This manuscript reported a new compilation of a variety of surface mass balance observations (stake, ice core/snow pit, GPR, ultrasonic sounders), and made it freely available. The compiled data were checked through vigorous quantitative criteria, and thus the relatively reliable data are left. The authors described the features of this new data set, and compare it with previously published observed SMB datasets. They also made a comparison with ERA5 outputs to test the representativeness of the data set for estimating climatic models.

This compilation is a big and complex task, and the resulting data set represent a huge data synthesis effort, which will facilitate a lot of new studies, especially for Antarctic mass balance studies and validation of climate models, and will be well cited. The data set is also interesting for glaciologist, climatologists, geographer, and so on. In my opinion, this manuscript is well written, and deserves to be published. Before publication, only following minor comments needs to be addressed by the authors.

1. This compilation needs many years of field work by the researchers from different nations, who provide the original data, i.e., the basis of the dataset. Therefore, their scientific contributions are indispensable for this study. However, future work probably simply cite the AntSMB dataset instead of the original studies. So provide the original citation information is important.

2. The authors mentioned a lot of Antarctic locations in the text, and some might not easily followed by the readers. So a map with the mentioned locations should be included. Alternatively, this figure is provided in the supplementary material.

3. In section 4, I know that the inter-comparison between different types of SMB measurements in the same locations is helpful. But before making the inter-comparison, please clarify how or based on what did you select these locations? And where are they on Antarctica?

Some other minor issues

Line 29: please update the reference

Line 35: add the reference

Line 44: add “the” between “ improve” and “ ice”

Line 89 : “The remainder” is English formulation?

Line 146: “100 kg m-2 yr-1” should be “ 100 kg m^-2 yr^-1”, and please use the same “unit” throughout the text. For example, in Figure 7, “kg m^-2 a^-1” was used, but use “kg m^-2 yr^-1” in the text.

Line135-149: There is no any reference in the review on SMB measurements from snow pits/ice cores. Please add the related references.

Line183: Change “which are composed of ” to “i.e.,”

Line 206:This sentence is confusing, and please rephrase it.

Line 310-311: Not necessarily, since it is due to the assimilation (or lack thereof) of satellite data, and these are mostly lacking before 1979, and thus please delete the sentence.

Line 320-375: please use italic “r” and “p”

Line 375: Change “r2” to “r²”

The authors made a remarkable work to update the current state-of-the-art Antarctic SMB observational dataset, including SMB observations at different temporal resolution and covering different time period, with quality control. This dataset is unique and crucial for evaluating atmospheric models, in the absence of reliable precipitation observations in Antarctica. It is all the more important than SMB is driving the interannual variability of Antarctic ice sheet mass change and that it's evolution under climate change is still highly uncertain, with potential significant contribution to sea level change during the 21st century. In this context, observational reference needed to improve the representation of SMB in atmospheric models is of much value for the whole earth system community.

Consequently, I strongly recommend this article to be published in ESSD.

I have a few suggestions to improve the article further.

Major suggestions

My main concern is about the number of observations given throughout the text. I think GPR numbers are not meaningfull because GPR is continuous and the number of points is arbitrary. However it covers large areas. Can you give more meaningfull numbers, such as the length or area covered? A solution would be to bin the GPR data points on a 1 km x 1 km grid and count the number of km2 covered by the data.

I thing you should revise the numbers for the following lines:

• p1 L14: "268913 individual multi-year averaged observations"
• p3 L76-78: "175373 individual GPR measurements"
• p7 L193-194: "including 265334 unique measurements by radar isochrones,"
• p11 L318-319 "Finally, 191816 multi-year averaged observations are left for model-observation comparison."

• p7 L194-198: "The majority of these observations (~ 69%) are derived from airborne snow radar measurements in the coastal zone of West Antarctica and Antarctic Peninsula, Ronne Ice Shelf, South Pole, Pine Island and Thwaites glaciers (Medley et al., 2013; Medley et al., 2014; Dattler et al., 2019). GPR data from two transects in East Antarctica account for 30% of all measurements in this subdatset." + Figure 2

I think it's not fair to compare stakes numbers with GPR numbers. It makes more sens to compare the areas covered, e.g. using a 1km x 1km grid or a 10km x 10 km grid (number of pixels covered)

• p11 L342-344 "A comparison of the density distribution of ERA5 precipitation minus evaporation (P-E) with the filtered multi-year averaged SMB observations reveals that the multi-year averaged dataset is representative of the entire P-E spectrum of the model at the continental scale (Fig.6a)."

Are the histograms for observed multi-year SMB including individual GPR points and other kind of data? As stated above, I think it's not fair to compare the number of GPR points with other kind of data, so the histogram might be biases with GPR datapoint numbers.

I suggest to (1) bin GPR data (2) differentiate GPR data from other data in the histograms.

