The paper of Zhang et al. simulates future climate projections over California using the Simplified Convection Permitting E3SM Atmosphere Model Version 0. Overall, the paper is well organized, and it can help in understanding future climate change in the California region, providing more detail due to the higher model resolution. However, some issues still need to be improved. Main concerns about this manuscript are listed below.

 

1.    Why does the author analyze the SSP585 scenario? The author mentions in section 2.1.5 that SSP585 is a worst-case scenario with a high probability that it will not occur. However, many studies show that the warming of the SSP370 scenario cannot be ignored either  (IPCC AR6). Of course, it is not necessary for the author to simulate the SSP370 scenario again, but it is necessary to explain it again in the article. I think just referring to the study of Tebaldi et al. is not convincing enough.

Reference:

Masson-Delmotte V P, Zhai P, Pirani S L, et al. Ipcc, 2021: Summary for policymakers. in: Climate change 2021: The physical science basis. contribution of working group i to the sixth assessment report of the intergovernmental panel on climate change[J]. 2021.

2.    Sometimes the resolution of the model described by the author is 3.25 km (Line 3), and sometimes it is 3 km (Line 47), which needs to be unified.

3.    Line 3: the author says that they have “developed” a CARRM model, but after reading the research, I think the author only applies the relevant model to the climate simulation field, so I suggest modifying the relevant expression. In my opinion, designing the new RRM grids and generating the model configurations cannot be considered as “developing” the new model.

4.    Figures 1-19: For each subfigure, it is recommended to add legends such as (a) (b).

5.    Line 82: The author should briefly introduce the difference between “Modern regionally refined model (RRM)” and convective-permitting model (CPM) in their research. In fact, CPM has been widely used in climate simulations in various regions.

Reference:

Prein A F, Langhans W, Fosser G, et al. A review on regional convection‐permitting climate modeling: Demonstrations, prospects, and challenges[J]. Reviews of geophysics, 2015, 53(2): 323-361.

Kendon E J, Ban N, Roberts N M, et al. Do convection-permitting regional climate models improve projections of future precipitation change?[J]. Bulletin of the American Meteorological Society, 2017, 98(1): 79-93.

6.    Section 2.1.5: In the historical period, is there any quantitative standard to describe the model's ability to simulate ENSO? This part is very important, because once the model cannot accurately simulate ENSO, the following years are not actually representative. In addition, the author should also emphasize how the simulation results related to future projections are related to ENSO.

7.    Line 329: 2-m temperature -> T2m

8.    Lines 342-345: Here, the author briefly analyzes the reasons for the biases of model precipitation simulation. In fact, regional climate models generally overestimate the meridional moisture flux. For example, the study of Gao et al. found that the WRF model overestimates southerly wind transport over eastern China. I suggest the authors cite this work to strengthen the reliability of the results

Reference:

Gao Z, Yan X, Dong S, et al. Object-based evaluation of rainfall forecasts over eastern China by eight cumulus parameterization schemes in the WRF model[J]. Atmospheric Research, 2023, 284: 106618.

9.    Section 3.2.2: In fact, for precipitation, the most obvious added value of the convective-permitting model is the simulation of diurnal variations and MCS. Can the authors show some figures related to diurnal variations in the supplementary material?

Reference:

Guo Z, Fang J, Shao M, et al. Improved summer daily and sub-daily precipitation over Eastern China in convection-permitting simulations[J]. Atmospheric Research, 2022, 265: 105929.

Yun Y, Liu C, Luo Y, et al. Warm-season mesoscale convective systems over eastern China: Convection-permitting climate model simulation and observation[J]. Climate Dynamics, 2021, 57: 3599-3617.

10.    Section 3.2.4: The author points out that the lack of marine stratocumulus clouds is a common issue in low-resolution model. In fact, for models with higher resolution but not enough to explicitly resolve the cumulus convection process, the simulation of cumulus clouds also has significant shortcomings. Cumulus clouds will release latent heat through condensation, affecting stratus clouds and ground temperature. Authors are advised to cite relevant work: 

Reference:

Chikira M, Sugiyama M. A cumulus parameterization with state-dependent entrainment rate. Part I: Description and sensitivity to temperature and humidity profiles[J]. Journal of the Atmospheric Sciences, 2010, 67(7): 2171-2193.

