the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Determination of the vertical distribution of in-cloud particle shape using SLDR mode 35-GHz scanning cloud radar
Patric Seifert
Alexander Myagkov
Johannes Bühl
Martin Radenz
Abstract. In this study we present an approach that uses polarimetric variables from a scanning polarimetric cloud radar MIRA-35 in the 45° slanted linear depolarization (SLDR) configuration, to derive the vertical distribution of particle shape (VDPS) between top and base of mixed-phase cloud systems. The polarimetric parameter SLDR was selected for this study due to its strong sensitivity to shape and low sensitivity to the wobbling effect of particles at different antenna elevation angles. For the VDPS method, elevation scans from 90° to 30° elevation angle were deployed to estimate the vertical profile of the particle shape by means of the polarizability ratio, which is a measure of the density-weighted axis ratio. Results were obtained by retrieving the best fit between observed SLDR-vs-elevation dependencies and respective values simulated with a spheroid scattering model. The applicability of the new method is demonstrated by means of three case studies of isometric, columnar and oblate hydrometeor shapes, respectively, which were obtained from measurements at the Mediterranean site of Limassol, Cyprus. The identified hydrometeor shapes are demonstrated to fit well to the cloud and thermodynamic conditions which prevailed at the times of observations. Some observations reveal that in mixed-phased clouds ice particle shapes tend to evolve from a pristine columnar or dendritic state at cloud top toward a more isometric shape at cloud base. Either aggregation or riming processes contribute to this vertical change of microphysical properties. The new height-resolved identification of hydrometeor shape and the potential of the VDPS method to derive its vertical distribution are helpful tools to understand complex processes such as riming or aggregation, which occur particularly in mixed-phase clouds.
Audrey Teisseire et al.
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2022-1431', Anonymous Referee #1, 17 Mar 2023
In this study, the authors show how the vertical distribution of particle shape can be retrieved from the SLDR configuration. Three cases were presented for demonstration and the presented method is shown to be skillful in identifying ice shapes. The method seems to be reasonable, however, the manuscript is difficult to follow. I feel that this manuscript is suitable for readers who are super familiar with this type of work, but is not friendly for people in other fields. I have been working with cloud radars for many years, but it took me very long time to figure out the innovative point of this manuscript. Therefore, I believe the authors should carefully discuss the research progress of this field and highlight the contribution of this work, instead of simply referring to for example Myagkov et al. (2016a). Different observing modes were mentioned in the manuscript, but no explanations were presented. It would be beneficial to summarize the characteristics of different modes in a table, and show why and how this mode is superior to others.
Please see my comments below.
Major comments
1. I am frustrated in connecting the symbols used this work to existing works. For example, SLDR is denoted by delta_s, rho_hv by rho_s. And polarization ratio, orientation…I took very long time to connect them to what have been used by Myagkov 2016. Other readers may have similar feelings… I would suggest keeping consistent with Myagkov who is the coauthor of this work.
In addition, I feel the first paragraph in section 2.3 is poorly organized. Please elaborate the used variables before analyzing the results in Fig.2.
2. I believe that the Rayleigh condition at Ka-band should be considered. The theoretical basis is based on the assumption of Ka-band Rayleigh scattering which is satisfied for pristine ice. However, in presence of large aggregates which usually occur at -5 to 0 C or -15 C, this assumption is violated and the retrieval should be made with caution.
3. L160-163. the concept of orientation should be well discussed. The reasons why these assumed K values are applicable should be elaborated.
4. The authors claim that rimed or aggregated ice particles are isometric and speculate that the retrieved polarization ratio around 1 indicates riming or aggregation. I feel that this statement should be used with caution. Firstly, the method may be limited to pristine ice at ka band. The Rayleigh condition at Ka band may not have reached, for example at 1.6 km in figure 10. Therefore, the basic assumption is broken. Secondly, very heavily rimed ice can be isometric. Lightly rimed ice or aggregation has a characteristic aspect ratio.
5. I understand that the authors do not have aircraft observations at hand for validation. Why the retrieved results are consistent with cloud physics should be discussed and the weaknees that no direction comparison was done should be discussed.
Technical issues1. L109. Do you mean that the spectral line with the highest Scx was used? If so, a figure illustrating the spectral processing is needed, since this has rarely been done. In addition, have you quantified the impact of spectral broadening which can effectively smooth the spectral peak?
