General

As the authors illustrate through the literature that they cite, estimating sublimation is very important for the annual water balance, and few studies have examined multiple (>3) years, which they do.

Overall this is a good paper, but there are a number of steps that are not well explained and thus the methods are unclear. All the details are listed below; here are several examples, 1) 30% of the EC data are gap-filled. How? This is apparently in one, or all, of the Euskirchen papers; 2) section 3.3 presents “standard” statistical methods to evaluate the relationship between sublimation rates and meteorological and environmental variables. To what end is this done? Is this the same as what is stated at the beginning of section 4.2?

We often do not have enough data, i.e., no EC measurements, to adequately estimate sublimation rates. The authors correlate EC-estimated sublimation with hourly and daily meteorological data (Air Temperature * VPD * Net Radiation * Temperature Gradient * Wind Speed). In Table 6, it looks as if these are multiplied together. The daily correlation is quite high (mean R^2 of 0.81 for the Lowland Boreal). The authors could consider doing a split sample analysis, i.e., leaving out 2 years to create the model, and then evaluating the model on the years left out. Further, are all five variables used in the MLR necessary? There can at least be a discussion of this.

Section 5.1 presents a discussion of uncertainty. The focus is underestimation due to blowing snow sublimation and “data processing.” The former is informative. The latter is attributed to an overestimation of sublimation that is actually melt-evaporation. This is interesting. However, under the umbrella of “data processing,” the authors should at least mention measurement errors and uncertainty (e.g., Hultstrand and Fassnacht, 2018; https://doi.org/10.1007/s11707-018-0721-0).

The MLR (section 4.4) and section 5.2 present and discusses sublimation rates as a function of meteorologic data. Some readers will not see the point of this analysis. I think that it is useful, as we often do not have EC data. However, we often have meteorological data from a regular weather station. The authors should consider using the bulk flux method as a comparison (as was done in various of the papered cited). At least provide a more thorough discussion of why the MLR is relevant here (and elsewhere). Also, consider which variables in the MLR is not readily available at a regular weather station, i.e., Temperature Gradient and perhaps Net Radiation. How would the MLR model degrade if only available at a regular weather station were used? This should computed (or at minimum discussed), since we typically don’t have “Temperature Gradient,” as defined in this paper.

Specific

• Line 22: “phase change from ice crystals in the snowpack” – technically they are snow grains and not ice crystals (see Fierz et al., 2009; IACS Guide, etc.)
• Line 30: “errors associated with solid precipitation measurements in the Arctic (Goodison et al., 1998)” – while this is a good reference. The authors should also consider the numerous recent papers (last decade) related to WMO-SPICE
• In the Introduction, consider the paper by Herrero and Polo, 2016 The Cryosphere) as they also compare various sublimation estimation methods.
• Line 33: “eddy covariance (EC) method … continuously measure latent heat fluxes” – this is mostly true. Sexstone et al. (2016) illustrated some of the uncertainty with EC estimates of the latent heat fluxes.
• In Figure 1, if any of the photos correspond to the Ameriflux sites (Figure 2) add those labels at least to the captions, but preferably in the figures with “b),” etc.
• Figures 1 and 2: perhaps combine these two figures as they both relate to the study sites
• Lines 81-82: the “mean annual precipitation (MAP)” was estimated as “140–270 mm, with 60% of that occurring as snow” by Euskirchen et al. (2017). Without reading that paper, I am curious if/how the precipitation was adjusted for undercatch. This is relevant and should perhaps be stated here. The estimate of precipitation impacts the computation of the % of sublimation to precipitation
• Line 82: “air temperatures are below freezing” – use the term “colder than freezing” instead of “below freezing,” as below has an elevational context
• Line 85, Line 122, and Figure 1f: How big are the “low stature” plants? I don’t expect them to have any noticeable amount of canopy interception, but due to the thin snow cover (maximum of 50 cm), it is likely that the vegetation is exposed for at least part of the winter. How does this impact the aerodynamic characteristics across the snow surface?
• Line 88: instead of “can top 20 m s-1,” use “can exceed 20 m s-1”
• Lines 92-94: consider adding citations to these three sentences
• Section 2.1: This is not crucial, but a monthly summary of mean temperature, total precipitation, mean and maximum wind speeds for the two sites (tundra and boreal) would be informative so understand the climate of the study sites
• Line 130 and prior: “gap-filled data (30%)” – this is a lot. The methodology used to fill in the gaps is apparently presented in one of the four papers by Euskirchen et al. (2012, 2014, 2017, 2020). However, since 30% are gap-filled, the method used needs to be presented in the paper, at least in an Appendix or Supplementary Information.
• Line 145: are station pressure data not required?
• Lines 161-162: why are “[s]tandard statistical methods … applied to evaluate the relationship between sublimation rates and meteorological and environmental variables?” To what end. Explain what was specifically done. I assume that this is to compute the % of total or winter precipitation lost to sublimation? If not, then why is this done?
• Line 165: No one cares that Arc was used to create the maps, since spatial data are not used in the analysis
• Line 170-171: what is meant by “mean rates are 5–7% of the maximum daily rate?”
• Figure 4: needs the same legend that is in Figure 3, unless you combine those figures (not necessary). Consider adding 2010 to Figure 4b, even with no results, so that the reader can visually line up the year across the two study domains. Consider putting the gridlines between the years, instead of the mid-point, so we know which bar belong to which year when there are sites missing for a specific year (e.g., Figure 4a 2014, 2015, 2016 and Figure 4b 2014 and 2016. As stated above, it would be useful to show some annual or winter summary data, at least peak snow depth, peak SWE and winter precipitation totals
• Lines 194-195: this sentence is part of the Methods and should be moved there
• Lines 204-205 and the next sentence: how were the “winter solid precipitation increase[s] at the lowland boreal sites (p value = 0.02 and r2 = 0.39)” computed? Explain the method used
• Table 2: the standard deviation (SD) seems large compared to the mean sublimation and % of solid precipitation sublimated. Are these distributions skewed, i.e., are the SD values misleading?
• Table 3: consider adding a sentence that the condensation is minimal compared to ET, but that deposition (downward sublimation flux) is not minimal compared to sublimation (away).
• Figure 5: consider adding what the other components of the box and whisker chart are, beyond the mean (red dot)
• Table 4: From the text, these are correlations to daily sublimation. Add this to the caption.
• Table 4 “temperature gradient”: needs to be explained better, as this is the gradient from the sensor through the air and the snowpack to the soil interface.

