This study presents a high resolution (1/16-deg and 1-hr) DOISST using  SEVIRI satellite as source and a model SST as first-guess, which will no doubt have many applications. Their analysis demonstrated that DOISST well represent the diurnal cycle with a low mean bias and RMSD of 0.4ºC. The seasonal features of SST diurnal cycle in MED area are described and compared with independent buoy SST and model simulation. My major concern is how the model SST at 1 m depth is used as the first-guess field. It is not clearly stated what depth the DOISST represent, and how the DOISST is validated with buoy SSTs at 0.2 m depth. The ideal case is that all three components are compared/generated at the same depth level. I recommend accept the manuscript after a major revision.

L14, it is not clear why the model analysis is used as the first-guess.

L15-16, it is not clear what is “any diurnal cycle”. “differences between satellite and model SST are free, or nearly free”. If this is the case, then why do we need DOISST analysis?

L35, Does this mean, the SST analysis will be absent when it is rain?

L71, “slightly less than that” => approximately?

L155, SST at 1 m level. What depth does the satellite SST in section 2.2 represent? How is the model SST at deeper level used as a first-guess of the satellite SSTs near the skin level?

L158, how is the buoy SST at 20 cm level used to validate DOISST

L166, delete an extra space.

L169, “between” => among?

Table 1,  sub-skin SST, What is the level of sub-skin? please add a depth level.

L188-190, the statement is not clear, and need to be clarify, particularly “allowing to interpolate SST anomalies using satellite data”. Is the “anomaly” referenced to an hourly climatology, how is the climatology is defined?

L201-202, what is the difference between L3C SST and L3C sub-skin SST? How are the SSTs at different levels blended in DOISST?

L212, f(r,dt)=f(r)*(dt) may not be appropriate. Delete “f(r)*(dt)=”?

L224, “no first-guess data are used”, how is this possible as described in L227-234?

L240, how is “co-located” defined, interpolated to the in situ location and time or rounded to a certain spatial and time resolution?

L272-273, How is the uncertainty of RMSD (±number) calculated?

L295, is it possible the biases result from the first-guess of the model SST at different level?

L337, define DWA earlier.

L347-347, will the underestimation of DWA in model affect the performance of DOISST since it is used as the first-guess?

L385-386, it may be better to explain the reasons.

L422, these depth information should be presented much earlier

L429, In MED area or over the global oceans?

The paper describes a new sea-surface temperature product merging SEVIRI data and results from a data-assimilative model. The important novelty is that this product resolves the diurnal cycle and provides full fields without gaps (level 4). The authors also include a quite detailed comparison with in situ observations. While this paper describes the results for the years 2019 and 2020, this data set is continuously updated and the results for the year 2021 are also available.

The main questions that I asked myself while reading the manuscript are:

A. As there are different depths for the different types of SST (skin-temperature, bulk temperature, foundation temperature) which depth level is the DOISST targeting by this product? I understood that the model and SEVIRI data have different reference depths. Should there not be first a conversion/adjustment, so that the temperature is comparable? Maybe interpreting some of the conclusions within this context would be useful.

B. Comparison: I would have expected a comparison to show that DOISST is better (compared to in situ observations) than other observational products. However, the author compared the new  product to a model solution. Are there other data L4 products available (resolving the diurnal cycle) based on SST data from geostationary satellites? In any case, the authors also compare the accuracy of their product (relative to drifters) to the accuracy of the SEVIRI data (at exactly the same location) which already shows some quite favorable results.

I recommend publications after minor revisions.

Minor comments:

1. line 106: assessment of the MED DOISST product covers two complete years (2019-2020)

Please clarify earlier in the manuscript the time coverage of the data product and the time coverage of the assessment. 

2. degree K (line 19, abstract) or degree C (line 39, introduction). Can you please use the same units?

3. an overview table with all products would be useful, including resolution (time and space) and coverage (time and space) and reference depth (e.g. skin, subskin, foundation temperature,...), even if the study uses a subset of the input data set. This table could also include the new dataset.

4. typesetting of the equation should be improved and follow the style of other Copernicus papers.

