This research study, while a worthy topic for investigation, fall far short of its intended goal, i.e. to improve our understanding of the radiative forcing by contrails, contrail cirrus and cirrus by using satellite measurements validated by in situ observations. Although the authors have devoted a fair amount of effort to analyze the satellite observations to identify contrails and contrail cirrus, they conclude in the end that it is impossible, so instead they take three passes from in situ measurements and three passes from airbone lidar to conclude, and this is my paraphrasing, "We can't identify contrails or contrail cirrus from the satellite measurements in the region where the aircraft measurements were made, but since we think we might have identified contrails and contrail cirrus with the airborne measurements, we will generalize and assume that there must also be such clouds observed by the satellites". This is not convincing.

I am unconvinced by the arguments that are made by the authors. Whereas case studies are an acceptable means for studying cloud microphysical processes when the data sets are limited and hard to obtain, this study does not fall in that category. There must be thousands of measurements by the DLR Falcon and Halo in contrails and contrail cirrus that could be used and yet the authors have chosen one day with only three passes, with no justification for why this day was chosen. In addition, the lengthy descriptions of the data are overly detailed with unnecessary discussions of irrelevant features. Every sentence has to be written with information that coveys succinctly the point the authors wish the reader to see and understand. There are too much speculations, i.e. "might be", "could be", possibly", with little concrete data that the reader can use to understand what the authors are trying to convey.

There are many other aspects of this manuscript that fall short of my expectations, but rather than  address them in this review, I will wait for what I hope is a more comprehensive (and convincing) study that has more in situ measurements in co-located satellite measurements. I will also expect to see a detailed discussion of how the in situ measurements were processed, including an engineering error propogation that includes the expected uncertainities in derivde quantities, time offsets in the cloud, NO and RH measurements, quantification of the polarization ratio (not perpendicular to forward but perpendicular to sum of perpendicular and parallel), etc. 

Summary of paper:

The authors use a combination of aircraft and satellite measurements, and radiative transfer modeling to analyze a band of thin ice cloud in the North Atlantic air corridor on 26 Mar 2014. (Young) contrails, contrail cirrus, and natural cirrus within the cloud layer are distinguished by in situ measurements of ice particle number concentration and NO gas concentration. The optical thickness, effective radius, and radiative forcing are computed for each cloud type within the cloud band.

General comments:

The goals of the paper are scientifically important and worthy of study, but the authors do not characterize the three cloud types in a convincing manner. The cloud types are defined from in situ measurements, but the authors appear to conflate individual contrail properties to the entire layer at the point of observation. It is not clear what distinguishes a (young) contrail from contrail cirrus, even without the context of the overall cloud band. Several if not most of the contrails appear to be at least two hours old and would likely be visible in the satellite imagery, yet I could find no attempt by the authors to use MSG/SEVIRI satellite observations to classify (or determine the history of) any possible contrail cirrus cloud, In fact, the authors seem to claim that such a distinction is not possible, with an example of an ambiguous contrail encounter between 0843 and 0845 UT to demonstrate the current difficulties in discriminating between young contrails and contrail cirrus. Thus, it seems as though the separation of the cloud observations into different types is essentially meaningless. Add on top of that the difficulties in assigning the properties of individual contrails to the entire cloud layer, the overall usefulness of classifying different points of the cloud as contrail, contrail cirrus, and cirrus is minimal.

Although it is clear that new and better definitions of aviation-induced and -influenced ice clouds are necessary, I’m not sure how the authors can proceed to strengthen the paper. Perhaps a more careful study of the numbers and ages of the contrails within a layer may allow for a more useful definition of how much a cirrus layer is influenced by aviation.

Specific comments:

The exposition of the research in the paper is not always easy to follow. For example, it is hard to see the details of the flight path in Figure 2, especially in the blue lines in the lefthand RGB-composite images. The lack of clarity makes it difficult to compare the flight path to the lidar data from Figures 3 and 4. Crucially, the authors never directly inform the reader about the flight path details (including three lidar legs and three in-situ legs) until Figure 6, leading to much confusion for the reader in Section 3.2. Several of the following comments highlight similar difficult-to-follow text.

Line 235 (Figure 2): It is suggested here, but not entirely clear, but have the blue flight segments in Figure 2 been adjusted to account of the 12 minute difference between the nominal satellite time and the actual time of the image acquisition?

