the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Oct 2021
Atmospheric tomography using the Nordic Meteor Radar Cluster and Chilean Observation Network De Meteor Radars: network details and 3D-Var retrieval
Alexander Kozlovsky
Zishun Qiao
Masaki Tsutsumi
Chris Hall
Satonori Nozawa
Mark Lester
Evgenia Belova
Johan Kero
Patrick J. Espy
Robert E. Hibbins
Nicholas Mitchell
Download
- Final revised paper (published on 08 Oct 2021)
- Preprint (discussion started on 17 May 2021)
Interactive discussion
Status: closed
-
CC1: 'Comment on amt-2021-124: On vertical velocity estimates', Jorge Luis Chau, 26 May 2021
The authors present an interesting technique to study gravity waves (GW) scales in the mesosphere and lower thermosphere (MLT) using multistatic meteor radar geometries, either by combining closely located monostatic systems or using additional receivers to existing monostatic systems. Given that the concept is relatively new and has a lot of potential, this and other techniques that resolve wind fields in smaller spatial scales are needed to explore MLT GW scales.
As the authors mention in the paper, usually meteor radars have not fitted for the vertical velocity. They present interesting results using monostatic systems by using a regularization scheme (e.g., spatio-temporal Laplace filter). The results are indeed interesting, since my colleagues and I do obtain larger variability on the residuals (or apparent vertical velocity) when we apply a gradient method (Chau et al., 2021, Conte et al., 2021). Note that the variability without fitting for gradients would be even larger due to horizontal divergence contamination as clearly shown in Chau et al. (2017). I tend to attribute the large difference on the distribution of residuals to the regularization the authors are using.
However, in lines 169-175, it appears that the authors attribute the estimated vertical velocities by Chau et al. (2021) and Conte et al. (2021) to not implementing the WGS84 geometry,
"The implementation of the WGS84 Earth geometry turned out to be important for all meteor wind analysis and essentially reduces the projection and altitude biases of the observed meteors. In particular, at mid- and high latitudes these corrections are rather significant and lead to substantial improvements of the wind as well as the vertical profile of ambipolar diffusion and, thus, has an impact on the temperature estimates (Hocking, 1999; Hocking et al., 2001). Furthermore, the vertical wind mean bias and variances are significantly reduced compared to other studies (Egito et al., 2016; Chau et al., 2021; Conte et al.,2021) without including additional damping terms or regularization constraints to the fitted vertical wind."
Could you please provide evidence that this is the case? For example, if you do not apply WGS84 the resulting vertical velocities are comparable to those reported by Chau et al. (2021) (at low latitudes) or Conte et al. (2021)? Although we are using WGS84 and the proper geometrical considerations (e.g., Bragg vector with respect to the local geographic coordinates), it would be good to know if the distributions of residuals in Figure 2, that are much narrower than reported in other works, is due to implementing WGS84 "without any damping terms or regularization constraints to the fitted vertical wind" or due to the regularization explained in lines 125-140.
In the case of vertical velocity estimates obtained as part of 3D wind field estimates (e.g., Figure 13), I do expect them to be less contaminated by horizontal velocity spatial inhomogeneities, since these inhomogeneities are resolved (or better resolved). The control simulations of one of my students (paper to be submitted) indicate that these horizontal velocity spatial inhomogeneities are the source of the observed variability on vertical velocity estimates (i.e., provide apparent vertical wind). However, they are not responsible for the relative high values lasting for a few hours above 95 km.
Again, I want to reiterate that this additional technique to resolve wind fields is interesting and needed to attack MLT GW scales.
Koki Chau
References
- Chau, J. L., G. Stober, C. Hall, M. Tsutsumi, F. I. Laskar, and P. Hoffmann, Polar mesospheric horizontal divergence and relative vorticity measurements using multiple specular meteor radars, Radio Sci., 52, doi:10.1002/2016RS006225, 2017.
- Chau, J. L., J. M. Urco, J. Vierinen, B. J. Harding, M. Clahsen, N. Pfeffer, K. M. Kuyeng, M. A. Milla, P. J. Erickson, Multistatic specular meteor radar network in Peru: System description and initial results, Earth Space Sciences, 7, https://doi.org/1029/2020EA001293, 2021.