Gao Z, Zhao C, Yan X, et al. Effects of cumulus and radiation parameterization on summer surface air temperature over eastern China[J]. Climate Dynamics, 2023, 61(1-2): 559-577.

11.    Line 498: Although this is a commonly used variable, give the formula for calculating the short-wave radiative forcing.  

12.    The author calculated multiple variables in the California region in the research. I would like to know whether these variables have an impact on each other? Or, how are they related? Can the author give a schematic diagram like the following article?

Reference:

Wang X, Chen D, Pang G, et al. Effects of cumulus parameterization and land-surface hydrology schemes on Tibetan Plateau climate simulation during the wet season: insights from the RegCM4 model[J]. Climate Dynamics, 2021, 57(7-8): 1853-1879.  

Summary

Zhang et al. in “Leveraging Regional Mesh Refinement to Simulate Future Climate Projections for California Using the Simplified Convection Permitting E3SM Atmosphere Model Version 0” evaluate the historical skill and future projections of the U.S. Department of Energy’s new cloud-resolving model, SCREAMv0.  The authors evaluate a historical 5-year (2015-2020) period produced by SCREAMv0 under an AMIP experimental protocol with regional refined mesh capabilities (RRM) focused over California (CARRM) compared against a conventional ESM simulation (E3SMv1) and observation-based gridded products (PRISM and UofASWE).  The variables evaluated are temperature, precipitation, and SWE.  The authors then evaluate projections of three, five-year periods (2029-2034; 2044-2049; and 2094-2099) in the future under a high emissions scenario, SSP585.  The future climate simulations project much stronger extreme heat days, more extreme precipitation rates, and substantial losses in winter-to-spring snowpacks.

Overall, I think the paper fits well within the scope of GMD and could be, given more work, a valuable contribution to the scientific community.  The findings have both scientific and societal impact as CARRM is one of the first applications of SCREAMv0 to evaluate its historical skill and investigate projections of future climate change.  CARRM also represents a state-of-the-art methodology leveraging an ESM with RRM capabilities.  This presents an exciting new advance in high-resolution climate modeling (e.g., convective resolving) as well as new capabilities for RRM simulations (e.g., inner grid nudging capabilities).  The experimental design description and methods section is also very clear and would enable reproducibility.  I respected that the authors even provided model error messages when running the simulations, which is both refreshing and uncommon in the literature.  I also appreciated that the authors provided all of their code/post processed simulation data in the “Code and data availability” section.

With that said, I encountered frequent issues with the manuscript’s sentence structure (e.g., incomplete sentences, run on sentences, logical flow, etc.) and imprecise description of results.  I also completely understand computational limitations in running cutting-edge RRM simulations, but given the large interannual variability of the California hydroclimate, I’m a bit worried about only using 5-year simulations to interrogate historical skill and future projections of precipitation, temperature, and snowpack (especially in the California Sierra Nevada).  The authors also stated quite strong conclusions from their results, without caveating those findings with any sort of observational and/or internal variability uncertainty estimates.  In my experience, simulations would need to be run for 15-20 years to even begin to see convergence in California hydroclimate summary statistics (i.e., mean, median, IQR).  At the very least, I would like to see the authors do a more rigorous job of setting the context of the CARRM simulations in the broader literature of GCM, RCM, VR, and RRM simulations conducted thus far and utilize more in-situ observations and, at least, one other observation-based gridded product.  This will, at least, better contextualize the uncertainty in leveraging 5-year simulations (internal variability).

Therefore, I think there are still substantial and major revisions needed prior to this paper being accepted in GMD.

Comments and Suggested Edits

Line 16 – Delete “In addition to the countless smaller valleys and mountain ranges, California’s complex topography” and add “This complex topography”

Line 18 – Change “accumulated water” to “acting as a key natural store of that precipitation”

Line 19 – Delete “extremely”

Line 20 – Change “energy sector” to “energy supply”

Line 15-25 – the authors use very dated citations and might consider citing some of the more recent literature on California hydrometeorology/hydroclimate

Line 28 – What does “14% generation” mean?  Do you mean “provide 14% of the energy supply”?