5. page 9: "All these parameters have been deduced from a statistical analysis of the satellite SST data"

Please give more information about how you choose the particular parameters (a, c, d, decorrelation spatial length R, decorrelation time length T). In particular, what objective criterion was used to decide that these parameters are appropriate? 

6. page 10, line 250:  "At each step of decreasing n, data that falls out of the interval I = [mean(delta) - n sigma, mean(delta) + n sigma] are flagged. The process starts for n=10 and stops at n=3." If the data is outside of the interval for n=3, why would one also check for n=10? But I guess that delta (the difference, and the mean and standard deviation) also depends on n by selecting a different subset for different n. I think that this should be clarified in the proposal.

7.  line 295: "The two diurnal cycles are practically coincident between 17:00 and 06:00, while they are biased by ~0.1 K between sunrise and 16:00, coherently

with the DOISST bias oscillation (Fig. 3). This bias could be related to skin SST getting warmer faster than 20 cm temperature"

I suggest you replace "20 cm temperature" by "temperature at 20 cm depth".

I am not sure if "coincident" is the right word. What about saying that the bias is close to zero (DOISST and drifter temperature) as you do not show the diurnal cycles of DOISST and drifter temperature individually.

The authors present a new Mediterranean Sea regional SST product that reproduces the

diurnal cycle. For this, the authors merge the SST from the CMEMS Mediterranean Sea

Physical Analysis and Forecasting product with the SST measurements from SEVERI

remote sensor, and they apply a methodology that is presented in Marullo et al. (2014). For

assessing the actual capability of the resulting SST product to properly capture the skin SST

variations, the authors use a set of drifting buoy SST measurements that are typically

acquired at 20 cm depth. This is a clear limitation of the assessment of, not only this product,

but all the satellite products that aim at reproducing the skin dynamics, there is not in situ

data to compare with. In the absence of in situ skin SST measurements, the quality

assessment that the authors present here is clear, and they provide evidence that the

product is properly capturing the diurnal cycle, or at least that it is capturing it better than the

model. So, I think the manuscript deserves its publication in the Earth Science System Data

journal.

I have some minor comments /questions to the authors.

Line 15-16: “The differences between satellite and model SST are free,or nearly free, of any

diurnal cycle”->I don’t understand this I though model does not reproduce the diurnal cycle

while the satellite does

Line 17: I’m wondering whether these drifting buoys are assimilated in the model or not.

Line 93: It would be interesting for the reader a comparison between the performance of this

skin SST OSTIA and MED DOISST.

Line 106: Do the authors plan to extend the temporal series backwards?

Lines 128-130: I don’t understand this. Why are the differences between SEVERI SST and

drifters larger during nighttime than in daytime? I would expect larger differences during

daytime because drifter measurements are acquired at 20cms depth and SEVERI

measurements are provided in the first mm. Are these differences reflecting in first order the

radiometric errors of SEVERI?

Line 166: Delete “ “ before “.”

Lines 188-191: I don’t understand this paragraph: 1) Why are you using differences between

satellite and model instead of satellite measurements directly? I don’t understand the point of

the reduction of one order of magnitude of the difference.. 2) Do you mean that for

generating hourly products you are considering all observations around the model in +/- 24

hours? Have you assessed the impact on the final product of considering different (reduced)

temporal windows?

Line 203: I would specify here also the model spatial grid.

Line 204: I would specify at which grid the regridded is performed.

Line 256: Estimates of the correlation with in situ may also provide useful information.

Line 258: Have you assessed SEVERI SST? It would be interesting for the reader the

comparison between SEVERI and MED DOISST performances (not only in the DWA).

Line 262: I would say pointwise difference.

Fig. 2: Perhaps it would be interesting to separate the map into daytime and night time.

Line 265: “tendency”->”predominance”

Fig 6. Is interesting that although the dispersion of DOISST DWA around Drifter DWA has

been significantly reduced with respect to the one of SEVIRI DWA, the maxima DWA eventsseem to be better captured with SEVERI than with DOISST (that they seem to be a bit

underestimated).