- Conte, J. F., J. L. Chau, J. M. Urco, R. Latteck, J. Vierinen, and J. O. Salvador, First studies of mesosphere and lower thermosphere dynamics using a multistatic specular meteor radar network over southern Patagonia, Earth and Space Science, 8, e2020EA001356. https://doi.org/10.1029/2020EA001356, 2021.
-
AC1: 'Reply on CC1', Gunter Stober, 28 May 2021
We thank Jorge L. Chau for his public comment on the 3DVAR retrieval.
Vertical velocities are an important physical quantity to measure and often part of scientific disputes. In the case of meteor radar observations, it was known since the first systems were deployed that the wide field of view and least square fitting techniques won’t provide reliable results for various reasons. For the analysis of reliable vertical winds, there is nothing unimportant e.g., phase calibration, antenna positions, surface motions (frost), mutual antenna coupling, etc. they all play a role and are nearly impossible to handle. Retrievals that offer a mathematically elegant way involving a priori knowledge to reduce the impact of all these quantities by giving preference to solutions that are physical more likely, which also provides an opportunity to reduce biases that are intrinsic to the least square fitting for certain mathematical problems.
The results presented herein are obtained by applying advanced mathematics with non-linear error propagation, a spatio-temporal Laplace filter, and WGS84 geometry. All three elements provide key contributions to the presented results. The geometry is important for the non-linear error propagation, which defines weights for the spatio-temporal filter between the iterations (max 8). Thus, we use complicated cost functions with several different weights and only the WGS84 geometry is always the same. Removing the WGS84 geometry certainly degrades the solutions, which is demonstrated in Figure 2 (Stober et al., 2018, AMT). Due to the radial nature of the observations any change of the horizontal winds also impacts the vertical wind solution (eq. 1, this paper). However, we did not explicitly evaluate how much each part contributes to the final solution for the meteor radar, but we did perform such a comparison for the Middle Atmosphere Alomar System in Gudadze et al., 2019 and there it turned out that the vertical velocities become even smaller when considering a WGS84 geometry compared to the case without.
There is no doubt that the WGS84 geometry is important in retrievals to obtain the results, but it is certainly not the only factor. However, we are confused about the statement in the public comment that Chau et al. 2021, and Conte et al, 2021 include the WGS84 geometry. We carefully read the papers and found no statement or reference or citation that suggests that such an important aspect is included. Neither is a WGS84 geometry presented in Chau et al., 2017, which uses spherical coordinates for the volume velocity processing (VVP, Waldteufel and Corbin, 1979). We also found no statement in Egito et al., 2016 about the implemented geometry for the fitting. Due to the missing information about the geometry, we assumed that this is the likely cause for the substantial deviations.
The second aspect in regards to the comment “without including additional damping terms or regularization constraints” which presumably refers to previous retrievals (Stober et al., 2017, Wilhelm et al., 2017), which included a strong damping of the solution towards w=0 m/s, which is comparable to a damped least square. However, this part was removed from the retrievals in this work. There is also no regularization included to cure a potential rank deficiency for the monostatic case. Nethertheless, it is true that our cost function tends to have a preference to smaller norms (giving preference to smaller errors for the vertical velocity), which appears to be related to small values of w, although a small value of w is not required and large values that are statistically significant should be still retrievable. The Sodankyla meteor radar indicates a certain skewness towards larger positive values, which are 3 times larger compared to Kiruna. This approach is also described for the 3DVAR retrieval, but with a different implementation. Currently, we are working on new retrievals with even more complicated cost functions to further improve this aspect and to perform certain benchmark tests.
Furthermore, we have to note that we were not able to retrieve vertical velocities with a smaller statistical uncertainty compared to its absolute value, which suggests that the effective measurement response for an instantaneous vertical measurement is negligible and no statistically significant values can be obtained. This led us to the conclusion to refer to these values as residual bias.
More importantly, we tested whether the retrievals are meaningful concerning existing model physics. Therefore, we performed an initial comparison of the meteor radar data using the UA-ICON model and simulated a meteor radar beam to extract vertical velocities. A comparison of the horizontal winds is presented in Stober et al., 2021 (https://acp.copernicus.org/preprints/acp-2021-142/). The UA-ICON run is published in Borchert et al., 2019 (https://gmd.copernicus.org/articles/12/3541/2019/). The histograms presented in the supplement Figure below are obtained using all available data from the meteor radars at Sodankylä and Kiruna and the complete UA-ICON run to ensure a sufficient statistics.
