Improvements to a long-term Rayleigh-scatter lidar temperature climatology by using an optimal estimation method

Jalali, Ali; Sica, Robert J.; Haefele, Alexander

doi:https://doi.org/10.5194/amt-11-6043-2018

Articles | Volume 11, issue 11

https://doi.org/10.5194/amt-11-6043-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/amt-11-6043-2018

© Author(s) 2018. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 11, issue 11

Research article

|

08 Nov 2018

Research article |

| 08 Nov 2018

Improvements to a long-term Rayleigh-scatter lidar temperature climatology by using an optimal estimation method

Ali Jalali, Robert J. Sica, and Alexander Haefele

Download

Final revised paper (published on 08 Nov 2018)
Preprint (discussion started on 17 May 2018)

Interactive discussion

Status: closed

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

- Printer-friendly version

- Supplement

RC1: 'Comment on "A middle latitude Rayleigh-scatter lidar temperature climatology determined using an optimal estimation method" by Jalali et al.', Anonymous Referee #1, 10 Jun 2018
- AC1: 'Reply to Referee 1', Robert Sica, 27 Jul 2018
RC2: 'Review of “A middle latitude Rayleigh-scatter lidar temperature climatology determined using an optimal estimation method” by Jalali, Sica and Haefele', Anonymous Referee #2, 11 Jun 2018
- AC2: 'Reply to Referee 2', Robert Sica, 27 Jul 2018
RC3: 'Review of the manuscript “A middle latitude Rayleigh-scatter lidar temperature climatology determined using an optimal estimation method” by Jalali et al.', Anonymous Referee #3, 12 Jun 2018
- AC3: 'Reply to Referee 3', Robert Sica, 27 Jul 2018

Peer-review completion

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

AR by Robert Sica on behalf of the Authors (27 Jul 2018) Author's response Manuscript

ED: Reconsider after major revisions (29 Aug 2018) by Gerd Baumgarten

AR by Robert Sica on behalf of the Authors (05 Sep 2018) Manuscript

ED: Referee Nomination & Report Request started (06 Sep 2018) by Gerd Baumgarten

RR by Anonymous Referee #1 (16 Sep 2018)

RR by Anonymous Referee #4 (24 Sep 2018)

Suggestions for revision or reasons for rejection

The paper of describes a comparatively new method for temperature calculations from Rayleigh lidar backscatter measurements, namely making use of the Optimal Estimation Method (OEM). The OEM is applied to a data set of 519 nights, and the seasonal and altitudinal variations of temperatures are derived. The method itself has been described elsewhere before (Sica and Haefele, 2015) as well as a similar temperature climatology based on the conventional Hauchecorne and Chanin method (Argall et al., 2007). The authors claim in the Interactive Discussion that they emphasize on the applicability of the OEM method to a much larger and more variable data set than done by Sica and Haefele. Here the revised version is reviewed.
I think it is a worthwhile approach to apply a new retrieval method to a larger data set and describe the results in a technical paper. Especially, since this technique allows much more detailed descriptions of the error budget of the data set compared to the established method. Nevertheless, I have some major concerns that the present paper fulfills this approach successfully. Before providing some specific (partly minor) comments, I will give some general comments on the structure of the paper, keeping the above mentioned intention of the authors in mind.

General comments:
It is reasonable to repeat the basic content of the OEM and the HC method in Section 2. This allows the reader to get a first overview with more details given in the cited literature, and at least it introduces the data set, abbreviations etc.
Regarding the results, I doubt that starting with the climatology is optimal for the intention of the paper. For the climatology (Fig. 2-4), data are partly averaged, gaps are filled and the whole set is strongly smoothed etc. It remains open how Fig. 3 and 4 contribute to the paper. It is well established that CIRA-86 provides only a coarse approximation of the true temperature field. The geophysical variability has been shown before. It is partly large (often around 9 K), but unfortunately ignored later when small differences between data sets are discussed. My first idea when comparing two large data sets (here the OEM temperatures and the HC temperatures based on the same raw data) are scatter plots and (only) second is plotting differences based on individual events. This allows 1:1 comparisons without being “disturbed” by geophysical effects like seasonal variations and the temperature change with altitude. Unfortunately, the authors concentrate on seasonal variations only, which to some extend looks like the repetition of earlier results.
The Figures 5 – 9 are interesting for the intention of the paper and well described. Even if the statistical uncertainty is typically exceeding the systematic errors, the extensive description of systematic errors and their variability is appreciated.
The authors stress a lot on the extension of the altitude range and the comparison with sodium lidars. I have the feeling that this is somewhat biased towards the OEM. Regarding the altitude coverage, the H&C method often provided results up to 95 km, but is limited to 90 km for the present study (p. 7, l. 31). On the other hand, with the OEM the 90% limit is typically reached around 100 km. Therefore some extend of altitude coverage is acknowledged, and it may depend on situation whether it is 5 km or 10 km. The comparison with sodium lidars revealed differences that are slightly smaller for the OEM than for the HC data set. Nevertheless, the improvement is small and the remaining, larger differences between OEM and Na as well as within the Na data sets need to be examined first. Figures 12-14 are only quickly introduced but hardly discussed. Maybe one should be kept, but the other seem to be redundant and mainly stress the difficulties of the comparison (valid for both Rayleigh methods). With respect to this, I wonder whether solar cycle effects on the temperature above 80 km are acknowledged. Four reasons for differences between the data sets are given, with the most important one unfortunately only in the last sentence of the fourth item: Large scale waves and natural variability. To improve the comparison in this altitude range, I suggest to try to do “closer” comparisons, e.g. with satellite data, instead of using data sets that only partly overlap in time and are separated in space.
In order to have a more concise but better focused paper, I suggest deleting Fig. 3 and 4, as well as two out of Fig. 12-14. Instead, the authors should provide scatter plots with direct OEM-HC comparisons (or other appropriate figures) as well as direct comparisons with other instruments above 90 km.

