The manuscript tries to evaluate low-level jet (LLJ) characteristics over the southern part of the Baltic from ship-based measurements. In principle, this is a good idea. Such information is needed for wind energy purposes and for model evaluation.

Unfortunately, the manuscript and the used datasets have some unfavourite properties which lead to an unsatifying result. Thus, I am sorry to suggest a rejection of this manuscript in its present form. Especially the revision issues 2 and 3 listed below are difficult to heal as they do not originate from the direct influence of the authors. A chance could be to find an attractive scientific question which could be addressed and answered with the available data. But this would mean to write a quite different paper.

There are three major revision issues:

Issue 1:

Astonishingly, the seminal works of Smedman and co-workers on LLJs over the Baltic Sea in the 1990s have completely been ignored.

Two mechanisms are given in the manuscript for the formation of LLJs over the Baltic: (1) advection of nocturnal jets formed over land, and (2) baroclinicity. But one decisive mechanism is missing: the flow transition taking place when air moves from the land to the sea. Especially when warm air moves from rough land to a colder and much smoother sea, a sudden acceleration due to the sudden reduction of surface friction sets in. Smedman and co-workers based their data interpretation on this mechanism.

I suggest that in any new version of this manuscript the section on LLJ formation mechanism is re-written starting with the papers and ideas of Smedman et al. Also a look at a very recent overview paper (most probably it came out after the authors finalized their manuscript) by Schulz-Stellenfleth et al. (2022, DOI: 10.1127/metz/2022/1109) might be useful.

Issue 2:

Evaluation of the lidar data in this manuscript is very much biased by two facts: (1) by the limited height of 300 m of the lidar measurements, and (2) by the ferry time schedule which allows for measurements at certain sections of the ship track at very few hours of the day only. The ship is always at the same place at the same hour of the day. Due to these two facts, neither a proper diurnal variation of the LLJ characteristics nor a proper frequency, core height, and core speed analysis of the LLJs can be made. Due to the second deficiency, the advantage of a moving lidar (compared to those in fixed positions) nearly completely disappears.

Issue 3:

The numerical models used in the manuscript have their own intrinsic deficiencies (in this context, the work of Sandu et al. (doi:10.1002/jame.20013) should be read and cited). 

At the end of the day this leads to a comparision between limited measurement data and limited model data which does not really makes sense.

Given the three above mentioned issues, it is not clear to the reviewer what is the actual purpose of this publication? This publication merely gives a record of lidar measurements onboard a ferry. The above mentioned limitations are partly addressed in the manuscript, but no conclusions are drawn from these facts.

A few minor comments:

l 38        „as“ instead of „us“

l 44/45  extension: lateral or vertical? (If lateral, it seems very small; if vertical, it seems very large)

l58-61   the paper Wagner et al. (2019) should be mentioned here again as it is already listed in the list of references

l 138-149            “newest”: at least a year must be given or even better a citation in order to properly identify the version of ERA5 data (the hint to the ECMWF webpage does not help either as webpages may be updated in future)

l 182-205            the LLJ detection algorithm can only work, if the height of the LLJ core is much lower than 300 m. What happens, if the core height is closer to the uppermost measurement level? This issue has to be discussed. Fig. 8c proves this problem.

The manuscript presents three months of lidar measurements made from a commercial ship ferry at a regular route. All types of offshore surface-based measurements are potentially interesting due to the scarcity. Especially wind field measurements in the Baltic Sea are of importance due to the upcoming expansion of offshore wind energy in the region. The presentation of the manuscript is overall quite good with clear figures. I noticed some language issues and some places where the references are presented in an unusual way.

This data set is tricky to analyze. It is too short to do climatological studies or analyze seasonal variation. Additionally, the variation in space is also a challenge. Although the authors make a good effort to address the latter, the data-set is biased in the way the measurements always seem to be from the same location at the approximately at the same time of day. This makes analyzes of temporal variation from one point not possible.