Although the agreement is impressively good, we don’t want to claim that our retrievals are representing ‘true’ vertical velocities. Our retrievals appear to provide statistically sound solutions to the meteor radar data, but certainly if one of the mathematical aspects turns out to be not applicable or not generally true, this might change. However, we are glad that our paper triggered such a good discussion and brought that much attention to the topic.
References:
Wilhelm, S., Stober, G., and Chau, J. L.: A comparison of 11-year mesospheric and lower thermospheric winds determined by meteor and MF radar at 69 ° N, Ann. Geophys., 35, 893–906, https://doi.org/10.5194/angeo-35-893-2017, 2017.
Stober, G., Matthias, V., Jacobi, C., Wilhelm, S., Höffner, J., and Chau, J. L.: Exceptionally strong summer-like zonal wind reversal in the upper mesosphere during winter 2015/16, Ann. Geophys., 35, 711–720, https://doi.org/10.5194/angeo-35-711-2017, 2017.
Stober, G., Chau, J. L., Vierinen, J., Jacobi, C., and Wilhelm, S.: Retrieving horizontally resolved wind fields using multi-static meteor radar observations, Atmos. Meas. Tech., 11, 4891–4907, https://doi.org/10.5194/amt-11-4891-2018, 2018.
Gudadze, N., Stober, G., and Chau, J. L.: Can VHF radars at polar latitudes measure mean vertical winds in the presence of PMSE?, Atmos. Chem. Phys., 19, 4485–4497, https://doi.org/10.5194/acp-19-4485-2019, 2019.
Borchert, S., Zhou, G., Baldauf, M., Schmidt, H., Zängl, G., and Reinert, D.: The upper-atmosphere extension of the ICON general circulation model (version: ua-icon-1.0), Geosci. Model Dev., 12, 3541–3569, https://doi.org/10.5194/gmd-12-3541-2019, 2019.
Stober, G., Kuchar, A., Pokhotelov, D., Liu, H., Liu, H.-L., Schmidt, H., Jacobi, C., Baumgarten, K., Brown, P., Janches, D., Murphy, D., Kozlovsky, A., Lester, M., Belova, E., and Kero, J.: Interhemispheric differences of mesosphere/lower thermosphere winds and tides investigated from three whole atmosphere models and meteor radar observations, Atmos. Chem. Phys. Discuss.
-
CC2: 'Reply on AC1', Jorge Luis Chau, 28 May 2021
Thanks for explaining and clarifying what is shown as vertical velocity estimates in the paper. It is clear now that the statement
"Furthermore, the vertical wind mean bias and variances are significantly reduced compared to other studies (Egito et al., 2016; Chau et al., 2021; Conte et al.,2021) without including additional damping terms or regularization constraints to the fitted vertical wind."
is a guess, since in the case of meteor radar estimates the contribution of not including WGS84 geometry has not been separated from the other considerations described (e.g., advanced complicated mathematics, a spatio-temporal Laplace filter, preference to solutions that are physical more likely, etc.).
We indeed used WGS84 geometry in our estimates. Although not explicitly mentioned in Chau et al. (2021) and Conte et al. (2021) since we considered trivial after our previous works including Stober et al. (2018), here is some implicit information to consider:
- a) WGS84 is explicitly mentioned in Chau and Clahsen (2019) which is referenced in Chau et al. (2021).
- b) In Conte et al. (2021) after EQ 1 it is written "where u = (u, v, w) is the neutral wind vector, with u, v, and w being its zonal (east-west), meridional (north- south), and vertical (up-down) components, respectively. k = (ku, kv, kw) is the Bragg wave vector (scattered minus incident) in the meteor-centered east-north-up coordinate system", i.e., we use local ENU coordinates.
The gradient method is implemented in spherical coordinate systems, but using the local distance to the center of the earth (local radius) calculated using WGS84. Bragg vectors are all calculated respect to the ENU of each meteor.
In our works, besides WGS84 we also consider error propagation and geometrical aspects, however we have preferred to avoid regularizing the data with expected physical values for the vertical wind. Instead we focus on understanding what we get. As I mentioned before 3D wind field analysis like yours are very welcome since, among other aspects, they would help reduce the observed variability of vertical velocity estimates.