2. Fig. 3. add discussions about orientation.
3. Fig. 6. Orientation information is missing
4. L174-175. I do not see the discussions about dendritic in your figures.
5. L178-180. The discussion on isometric particle shape class is lack of ground.Citation: https://doi.org/10.5194/egusphere-2022-1431-RC1 -
RC2: 'Comment on egusphere-2022-1431', Anonymous Referee #2, 13 Apr 2023
The goal of the manuscript is to provide information about the vertical evolution of ice particle shape class (prolate, oblate, isometric) inside clouds. The method is based on a previous approach using RHI scans with a cloud radar operating in slanted LDR mode. The method is applied to three case studies of clouds located in different temperature regimes. Based on the temperature regime, the retrieved particle shape classes are “evaluated”.
Overall I have to say that I found the manuscript to be not well written and very difficult to follow. I can understand that a first author, who is deeply into the topic and who has developed such a complex method might have some difficulties describing the method with sufficient clarity. However, I am disappointed by the fact that the co-authors, who I know as excellent writers of scientific contributions and who are renowned and well-experienced experts in this field are not better at helping the main author to compile a clear and readable manuscript before it is submitted to a journal.
In its current form, the manuscript cannot be published in my opinion. I see a number of major weaknesses in the current manuscript which I think need to be carefully addressed by the authors and the entire manuscript need to be substantially reworked before it can be published.
- The authors should make more clear which parts of this work are substantially different from the previous approach by Myagkov et al., 2016a,b. I understand that the hybrid-mode Ka band radar in Myagkov et al., 2016 is different from the SLDR mode radar in this study. But as mentioned in Myagkov et al., 2016a their approach can also be applied to SLDR mode radars. So what are the major changes and improvements in this work? It would be also a great help for the reader if you could use more “common” notation for variables or make them at least consistent with the one used in Myagkov et al. Ideally, the method from Myagkov et al., 2016 should be directly compared to your new method using an identical case study, but I don’t know whether such a dataset exists. I also think the authors need to describe better the state-of-the-art of different SLDR approaches and also provide all necessary references to previous work (SLDR work done by Reinking, Melnikov, Matrosov, etc.) to the reader (see comment to the introduction below).
- The method described here and also the one in Myagkov et al., 2016a,b are based on Rayleigh scattering approximation. To my knowledge, it has not yet been proven (for example with DDA scattering simulations) that this approximation is a valid assumption for the polarimetric quantities used in this work or what the uncertainty range associated with the Rayleigh assumption is. Previous studies such as Schrom and Kumjian, JAMC, 2018 found quite large deviations in polarimetric variables for plate-like particles even when using T-Matrix and even at low radar frequencies. The fact that the retrieved values are consistent with laboratory data (as shown in Myagkov et al., 2016b) is in my opinion not sufficient as the agreement could be caused by compensating errors in the retrieval itself. I would have expected that the authors do at least an attempt to test the general consistency of the Rayleigh assumption with existing DDA scattering datasets (for example Lu et al., AMT, 2016). I also realized that the PICNICC project is part of a bigger German research collaboration on radar polarimetry (PROM). Wouldn’t it be possible that some of the collaborating projects provide you with DDA calculations of plates and columns in order to check the validity of the Rayleigh assumptions which are the core of the method? This would have been a clear improvement or at least a confirmation of the previous approach.
- As you mention in the text, one of the strong assumptions of all RHI based SLDR retrievals is that particle populations are homogenous at a certain height level within a few km horizontal distance. I was hoping that you provide some objective criteria for estimating the homogeneity of particles. Your method seems to still depend on the “experienced-eye” selection of RHI scans (L. 191). In my opinion, this is completely impractical for compiling statistics (which you mention in the conclusions L. 415 ff).
- The biggest weakness in my view is the lack of an independent evaluation of the method. The case study analysis which you present is at the most a consistency check. The fact that the retrieved shape is roughly matching the expected shape in this temperature region cannot be counted as an evaluation of the method. How can you be sure that only one particle type with a distinct shape is present in your volume (the cloud in your second case is roughly 1.5km thick)? How can you know that your crystals are not partially rimed in an environment which Cloudnet classifies as mixed-phase? I think the case studies shown here are also less convincing than the comparison shown in Myagkov et al., 2016b where the authors selected only thin and clearly liquid-topped mixed-phase clouds.