However, my main concern is that the main results of the study is the comparison with the reanalyses products. It is not clear what is really novel here that hasn’t already been published in similar studies from the same region using the same reanalyses products, which you also cite in the manuscript e.g. Witha et al. 2019 and Hallgren et al. 2020. To be able to accept this manuscript I would like to see some more, other type of analysis trying to get a deeper understanding of the results from the comparison such as: During what conditions do the models perform better/worse? Also adding more evaluation metrics could be useful in this sense. How can one use these results to improve the models?  Discussion of the benefits of using ferry based Lidar would be useful and give examples of these. Illustrative case studies could also be useful e.g .perhaps for some specific synoptic situation where the analysis would benefit from a moving platform. Is it possible to use this type of platform to evaluate models for internal boundary layer? These are just some examples, but this study would require some more along these lines.

Specific comments:

Line 3: it is stated that the objective is to evaluate performance of the ship-mounted lidar to investigate LLJ properties along the ship track. However, I can’t see that this is presented in the manuscript. The LLJ properties from Lidar measurement are presented, but the performance is not evaluated in any formal sense.

39: “results are insufficient”

51: LLJs in the Baltic Sea have been studied also before the mentioned references. 1984 Högström and Smedman present a first paper where the LLJs formation mechanism is described as an “analogy in space to the classical Blackadar nocturnal jet frequently observed in continental areas”. This mechanism is missing in this section. Other studies also followed from the group e.g. Smedman et al. 1995: Spectra, variances and length scales in a marine stable boundary layer dominated by a low level jet, BLM, 76(3):211–232.

61: “sloping topography” (not sloppy)

63: Concerning the Stensrud 1996 reference: I think this was first presented in Holton 1967: The diurnal boundary layer wind oscillation above sloping terrain. Tellus

69-70 a detail but is there support to say that NEWA is one of the most frequently used re-analyzes products? ERA-5 is for sure one them though.

116: “likewise in any” replace with something like “and like any”

118: why was -23 DB limit chosen?

121 I suggest replacing “filtered” with “rejected”

122 how is this 70% limit different from the 80% limit mentioned on line 121?

125 replace “capture” with e.g. “simulate”

134 and 139 : correct reference for ERA-5 Hersbach et al. 2020 https://doi.org/10.1002/qj.3803

146-147 how did you deal with this (mismatch between cycles)

155 “spin-off” replace with “spin-up”

179 “concentrates” do you mean “conserves”?

201 Reference Kalverla: The handling of references should be to place the parenthesis around the year only. This needs to be checked at several places in the manuscript

202 “extended” replace with “extending”

217 comparison with Witha et al: you are using essentially the same data set as Witha et al., please comment on why the results are different.

247 “misestimation” replace with “undererstimation”

249 Cheinet at al. year missing

263-264 the onshore daily cycle is well studied as mentioned previously in the manuscript.

267 Can you motive the choice of these four locations?

Figure 8: does this figure show comparison of co-located model-observations pair in time also, or just the location co-location?

281 “appearance” replace with “occurrence”

Section 3.2.2. you use the term “inshore”, “near shore” and “onshore” to describe the same locations, please be consistent.

282-283  not following here, in the previous sentence it is stated that the frequency is overestimated in ERA 5 140% and then it is stated that it is underestimating in this sentence (?). Please clarify.

294-295 “The increase in the wind profile” do you mean the extension of the analyzed wind profile height in the models?

Figure 9 Here you could add the lidar measurements in the shadowed areas. Additionally, this type of analysis would benefit from some more statistics: do the means differ from each other significantly? What is the spread around each averaged point?

Table 3: Why not add the Lidar measurements in the comparison and present a similar analysis as in Figure 9?

Figure 10: Is there any correlation between frequency bias and the forecast length? Are the means statistically significantly different?

359 “alarm” the correct term is “false alarm”, you need to correct this at several instances in the manuscript.

Table 4: spelling  “mises” --> “misses”

382 Last sentence: please clarify, it's hard to follow the reasoning here.

Figure 11: Why not include ERA and NEWA in the same plot? This would make the comparison easier.

Figure 12: Other options are also available and should be commented: e.g. interpolate the nearby model data to the measurement location or combining a spatial and time window.

491  One way to investigate how successful the motion correction is would be to study the spectrum of the velocity measurements. If the motion correction algorithm is successfully implemented the peak around the mean wave period should been removed. Although this requires access to the raw turbulence data from the lidar which might not be the case here (?)