Section 3.1: The peaks in backscatter during Leg 2 look like individual contrails. It is not clear how the lidar observations compare with the HALO aircraft flight path. Figure 2 suggests that most of the flight legs are perpendicular to the NAR corridor traffic but some legs around 0800 are parallel to NAR corridor traffic. What direction is HALO flying relative to NAR corridor during Legs 1, 2 and 3 in Figure 3?

Line 311: “properties collected during the three legs.”  Which three legs?  The legs described in Figures 3 and 4? Don’t the authors state that those are WALES measurement legs and thus “can neither be directly inter-compared nor directly compared to in situ observations taken in between”? How is the reader to know that Figure 5 possible, unless the authors tell the reader beforehand that there are 3 lidar legs and 3 in-situ legs?

Figure 6: The reader cannot discern any (young) contrails (blue color) in Figure 6d. I suggest this be removed from the figure. Lines 404 through 406 state that only 1 percent of the observations are (young) contrails.

Line 400: The discussion about number concentration (N) at this point appears muddled. “Ncas occurrences decrease by more than 2-3 orders of magnitude from 0.03 to 0.78-0.84 cm^{-3}.” Shouldn’t this read occurrences increase by more than 2-3 orders of magnitude”?

Section 3.2.1: The discussion in this section implies that most of the contrail cirrus observations are from contrails at least 2 h old. How old are the (young) contrails estimated to be?

Lines 407-409: “We finally remark that MSG/SEVIRI satellite observations are left unused for this classification since the distinction between contrail cirrus and natural cirrus from satellite observations is inherently difficult due to the typical characteristics of young contrails - large N - cannot be measured by passive sensors.” This statement conflates (young) contrails with contrail cirrus, and would thus make all of the previous discussion from Table 1 classifying each cloud type meaningless.

Section 3.2.2: This section is poorly worded and misleading. We are not simply looking at the temporal evolution of cloud properties, but variables changing in time and space. The following sentences explain that and thus contradict the beginning sentence. The description of how the SEVIRI measurements are classified according to the HALO observations is a bit unclear. Given that the SEVIRI and HALO measurements might be displaced by as much as 7.5 min, and the total time of the (young) contrail observations is around 110 s (1 percent of 3 h), it it not surprising that the Reff measurements between HALO and CiPS are not correlated, and that no significant difference between the IOT between contrail snd contrail cirrus was found. Even without the fall streaks, it seems unlikely that the properties of individual (young) contrails can be determined from the SEVIRI data generally.

Lines 453: Most estimates of contrail optical thickness from polar orbiting IR sensors are from clouds at least 2 h old, and thus may not the the young contrails that are implied here. The estimated age of the contrails is not mentioned until line 476 after much discussion about IOT estimates of “contrails”. The terms contrail and contrail cirrus are being mixed together and it is unclear what the authors are talking about in this section.

Lines 478-479: “From the point of view of optical thickness, the entire cloud seems to be homogeneous without remarkable differences among the cloud types defined in Sect. 3.” This statement reinforces the overall lack of utility of the cloud types.

Figure 8: The caption in this figure is not helpful. The cloud properties measured by CiPS are necessarily “at SEVIRI spatial resolution” while the Reff measured by HALO are concurrent and collocated aircraft measurements.

Line 437 (Figure 8): What times is HALO in the northern part of the race track? Why make the reader determine these times on their own from Figure 6, but not the southern part of the race track?

Lines 528-529: “Finally, we jointly assess radii variations of natural cirrus and adjacent contrail cirrus from CiPS with simultaneous HALO measurements for this NAR case.” Please tell the reader that this section refers to Figure 8.

Lines 535: “The temporal variability of CiPS Reff along the flight path…” The authors again appear to neglect that even satellite measurements are functions of both time and space. Simply say “The variability of CiPS Reff…”. Also, say “than that of collocated in situ Reff” instead of “than that of simultaneous in situ Reff”. If the data are averaged over the MSG/SEVERI pixels, they can’t be simultaneous. One quantity is time averaged while the other is not.

Section 4.4 and Figure 13:  Why include this section? Isn’t this redundant because the authors have already compared collocated satellite and HALO observations? Why are the various regional quantities computed until 18 UT when only HALO and SEVIRI observations from 0830 to 1230 UT were presented earlier in the paper? Why does a positive vertical velocity imply “the local downward motion of airmass to warmer temperature layers”. Doesn’t positive vertical velocity mean upward motion of airmasses? 

Typographic errors and other minor issues:

Lines 162-163: “For CTHs larger than approx. 8 km, CTH has an absolute percentage error of 10%, with underestimation for CTH > 10 km at 50° N and overestimation for CTH < 10 km at the same latitude.” What does this mean? That CTH is underestimated when the measured CTH > 10 km but overestimated with the measured CTH < 10 km?

