PAPER SYNOPSIS

Fang and co-authors present an intercomparison study to quantify uncertainty across commonly-used snow water storage products in the midlatitude American Cordillera. They focus on how these multiple products map peak snow water equivalent (SWE) across the western United States (WUS) and the Andes. This includes 10 SWE datasets in WUS and 8 datasets in the Andes, with varying spatial resolutions (high resolution, HR: < 1 km; moderate resolution, MR: 1-10 km; low resolution, LR: > 10 km) and source methodologies (e.g., data assimilation in the reference dataset, station interpolation as in UA SWE, model only as in GLDAS, etc.). As a reference dataset, the authors utilize the recently developed snow reanalysis products from their research group, which is based on data assimilation and remotely sensed snow cover. The analysis first compares SWE volume climatology time series and peak SWE climatology maps in the two study regions, which generally show higher SWE in the HR products and lower SWE in the LR products. Spatial patterns are further assessed by latitude and elevation, with a focus on rain shadow patterns which tend to not be captured by LR products. Interannual patterns in SWE and in the source precipitation, snowfall, and ablation data (all products except the reference product), and the effects of land surface model and spatial resolution are assessed on these driving variables.

OVERALL RECOMMENDATION

This is a novel paper that should be published following revisions. I think this has high potential utility for the community, as it provides a much-needed intercomparison between commonly-used SWE datasets. I offer a few major suggestions for revisions which hopefully may improve the utility of the paper for the community.

MAJOR COMMENTS

1.    The paper’s goal is to characterize uncertainty in snow water storage. This is more feasible when done across all scales (10 products in WUS, 8 products in Andes), however I think there are too few datasets to assess uncertainty within a given spatial resolution (e.g., HR and MR). In other words, having only 1 or 2 models in a given resolution does not make it possible to quantify uncertainty with confidence (i.e., it becomes one model versus another). This is not the fault of the authors per se, as they are generally using what is readily available. While recognizing the significant work that has already been done, I might suggest adding additional HR and MR datasets as feasible. For WUS, two readily available and well-known daily SWE datasets are SWE reconstruction at 500 m (e.g., Bair et al., 2023) and DayMet SWE at 1 km (e.g., Thornton et al., 2022). Including other SWE datasets such as these would help to better characterize uncertainty at finer spatial scales and would represent more distinct approaches that are not currently represented in the study (e.g., SWE reconstruction). I would argue that having a more comprehensive sampling of existing SWE datasets would elevate the utility of the paper to the community.

2.    The SWE product intercomparison focuses on the “snow accumulation season” (L. 135-145), which is defined as the period before peak SWE (L. 139-140). However the accumulation season is not always well-defined in all years, locations, and spatial scales. For instance, snow may be more intermittent in lower elevations, in drier years, and/or at coarser spatial scales. Notably, the timing of peak SWE varies in these cases (as across the products in Figure 2), which suggests that the uncertainty in snow water storage may be larger at other times in the year (e.g., March 1 in WUS). Hence, I am wondering about whether peak SWE is necessarily the most robust way to characterize uncertainty across snow products? In addition to the analyses presented, it could be helpful to characterize the uncertainty in time (e.g., by dowy) rather than just by a fixed point (e.g., peak SWE).

3.    Section 4.1 analyses spatial variations in peak SWE across the study regions and with respect to windward/leeward basins. One aspect that would be useful to analyze and compare across products is the lapse rate in peak SWE across the windward/leeward sides. While peak SWE is lower in the LR products and higher in the HR products (Figures 5-6), it must be remembered that the LR products have less variation in elevation than the HR products. As such, I think this should be normalized in order to assess how the elevation gradients in peak SWE compare across products. This would be potentially important to know for certain applications (e.g., downscaling a LR product to higher resolution).

GENERAL COMMENTS

-  Please make consistent use of the acronym for the low/coarse resolution products. Sometimes it is “CR” and sometimes “LR”. Please select one convention only and use it consistently. 

LINE COMMENTS

- Line 57: It may be worth noting the mountain ranges are also disparate with respect to elevation.

- Line 75: Delete “shows that”.

- Line 112: Add “satellite snow cover” before “observations”.

- Line 131-134: I think this climatological analyses could be of interest, and would request their inclusion in the supplement document.

- Line 171: Delete “choose to”.

- Line 179: I would think that all three resolutions (HR plus MR and LR) may straddle both windward and leeward watersheds rather than just the MR and LR resolutions. I assume you would see this if you zoomed in more in Figures S3a and S4a. Also, replace “CR” with “LR” here?

- Line 305 and 330: I find the titles for sections 4.3.1 and 4.3.2 to be odd. Consider reducing and rewording.

- Line 310: Replace “more” with “higher”.

- Line 326: Replace “is” with “are”.

- Lines 359-364: I would suggest elaborating a little more here on model differences.

- Line 400: Replace “less” with “fewer”.

