and storm surge in the present-day observation? This question is critical, as it provides information about the performance of the authors’ approach, i.e., is the method valid for the present-day climate?
3. How well does the CESM LIG simulation match proxy data regarding temperature and precipitation? The authors’ results suggest that the large-scale circulation is an important driver of the storm surge changes. However, based on the results presented, it is unclear how well the GCM simulation reproduces the proxy-suggested large-scale climate change, and therefore, it is unclear how well the results in storm surge are.
4. New analysis should be added on the mechanisms linking large-scale circulation and the storm surge changes. Hoes does the large-scale atmospheric circulation impact regional storm surge? Does the mechanism depend on timescale? In other words, do we see similar mechanistic connection in present-day observations at short timescales (e.g., interannual and inter-decadal timescales)?