As in your case, we also do not claim we are getting the true vertical winds, and definitely our philosophies to attack the interesting topic of MLT vertical winds are different. I find these differences positive since they enrich our research activities.
Looking forward to your published paper.
Koki Chau
References
- Chau, J. L., and M. Clahsen, Empirical phase calibration for multi-static specular meteor radars using a beam-forming approach and arbitrary interferometric configurations, Radio Sci., https://doi.org/10.1029/2018RS006741, 2019.
- Chau, J. L., J. M. Urco, J. Vierinen, B. J. Harding, M. Clahsen, N. Pfeffer, K. M. Kuyeng, M. A. Milla, P. J. Erickson, Multistatic specular meteor radar network in Peru: System description and initial results, Earth Space Sciences, 7, https://doi.org/1029/2020EA001293, 2021.
- Conte, J. F., J. L. Chau, J. M. Urco, R. Latteck, J. Vierinen, and J. O. Salvador, First studies of mesosphere and lower thermosphere dynamics using a multistatic specular meteor radar network over southern Patagonia, Earth and Space Science, 8, e2020EA001356. https://doi.org/10.1029/2020EA001356, 2021.
- Stober, G., J. L. Chau, J. Vierinen, C. Jacobi, and S. Wilhelm, Retrieving horizontally resolved wind fields using multi-static meteor radar observations, Atmos. Meas. Tech., 11, 4891-4907 https://doi.org/10.5194/amt-11-4891-2018, 2018.
- AC2: 'Reply on CC2', Gunter Stober, 28 May 2021
-
CC2: 'Reply on AC1', Jorge Luis Chau, 28 May 2021
-
RC1: 'Comment on amt-2021-124', Anonymous Referee #1, 21 Jul 2021
Review of "Atmospheric tomography using the Nordic Meteor Radar Cluster and Chilean Observation Network De Meteor Radars: network details and 3DVAR retrieval" by Stober et al.
The paper introduces a new technique to derive 3D wind fields from networks of meteor radar stations in a tomographic approach. The performance of this 3DVAR retrieval is demonstrated for two meteor radar networks, the Nordic Meteor Radar Cluster in Northern Europe, and the CONDOR network in South America.
Based on several observed events, the benefits of this approach, and the different characteristics of the two radar networks are discussed. As a diagnostic parameter, the Shannon information content is derived. It is found that, as a consequence of its linear arrangement, the CONDOR network is more sensitive to meridional winds over the line connecting the stations, and more sensitive to zonal winds in two patches of enhanced sensitivity parallel to this line.
It is shown that both radar networks are capable to resolve the counter-rotating vortices of breaking gravity wave events that are important for the excitation of secondary waves, which is currently a hot topic in atmospheric dynamics. Horizontal wavelength spectra are derived, and the impact of a minor sudden stratospheric warming in December 2019 is investigated in a keogram analysis.Overall, the paper is very well written and fits in the scope of AMT.
Publication of the paper in AMT is therefore recommended after addressing my minor comments.
MAIN COMMENTS:(1) The difference between the absolute wind speeds in Fig.4, left, and Fig.4, right, should be calculated and discussed. This difference can be used as further diagnostics and measure of errors.
(2) The Data Availability section of the paper is missing.
SPECIFIC COMMENTS:(1) l.29: The reference de Wit et al. (2017) should be included here as an earlier reference for the occurrence of secondary gravity waves over South America.
de Wit, R. J., D. Janches, D. C. Fritts, R. G. Stockwell, and L. Coy (2017),
Unexpected climatological behavior of MLT gravity wave momentum ï¬ux in the lee of the Southern Andes hot spot,
Geophys. Res. Lett., 44, 1182-1191, doi:10.1002/2016GL072311.
(2) l.37: There are also imaging satellite instruments that provide spatially resolved 2D, or even 3D observations of gravity waves. These should be mentioned here:Randall, C. E., et al. (2017),
New AIM/CIPS global observations of gravity waves near 50-55 km,
Geophys. Res. Lett., 44, 7044-7052, doi:10.1002/2017GL073943.Ern, M., L. Hoffmann, and P. Preusse (2017),
Directional gravity wave momentum ï¬uxes in the stratosphere derived from high-resolution AIRS temperature data,
Geophys. Res. Lett., 44, 475-485, doi:10.1002/2016GL072007.