Line 188: I don’t think that “detailly” is a valid word. Perhaps “in detail” would be better here, or simply say that both water and ice clouds are represented in the model (It is assumed that they would be represented realistically as possible by the model.)

Line 199: Change “transit to” to “transition into”.

Figure 2: Time series of contrail cirrus… sounds better than “Temporal variation of contrail cirrus” in the figure title.

In this paper, the authors investigate an important problem, how to distinguish naturally formed cirrus from contrail cirrus. They use a set of HALO measurements from a flight in the ML-CIRRUS campaign to measure in-situ cirrus properties and gases. This is compared to SEVIRI observations and used combined with a radiative transfer model to estimate the radiative properties of the different cirrus types.

This is difficult problem and one of interest to the readers of ACP. The authors have made a good attempt to address this problem, but I would suggest there are some aspects that should be improved before publication.

Main points

The results on this work are based on three transects from a single flight. ML-CIRRUS flew through many contrails during the campaign, why is only this set chosen (and could the results/method be easily expanded to other flights?). It is noted that the control NO threshold varies, but is this simple to generalise? I don't think it has to be set manually.

A related point, but a lot of the statistics are given in counts, but it is not clear what a count is? Is each one an individual contrail, on SEVIRI pixel, or a second of aircraft time? These will all give different results for the accuracy of any method.

The authors spend a considerable amount of time looking at Reff from CIPS. Looking at Strandgren et al (AMT, 2017), it doesn't appear that Reff is validated in that paper. In addition, the comparison to HALO Ref values (Fig. 8c) makes it look like CiPS doesn't have the variability to represent Reff. Does CiPS have the capability (or information) to retrieve Reff?

I am unclear if the extent to which these contrails can or should be considered as temporal evolutions. Fig 8 suggests that they could be a temporal evolution, but around line 300, it is suggested otherwise.

This is more of a style thing, but I found the text could be broken up more (into paragraphs for example) to help the reader. There are several cases were a paragraph spans most of a page (e.g P16), which is too long.

Minor points

L21 - consistency in the ordering of the cloud types would be nice (perhaps throughout)

L163 - This would suggest the CTH is biased towards returning 10km? Does this affect the results?

L193 - I would not start a sentence with 'and'. Libradtran recommends this, I assume that is what you used?

L207 - Presumably this could be checked by looking at the contrail evolution in SEVIRI data

Fig 1. - This should indicate the study region. It is almost coincident with a MODIS overpass, which could be used for a high resolution check of the contrail properties.

L216 - What is the SEVIRI resolution at this location?

L226 - The first use of NAR?

L273 - Do these contrails line up with those observed in SEVIRI? That could give more confidence in the identification

L279 - I might have said the ice supersaturation was 'occasional' - the third flight has almost none (if I am reading Fig. 4 correctly).

L300 - I would make the temporal comparison (or lack of it) clear earlier (maybe in the flight description).

L306 - aircraft

L327 - Grammar. Also, is this expected? Could it be due to errors in the RH retrieval (or reanalysis)?

L333 - The previous sentence just noted that different aircraft might produce different NO amounts.

Eq 1 - Using min would also include an impact of instrument noise. Have you thought about using a different measure, perhaps a statistic/algorithm that can remove outliers instead (e.g. RANSAC) for identifying the background?

Fig. 7 - I like the reduction in aspherical fraction in the contrail region, but is this a consistent effect, or just observed in one case?

Fig. 9 - Given the retrieved Reff has an impact on the optical depth, does the lack of sensitivity to Reff also imply that CiPS is performing poorly when retrieving the IOT? That could potentially explain the difference in optical depths from the expected distribution?

L468 - fast -> quickly

L498 - north

L598 - derived how?

L604 - I was initially skeptical of this, but looking further at CiPS, this doesn't seem so unreasonable. For readers unfamiliar with CiPS, you might want to note that the CiPS retrieval is only dependent on thermal IR channels (which makes it independent of the surface/low cloud properties).

L618 - What is done for these situations? DO they occur often? Does it impact your results?

L635 - Is this likely? Perhaps some indication of windspeed at this time would be useful?

L643 - I don't understand this measure of uncertainty or how it is applied here.

Fig. 13c - Is this vertical velocity relevant?  Can ERA5 simulate the cirrus vertical velocities at the small scale required for ice processes?