Dear Colleagues,
I am now in receipt of a third review on your paper. The original reviewers are unavailable and I would like to move this process forward so I am raising some important issues in addition to those raised by reviewer 3.

In general I have strong concerns about the approach, and the lack of sampling statistics. Reviewer 3 has offered one option - namely to pursue either N retrieval or adiabatic retrieval but to do so more rigorously. I would like the authors to weigh their options. As it stands the manuscript is not acceptable for publication.

I offer some detailed comments and ideas below in the hopes that they will help the authors improve the paper.

Graham Feingold

Main comments:

1) I believe that caution is warranted when using a ground based radar/microwave radiometer method as reference and comparing with a passive approach (SEVIRI, MODIS). The methods differ not only is their spatiotemporal sampling but also in their sensitivity to the drop size distribution. I know the authors are well aware of this (comments to this effect are peppered along the way) but this then leads one to ask whether one should expect good comparison between the space- and surface-based approaches. Moreover, if the comparisons are good is it for the right reason. With such a small sample (4 cases) one cannot draw conclusions.
Some papers that might be helpful are
doi:10.1029/2004JD005648 and doi:10.5194/acp-12-1031-2012.
I raise these papers not to simply suggest reference to them but because I think that important lessons can be taken from them.

2) The use of radar and microwave radiometer to retrieve drop concentration is a very risky proposition given the disparate moment-weightings (0 vs 3 and 6). Frisch and colleagues have discussed how the ratio of the moments in Eq. 10 can become very unstable, which explains why you sometimes retrieve very high N (line 598). In fact in later work they resorted to using fixed (climatological N; Frisch et al. JTECH 2002). It’s one thing if one retrieves Reff profiles with radar and microwave radiometer but another when one attempts N. In this regard I am in agreement with reviewer 3 that the attention to retrieval of N is not very rigorous - not withstanding the effort put into OE.

3) Eq. 14 for the radar derived adiabatic measure also raises questions. If the authors have access to aircraft measurements - e.g., from colleagues at TROPOS they could test this sensitivity. The authors should look at the sensitivity of this value to radar thresholds; I suspect it will be quite sensitive.

Other points:

1) Lines 122-123: Can you support this claim? What are the relevant spatiotemporal scales for radiative forcing?

2) “However, when the results of both approaches are in agreement, it is likely that the correspond- ing cloud layers are well suited for the investigation of key factors determining the first indirect effect.” Could you explain why? Is this logical? What about compensating errors?

3) “Although rain might be a possible explanation for higher QL observed with the ground-based microwave radiometer, there are no are no signs for precipitation in both radar signal and satellite observations.” Could you explain why? Is this because of wetting of the window? Is the QL highly variable as well? (typical in rain).

4) 4.2.2: There is a great deal of discussion about differences, but given the small sample what statistical significance does this have?

5) “Due to the N ∝ Reff^−2.5 relationship (see Eq. 5)”
But tau in eqn 5 is also dependent on Reff…

6) “ In our study we filtered out drizzling profiles as well as possible, but the radar reflectivity still remains very sensitive to few larger droplets in a volume, which can not totally be ruled out. Therefore also the correct radar calibration is an issue.”
This statement should have appeared in the beginning, and frankly, it might have been a point at which it which the authors decided to stick to optical measurements + microwave radiometer - particularly for the comparison. In this way they could focus on averaging issues in the comparison, rather than a mix of issues.

7) “captured surprisingly well for some cases. We discussed the large uncertainties that may occur depending on the observed scene and observation geometry.”
I believe that this statement should be revisited given the poor statistics. Is the comparison good because of compensating errors, pure chance, or because it really is good.

Minor:

1) Throughout the text there is reference to “indirect effects” when the authors really mean the microphysical response to a change in aerosol. The indirect effect is a radiative response and this is not addressed here.

2) Throughout the paper, there tends to be a lack of recognition to early work on various topics.
Line 179: Baker and Latham discussed mixing in the 1980s.
Line 250, Frisch et al. (1995) were I believe the first to perform radar/microwave radiometer retrievals of Reff profiles and N. Line 558: Warner showed in the 1950s that shallower clouds are more adiabatic.