(3) Caption of Fig.4 is incomplete. Suggestion:
...Cluster and a Cartesian geographic grid.
->
...Cluster using a Cartesian grid (left) and a longitude/latitude geographic grid (right).
(4) Fig.4:
Please show in an additional panel the wind strength difference between the two representations, and add some discussion. This would give the reader an impression of the robustness of the results. At least in the regions of high measurement content the differences should be small.
(5) Fig.5: Please comment!
Is this a typical event, or a particularly strong event?
(6) l.354: Please add this information:
Do you assume vertical wind to be zero for the data assimilation mode?
(7) Caption of Fig.8 does not match the figure!
Shown is the measurement response, not the wind fields.
(8) l.408/409: Here you state that gravity wave activity would be enhanced at 69-70N. Please be more specific!
Does this statement refer to Fig.10 where stronger variability is seen in the two panels on the right hand side?
However, the left two panels are domain-averages and should therefore be much smoother, anyhow.
Or does this refer to Fig.11?
In Fig.11 the semidiurnal tide is the strongest mode of variability, and other fluctuations are difficult to see. Could you therefore provide some more guidance to the reader where exactly one should see this effect?
(9) Fig.14: Please add information!
The wind fields are very different. Please state whether this is an effect of the semidiurnal tide.
(10) Fig.14: Question: are the "92km" for the upper left two panels wrong?
Should it read "90km"?
(11) The "Data availability" section is missing!
TECHNICAL COMMENTS:caption of Table 1: "ALO" does not belong to the Nordic meteor radars.
Nordic meteor radars -> Nordic and ALO meteor radars
l.251: to included -> to include
l.418: periodigrams -> periodograms ???
(periodigrams is rarely used)
l.486: Peruian -> Peruvian-
AC3: 'Reply on RC1', Gunter Stober, 11 Aug 2021
We thank the reviewer for assessing the submitted publication on the 3DVAR retrieval algorithm and the constructive suggestions to improve the manuscript. The final revision will include a point-by-point reply to each raised concern.
Data availability:
The 3DVAR retrievals are implemented to permit routine operation for both radar networks. The data shown in this paper is available upon request through Alexander Kozlovsky at SGO for the Nordic Meteor Radar Cluster and from Alan Liu for the CONDOR network. There was a plan to later upload these files to the ARISE database. However, current funding limitations have delayed this project.Difference plots:
Difference plots between both retrieval grids shown in Figure 4 are doable. However, the interpolation to a common grid introduces additional errors. We will rerun both domains to ensure and diagnose whether different meteors enter the retrieval at the domain boundary (this is certainly the case at Sodankyla), which could introduce additional systematic offsets. The revised manuscript will include a figure illustrating these differences.Reference suggestions:
We thank the reviewer for pointing at additional references, which are going to be included in the proposed paragraphs in the manuscript.Other comments:
There are a couple of other suggestions mentioned by the reviewer, which we will consider as proposed in the revision. We are going to expand the discussions and add more details to explain the Figures. However, we have not yet performed a systematic seasonal analysis for all the different events. Our main intention of the submitted manuscript is the retrieve and demonstrate that various meteorological events can be found and analyzed from the data. We are also grateful for correcting typos and other mistakes in the figures and figure captions.
-
AC3: 'Reply on RC1', Gunter Stober, 11 Aug 2021
-
RC2: 'Comment on amt-2021-124', Anonymous Referee #2, 02 Aug 2021
The paper introduces a new algorithm to analyze the MLT observations by the radar networks and utilizes two the data from two radar networks to demonstrate the results based on this data retrieval algorithm. The new and improved capability to obtain high quality horizontal mapping of zonal and meridional winds by this technique is quite impressive. The paper has demonstrated that this work is high quality and provide a new tool to study various important dynamic topics in the mesosphere and lower thermosphere in different scales. The figures are clean and clear with proper captions. I understand this is more like a technology journal, but it would be good for illustrate how some of the dynamic parameters was calculated. For example, the body force. The section of discussion reads like a summary of this work.
I only have some minor technical comments. Please see below.
