In this paper the authors convincingly demonstrate the capabilities of the IPDA-lidar technique in terms of measuring column-average concentrations of CO2 under various weather-conditions. Not only do they demonstrate that their air-borne ASCENDS demonstrator is capable of accurately measuring XCO2 over clouds, but at the same time also the possibility to obtain partial columns in case of broken cloud layers, which allows them to even obtain quasi range resolved measurements.

The paper is well written and the results/claims are illustrated/supported by informative plots and illustrations. I see no obvious major issues with the contents and the presentation and therefore recommend publication after considering some minor comments that are listed in the following.

General remarks:

p. 6, l. 5-8: It is not totally clear what the authors want to state with this sentence. It is intuitive to assume that the SNR increases with more averaging, but not obvious why at the same time noise is supposedly increased by lidar range variations. I can imagine a worse resolution and potentially also a measurement bias. Also, the reference given in the text did not help me understanding the statement that is made here. Maybe this can be reformulated to make it clearer.

Comments regarding language:

p. 2, l. 23/24: This paper describes this lidar XCO2 measurements made to cloud tops during the summer 2017 ASCENDS/ABoVE (…) campaign … -> … describes the lidar …?

p. 7, l. 23: below DC-8 aircraft -> below the DC-8 aircraft?

Referee comments on "Airborne Lidar Measurements of Atmospheric CO2 Column Concentrations to Cloud Tops made during the 2017 ASCENDS/ABoVE Campaign" by Jianping Mao, James B. Abshire, S. Randy Kawa, Xiaoli Sun, and Haris Riris

The authors have been measuring column-integrated CO₂ concentrations using an aircraft-based lidar for a number of years now.  In this paper, they report on results obtained on flights made in the summer of 2017 as part of an ASCENDS/ABoVE campaign.  The remote measurements made by the lidar are compared to the equivalent quantity derived from in situ CO₂ measurements taken from nearby vertical profiles made by the aircraft spiraling down to the surface and back, weighted appropriately in the vertical.  The lidar CO₂ estimates are made not only using the signal bounced off of the Earth's surface, but also using the signal bounced off the tops of clouds at various heights; they note that such above-cloud CO₂ measurements could result in a large increase in data volume, should an equivalent instrument be flown on a satellite, compared to passive satellite sensors, which cannot make useful CO₂ measurements in cloudy conditions (due to not having a ranging capability to know the path length of the signal accurately).  The accuracy of the CO₂ estimates was 2-3 times better than on previous flights, due to the use of a wavelength step-locked laser diode source and a high-efficiency detector in this campaign.  The random-error uncertainties on CO₂ estimated over clouds are now down in the 0.6 to 0.9 ppm range, when integrated over 10-second spans, a level that would make above-cloud CO₂ estimates useful, especially if achieved on-orbit.

This is important work that should be published in this journal, after the points noted below are addressed.  The main point of concern is that the definition of the column-integrated quantity of interest -- "XCO₂" -- appears to be different here than from how it is normally used.  When the term "XCO₂" is used in the satellite CO₂ measurement community, or by those researchers who use Earth-based Fourier transform spectrometers (e.g. as part of the TCCON and COCCON networks) who validate such measurements, it is defined as the mass-weighted dry-air mole fraction: moles of CO₂ per moles of dry air.  The contributions to the column integral come equally by mass from all layers of the atmosphere.  The sensitivity of the instrument to the different layers (as defined by the averaging kernel vector) might vary, but contributions from an a priori estimate of the CO₂ profile are used to make up the difference: if the instrument has low sensitivity to CO₂ in the upper part of the column (values of the averaging kernel vector close to zero), then most of the information on CO₂ for that part of the column integral is taken from the prior; conversely, if the opposite is true near the surface (values of the averaging kernel vector close to one), then little information needs to be taken from the prior.  Here, however, the column integral addressed does not appear to have equal weight (by mass) from each layer of the atmosphere, when computing the average.  Instead, it appears that the weight of each layer is taken directly from the averaging kernel vector, normalized so that the total across all layers sums up to one, for whatever portion of the total column is being measured.  No contribution from an a priori CO₂ profile is used to make the column-integral mass-weighted.  There is nothing wrong with using this flavor of vertical weighting to present the results here.  However, if the authors want to assign the name "XCO₂" to it, they must at minimum note how their definition differs from the one commonly used.  In my opinion, it would be better to use some other symbol to denote the vertical weighting used here.  [If I am wrong and the authors are actually using a flat mass-weighted vertical integral, then they must do a better job of describing their method in the manuscript, so that the reader can understand what they are doing.]  The second point that the authors should address is how they remove the effect of water vapor from their column integral.  Since XCO₂ is usually defined using the mass of DRY air (water vapor removed), then the contribution of water vapor to the mass of each layer, when doing the vertical weighting, must be removed.  The authors should address this in their method discussion.  If they do not remove the effects of water vapor, then they should note this when they define the vertical integral that they are using (i.e. what they call "XCO₂" at the moment).  This point is repeated below, to some extent, in the detailed comments.