General Comments: 

The manuscript presents an interesting study on snow sublimation quantification in tundra and boreal forest sites in Northern Alaska. Snow sublimation was calculated using latent heat data from eddy covariance measurements over ~12 years at six sites differing by snow classes, vegetation communities, and permafrost. As stated in the manuscript, few studies have quantified sublimation in such environments by direct measurements. However, the knowledge gaps this study aims to address compared to previous studies, which are only briefly mentioned, need to be better highlighted in the Introduction. 

Next to quantifying the magnitude of snow sublimation, the study aimed to assess the sublimation spatial and temporal variability. This assessment can benefit significantly from the availability of the time series length at different sites. However, this variability is not enough described and discussed, for example among sites within the same snow class as shown by the mean monthly sublimation analysis. The relation between sublimation rates and snow cover length shows that the strength of the connection between the two variables can be quite different among sites for the same snow class, too. Differences in the interannual variability of cumulative annual sublimation between snow classes are also not described and further explored. Additional investigations and links to the effect of meteorological variables and, or site characteristics will be beneficial to improve this assessment. 

Similar considerations apply to the analysis between sublimation rates and meteorological drivers. No figures with the spatial and temporal variability of the meteorological data are shown, nor figures with correlations between variables. These figures need to be included to support the study results and discussion. Some variables, e.g. vapour pressure deficit and temperature gradient, need to be introduced, and the information they provide regarding processes at the different sites needs to be explained. The reasons why some correlations are more robust than others depending on the site and why correlation strength increases with the time scale should be adequately discussed. 

The discussion of the study’s results compared to other studies is valuable. However, more clarity is necessary on how this study stands compared to the other studies, for example, in the vegetation and meteorological controls paragraphs.  

Consider addressing all the points mentioned above and in the following detailed comments to enhance the presented work's clarity, readability, and strength.

Detailed Comments:

Line 12: What does it mean "with differing permafrost conditions"? Please clarify.

Line 17-18: What is the key finding of this analysis? Please add. 

Line 25: Only one reference?

Line 26: Please elaborate more on the challenges of measuring snow sublimation.

Line 30: Please elaborate more on the error associated with precipitation measurements.

Line 34-35: Please add references.

Line 54: This could be moved in the study area section.

Line 69: Please mention what is Ameriflux.

Line 116: The function of this section here needs to be clarified. Types of sublimation can be mentioned in the Introduction, and details on the sublimation estimated by EC in section 3.1. 

Line 127: If the instrument measures latent heat fluxes at 2.5-3m, is it below or above the vegetation canopy? In line 121, it is stated that the canopy sublimation term does not apply for five out of six EC sites; what about the site with tall vegetation? What is the effect of canopy sublimation there? Please clarify. 

Line 129-130: This sentence needs to be clarified. What are filtered latent heat measurements and gap-filled data? How are gaps filled? It is not enough to mention that information is described in the references as this is the primary measuring method used. Please provide essential information. 

Line 163: This variable has yet to be introduced; please specify what it is and which measure it provides. 

Line 164-165: Please add references, explain the purpose of using a post hoc Tukey test and add the significance level.

Line 177: Please clarify what is meant for “an average winter”.

Line 180-181: What do these sublimation rates represent? The mean of the three sites? Please clarify.

Line 181-182: See previous comment.

Line 189-190: The relative change among a region's sites is quite variable over the years. Additional analyses are needed to support this statement. 

Line 220: Please specify whether significant differences refer to the mean of the annual sublimation rates at the different sites. 

Line 231: Please state the two variables of the correlation. Please specify the type of correlation coefficient. 

Line 251: Are the same explanatory variables providing the highest r² for the lowland boreal forest and tundra sites? It would be informative to show how the r² increases by including stepwise the explanatory variables. 

Line 235: This sentence needs to be clarified. Please refer to where information about “higher wind speed” and “lower relative humidity” can be found in the manuscript.

Line 272: Please elaborate more about the findings from the literature and this study.

Line 285-286: This sentence needs to be clarified. How is it possible to deduce this statement from Tables 4 and 5? 

Line 301: Which lower latitudes? Please clarify. This terminology is used in line 312, too, making the context unclear.

Line 309-310: This sentence needs to be clarified. 

Line 319-320: Is it not possible to advance any explanation about these results?

Figure 4: Legend is missing; please add it.

Figure 5: Please add some explanation in the caption about significant differences.

Figure 6: In the manuscript, the range reported for r² is 0.4-0.8. Please check consistency. Please add which r² belongs to which relation and define what r² is in the caption.

Table 5: Please define r² in the caption.