Looking at Fig.1c), one can also see that low-elevation datapoints are from GPR and are concentrated on WAIS. I suggest you separate Fig 6 into WAIS and EAIS to discuss the difference of coverage between the 2 basins, and change the conclusion accordingly.

p11, Section 6.2.2 Annual resolved SMB observations

→ Can you separate ice cores and other data to assess if there is a real problem specifically with SMB interannual variability in ice cores? It is of importance as annual SMB observation from ice cores have been used to create the widely-used SMB reconstruction from Medley and Thomas (2019).

→ Can you also check for interannual variability (std of annual values) : ice cores vs. other observations vs. ERA5?

Section 6.2 A subset of data used for the comparison with ERA5

→ You should add a comparison with AWS, at least at some key sites, to assess the usefullness of this dataset.

→ You could add the comparison with ERA-Interim too, to assess the difference of performance compared to ERA5.

Minor comment and typos

Abstract

• p1 L15: 78968 records at daily resolution from 32 sites across the whole ice sheet

A number of years covered would be more meaningful.

1. Introduction

• p1 L28-29: (Wouters et al., 2013) (Church et al., 2001)

Can you update these references with more recent publications?

2 Description of the AntSMB dataset

2.1 Data collections and sources

• p3 L83: annual

→ annually (everywhere)

• p3 L85: high-resolution ultrasonic sounder observations

Which density was used to convert to accumulation?

2.2 Selection criteria

• p4 L97: public

→ publicly

• p4 L110: "on both stable isotopes and chemical markers, and natural radionuclide"

Can you rephrase? You mean datation based on both stable isotopes and chemical marker?

• p4 L111-113 : ". The available GPR-based snow accumulation rate data are included, because their uncertainties are <5% at a firn depth of 10 m, and decrease with the increase of the depth (Spikes et al., 2004; Eisen et al., 2008)."

Does it depend of the post-processing quality? Interpretation of reflectors, proper calibration with ice cores, correct density estimates?

• p4 L113-116 "SMB records of annual resolved ice cores should be either cross-dated or layer-counted. Their chronology should include at least two age control points, with one near the youngest part and another near the oldest part of the time series. Also, they must be confirmed by the data generator. Furthermore, ice core SMB records are corrected for the impact of firn density and the vertical strain rate profile"

Add reference.

• p4 L116-118: "The preliminary quality control for AWS snow accumulation data has been performed by data owner by means of removing the null measurements and physically anomalous snow accumulation data (i.e., data outside of the initial and final accumulation values)."

Add references.

• p4 : Third paragraph

This paragraph is difficult to read. It misses a small sentence stating that it is related to quality check. In addition the transition is abrupt between the different kind of data, maybe you could use bullet points? It is all the more true that the different kind of data are described afterwards in Section 1.3

2.3 Types of data measurements collected in the AntSMB dataset

• p5 L127-128 : "spatial representative of a single stake records is very limited due to small-scale disturbance from post-depositional effects such as the interactions between the stake and local wind."

spatial representative → spatial representativity

Not only this but also because of large natural spatial variability .

• p5 L146 "with low accumulation of <100 kg m-2 yr-1,"

→ with accumulation lower than 100 kg m-2 yr-1 (everywhere)

• p5 L135-148: 2.3.2 Snow pits/ice cores

Add references.

• p6 L166-167: "The resulting uncertainties were estimated to be about 4% of the calculated SMB at a firn depth of 10 m, and about 0.5% at the depth of 60 m (Spikes et al., 2004)."

I am surprised by such good results. How much does it depend of the calibration?

2.4 Structure and metadata

• p6 L183-185: "The AntSMB dataset includes three subsets which are composed of multi-year averaged SMB observations from stakes, ice cores and GPR measurements, annual resolved SMB measurements by ice cores, stakes and stake networks, and AWS daily snow height measurements."

Suggestion for readability: "The AntSMB dataset includes three subsets which are composed of (1) multi-year averaged SMB observations from stakes, ice cores and GPR measurements, (2) annually resolved SMB measurements by ice cores, stakes and stake networks, and (3) AWS daily snow height measurements."

• p7 L190-191 "have been detailly discussed"

→ have been discussed in detail

• p7 L201 "Annual resolved SMB"

→ Annually resolved SMB

• p7 L209-214: Paragraph on AWS: add the total number of AWS and the number by region. The last sentence of the paragraph is a repetition of "section 2.3.4 AWS"

3. Spatial and temporal analysis of the AntSMB dataset

3.1 Spatial coverage of SMB records

• p7 L218-220 "AWS snow accumulation measurements were obtained at 32 sites, of which ten are located at Dronning Maud Land, seven at the Ross Ice Shelf, and four along Chinese transverse route from Zhongshan Station to Dome A"

Move the numbers to previous section (consistently with other datasets)

• p8 L223-224 "with the accumulation of < 70 kg m-2 yr-1 (Frezzotti et al., 2007; Frezzotti et al., 2013)."