Other general comments:
I have two major criticism regarding the introduction:
a) In my understanding, an introduction should include an overview of the previous work done on the topic and also describe the current state-of-art. I miss a number of studies related to SLDR which can be easily found by a quick internet search. Currently, the introduction is heavily weighted toward the work of the authors and their working group. I strongly suggest that the authors provide a more thorough and less self-biased literature overview. For example in the paragraph starting at L. 44 the only two studies related to SLDR that you introduce are from Matrosov et al., 2012 and Myagkov et al., 2016a. This is not really complete considering the previous work done on this topic. No need to discuss every study but at least a complete list of references should be provided. Just a few examples of missing studies related to SLDR:
Reinking et al., JTECH, 2002 (Used SLDR to distinguish drizzle, crystals and irregular ice particles)
Matrosov et al., JAS, 2005 (Fall attitudes of dendrites derived using SLDR and HLDR)
L. 41: The recent study by Luke et al., PNAS, 2021 where the authors used Doppler spectra and LDR to separate different hydrometeor populations is missing.
b) My second criticism is related to the motivation of the relevancy to retrieve shape information of ice particles. Just to avoid misunderstandings: I agree that information about ice particle shape is valuable but I find your discussion and argumentation of why this information is relevant for models or our microphysical understanding to be quite vague. For example, in L. 29-31: You argue that the shape information can be used to distinguish pristine ice from aggregates or rimed particles. Many aircraft studies showed in the past that most ice and mixed-phase clouds are dominated by irregular particles. Can your method distinguish between irregular particles and aggregates/rimed particles? What would happen if you probed a mixture of columnar and plate-like particles? They could be both “pristine” but probably would look like aggregates. As you write in the introduction, if you are looking at the top of a mixed-phase cloud, you can assume that all particles grow into a particular shape due to the similar temperature regime they were nucleated. But for deeper clouds, I highly question whether an approach indicating a change in shape can really distinguish between a mixture of ice crystal shapes, aggregation or riming. In L. 37-40 you also mention that the shape information could be useful to “support the improvement of these processes in numerical models”. Can you explain better how you think a vertical profile of shape can be useful for model development? I am very skeptical since most bulk schemes assume a constant ice particle habit throughout the entire temperature range. Only very few experimental models currently exist which take particle habit into account.
Section 3 with the description of the methodology needs to be substantially reworked. I found the sentences and formulations (for example Sec 3.2) to be extremely confusing and hard to follow (see several specific comments below). I also found it disappointing that the retrieval code is not made publicly available to the community for example on github/gitlab. This should be a standard nowadays and would also help the community to further develop the approach.
Specific comments:
L. 4: I find the formulation “wobbling effect of particles” unclear. I guess you want to say that SLDR is less sensitive to particle orientation as already shown by previous studies, correct?
L. 7: “SLDR-vs-elevation” maybe a bit unconventional. I suggest writing it out.
L. 9: “columnar and oblate” why not prolate and oblate or columnar and plate-like? This would be more consistent.
L. 12-13: How can you be sure that the shape is not primarily changed by sublimation or depositional growth in different regimes (for example: first columnar and then plate-like resulting in capped-columns)?
L. 19: I think a textbook reference would be more appropriate here for this first sentence.
L. 20: Add “the formation of” before precipitation.
L. 41-43: Another limitation of the spectral approach is that hydrometeors with similar terminal fall velocities (for example drizzle and small ice) cannot be distinguished in the spectrum. I suggest mentioning that several polarimetric variables can also be resolved spectrally. Would your method be principally applicable in a spectral approach?
L. 56-58: That sentence contains some redundancy, or?
L. 82: Maybe more general plate-like particles?
Table 1: Please extend the table and add information on sensitivity at a certain range, beam width and co-cross-channel isolation.
L. 95: figure axes instead of caption?
L. 115/Eq(2): Explain meaning of <>
L. 116: “subject to noise artifacts”. Unclear what you mean here. I thought only the maximum peak is used…
L. 150: “gradients of rho_s against the elevation angle” What do you mean here? I can’t see a gradient against an elevation angle as you only show 90° and 150°.
L. 151: “giving an idea about the shape” please rephrase.
L. 143: “assumes Rayleigh scattering” I suggest adding some sentences in the discussion that the Rayleigh assumption of the current technique should be checked with more realistic scattering methods. For example, Schrom and Kumjian, JAMC, 2018 show that typical radar polarimetric variables of plate-like particles calculated with T-Matrix deviate quite substantially from DDA calculations even at low weather radar frequencies. In my opinion, those results should be a warning that similar uncertainties might also affect SLDR methods. Ideally, this manuscript should provide a consistency check with existing DDA scattering databases.
Fig. 2: If find it quite confusing to denote a “domain” by single symbols. Why do you not include some dashed lines separating the different domains and add some letters A, B, C for referencing to those regions? Add delta= or Elevation= on the top of the columns in front of the number.
L. 187: I see the point that only using two elevation angles reduces the total time needed for a scan. But on the other hand, aren’t you also losing information by reducing the number of elevation angles so drastically? Can you provide an evaluation of what is the best compromise between scan speed and number of elevation angles included?
L. 190: I agree that the need to assume horizontal homogeneity is a very strong assumption limiting the applicability of the retrieval. But how broadly applicable is your retrieval method if one needs an “experienced-eye evaluation” of each RHI scan? I think this is a big limitation and you should better provide some more objective criteria which can be used.