Line 29-30 – this sentence does not make sense and needs to be removed or revised for clarity “The predicted short-term and long-term effects due to energy changes of future wind and radiation on Central Valley temperatures have also received attention”

Line 32-40 – I think this paragraph should be the first paragraph and portions of the 1^st paragraph should be integrated into this paragraph

Line 48 – Change “sometimes rapidly” to “can rapidly”

Line 51 – I think you mean “snow-rain transition” rather than “snow albedo feedback”

Line 53-54 – Change “these microclimatic nuances … accurately project and understand of the state’s complex climate patterns” to “The processes that give rise to these microclimates … understand their interactions and project how they may become altered under climate change”

Line 55 – Delete “hazard”

Line 60 – this was also recently evaluated in RRM-E3SM at 14km vs 7km vs 3.5km horizontal resolutions - https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2023MS003793

Line 66 – Change “simply disappear with running with a resolution of 3km” to “are significantly reduced when run at 3km horizontal resolution”

Line 74-75 – Change “with the main purpose of dynamically downscaling…” to “allowing for low-resolution boundary condition data to be dynamically downscaled to high resolution over regions of interest”

Line 77 – Change “predictions” to “projections”

Line 77-78 – this sentence does not make sense and needs to be removed or revised for clarity “The sub-GCM grid-scale details are represented by downscaling capability”

Line 79 – Change “global GCMs have also undergone a history of…” to “GCMs now have the capability to employ variable resolution grids and regionally refined meshes by capitalizing on unstructured grid development”

Line 82 – Change “GCM scale” to “Synoptic scale”

Line 88 – Change “more comprehensive simulations” to “AMIP and fully-coupled simulations”

Line 113 – change “regionally refined model” to “RRM” (already defined earlier in the manuscript; check in other parts of the manuscript)

Line 117-118 – SCREAM has already been defined earlier in the manuscript

Line 119-120 this sentence does not make sense and needs to be revised for clarity “SCREAM leverages DOE’s leadership in computationally intensive frontier designed for convective permitting scales and does not parameterize subgrid-scale deep convection since it uses a resolution of 3.25 km globally”

Line 125-128 – The authors should remove mention of “Amazon precipitation, tropical precipitation, etc.” when making the case that it is well suited for application over California.

Line 140 – Change “towards improving marine…which is important for the climate off the California coast” to “that improved marine…which is important for representing the California coastal climate”

Line 143 – Change “routine” to “routing”

Line 152 – “transition zone between them” it is hard to tell if there is, at least, a 6 dx grid cell transition between refinement regions.  Is this the case?  If not, do you think this would present any numerical artifacts in the simulation?

Line 157 – Change “computation” to “computational”

Line 158 – Delete “passing westward” to “originating”

Line 160 – FWIW the “sensitivity of the size of the refined mesh for the simulation of atmospheric rivers” was (partly) explored with CESM already - https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019JD031977

Figure 1 – can the authors standardize the color bar range across all four subpanel plots of topography?  The authors should also add subpanel plot identifiers a) … f) and reference in the main text.  Also, any acronyms should be defined in the figure caption to make the figure “standalone” from the main text.

Line 165 – reference for “tricolor function”?

Line 167-169 – cite this statement (nominal vs effective resolution)

Line 173 – “hoirzontal” to “horizontal”

Line 174 – “highorder” to “higher order”

Line 177-182 – cite this statement about “good grid properties” and explain what is meant by “good”

Line 177-182 – break this sentence up into several (run on sentence)

Line 200-204 – I applaud the authors for putting this in their manuscript (esp. for reproducibility).  FWIW (I believe) these errors have also been reported in detail here - https://github.com/E3SM-Project/E3SM/issues/5089

Line 208 – Change “drive coarse grid” to “provide coarse scale fields that drive CARRM”

Line 213 – Define “CICE”

Line 214 – Change “was” to “were”

Figure 2 – can the authors make sure that text does not run onto the figures.  Please define all acronyms to make the figure “standalone” from the main text.  Change “with no nudging is applied in red area” to “where nudging is not applied in red areas”

Line 229 – Delete one of the “the”

Line 238-239 – rather than making two rows of angled maps, couldn’t the authors simply difference E3SMv1 and SCREAMv0 CARRM IVT fields and show the difference plot?  This would also allow you to “see” any potential “artifacts” from the various nudging coefficients.