- Lines 448-452: I think these sentences are not well justified and need to either be removed or better connected to the study. The study does not suggest why future/new spaceborne data are needed to assess SWE in these mountain ranges. This conclusion might have been better motivated if an existing spaceborne sensor that maps SWE (e.g., passive microwave) had been included. Multiple SWE datasets in this paper utilize existing spaceborne snow cover data (e.g. reference and SNODAS) and appear to capture certain spatial patterns like the rain shadow effect.

- Lines 473-492: It appears the ERA5 paragraph (Lines 473-485) needs to be swapped with the ERA5-Land paragraph (Lines 487-492) based on their resolutions (ERA5-Land is a MR product, ERA5 is a LR product).

TABLE AND FIGURE COMMENTS

- Figure 1: Suggest labeling the Cascades in the WUS map since the text references them in multiple places. 

- Figure 7b-c: There appears to be an interesting outlier year where UA and ERA5-Land have much lower peak SWE than WUS-SR. This appears to be a high snow accumulation year. Can you please identify which year this is in the text and provide a brief discussion point about it? These products have greater correspondence to WUS-SR in most other years, so this year may be negatively skewing the error statistics.

- Figure 7 caption (line 296): Replace “is” with “are”.

- Figures 7j and -8a: It seems for the heat maps, a calculation of the spearman rank correlation would be useful to assess the agreement in dry to wet years for each product vs. the reference.

- Figure 10: It would be helpful to include a dashed line for the t_peak (DOWY) of the reference data.

REFERENCES

Bair, E. H., Dozier, J., Rittger, K., Stillinger, T., Kleiber, W., and Davis, R. E.: How do tradeoffs in satellite spatial and temporal resolution impact snow water equivalent reconstruction?, The Cryosphere, 17, 2629–2643, https://doi.org/10.5194/tc-17-2629-2023, 2023.

Bair, Edward (2023), SPIReS-MODIS-ParBal snow water equivalent reconstruction: Western USA, water years 2001–2021, Dryad, Dataset, https://doi.org/10.25349/D9TK7H.

Thornton, M.M., R. Shrestha, Y. Wei, P.E. Thornton, S-C. Kao, and B.E. Wilson. 2022. Daymet: Daily Surface Weather Data on a 1-km Grid for North America, Version 4 R1. ORNL DAAC, Oak Ridge, Tennessee, USA. https://doi.org/10.3334/ORNLDAAC/2129

The authors present an inter-model comparison of snow water storage over the western US and the Andes. A newly developed and well-verified (high resolution) snow reanalysis product is used as the baseline reference for the comparison to global and regional snow models at high, moderate, and low resolutions. The authors report that models with spatial resolutions <10 km are in much better agreement with the baseline product, with large underestimates by coarse global products. They show that uncertainty is much higher in the Andes than in the western US, largely due to greater precipitation uncertainty in the Andes.

The authors show that models at sufficient resolution that assimilate in-situ precipitation and snow observations tend to improve snow representation compared to models that do not include such observations. High resolution products (and presumably those constrained by measurements) better resolve orographic spatial patterns in snow volume caused by the prevailing trajectory of storm-tracks and rainshadow effects on leeward slopes.

The paper is well-written and clearly structured. The references cited are complete and up to date. The results are relevant to local, regional, and global snow water resources that are under duress and changing rapidly due to climate change.

I have two general comments that are aimed to improve the clarity of results and their implications.

1.  It is difficult to appreciate the water storage units of cubic kilometers and to put the climatological peak and uncertainty metrics in the context of water resources. It seems that all reservoirs in the contiguous US hold 600 km3 of water (Steyaert et al., 2022). This suggests that the climatological average snow water storage in the western US is 269/600 or 45% of all reservoir storage in the contiguous US (much of which is in the western US). While this is a compelling number, the more compelling result, in my opinion, would be expressing the uncertainty of global models relative to this US reservoir storage estimate. My quick assessment (check this) is that the low-resolution products underestimate snow volume by nearly 24% of all the water held in these US reservoirs. That astounding fact is likely not appreciated by most users of those (commonly used) data.

Steyaert, J.C., Condon, L.E., WD Turner, S. and Voisin, N., 2022. ResOpsUS, a dataset of historical reservoir operations in the contiguous United States. Scientific Data, 9(1), p.34.

2.  Please discuss the implications of snow model uncertainty in coarse scale model (> 10 km) applications on the topic of snow volume sensitivity to warming. For example, Siirla-Woodburn et al. (2021) sited in this paper uses coarse-scale model output and concludes a dire water resource scenario for mid-century. Might results of such studies be different and arguably more accurate if models were run at finer spatial resolution?

Siirila-Woodburn, E. R., Rhoades, A. M., Hatchett, B. J., Huning, L. S., Szinai, J., Tague, C., Nico, P. S., Feldman, D. R., Jones, A. D., Collins, W. D., and Kaatz, L.: A low-to-no snow future and its impacts on water resources in the western United States, Nat Rev Earth Environ, 2, 800– 819, https://doi.org/10.1038/s43017-021-00219-y, 2021.

Detailed Edits:

Line 200: To make the comparison clear, perhaps add “in the Andes than they do in the WUS”.