- Line 38. The Na Doppler lidar has been an important ground-based instrument, and has many important contributions to the MLT dynamics, due to its capability of day and night time simultaneous measurements of temperature and horizontal winds [Krueger et al., 2015]. Due to its horizontal wind capability, it can derive the intrinsic properties of GWs. It is unfortunate that the author misses this important instrument.
- Line 44-45. There are actually many of this collaborative investigations between the Na Doppler lidar and airglow instrument. So, I suggest the author replace “only a few” with many. The problem with such Na Doppler lidar – Airglow investigations is that they are mostly focusing on single case studies, such as Yuan et al. 2016, Cai et al., 2014, and, thus, cannot provide statically large numbers of cases to build robust database of these intrinsic properties of GWs.
- Line 2. Replace “can be” with “is”
- Line 60, to investigate
- Line 73, delete “are going to”
- Line 224, demands
- Line 231, delete “thus”
- Line 382, replace “possibilities” with “capabilities”
- Line 384, delete “very”
- Line 401-409, the semidiurnal tidal activity is not quite clear in Fig. 10. I wonder if the author can derive and show the semidiurnal amplitude variations instead of the hourly winds.
- Line 440, I am not quite clear what depend on the choice of… Please specify.
- Line 440 – 443. This is a very long sentence, please consider to revise.
- Line 521, I do not see any “diurnal tidal pattern” is discussed in the paper.
-
AC4: 'Reply on RC2', Gunter Stober, 11 Aug 2021
We thank the reviewer for the valuable comments on our paper. The manuscript will be revised according to the suggestions. A more detailed point-by-point reply with all the changes is going to be prepared after the public discussion is closed.
Comment:
Line 38. The Na Doppler lidar has been an important ground-based instrument, and has many important contributions to the MLT dynamics, due to its capability of day and night-time simultaneous measurements of temperature and horizontal winds [Krueger et al., 2015]. Due to its horizontal wind capability, it can derive the intrinsic properties of GWs. It is unfortunate that the author misses this important instrument.
Reply:
Some Na - Doppler lidars are indeed capable of measuring horizontal winds but often assume w=0 m/s. We will expand this part of the introduction and add more specific information on different Na- lidars and include citations to the proposed publications. However, we have to note that due to the imaging capability of the 3DVAR algorithm there is no need to measure temperature to derive the intrinsic gravity wave properties. The intrinsic horizontal wavelength is directly assessable from the images and the phase velocities can be inferred from successive images (see also Stober et al., 2014, Stober at al., 2018). Gravity wave properties from lidar observations are often analyzed applying a hodograph analysis, which requires temperature measurement to derive unambiguously the intrinsic gw properties.
Furthermore, NA-lidar winds could in principle be used and included similar to all other types of radial observations. The 3DVAR algorithm has a build-in tool to permit the usage of different data sets.
Comment:
Line 44-45. There are actually many of this collaborative investigations between the Na Doppler lidar and airglow instrument. So, I suggest the author replace “only a few” with many. The problem with such Na Doppler lidar – Airglow investigations is that they are mostly focusing on single case studies, such as Yuan et al. 2016, Cai et al., 2014, and, thus, cannot provide statically large numbers of cases to build robust database of these intrinsic properties of GWs.
Reply:
Thanks for the recommendation in rewording the issue. The revised manuscript is going to be changed and the suggested reference is going to be added. However, we have to point out that the 3DVAR retrievals are 24/7 observations. Case studies will be possible, but the scientific emphasis is certainly more towards statistical GW effects and their seasonality. We are going to rephrase this aspect and use more precise wording to express these differences.
We thank the reviewer also for all the technical corrections, which are going to be included in the revised manuscript.
References:
Stober, G., Sommer, S., Rapp, M., and Latteck, R.: Investigation of gravity waves using horizontally resolved radial velocity measurements, Atmospheric Measurement Techniques, 6, 2893 – 2905, https://doi.org/10.5194/amt-6-2893-2013, http://www.atmos-meas-tech.net/6/2893/2013/, 2013.
Stober, G., Sommer, S., Schult, C., Latteck, R., and Chau, J. L.: Observation of Kelvin–Helmholtz instabilities and gravity waves in the summer mesopause above Andenes in Northern Norway, Atmospheric Chemistry and Physics, 18, 6721–6732, https://doi.org/10.5194/acp-18-6721-2018, https://www.atmos-chem-phys.net/18/6721/2018/, 2018b
Peer review completion
