Summery

The article focuses on assimilation of sea ice thickness (SIT) in a coupled climate model. The novelty of the article is that it presents an assimilation of ENVIAT SIT observations in addition to the commonly assimilated observations from CryoSat and SMOS combined in the CS2SMOS product. The assimilation is done by employing an ensemble Kalman Filter and the study investigates the skill of two assimilation runs in comparison to a free model run. The two assimilation runs are: One run called CRLT, assimilating sea ice concentration and sea surface temperature and hydrographic profiles, and the second run called +SIT assimilated in addition to the variables in CRLT also SIT. The performance of the assimilation is validated against independent and assimilated data and their forecasting skill is assessed, well in line with the standards of the field. The model it self has, as the author points out, a rather large sea ice thickness error, which is reduced by both assimilating CS2SMOS and ENVISAT. Further the study investigates the skill of the assimilation to predict the sea ice edge and extent. Better results are found for predicting the sea ice edge than for the sea ice extent.

General comments

Overall the article is of good scientific quality and presents a significant body of work as well as citing current and relevant literature, with the assimilation of ENVISAT SIT being the main novelty. I currently though see two main points in need of improvement:

1) Currently the article is missing a clear discussion of the differences in skilfully predicting the sea ice state between assimilating ENVISAT and CS2SMOS. The title, summery and aim in the introduction (line 60) give the impression that the article mainly addresses new insight from assimilating ENVISAT in comparison to only assimilating CS2SMOS, however the article only discusses the difference between the reduction of bias and bias free root mean square error for both satellites, not the difference in forecasting skill. Keeping the large SIT bias of the model in mind, the bias correction is to be expected. Either additional analysis should be added, or this should be made clear in the title, introduction and summery and the choice to not analysing the difference in skill between assimilating only CS2SMOS and assimilating both SIT products should be motivated. For example does the summery state that the article focuses on the skill of seasonal prediction. From reading this I would expect that the article analysis the differences in skill from assimilating ENVISAT vs. CS2SMOS, however this is not the case. This is unfortunate, since the assimilation of ENVISAT is the main novelty of the study.

2) It is clear that a lot of work and testing has gone into the assimilation set up, which is very well done. However the method section is currently a bit confusing and lacks some important information for the study to be reproducible: 1) how is the SIC updated by the SIT assimilation, in one categories, in several categories, etc? 2) How does Full filed assimilation differ from anomaly field assimilation? What effects on the results are expected of mixing them? 3) Is there a reason why you choose to not assimilate both SIT observations during their overlap? 4) Please add version numbers to the model components where applicable. 5) Which variables are in the state vector. For the Ocean this is clear, but how about the sea ice?

Minor comments

line 25-27: please add reference

line 234: first occurrence of acronym RHS

line 280-281: add citation

line 315 which visual agreement is referred to?

table 2: it would increase readability to use the same exponent within one column (not the case for SIV of +SIT)

line 319: +SIT has a strong discontinuity only in SIV, the SIE is actually in line with observations, or? (figure 5)

1) line 320-325 which table figures are you referring to?

3) figure 3: what is the criteria for not enough point to calculate a yearly average?

2) figure 7: please add a accessible colour bar varying between two colours.

Line 408-409: Which section is this based on? From figure 5 it looks as if the sea ice extent is only improved in September, not in March during the EVISAT period.

Line 410: CS2SMOS is only available from 2010, so it can not improve anything before.