Overall evaluation

------------------

A very relevant, interesting and well written paper perfectly in the scope of AMT. I recommend for publication after the following comments have been adressed:

General comments

----------------

- Here we only look at cases were we know fog develops. How many false alarms for fog onset to expect? Can you please comment on this? 

- The commercial MWR have a surface temperature sensor incorporated. With this cheap addtion, the retrieval can be constrained at the lowermost levels. Excluding these is scientifically possible, however, it does distort reality if we consider the instrument as a whole (MWR plus meteo station). Please account for this in your manuscript, not just in the appendix. You have done all of this in appendix B, and I found it relevant enough to include Fig B1 and B2 to the main manuscript in place of Fig 5 and 7.

Specific comments

-----------------

- l. 35:  on clear evenings with light winds -> on clear evenings with no or light winds (otherwise one might think that wind is needed)

- l. 54:  The DIAL by Vaisala claims to have a lowest usable range gate at 50 m (see e.g. Mariani et al. 2021). Anyway, I agree that in the end this does not change a big deal for the usability for fog.

- l. 66:  I don't like the +/- here as the direction of the sign of the change has opposite influence on the melting potential.

- l. 68:  It is an over-simplification to reduce MWR to K- and V-band. What about "... that measure downwelling radiation. Commercial sensors for temperature and water vapour profiling typically operate 14-35 spectral channels at 22-31 GHz and 51-58 GHz.

- l. 77:  "Despite the promise of MWRs to improve fog forecasts". Please re-word the beginning of this sentence. I judge that the above-cited references (Martinet, ...) demonstrated that the improvement is more than just a "promise" that might not be fulfilled. 

- l. 104: your study only assess the situation of supercooled radiation fog. Would you expect your results to also be representative for fog>0°C? Why (not)? Please elaborate in the manuscript. Ok, see you mention it is not guaranteed to be representative in Sect. 4, but maybe you want to give an indication here already.

- l. 173: here you study only cases without cloud. Please comment on what effect clouds above fog would have on AERI observations.

- l. 249: I undertand that you are limited to max altitude 10m by the height of the tower. However, this comparison is most probably artificially penalising MWR which, with its lower vertical resolution, might perform better when comparing to 20 or 50 m. And having a somwhat thicker layer might also make sense from a physics point of view. Please comment on this.

- l. 317: "This suggests that most of the information ... comes from the prior" -> It would be interesting to see the measurement response (i.e. the sum of the averaging kernels). This would allow us to judge how much info comes from the prior rather than just guessing.

Fig. 8:   Please specify the extent of the central box in the legend

Appendix B: Can you also show the corresponding figure for the 0-10m lapse rate with and without surface temperature. I would suggest to use this instead of Fig. 6

- l. 388: ... Results of Appendix B... --> you probably wanted to add a reference to Appendix B

- l. 427: "AERI contain more information about the temperature near the surface than the MWR measurements". you might add something like "what indicates the importance of including surface observations to the retrieval if available from the MWR's weather station"

Comments to AMT-2022-12: Passive ground-based remote sensing of radiation fog

By H. Guy et al.

Overview:

The paper discusses the ability of an IR ground-based sensor (AERI) to detect the onset of radiation fog in Greenland and compares it with a microwave radiometer. A dataset of measurements from microwave and infrared sensors, ceilometer, radiosondes, and a cloud radar is selected, retrievals of temperature and humidity profiles and liquid water path are developed and used to identify times of fog and results are presented and discussed.

General comments:

I found the paper interesting and well written, and I recommend it for publication. The figures are of good quality and the presentation is clear. If I may put forward a general comment for a minor modification, in my opinion the discussion part would benefit from a more balanced approach where, instead of trying to prove that the AERI is “better” than the MWR to detect radiation fog (which was not so clear to me at the end of the discussion), the advantages of using both instruments   to improve fog forecast under a broad range of conditions are discussed.

To explain better this point, I’ll mention here the 3 criteria used in the paper: Accurate temperature and humidity retrieval, characterization of shallow surface inversions, detection of small changes in LWP.

The temperature and humidity retrievals (section 3.2), at least in the dataset analyzed, appear reasonable from both instruments. I agree that the bias in the MWR channel is an issue that needs to be addressed, even in assimilation, but it can be addressed. The problem of detecting surface inversions (0-10 m) from the MWR (section 3.3) is easily overcome with the help of surface temperature measurements. Obviously, if using only brightness temperatures, the MWR won’t be able to resolve the two heights because measurements at 0 and 10 m are highly correlated (that’s why you get a constant difference in Fig. 6a). However, that is the reason why the instrument is equipped with surface sensors.

It therefore appears that the only real advantage of the IR measurements is the higher sensitivity to small changes in LWP (section 3.4). Even here however, this advantage is true in very controlled conditions such as those carefully selected for this specific comparison, i.e. when the sky is clear, no ice, LWP < 40 g/m², and the cloud (fog) base is accurately known (as shown is section 3.1). It is not entirely obvious to me that a correct identification of all these conditions is easily achieved in NWP models and how the mischaracterization of any of these conditions could affect the results. When everything is put together therefore it almost seems that the advantages of using IR are not that clear.

I understand that this paper is an initial step and that a more holistic assessment of fog detection with both instruments is out of its scope, but perhaps some ideas on how to use the synergy of both instruments to obviate shortcomings in each single one would be useful. For example, could having both estimates of fog (MW and IR) help identify cases when one of the two is not acceptable for any reason? Or to identify false positives or missed cases?