Line 240 – Change “Experiment” to “Experimental”

Line 241-244 – Change “radiative forcing path close to the highest…” to “which is comparable to the radiative forcing path of the highest representative …” Change “due to policy” to “due to policy interventions that promote carbon emission mitigation and sequestration”

Line 241-252 – if the authors are worried about the current debate in the climate literature regarding “hot models” (equilibrium climate sensitivity) and/or “worst case scenario” (climate “doomism”) they could also cast their simulations as being run according to the global warming levels of interest to the IPCC AR6 reports (i.e., +1.5 degC, +2.0 degC and +3.0 degC)

Line 249 – Delete “to some extent”

Line 258-261 – as mentioned earlier, this might be a good time to provide the global warming levels provided by E3SMv1 (relative to an early 19^th/20^th century 30-to-50 year baseline)

Line 263 – Change “to reanalysis” to “with a reanalysis dataset”

Line 264 – Change “regional model to historical” to “with historical”

Line 268 – “WRF” has not yet been defined and should probably be mentioned in the following way “like commonly employed regional climate downscaling approaches such as the Weather Research and Forecasting, WRF, model”

Line 240-270 – do the authors allow for a “spin-up” of the model prior to analysis?  Notably this will be important for the land surface (e.g., soil moisture and, potentially, snowpack).

Figure 3 – provide the global and/or regional warming levels for each of these 4 time periods

Line 285-287 – given the large interannual variability of snowpack (i.e., snow depth, SWE and snow covered area) and mismatch/uncertainties across gridded SWE products I would suggest also comparing to the WUS snow reanalysis - https://nsidc.org/data/wus_ucla_sr/versions/1#:~:text=This%20Western%20United%20States%20snow,water%20years%201985%20to%202021.

Line 319 – “…simulated time period does not coincide with observations…” I don’t think this is true.  I think what you’re trying to say is that the CARRM simulations are AMIP-style experiments (i.e., only prescribes observed lower boundary conditions in SST/sea-ice conditions) and should not be expected to exactly recreate the 2015-10-01 to 2020-09-01 period?

Line 321-329 – the last sentence of this paragraph stood out to me.  A common and systemic issue that I’ve seen across models is a cold bias in mountains, especially during winter, this appears to be the case in CARRM too (visual comparison of CARRM to PRISM in the highest elevations of the California Sierra Nevada).  Can the authors state how much colder the CARRM simulations are in the California Sierra Nevada?  I would also make a note that the simulations represent only 5-year averages whereas PRISM represents 30-year averages.  This is especially important given the large interannual variability in California temperature/precipitation (especially compared to other regions around the world) and might obscure “warm” or “cold” biases (and precipitation/snowpack too).

Figure 4 – I would suggest putting “left column”, “right column” … and “top row”, “middle row” …  Also, TREFHTMX, TREFHT and TREFHTMN are not defined explicitly and I would just delete those acronyms from the figure entirely.  I would eliminate redundancies in subpanel plot titles too (e.g., only define E3SMv1: 100 km in one subpanel plot).  The authors use CARRM throughout the text, but in this figure they put SCREAM CA RRM, I would either use SCREAM CA RRM or CARRM throughout.  Last, could you change “degC” to “^oC”

Line 330 – building off my previous comment, this paper nicely highlights the large interannual variability of California precipitation and the (potential) issues with comparing 5-year vs 30-year normals - https://www.mdpi.com/2073-4441/3/2/445 (see Figure 2)

Figure 5 – PRECT in the plot label(s) is not necessary and not defined in the caption either.  Also, PRISM is not an observation, but rather an observation-based gridded product.  Also, why is CARRM not shown?  I would show the CARRM 5-year normal, wettest year, and driest year (for visual comparison).

Line 340 – please be consistent with naming conventions “RRM” should be “CARRM”.  Also, this point about “exceeding the wettest years of PRISM” is important to note, especially since you only simulated 5-years.  If CARRM simulated 30-years, would this bias amplify?  I would argue given the large interannual variability of temperature/precipitation in California that the authors need, at least, 15-20 years of simulation to know if the CARRM’s temperature/precipitation/SWE mean, median, IQR summary statistics are converged.  Notably, PRISM (and other observation-based gridded products, such as Livneh, etc.) has been known to underestimate extreme precipitation (particularly from ARs) as discussed in - https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2023MS003793 and -  https://journals.ametsoc.org/view/journals/bams/100/12/bams-d-19-0001.1.xml .  Given this issue in PRISM (and other products) and the potential to “falsely” attribute an over precipitation bias (especially with only five simulated years), could the authors compare CARRM to this observation-based gridded product for extreme precipitation - https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/LULNUQ (more details in this publication - https://link.springer.com/article/10.1007/s00382-019-04636-0#Ack1 )

