Thanks for the opportunity to review the manuscript. My comments are attached.

Review of “New Global Mean Dynamic Topography CNES-CLS-22 Combining Drifters, Hydrography Profiles and High Frequency Radar Data” (essd-2025-429)

General Assessment

The paper presents the latest CNES-CLS Mean Dynamic Topography, an essential altimeter reference surface needed to reconstruct Absolute Dynamic Topography, and, under the geostrophic assumption, ocean surface currents, from the altimeter Sea Level Anomalies. The new calculation is essentially based on the same methodology than previous versions, but includes more recent Mean Sea Surface and geoid solutions together with updated in-situ data and a new, regional dataset of HF radar velocities for the Mid Atlantic Bight.

The continuous improvement of the Mean Dynamic Topography is key for the optimal exploitation of altimeter measurements, all the more in the new era of wide-swath altimetry. The main challenges in this context are to resolve shorter scales than before, improve the quality of the solution in coastal areas where altimeter data quality is lower and in-situ measurements are lacking, as well as at high latitudes.

The present CNES-CLS22 MDT, despite tackling these different challenges, does not depart significantly from the previous CNES-CLS18 version, apart from the Arctic area where clear improvements are obtained.

The paper is well structured and well written but some methodological steps are only superficially described and the obtained results should be discussed more critically.

 I recommend publication after minor revisions.

Major comments

• The method used to calculate the new MDT is essentially the same as described in previous papers. One improvement is mentioned in the abstract, then briefly at the beginning of section 3.1 and finally in the conclusion, i.e. the use of a lagrangian filtering at the coast to reduce the intensity of currents normal to the coast when calculating the first guess, but the method is not further described, and the impact of the approach is only qualitatively and superficially discussed in the case of the Mid-Atlantic Bight area. Details about the filtering method shall be provided, together with a more thorough discussion of the impact of the approach for improving the accuracy of the obtained field along the coast.
• The MDT calculation is based on the use of a multivariate objective analysis, which “takes as input the a priori knowledge of the MDT variance and zonal and meridional correlation scales ». An important input information, the observation errors, is not discussed at all. The authors shall provide clear information on how the observation errors are calculated.
• This is particularly important for the Mid-Atlantic Bight Area where different datasets are used simultaneously: T/S profiles, drifter measurements and HF radar. I expect the relative errors associated to these different observations to have a strong impact on the final solution and this shall be discussed. In addition, as the inclusion of the HF radar is another novelty of the paper compared to previous work, an analysis of the added-value of including this information of the final result is needed. For instance, what is the impact of using the HF radar information together with the T/S profiles and the drifters, compared to using only the T/S profiles and the drifters as done in the past?
• Regarding Figure 5, I am surprised to see the difference in energy at scales larger than 125km between the CNES-CLS22 MDT and the first guess. This means that the synthetic observations also bring information at large scale (from the plot, at least until 600km wavelength -> 300km resolution). This shall be discussed.
• The validation results shown in section 4 are not showing significant improvement of the new solution compared to the previous one, except for the Arctic Ocean. The results shall be discussed more critically, also pointing out the limitations of the current approach and the ways of improvement in the future.
• In particular, results of section 4.2.2.1 are surprising: It is not clear why the CNES_CLS MDT22 would be less accurate in such areas? You use more data than previously, but still average them on 1/8, similar to what was done for CNES-CLS18. You use the same covariance (?)...Maybe you use different error levels? Line 381-382 it is mentioned that the “CNES-CLS MDT22 offers improvements in representing the large-scale circulation more realistically”. Not clear which result support this statement? Also because from Figure 10a, 10e and 10f, the difference between the 2 MDTs is rather large scale and it is clearly this large scale gradient difference which causes the large discrepancy in transport. As the CNES-CLS22 MDT large scales come from the first guess, it would be interesting to look at the first guesses, and calculate the mean transport as seen in the first guesses.

Minor comments

Attach you can find an annotated pdf files with typos and comments.