Statistical downscaling of water vapour satellite measurements from profiles of tropical ice clouds

Carella, Giulia; Vrac, Mathieu; Brogniez, Hélène; Yiou, Pascal; Chepfer, Hélène

doi:https://doi.org/10.5194/essd-12-1-2020

Articles | Volume 12, issue 1

https://doi.org/10.5194/essd-12-1-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/essd-12-1-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 12, issue 1

Peer-reviewed comment

|

03 Jan 2020

Peer-reviewed comment |

| 03 Jan 2020

Statistical downscaling of water vapour satellite measurements from profiles of tropical ice clouds

Giulia Carella, Mathieu Vrac, Hélène Brogniez, Pascal Yiou, and Hélène Chepfer

Download

Final revised paper (published on 03 Jan 2020)
Preprint (discussion started on 04 Jan 2019)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'review about the article “Statistical downscaling of water vapor satellite measurements from profiles of tropical ice clouds” by G. Carella.', Anonymous Referee #1, 20 Feb 2019
RC2: 'Review of manuscript ESSD-2018-138', Anonymous Referee #2, 07 Mar 2019
AC1: 'Response to reviewers', Giulia Carella, 01 May 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Giulia Carella on behalf of the Authors (04 May 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (12 May 2019) by Giulio G.R. Iovine

RR by Anonymous Referee #1 (13 Jun 2019)

Suggestions for revision or reasons for rejection

Ok, I really have a problem with the answer to my first question. The question was "why did you choose CALIOP?" And the answer is “there is indeed no connection between backscatter at 532 nm and 1064 nm and the water vapour concentration”.

This is followed by a variety of interesting articles as a proof for this connection. The problem: none of them gives proof any good connection between water vapor and a lidar.
Actually, they proof more the opposite.
For example, Jansen (1996, 2001) mainly explain the dehydration during ice cloud formation. But this paper also shows, that you can easily have variations of 100-160% humidity without cloud formation. Such a change in humidity would be visible in a microwave system, but I doubt there would be a signal in the lidar. Because lidar reacts only to scattering, means ice particles.
The other thing that becomes obvious, for example in Jensen et al., 2001: Due to wind advection you can have large changes in ice water content, even if the water vapor mixing ratio is constant (see for example Figure 5 of the reference above). The result: your lidar actually reacts differently, but microwave does not.
Luo & Rossow (2004) is not really relevant here, if you ask me. They looked more into life cycles. And the left part of Fig 14 in that paper proofs actually the opposite again: IWP is changing over time, extra water vapor is changing. So, your Lidar would have a constant signal, WM would see a variation. You could argue, that you observe only steady state (the right part of the Figure). But CALIOP is a polar orbiter, so you would have to justify that somehow.
Eguchi & Shiotani (2004) are focusing scales of Rossby- and Kelvin-waves. Your downscaling is on a kilometer scale. If you look at Fig. 4 and 5 there, then you can see that there is no exact match of cirrus frequency and water vapor mixing ratio. Figure 7 makes it even more obvious. And that is on a 1-degree scale. Is there a reason that you assume it works on a km scale?
Your best argument is Cesana and Chepfer (2013), where you make the case that there is an “indirect correlation” between RH and ice particle (shape, size,…)
However, it is also another good example for the opposite argument: when looking at Figure 13. water vapor and cloud fraction do not overlap

Ok, I am not going through all the other articles, the outcome is similar: None of them show a perfect correlation between water vapor and cloud properties (fraction, particle size, shape, ...).

Don't get me wrong. I think you made a good argument with Cesana and Chepfer. But you have to exploit it and explain it. You have to point out that water vapor is not measured directly by CALIOP. You have to explain that cloud properties (fraction, particle size, thickness, phase) will be the dominant signal in this approach.
And I propose, you have to state this very obvious and in the beginning. It makes sure that people will not use your “water vapour product” to proof a correlation with cloud properties, for example for CALIOP. (People might still do it afterwards, then they cannot be helped.)

My proposal: Include a separate chapter after the introduction. Explain your approach and the uncertainties connected with it in more detail. Explain, that water vapor is not measured directly by the lidar. Explain your assumption more detailed: small scale cloud variations in the cloud (particle size, phase, …) are the result - or lead to - small scale water vapor variations. Cause and effect are a matter of debate here and you can actually find citations for both sides, if you look into it. But they are more in the direction of cloud microphysics papers, not in the large scale dynamics, which you cited before. And last but not least: point out, when this product is not useful (e.g. doing a comparison of water vapour and lidar cloud properties)
You spend a lot of time later to describe your mathematical approach, so one little chapter about the physical idea would add a lot of clarification for possible interested users.