Line 350-358 – there is a rich literature in RCMs, GCMs, VR and RRM simulations of California precipitation.  I would compare and contrast more studies here (particularly newer studies) to the CARRM results.  This will further add to the richness of this study.  Also, can the authors produce a plot of total precipitation, convective precipitation, stratiform precipitation, and the % contributed by convective precipitation?  This will provide more precision in highlighting which aspect of the CARRM precipitation simulation is driving the “error” or “mismatch” compared to PRISM (which could inform future CARRM simulations and/or where to target subgrid-scale parameterization development).

Line 355 – do the authors think that simulations at 0(100 m) are truly needed in every hydroclimatic region (i.e., is this estimated convergence point in precipitation the same for the Amazon as it is for the California Sierra Nevada)?  Wouldn’t this 0(100 m) mostly be for regions where precipitation is dominated by convection rather than stratiform precipitation (like in California)?  I think these studies makes the case that, in the midlatitudes, nonhydrostatic simulations at convective resolving scales may have diminishing returns in enhanced skill - https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2021MS002805 - and - https://agupubs.onlinelibrary.wiley.com/doi/10.1002/2016JD025287

 Line 359 – why is snowpack the most “prominent” quantity?  Also, SWE stands for snow water equivalent (i.e., the amount of water that would be produced by the snowpack if it were instantaneously melted).

Figure 6 – what are the units of each of these quantities?  Delete PRECT and SNOWHLND.  Why are 30-year normal and 40-year normal used?  Also, what are the Sierra Nevada only average values for SWE in CARRM vs SWE UA (a statewide average incorporates a lot of 0s into the averaging)?  Or, alternatively, what is the km^3 of statewide SWE storage (this is a more useful metric for water resource managers anyways)?

Line 361-362 – what is meant by “essentially captures”?  How is that quantified?  Also, I would argue that the authors need to utilize, at least, one other SWE reanalysis product given the large uncertainties in spatiotemporally estimating SWE combined with the fact that there is large interannual variability of SWE in the California Sierra Nevada.  I would suggest this product for comparison - https://www.nature.com/articles/s41597-022-01768-7

Line 365 – “four segments” or “four time periods”

Figure 7 – delete TREFHTMX, “degC” to “^oC”, eliminate redundant titles.  Why is JJA only shown?  Add DJF plots too.  Again, inconsistency in using SCREAMv0 CARRM vs RRM vs CARRM, pick one and be consistent throughout.  Can the authors provide difference plots between the five-year periods, put them in the supplemental, and highlight them in the main text?

Line 373-374 – this is an incredible increase in temperature in the Sierra Nevada.  Is this for JJA?  How does that compare to other simulations produced for California?  What about DJF?  In fact, why isn’t DJF (or other seasons) discussed?

Line 377-378 – citation?

Line 380-381 – “which is less dependent on intricate model physics”.  I’m not sure what is meant here because temperature is shaped by land-sea contrasts, lapse rates in complex terrain, cloud spatiotemporal characteristics, land-atmosphere interactions (fog formation), solar radiation seasonality, etc.  I would consider these processes “intricate”.

Line 384-386 – this single agricultural example is a little random (i.e., why are grapes chosen and not other agricultural commodities) and simplified (i.e., wine grape growth surely responds to more than just 10 °C temperatures).  I would add a few other agricultural impact examples here, if kept.  Also, in my opinion, an even more important issue than wine grapes would be the working conditions that agricultural workers/animals will face in a warmer world?  See this recent study for more details/metrics to employ - https://www.nature.com/articles/s41467-023-43121-5

Figure 8 – add the observation-based gridded product(s) here too.  Since the observation-gridded products have 30-40 years, I would also add +/- standard deviation or confidence intervals vertical lines to the histograms (especially to see if the 5-year simulated periods fall within interannual variability).  I’d do this for all histogram/bar chart plots.  Change “TREFHTMX”, “TREFHT”, etc. to their actual names (no need for random acronyms that are not used in the main text or caption).  Please be consistent with model naming conventions.  “degC” to “^oC”

Line 388-391 – why not use lat/lon locations that have an observation station and compare to the CARRM simulations?  Then you could add these true observations to Figure 9 (and others in the study).  This would also provide an additional comparison to the observation-based gridded products which may/may not include some of these stations and highlight uncertainties from the gridding process/statistical co-variate assumptions employed in these products.