Specific comments (l. x/y means page x, line y):
l. 2/14-17: I suggest to write “resonance lidars”. Beside sodium lidars there are also powerful temperature lidars using iron or potassium.
l. 2/21-3/3: Here is a long section introducing different climatologies and the seasonal variation of temperatures in the MLT. This erroneously draws the attention of the reader to geophysical problems rather than the intended methodic comparison. If different mid-latitude lidar temperature data sets shall be introduced, I suggest adding Gerding et al., ACP, 2008, because there Rayleigh and resonance data are combined, overcoming the limitations from unknown seed temperatures.
l. 2/31: I assume that the two beam technique is mainly for wind measurements rather than for daylight capability.
l. 7/26: This sentence needs revision.
l. 7/31: I do not understand how (later) data can be compared to Na data above 90 km, if the HC method is limited here to altitudes below 90 km. On the other hand, also Fig. 10 and 11 show data above 90 km. Maybe I do not understand the method here.
l. 9/19: I do not see any white spots in Fig. 4. They seemed to be filled blue (0-3 K) now. Please clarify.
l. 10/2: I suggest to write “temperature variability” instead of “temperature change”.
l. 10/3: I do not see the maximum between 80 and 90 km. The largest variability in this period is observed above 100 km, which maybe should be ignored (but the 80/90% lines are missing). Even then, the variability is generally increasing with altitude (with a lot of temporal variability). You should further repeat which type of waves you expect to see in this plot. But as written above, this figure is not essential for this paper.
l. 16/15: Maybe I missed the point, but please explain why the LLR channel is not used for the HC method.
l. 16/17: Please provide an altitude range for “bottom range of measurements”.
l. 16/22: Please explain how you derived the +/- 0.05 K difference and why merging the data will provide in a larger error.
l. 16/26: Why is HLR data lacking in the winter months?
l. 17/14 and l. 22/26: Please explain why Na lidars are most sensitive between 90 and 100 km. I would expect their best performance around the maximum of the Na layer at 90 km (i.e. at 85-95 km).
l. 18/14-15: I am not sure that I understand this sentence. Do you mean the >10 K difference above 100 km? a) This occurs also in summer, b) this is above the 90% line, c) it is even smaller in Fig. 12 (URB). Reference to AS2007 is confusing, because there the winter mesopause is explicitly not covered (“The winter mesopause is above the top altitude of the PCL climatology“).
Section 4.2: I see the whole comparison with the Na climatologies problematic. If I understand correctly, the Na data are define as “truth” and a smaller deviation of the OEM data compared to the HC data suggests the OEM data being better. To some extend this is OK, even if the larger differences between the Na data sets limit the validity of this approach. Any closer discussion of differences is – from my point of view – problematic, because the natural variability of all data sets needs to be acknowledged first. Day-night differences in Yuan 2008; covered years disagree in all data sets (but overlap); differences due to incomplete sampling, … On the other hand the authors mainly aim for the comparison of the RMR retrievals, i.e. the problems of comparing climatologies are outside of their scope and the reader gets side-tracked.
l. 19/4-5: I do not see that the AS2007 comparison is worse than the OEM-Na comparison shown in Fig. 12-14. Please explain!
l. 22/22: I do not agree that OEM-Na is within Na-Na differences. Often it is larger. Sometimes OEM-uCSU is smaller, but the daytime data of uCSU has an unknown effect.

Hide

RR by Anonymous Referee #2 (24 Sep 2018)

ED: Publish subject to minor revisions (review by editor) (04 Oct 2018) by Gerd Baumgarten

AR by Robert Sica on behalf of the Authors (11 Oct 2018) Manuscript

ED: Publish subject to minor revisions (review by editor) (23 Oct 2018) by Gerd Baumgarten

AR by Robert Sica on behalf of the Authors (23 Oct 2018) Manuscript

ED: Publish subject to technical corrections (24 Oct 2018) by Gerd Baumgarten

AR by Robert Sica on behalf of the Authors (24 Oct 2018) Manuscript

Short summary

We use 16 years of lidar (laser radar) temperature measurements of the middle atmosphere to form a climatology for use in studying atmospheric temperature change using an optimal estimation method (OEM). Using OEM allows us to calculate a complete systematic and random uncertainty budget and allows for an additional 10–15 km in altitude for the measurement to be used, improving our ability to detect atmospheric temperature change up to 100 km of altitude.