The rest of the article was answered and clarified satisfactory. So, as soon as the “physical interpretation” is clarified, I have no problem with the publication later.

Hide

RR by Anonymous Referee #3 (24 Jun 2019)

Suggestions for revision or reasons for rejection

The primary goal of this paper is to downscale vertical profiles of relative humidity (RH) from the SAPHIR instrument from the native 10-km horizontal resolution to 90-m horizontal resolution. This approach relies on a profile of scattering ratio from the CALIOP lidar that is collocated to SAPHIR then clusters containing tropical ice cloud data only are used to infer ice cloud features within vertical profiles of RH. A statistical methodology is then used to obtain downscaled profiles of RH at very high spatial resolution (90 m).

The science justification for wanting to derive small scale RH structures is strong and additional structure in RH at small scales from satellite data would be very useful for atmospheric research. With that said, I found this paper very difficult to understand at times and why the authors made particular methodological choices without explaining them makes the presentation confusing. Some of the figures are problematic: figs 2 and 8 do not have x-axis scales, fig 5 uses a light shade of yellow that underplays the degree of scatter, and fig 9 needs lots of work to improve and virtually no discussion on the results is included. There are many studies in the literature relevant to this topic that are ignored and would be very helpful for informing the author’s methodologies and validation strategies. Below are some general comments that I hope the authors would consider to improve the clarity of their potentially promising, yet still immature results.

There was no “proof of concept” with independent validation or even something as simple as a statistical summary/ histograms of the small scale features in RH. Figure 9 is really confusing and very little effort is made to explain the final product of downscaled RH. Why not show some summary statistics: mean values, standard deviations, perhaps skewness varies with height? There is enough information in the literature to develop a fuller story to end the paper and make comparisons. Some basic summary statistics are mandatory in this reviewer’s opinion. There is no discussion on how one would interpret a 1 km vertical x 90 m horizontal observation (90 m in one or two dimensions? what about the second dimension across the lidar track? also the small scale patches in this figure looks coarser than 90 m. how do we interpret a highly asymmetric 3d product?) The authors ignored the literature on atmospheric sounding besides a cursory mention of Microwave Limb Sounder data. There was no mention of infrared hyper spectral sounders AIRS, CrIS, and IASI which are quite useful for sounding RH information within thinner cirrus clouds, as they have higher vertical resolution than microwave SAPHIR 183 GHz channels. There are many publications that exploit RH in clear sky and within ice/cirrus clouds and none are mentioned. The distributions of RHi are closer to 100% with lower standard deviations in cirrus compared to clear skies in proximity to cirrus, with much more variable RHi and more frequent supersaturation. To add to this, there are also many papers on in situ aircraft observations of RHi distributions in cirrus/clear sky and the scale dependence of whether temperature or water vapor controls the RHi distributions is inferred from HIPPO aircraft observations. No separate error propagation as a function of scale for temperature and specific humidity is attempted. There are no cited papers on power law relationships that highlight regime and height and cloud/clear sky dependence of variability as a function of scale, and the existence of scale breaks in the exponents and how that impacts the small scale RH. How does the statistical methodology used in this paper deal with known scale dependent variability in temperature, specific humidity, and cloud geometrical and microphysical properties? Does the scattering ratio contain enough information for all of these relevant geophysical factors? If the authors made plots/summary statistics of RH at different scales and compared to the published literature, that would go a long way towards proving the utility of this approach. Also, wouldn’t better process-relevant covariances be obtained from a profile by profile matching of SR to SAPHIR? It is not well explained why clustering was used (besides the issues of “mixed phase" brought up in a response to a reviewer, but for clouds colder than 240 K or so there should be very few of these clouds).

Hide

RR by Anonymous Referee #4 (29 Jun 2019)

RR by Anonymous Referee #5 (02 Jul 2019)

Suggestions for revision or reasons for rejection

Overall this manuscript presents an interesting analysis. My main concern is I'm not completely sure how to interpret the fidelity of the downscaling without some additional independent validation with high resolution measurements. The authors should be more clear on this point.