L. 199: What do you mean with “scan pattern”. Fig. 2 only shows two constant elevation angles, right? Or do you mean Fig. 1?
L. 217: I probably don’t understand the method correctly: I thought you measure at 90° and 150° elevation only (L. 187). What I don’t get is why you need to apply a polynomial fit to obtain the values at those angles? Also how can you apply a polynomial fit if you have only two points? Or are you using complete RHI scans to derive the fit which you can then apply to only two-elevation angle measurements? Please explain better.
L. 221: Unclear what you mean. In Fig. 2a I see that delta_s is very “straight” for prolate and isometric particles. For the oblate ones, I see that a non-linear fit is needed. Please rephrase. Again in the entire paragraph it is unclear to me what is fitted against what.
L. 231: Hard to follow without references. Is this valid for all cloud radars even at all different wavelengths? What are typical minimum values of delta_s of common cloud radar systems? The references in Sec. 2.1 are only valid for your MIRA system, or? How low was delta_s for the systems used in previous studies, for example the Ka and W band systems used in the US?
L. 233: Confusing. In Fig. 2c, d the minimum (dark blue) values of the colorbar are indicating -35 dB.
L. 241: What is the prediction interval of a polynomial fit? Please explain better.
Section 3.2 and Caption of Fig.5: In the second part of section 3.2 and especially in the caption to Fig. 5 I got completely lost. I tried to read and understand it now several times but I am totally confused. For example some formulations of caption to Fig. 5: “ intersection of observed delta_s in the data fields simulated with the spheroidal scattering model”. How can observed delta_s be simulated? “Isolines of the observed values in the model space” Isolines of observed values? “Overlay of intersections of the isolines” what is that? “As observational input, hypothetical values of typical oblate particles…” What is an observational input? I think the section and the caption need to be carefully rephrased. Please give it to somebody who is not an expert in order to ensure that an average experienced reader can follow.
L. 275: What is a “hypothetical case study” and what should it be useful for??
L. 281: “Stratification of ice particle shapes” again unclear what you mean by that. Maybe a vertical profile?
Figure 7: The temperature information used from model analysis in your Cloudnet classification has obviously some problems as the melting level jumps unrealistically up and down over time (for example between 13:00-13:30 and also later on). How does that affect your results? You should check the model fields or only rely on the melting layer estimated from the radar itself.
Section 4.1: It is still unclear to me how you can use modeled values of delta and rho based on pure ice particles for this rain case. In the previous chapter, you argue that zeta is “not strongly affected by the refractive index” but now you are using observed delta and rho values as well? By the way, as in most of the manuscript you are using an unscientific formulation “not strongly affected by the refractive index”. You should quantify what you mean by that: Less than 10%, less than 1% or something similar. And what is actually the problem of using refractive index of water for this case study? I guess the Rayleigh scattering parameters won’t take much time to be calculated.
Description of Fig. 9: In the plot you show linear fits which have by definition a constant gradient (derivative is the slope). If you want to indicate that the gradient changes over the range of elevation angles then you should add a curve that represents a derivative over a certain elevation angle range. Right now I don’t find this very consistent and physically sound.
L. 327: At 1300m I find in the blue box area of Fig. 7 temperatures far below 0°C. But again I think this is an artifact of the model temperature field used for Cloudnet.
L. 330-339: Sorry, but this interpretation of particle shape above the melting layer is completely invalid in my opinion. Your method implicitly assumes that the volume is composed of only one particular particle shape, correct? In such a relatively thick mixed-phase cloud it is almost certain that close to the ML you face a mixture of single crystals, aggregates and maybe even rimed particles. This mixture is not only comprised of very different shapes but also various densities.
L. 367: It is not true that between -10 and 0°C one expects only columnar particles. For example, between -8 and -10°C one expects plate-like particles.
L. 382-383: “rather dense” and dendritic particles can’t be true. Due to the branching of dendrites their density is much less than for example hexagonal plates.
Fig. 16: The distribution of rho_s does not look like following a linear relationship. What does this indicate and why not using also polynomial fit for rho as well?
L. 415: You claim that you developed an automatized framework but in L. 191 you say that the method needs “experienced-eye evaluation of the obtained RHI scan”. How can this method be automatized if you need to inspect each single RHI scan?
Typos:
L. 211: of the results
Citation: https://doi.org/10.5194/egusphere-2022-1431-RC2
Audrey Teisseire et al.
Audrey Teisseire et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 279 | 87 | 11 | 377 | 6 | 7 |
- HTML: 279
- PDF: 87
- XML: 11
- Total: 377
- BibTeX: 6
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1