Detailed comments:

page 2, Section 2:   The authors need to do a better job defining the vertical weighting that goes into "XCO₂": if it is indeed different from the usual definition (see comments above), the authors need to be clear about this; they may want to consider using a different name to describe their quantity, to avoid confusion.  In particular, the authors appear to use a vertical weighting proportional to the averaging kernel vector, instead of the flat mass-weighting usually used for XCO₂.

page 3, Line 18: It was not clear to me how "using a 15-m vertical layer thickness" relates to improving the SNR.  Maybe add some more detail?

p3 L47: "of atmosphere": What do you mean by this, exactly?  Atmospheric temperature, pressure, and water vapor?  Some subset of these?  Or additional variables, as well?  It would be clearer to spell this out more precisely.

p5 -- the definition of "XCO₂" and the use of the averaging kernel (AK) is not clear.  It appears that the AK defines the vertical weighting that is used in the average, and XCO₂ is a non-flat-with-pressure vertical average.  The assumed a priori CO₂ profile is not used in the calculation of the vertical average.  All this is fine, but it needs to be explicitly described, and the difference between the definition of XCO₂ here and elsewhere needs to be clarified.  If XCO₂ is indeed its usual flat-with-pressure definition, then we need to know the a priori CO₂ profile used.  Also, how is the effect of water vapor removed, if XCO₂ is versus dry air?

p7 L1-4:  These lines could be reworded to make the point clearer.  Maybe something like: "In addition, variability in the elevation of the cloud tops may degrade the accuracy of the XCO₂ retrievals. The cloud top altitude used to calculate the photon path-length in the retrieval is taken as the centroid of the cloud top altitudes calculated from time-averaged lidar range measurements.  These range measurements are not taken at the same time as the measurements across the center of the line, and are also averaged across a longer time (1 second); the difference in range that results causes an additional error in the XCO₂ retrievals that causes greater noise (Mao et al., 2018)."  I am not sure if I have captured the key point -- if not, please reword to make the idea clearer.

p8 L22-24: "This small 1.3 ppm difference between these two sets of measurements indicates slightly lower carbon below clouds likely caused by a small local CO₂ drawdown during summertime."  Actually, CO₂ at these Arctic latitudes is mainly dominated by the signal of terrestrial carbon and release from further south being transported up north.  I.e. it is not a local signal.  For example, at the Barrow site, the large seasonal variability there is not due to a strong local terrestrial signal, but rather to the strong seasonal terrestrial at lower latitudes in the northern hemisphere being transported to the north by winds and mixing.  Please reword to reflect this.

Section 4.3:  If these low-cloud measurements can be used to isolate CO₂ in the layer under the clouds, why have you not tried to use your measurements to calculate this?  I imagine this is because the large positive CO₂ values measured by the in situ sensor just close to the surface make assigning an average value for the thicker layer under the clouds less meaningful, but it would be useful for you to note what your reasons were.

p10 L13-15: "For future space-based lidar measurements, the higher orbit altitude and longer atmospheric path length to cirrus clouds should allow useful XCO₂ measurements to cirrus clouds as well."   It is not the orbit altitude or the path length that matters here, but rather the mass of the atmosphere (since CO₂ is well-mixed) between the top of the atmosphere and the 10 km flight referenced here.  For the Arctic, where the tropopause is low, that atmospheric mass is small, so that there might not be much of an improvement between what is seen from the satellite and these flights.  Therefore, I am not sure that the assertion that useful measurements above cirrus should be possible from the satellites is valid, at least not in the Arctic, where there is not much atmospheric mass above 10 km.

Editorial comments:

page 1, line 12:  add commas on either side of this phrase: "as well as to the ground"

p2 L5-6: add commas on either side of this phrase: "in addition to those to the ground"

p2 L23: for clarity, replace "this lidar" with either "this lidar's" or "these lidar"

p2 L30: why is "Sounder" capitalized?  Is it the official name of the instrument?

p3 L47: add a comma before "since"

p4 L12: add a comma after "Alaska"

p4 L25: add a comma after "conditions"

p7 L13: remove the period after "Fairbanks"

p8 L27: change to "provide"

p8 L32: add comma after "deviation"

p8 L36: add comma after "Canada"; remove "the" at the end of the line

p9 L9:  add comma after "10"

p10 L13: add comma after "short"

p10 L17: change "Inter Tropical" to "Intertropical" or "Inter-Tropical"?

p10 L21-22: put commas on either side of "along with those to the ground"

p10 L26: change the period after "2018)" to a comma

p10 L27: add "and" before "flux"

p10 L29: change "is" to "are"

p11 L3: "conflicts"

p11 L7: add a comma before "as well as"

p11 L17: "of" not "Of"

p21 L10: "maneuvers"

p23 L5: "left" not "Left"

p24 L6: add comma after "ground"

p25 L10: add comma after "right"

p26 L4&5: "left" not "Left"

References: put the journal names in italic script, and the volume numbers in bold?

Whole document: if the journal style permits, put Latin phrases like "in situ"

and "a priori" in italic script.

Whole document: remove the dash sign between number and unit?