Page 1, Line 12: Should be “will help”

Page 2, line 16: I’m not sure what “surface-blind” means in this context.

Page 2, line 31: “clouds” -> “cloud”

Page 3, line 9: “backscattered” -> “backscatter”

Page 3, line 12: delete “to”

Page 4, line 1: “spaceborn” -> “spaceborne”

Page 4, line 12: “to predict the (downscaled) water-vapour vertical structure using cloud profiles only” This statement is confusing since later in this paragraph the authors state that they are looking horizontally and that the current study is not considering the vertical heterogeneity. Please clarify the wording.

Page 14, line 9: Several studies have looked at biases in reanalysis upper tropospheric humidity and should also be cited here:

Davis, S. M., Hegglin, M. I., Fujiwara, M., Dragani, R., Harada, Y., KOBAYASHI, C., Long, C., Manney, G. L., Nash, E. R., Potter, G. L., Tegtmeier, S., Wang, T., Wargan, K. and Wright, J. S.: Assessment of upper tropospheric and stratospheric water vapor and ozone in reanalyses as part of S-RIP, Atmospheric Chemistry and Physics, 17(20), 12743–12778, doi:10.5194/acp-17-12743-2017, 2017.

Jiang, J. H., Su, H., Zhai, C., Wu, L., Minschwaner, K., Molod, A. M. and Tompkins, A. M.: An assessment of upper troposphere and lower stratosphere water vapor in MERRA, MERRA2, and ECMWF reanalyses using Aura MLS observations, J Geophys Res-Atmos, 120(22), 11,468–11,485, doi:10.1002/2015JD023752, 2015.

Page 14, line 23: “tow” -> “two”

Hide

RR by Anonymous Referee #6 (18 Jul 2019)

ED: Reconsider after major revisions (23 Jul 2019) by Giulio G.R. Iovine

AR by Giulia Carella on behalf of the Authors (17 Sep 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (25 Sep 2019) by Giulio G.R. Iovine

RR by Anonymous Referee #3 (02 Oct 2019)

RR by Anonymous Referee #5 (08 Oct 2019)

RR by Anonymous Referee #1 (10 Oct 2019)

Suggestions for revision or reasons for rejection

The authors did a pretty good job in answering my comments from the last iteration. The additional chapter “Physical approach and related limitations” describes very nicely the caveats of the approach. So, even if I might have still my doubts about the product, I cannot say that the authors are not honest about it.
I assume the data will get released sooner or later anyway. Then I would rather have this paper published, to make sure that users understand what they are dealing with.

So, I would propose to consider this article for publication.

I have just a few technical comments:

Abstract:
Add one more sentence after the second sentence (I copied it from chapter 3):
“Here, we present a method to downscale observations of relative humidity available from the passive microwave sounder SAPHIR at a nominal horizontal resolution of 10 km to the finer resolution of 90 m using scattering ratio profiles from the lidar CALIPSO. The small scale water vapour is hereby deduced from the indirect physical correlation between RH and the lidar observations.”

Cut the next sentence in two pieces to make it more readable:
“With the scattering ratio 5 profiles as covariates, an iterative approach applied to a non-parametric regression model based on Quantile Random Forest is used. It allows to effectively incorporate the high-resolution variability from cloud profiles into the predicted relative humidity structure”

Chapter 3:
“This work is rooted on the following physics”
Replace it by something like “We use the following assumption in our approach”

Same chapter, page 7, line 6:
There is a “d” too much : Correct to “To avoid the misuse”

Same page, further down - line 9:
I would replace “relevant” with “applicable”

I think, that should do it.
Regards,
-Mathias

Hide

ED: Publish subject to minor revisions (review by editor) (29 Oct 2019) by Giulio G.R. Iovine

AR by Giulia Carella on behalf of the Authors (07 Nov 2019)

ED: Publish as is (26 Nov 2019) by Giulio G.R. Iovine

AR by Giulia Carella on behalf of the Authors (01 Dec 2019) Author's response Manuscript

Short summary

Observations of relative humidity for ice clouds over the tropical oceans from a passive microwave sounder are downscaled by incorporating the high-resolution variability derived from simultaneous co-located cloud profiles from a lidar. By providing a method to generate pseudo-observations of relative humidity at high spatial resolution, this work will help revisit some of the current key barriers in atmospheric science.