Figure 9-10 - Change “TREFHTMX”, “TREFHT”, etc. to their actual names (no need for random acronyms that are not used in the main text or caption).  “degC” to “^oC”.  Please be consistent with model naming conventions.  The 5yr, 3mon, 30d should be put in the caption and the “seasonal” title deleted.  Add observation-based gridded product(s) here too as this will highlight where the 5-year simulations fall within the distribution of a 30-40 year observation-based gridded product.  This will also highlight if the authors should run additional simulation years. 

Line 391-392 – the authors need to define more clearly what is meant by 5x3x30 (this is oddly better defined in the figure than the main text).  Also, delete “approximate”.

Line 395-396 – “highlighting the extreme heat and cold” I think what the authors are trying to say is the “wide range of temperature spatiotemporal variability across California landscapes”?

Line 397-398 – cite Figure 1 when referring to topography underrepresentation in E3SMv1

Line 402-405 – 60 degC would be substantially higher than the historical all-time record reached this past year - https://www.sfgate.com/bayarea/article/what-is-hottest-temperature-in-death-valley-18254957.php - and might be important to highlight to provide context to readers.  Also, if locations with station observations were used, this shocking finding would be, in my opinion, even more impactful to readers.

Figure 11 – see previous figure comments about use of random acronyms, repetitive figure titles, etc.  Also, please provide difference plots, put into supplemental, and point to them in the main text.

Line 419-422 – “observed” or “projected”?  Also, these mm/day numbers need to be better contextualized.  For example, how does 5 mm/day translate to % of annual total precipitation?  Or, km^3 of additional precipitation or … 

Line 423-427 – does E3SM skillfully represent ENSO and its atmosphere-ocean teleconnection response to the coastal western United States?  For example, this study is a good lead for CESM - https://journals.ametsoc.org/view/journals/clim/36/1/JCLI-D-22-0101.1.xml  This might also be why there is weak relationship of the ENSO teleconnection response to warming.  Do the authors think a fully coupled simulation would lead to a different answer compared to an AMIP-style simulation (as used in this study)?

Line 426 – why are “heavy precipitation events” not shown?  This seems odd given the five-year simulations produced by CARRM to evaluate California’s hydroclimate are likely too few to evaluate annual to decadal scale behavior in the model. Yet, these five-year simulations, evaluated at the event scale, would have a larger sample size to compare/contrast statistics between reference datasets and historical vs future climate contexts.  I think the authors should discuss the spatiotemporal characteristics of (for example) the largest 5-10 storms simulated and compare to in-situ, observation-based gridded products and future warming simulations.  This shouldn’t be left to “future work”.

Line 431 – “observations” – again, the authors do not compare CARRM to actual observations (but should!)

Figure 12 – see previous figure comments about use of random acronyms, figure titles, etc.  Also, please provide observation-based estimate histograms with +/- standard deviation/95% confidence intervals to see if the historical experiments are within interannual variability range AND the warming level histograms fall inside/outside historical interannual variability.

Line 440-442 – citation on the importance of organized convective systems for California precipitation?  Also, similar to an earlier comment, please provide a plot showing the total precipitation, convective precipitation, stratiform precipitation, and % contributed by stratiform precipitation across all five-year CARRM experiments.

Line 442-444 – again, it is odd that the authors state that 6-hourly precipitation is important but say it will be covered in future work, especially since hourly or event scale analysis would be more fit for purpose when using 5-year simulations (much more robust sample size of storm events than annual/seasonal averages), particularly in the context of California hydroclimate/hydrometeorology

Line 439-447 – it is striking that mesoscale convective systems get significantly more mention in this entire section than atmospheric rivers (ARs) and extratropical cyclones (ETCs), yet ARs and ETCs are the dominant sources of precipitation for California…

Line 449-455 – I’m highly skeptical that the southwest monsoon/North American monsoon is driving these precipitation change signals in CARRM over California (Figure 13)… the largest changes in precipitation occur over mountains and, therefore, are likely orographic convection.  Also, to definitively make this statement with actual evidence, the authors would need to do feature tracking of MCS/monsoon events…

Line 460 – “compelling indicator”, how so and why more so than other climate variables?