3) Line 555-556 is an odd statement.

4) Rewrite “The maximum of the radar reflectivity in each profile did also not -20 dBZ…”

Dear Colleagues,
I have reviewed your revised manuscript and have reached a decision to accept subject to technical corrections. Please do make these corrections (particularly the one related to the surface Nd retrieval) since I believe they will improve readability and clarity.
Sincerely,
Graham

General

The manuscript would flow much better if you would define cloud properties like Nd, Reff, f_ad, Z, H, etc (as you do already) but then use these symbols in the text rather than repeatedly writing out “drop concentration, effective radius, adiabatic fraction, etc..” Sometimes you even redefine the cloud parameters.

You periodically talk about aerosol-cloud interactions but there are no aerosol measurements and the paper is about the retrieval of cloud properties, many of which are relevant to the aerosol-cloud interaction problem.
Please tighten the language throughout.




Abstract

1) Twomey effect can be removed since it is jargon this early in the paper.

2) “The adiabatic cloud model
is often applied and modified using a sub-adiabatic factor
to account for entrainment within the cloud.”
Change to “The typically applied adiabatic cloud profile is modified using a sub-adiabatic factor to account for entrainment within the cloud.”

3) While a synergistic method… (grammar)


4) “Low-level liquid clouds are found in many areas around the globe and play an important role in the energy balance of the Earth”

5) Han et al. (1998) proposed to investigate a column Nd which is the integral of Nd over H.

6) “While remote sensing observations from ground are al- ways local column measurements”

7) Line 115, remind the reader that this is a geostationary platform

8) Line 157, define QL as liquid water path. (Not sure you ever defined it.)

9) Line 200: This is an opportunity to say that while it would have been preferable to derive Nd from cloud optical depth (tau) and QL, you don’t have an MFRSR or similar. Point out here that this means that you are trying to derive the zeroth moment from the 3rd and 6th moments (rather than 2nd and 3rd in the case of tau and QL). This is the fundamental reason why Nd retrieval from the ground is so sensitive to distribution breadth.

10) Line 219: “may occur”

11) Line 214: “where Nd(z) stays constant in the vertical layer” is confusing since entrainment mixing is a local process and here you are talking about Nd remaining constant with height over the depth of the cloud.

12) Line 291: The output data is available at a unified …

13) Line 293: uses further information from a NWP model.
Please provide the key parameters (T, P?). Otherwise the list that follows sounds like it comes from NWP.

14: Line 372: “While the ceilometer-derived cloud base…”

15) Line 421: Define IQR at first use.

16) Line 427: “The adiabatic factor not only changes from case to case, but also varyies with time for individual days..”

17) Line 436: “is problematic” for what?
You might say: “could affect retrievals of cloud properties”.

18) Line 475: ACI is not defined.
19) Line 606: “look at”
20) Line 740: “to which”
21) Line 751: Again my comment about trying to retrieve the 0th moment from the 3rd and 6th, which explains the sensitivity to breadth.

22) Line 726: There are much earlier references that date to the beginning of precipitation radars - perhaps Louis Battan or Joanne Malkus/Simpson, or David Atlas.

23) Line 779: I think you may have missed the point here. Why not write something like: “Recognizing the difficulty in retrieving Nd from the 3rd and 6th moments, Frisch et al. (2002) used a cli- matological mean value for Nd in order to retrieve re. They reported an average Nd of 212 ± 107 cm−3 at the Southern Great Plains site for continental clouds, which is similar to the median value found for our example cases in Fig. 5.

24) Line 814: Uncertainties in..

25) Line 918: “limitations”

26)Line 918: change “current” to “our” since some current sites do have surface cloud optical depth measurements.

27) Line 919: you don’t look at aerosol-cloud interactions

28) Line 944: “upward”

29) Line 950: Again, the Nd retrieval from radar is poorly posed because of the moment disparity and the potential instability of the ratio in Eq. (10) as pointed out by Frisch.

30) Eqns (12) and (13): It would have been nice to have the Nd and H equations to refer to so that the error terms could have been easily identified. For example, I immediately looked at Eq. (9) for Nd but it is not formulated in terms of r_e and tau but QL and tau.