Here 70, above it was 100?

• p8 L229-231 "ranging from a minimum of 3 years to amaximum of 1000 years. The covered time periods are closely associated with the measurement method. AWS provides very high-resolution measurements of snow height changes, but the records generally span only a few years (1-18 years)."

it's contradictory : minimum of 3 years vs. (1-18 years)

• p8 L237 "resulting records in our dataset ranges from decadal to centennial"

Can't GPR have annual resolution as in Medley et al. (2014)?

• p8 L239 "For annual resolved SMB subdataset"

→ annually (everywhere)

• p8 L239-240 "of 183 time series from ice core and stake network measurements, 47 span the last 200 years (Fig. 3a)."

I don't know how to read this information from Fig. 3a ? It that the number of measurements in 1800? It seems there is a jump just in 1800?

• Figure 3 : add informations on the panels for readability : annually-resolved SMB, multi-year SMB, etc.
• p8 L246 "cover <20 years"

→ cover less than 20 years (everywhere)

• p9 L256-258 "Given that no existing observed SMB dataset can be used as an independent reference to the different types of Antarctic SMB observations, the inter-comparison of SMB determined by different methods at the same or near locations are made, as presented in Figure 4."

Did you compare the data for the overlapping time periods or by comparing multi-year averages over different periods?

• p9 L260-261 "In addition, no systematic errors between the different methods are found."

Aren't AWS giving generally lower SMB than ice cores/snow pits ?

• p9 L280 "continuous stake measurement"

What is a "continuous stake measurement" ?

• p10 L299-300 "(ERA-Interim) is likely to be the best or among the best reanalysis dataset for the representation of Antarctic precipitation (e.g., Bromwich et al., 2011; Wang et al., 2016)."

I am surprise as ERA-Interim is known to be too dry inland over the East Antarctic plateau?

• p10 L303 "(~ 31km and 137 pressure levels, respectively)"

Give the same information for ERA-Interim

• p10 L306-307 "Here, our main objective is to determine if the AntSMB dataset is also capable of representing main features of SMB in space and time, compared to ERA5."

It must be the reverse : want to know if ERA5 is able to provide a good SMB compared to the AntSMB observational dataset?

• p11 L325-327 "This can be confirmed by that ERA5 simulated individual records highly correlate with each other (r>0.70), but exhibit a variety of relationships with their corresponding ice core SMB time series at 35 cores on the DML plateau, including significantly negative, positive and insignificant correlations (Fig. 5a)."

I don't understand this sentence, could you re-write? On the DML plateau,

• p11 L328 "over the Berkner Island and Ronne Ice Shelf"

Place all locations on a map, at least in Fig. 1

• p11 L334-336 "Despite only one core at the top of four ice domes where the local noises are minor (Monaghan et al., 2006a), the records are not discarded in the estimate."

I don't understand this sentence, can you clarify?

• p11 L338-339 "If the records from a single ice core are confirmed to be less local noisy by data owners, we also don’t omit them."

What do you mean by "less local noisy by data owners" ?

• p12 L351-352 "This suggests a good representation of the major spatial patterns as presented by observations, such as coast-to-plateau SMB gradients."

The correlation can be equal to one and the slope equal to 0.5; correlation cannot be used to assess the good representation of SMB, all the more than the general trend of observations and model is driven by the topography.

• Figure 7 (a) Avoid using rainbow colormap: https://betterfigures.org/2015/07/10/a-welcome-development-for-matplotlib/

   

• p12 L352-352 "No systematic spatial bias is observed on the West AIS"

I can see a wet bias on the coast and dry bias inland?

And a dry bias over Ross ice shelf too?

• p12 L370-371 "we use the continuous time series of stake measurements along the JARE traverse route from Syowa station to Dome F."

Aren't these data included in the annually resolved SMB dataset? Why not include them directly in the annual dataset shown in Fig. 5a? Is there other annually-resolved stake lines in your AntSMB dataset?

• p12-13 L380-381 "These may result from the limited performance of ERA5 for the storm frequency related to synoptic-scale circulations, and sublimation because of circulation variations."

Maybe also because of unresolved processes in ERA5 such as drifting snow?

• p13 L394-396 "A temporal homogeneous climatology of SMB for the second half of the 20th century may be obtained by temporal rescaling of the multi-year averaged SMB subdataset against ERA5 outputs."

Can you clarify? Are you thinking about the kind of temporal rescaling as done by Medley and Thomas (2019) and Wang et al. (2019), as said latter in the paragraph?

• p14 L415 "The scientific community are"

→ The scientific community is