Line 460-463 – add citations to back these statements about “thickest during the spring season”, snow sensitivity to warming and that the Sierra Nevada will be “essentially devoid of snow by end of the century”  There is a rich literature (esp. in recent years) investigating all of these points made.

Line 465 – “6 degC” I would state “a local warming of 6 degC”

Line 473-474 – “SWE threshold of ~0.2 m”?  Do you mean to say snow-to-liquid ratio (i.e., the amount of water produced for a given amount of snow depth)?  If so, this snow-to-liquid ratio can vary substantially from season to season and across mountain regions, especially in maritime versus continental mountain ranges.

Figure 13 and 14 – see earlier figure comments about redundant titles, acronym usage, etc.  Also, please add observation-based gridded products for comparison.

Figure 15 – this figure seems unnecessary.  Why not just add these lat/lon grid cell locations (and those used for Figure 9, etc.) onto Figure 1?  Also, why not use lat/lon locations where in-situ observations exist and add them to compare to CARRM?

Figure 16 – see earlier figure comments related to box and whisker plots, etc.  Please add observation-based gridded products for comparison.

Line 479-486 – “rapid melting in spring” this is not shown.  Can the authors provide a water year xy plot of daily/hourly SWE accumulation/melt cycles that shows this spring snowmelt response?  This is especially important for water resource managers in California and would not show up clearly in the MAM average map plots and/or box and whisker plots.

Line 497 – “is scale aware should” to “is scale aware and should”

Line 512 – “very 500” to “every 500”

Line 515 – “dumped” to “is estimated to have dumped”

Line 511-512 – “Atmospheric River trends over California” by this time in the study, readers have forgotten that the authors have tracked atmospheric rivers using TempestExtremes.  Please remind them here right away.  Also, how many AR events are tracked in the CARRM simulations across each five-year period?

Line 512-533 – these five-year CARRM simulations/experiments are definitely not fit for purpose in evaluating an 1-500 to 1-1000 year event (ARkStorm)… so why is this an major stated goal of the study?  At best, these CARRM simulations would confidently estimate the 1-20 year storm (unless the authors were extremely lucky in sampling internal variability in E3SMv1 and/or SCREAMv0)?  With that said, I would advise the authors completely revise this entire paragraph to be within the flood return frequency context that the authors feel their CARRM simulations are fit for purpose in answering (e.g., 1-20 year event).  Even if the authors don’t discuss the ARkStorm, I think evaluating, for example, the 1-20 year event is still useful and important to the California Department of Water (DWR) as they will likely not overengineer their built infrastructure/management system to be resilient to the 1-500 to 1-1000 year storm ARkStorm like event.  For example, see CA DWR’s Central Valley Flood Protection Plan on the major flood events of the last 160 years for guidance on other potential events to evaluate CARRM skill (some of which are closer to the 1-20 to 1-50 year flood events)  - https://water.ca.gov/-/media/DWR-Website/Web-Pages/Programs/Flood-Management/Flood-Planning-and-Studies/Central-Valley-Flood-Protection-Plan/Files/CVFPP-Updates/2022/Central_Valley_Flood_Protection_Plan_Update_2022_ADOPTED.pdf

Figure 18 – see earlier figure comments related to box and whisker plots, etc.  Please add observation-based gridded products for comparison.  I would also either add more flood event markers in addition to the ARkStorm Ref.  I don’t think the CARRM simulations are fit for purpose in comparing to ARkStorm (would require E3SMv1 LE and/or many more simulated years of SCREAMv0). See CA DWR report mentioned in earlier comment to track down other flood events/reference markers.

Line 535-537 and Figure 19 – can the authors provide AR vs non AR contributed precipitation maps or stacked histograms or … to show readers this point about AR contributed precipitation.  Figure 19 is unnecessarily difficult to read and interpret.  For example, could the authors reproduce Figure 2 from this study - https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020GL089096

Line 538-547 – do the authors only attribute AR-contributed precipitation based on the immediate grid cells where AR conditions coincide with precipitation (i.e., in grid cells where TempestExtremes binary masks are at a value of 1)?  If this is the methodology employed by the authors, this would lead to a considerable underrepresentation of AR-contributed precipitation, particularly at the leading edge/peripheries of the AR masks produced by TempestExtremes.  Why don’t the authors instead use a shapefile/box region over California and associate any precipitation as AR-produced when an TempestExtremes AR mask exists over California?  This would avoid the issue in which an AR’s IVT plume depletes rapidly below 250 kg/m/s as the AR progresses across California’s complex terrain and would thus underrepresent AR-contributed precipitation during times/locations of heavy precipitation.

