Mean sea surface(MSS) has important applications in geodesy, geophysics, and oceanography. The manuscript constructed a new global MSS model  SDUST2020 with the resolution of 1'x1' from multi-satellite altimetry data, and evaluated its accuracy using several methods. Comparing with previous MSS models, several new altimetry missions and loner time-span data were included for  modeling SDUST2020. Generally, it is a good MS and provide a valuable dataset. The results are of scientific sense. I recommend a moderate revision and English expression need to polish. Please find detailed comments on the current MS below.

1. Whether the altimeter data were retracked? If so, what retracking method was used? And how coastal altimeter data were treated in this study?
2. What's the meaning of f(t) in equation (4) ?  It is suggested not to use the  same character for different quantities in equation (3)-(5). 
3. According to the comparison and validation, SDSUT2020 have better accuracy than CLS15 and DTU18.  Except the accuracy, is there any obvious improvement to reveal details features of MSS? It is suggested to compare these MSS models in some typical sea regions.
4. In section 3.1, T/P series data between 66°S and 66°N were used to calculate ocean variability correction for ERS/GM,HY-2A/GM, SARAL and Cryosat-2 which latitude ranges beyond 66°. It need to extrapolate. How does the polynomial fitting interpolation(PFI) perform to do the extrapolation?
5. In section 3.2, for crossover adjustment, did the author set threshold of time difference of two tracks?
6. In Figure 6-8, there are large differences in polar regions between MSS models. What’s the reason?

Technical corrections:

1. Line 22-23: ‘sea level contains a variety of variation information about time scale.’ should be rephrased.
2. Line 36-37: ‘are published’ ---> were published
3. Line 214: ‘decimeter magnitude to centimeter magnitude RMS’
4. Line 232: delete ‘since’
5. Line 242: 106m should be -106m.
6. Line 378: ‘that’ ---> those

The manuscript presents the construction of a new global MSS model  SDUST2020 with the resolution of 1'x1' from multi-satellite altimetry data, and evaluated its accuracy using several methods. Some of the novel features of the new MSS are longer timeseries and the use of J-3+S3A+HY-2 data. 

Upon reading the manuscript I felt that there are serious uncertainties in the manuscript related to the method used to derive the new MSS but also to the evaluation,  which needs to be addressed before it can be considered for publication. In many instances the authors Unfortunately, the manuscript suffers from very many sentences that are very difficult to understand which has made the review very difficult.  I decided to recommend that the manuscript is rejected due to the following major issues: 

Altimeter processing description: how were the altimeter data processed and potentially retracked. Were 1 or 20 hz data used for the derivation?. Which range and geophysical corrections were used?. Where the state-of-the-art tide model FES2014b is used consistently. 

Averaging technique. Compared to the CLS15 and DTU18 which has a well-known averaging date of 01.01.2003, the averaging technique outlined from line 110 onwards is problematid. Here  the 9 MSS models shifted slightly in time are averaged using the weighted average according to the reciprocal square of the estimated  SSH error. In this process, 9 grids that are heavily correlated are averaged. Firstly,  the correlations should be taken into account in forming the average. Otherwise, it will not be possible to determine an averaging time for the combined MSS.  Without such the MSS is very hard to use. Secondly, there is no indication of what the estimated SSH error is and how it was derived. 

The crossover adjustment is also severely problematic.  When the author writes that the crossover adjustment is carried out in two steps using 1) a condition adjustment method and 2) filtering and prediction along the track. I do think that the second step is a crossover adjustment and that the sentence is largely garbled. Also, the central quantity f is not defined in eq 4.

In the crossover adjustment, the authors select regions of 20 x 20 degrees.  Dependent on the latitude all wavelength longer than the region will be absorbed (the a0 term in Eq 4). At high latitude, this can be wavelength down below 1000 km and even lower dependeíng on the number of parameters used in the adjustment. Please explain how the signal longer than say 1000km is perserved in the solution. Especially for the MSS at high latitudes. My gut feeling is that the adjustment must have been made in a remove restore fashion with CLS15MSS and that the SDSU consequently becomes a correction to this MSS. This would also explain the pattern of differences at high latitudes.  Figure 5 onwards which shows that CLS15MSS and SDFU20 have the same voids at high latitude. If this is the case CLS15MSS should have been acknowledged. 

interesting enough I noticed  large discrepancies with other MSS models above 80N. This could fit with the fact that 80N is the northern limit of the 20x20 degree boxes, so data in the few degrees to the north of 80N are not adjusted?

The author process to perform something they call a self-crossover adjustment. I have never heard this word before. If what the authors perform is a  moon mission crossover adjustment, this is problematic by several standards. First, the adjustment should have been performed as a multi-mission adjustment with the reference tracks of Jasons. Secondly, what is the usage, and interpretation of this adjustment except for the obvious that the numbers reduces. In principle, this has nothing to do with the MSS derivation unless derived errors are used for the following step. Why does the authors not perform a multi-missioin adjustment with the reference tracks. 

Section 5.1 present the comparison with CLS15 and DTU18 models. Here the authors present the central table 5 which is used to infer the accuracy of the models from high to low. In my view, it only explains that the authors are doing something wrong in my view. First of all the DTU15 and CLS18 MSS are not different on average by 1.27 cm Many investigations (e.g., Pujol et al. 2019) show much smaller numbers. The differences in Table 5 between the model's present standard deviation of >29 centimeters are clearly not what other authors present. 

Also, the following figure 6 demonstrates that the standard deviation is far less than 29 cm. I think that the authors have potentially forgotten to apply the confidence mask in the Arctic Ocean and elsewhere. 

Consequently, the conclusions drawn in line 275 -280 are not correct because the numbers can be from a specific region (or even from land?). 

A little later the authors also present the average and RMS about the formal error (again garbled sentence) of 1 and 1.5 cm for SDSUT. What does this mean and how does it relates to Table 5. 

The comparison with tide gauges is questionable. First of all. Have the author included the formal error on the MSS in this comparison and does the MSS fit within this?. Secondly, have the authors ensured that the same version of the reference ellipsoid  (TOPEX vs WGS84/GRS80) has been used and that the version is employing the tide system?.  In this section the authors only present numbers but no interpretation of the results. Is it realistic that the differences range up to nearly a meter  (with a formal error of 1 cm claimed for the SDSUT. 

Finally, we are in 2022. CLS and DTU have both released 2021 versions of their models.

Please note, that throughout I do not disagree with the fact that the SDSUT might compare favorably in the various comparison. This is in my view somewhat expected as  longer time series are used in its derivation

The manuscript describes an approach of constructing the global MSS model, with witch the authors have added a new product (SDUST2020) to the market. Also, they have evaluated its quality by comparing it with two popular models (DTU and CLS). I personally support it’s publication in ESSD after properly addressing the questions raised by previous reviewers.

I do not have any additional major concerns after reading the comments from the other two reviewers. However, one problem is that I am not sure how the authors are going to improve their manuscript based on those comments. It seems to me that the authors are focusing too much on clarifying their method to just the reviewer him/herself, and did not say anything what has been done to prevent a similar question being raised by a common reader. I therefore suggest the authors update their manuscript lively after responding the reviewers if its possible. If update the manuscript constantly is not an option, at least, they should describe in details how the comments are addressed in the manuscript itself. 

After reading the authors reply to the comments I find that the authors have response fairly well. The authors response the reviewers comments fairly well but If a question has been raised or a misunderstanding has been made by an reviewer (expert) when reading the manuscript, then their is good chance that the same will happen to a other readers. Therefore, in their reply, the authors should focus more on describing how they plan to improve their manuscript rather than response to the reviewer him/herself.