Line 538-554 – This statement, “In summary, the greatly increased total precipitation in California by end of century is primarily due to greater amounts of precipitation falling from individual storms instead of a greater number of storms, which is dominated by larger precipitable water under the substantial warming scenario”, is, in my opinion, not backed up by evidence in the study and inconsistent with other studies in the AR literature.  If I read the Methods correctly, the authors didn’t actually isolate distinct AR events in TempestExtremes using, for example, StitchBlobs (i.e., estimate when each individual AR lifecycle starts/ends).  If that is correct, how could the authors make a statement about changes in single AR event precipitation totals vs sequential AR event precipitation totals (nor the number of AR events with warming)?  I’m also guessing the conservative approach the authors took in estimating AR-contributed precipitation (only immediate grid cells where TempestExtremes says AR conditions exist) would bias this statement.  For example, the core of the AR events (immediate grid cells where TempestExtremes says AR conditions exist) would cover larger and larger areas of precipitation as integrated water vapor increases with Clausius-Clapeyron.  This would likely erode the AR periphery issue of AR-contributed precipitation (i.e., more and more grid cells over California would be under AR conditions as integrated water vapor increases with Clausius-Clapeyron).  The alternate approach suggested in estimating AR-contributed precipitation (see previous comment) would fix the area over which AR-contributed precipitation is estimated and eliminate issues with changing areal extents of AR masks in warmer climates.

Line 555 – typically this subsection title is labeled “Discussion and Conclusions”  Also, if this is truly a Discussion, the authors needs to compare and contrast the findings made in this study with the broader literature more consistently throughout each paragraph of this section.  Please do so as this allows the reader to know if the author’s findings are consistent/inconsistent with the broader literature.  This will also elevate the novelty/richness of the study.

Line 556 and 578 – “SCREAM” again, inconsistency in naming conventions used throughout the study.

Line 561-562 – “Through the development history …” this sentence does not seem necessary.

Line 571 – “high internal variability” or extreme phasing of climate modes of variability (e.g., ENSO)?

Line 578 – delete “with SCREAM” and put “, respectively,”

Line 581 and 583 – “CA” to “California”

Line 584 – “SCREAM-RRM” again, inconsistency in naming conventions used throughout the study.

Line 585 – “observations” to “observation-based gridded products”

Line 592 – delete parentheses around “(positive)”

Line 597-599 – “This increase primarily stems from greater amounts of precipitation…” this statement is not backed up by the methodology.  See earlier comments about how the authors used TempestExtremes.

Line 607 – Change “meaning” to “context”

Line 609 – These CARRM simulations are not “predictions”.  These CARRM simulations are “projections” based on a possible global socio-economic development pathway (SSP585).

Line 610 – “warm bias”.  See earlier comment about DJF, especially in mountain regions.  I believe CARRM has a “cold bias” in that context.  I think its important to be nuanced when stating “warm bias”.

Line 613-615 – this is why I suggested that the authors provide plots of total precipitation broken down into convective and stratiform and then further by % contributed as it will allow them to be more precise on where/when the CARRM convective parameterization matters.  My hypothesis is that in California, convective precipitation is substantially smaller than stratiform precipitation (even with convective resolving simulations).

Line 616 and 621 – “SCREAM-RRM” again, inconsistency in naming conventions used throughout the study.

Figure A-C – see other map plot comments about redundant titles, delete variable acronyms, etc.  Also, for Figure A-B, it is odd to not include Canadian and Mexican (etc.) state boundaries (Hawaii too?)  Also, why not provide Figure A for all future climate simulations in CARRM too.  Please also provide difference plots.  This will allow readers to visually see the amplifying effects of climate change on IVT through the Clausius-Clapeyron relationship and (potentially) diminishing effects of climate change on IVT through shifts in the storm track/jet that shape AR latitudinal variability.