the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Mar 2025
Vertical Wind and Drop Size Distribution Retrieval with the CloudCube G-band Doppler Radar
Abstract. Macrophysical properties of clouds are influenced by underlying microphysical processes. In practice, there is often an observational gap in bridging the two. For example, our current understanding of aerosol-cloud interaction and cloud-climate feedback is hindered by a lack of robust measurements of the distribution of drop sizes within clouds, especially for the smallest drop sizes. Doppler radar measurements have proven useful in estimating rainfall drop size distributions (DSDs) but face an intermediate challenge of requiring a correction for the presence of vertical air motion. Recent advances in millimeter wave technology have made radar measurements at ever smaller wavelengths possible, allowing for analysis of particle size dependent scattering effects to back out estimates of vertical winds and thereby DSDs. This work demonstrates a method of deriving range-resolved DSDs using Doppler spectra at 238 GHz measured by the CloudCube ground-based G-band atmospheric Doppler radar. The observations utilized are of marine boundary layer clouds during March and April 2023 in La Jolla, CA, USA, taken as part of CloudCube’s participation in the Eastern Pacific Cloud Aerosol Precipitation Experiment (EPCAPE) campaign. This method first identifies notches in the velocity spectra and compares them to the theoretical notch velocities predicted by size dependent backscattering and terminal velocity models to estimate the range-dependent vertical wind. After removing the vertical wind, binned DSDs are retrieved from the zero-wind spectrum. Bulk properties of the precipitation are then derived including the number concentration, liquid water content, characteristic drop size, and precipitation rate. These bulk properties are relatively invariant to the assumptions made in the estimation of the full DSD retrieval, making large volumes of such retrievals useful tools in assessing physical models of drizzle.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Measurement Techniques.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.- Preprint
(1599 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 03 May 2025)
-
RC1: 'Comment on egusphere-2025-618', Anonymous Referee #1, 11 Mar 2025
reply
This manuscript describes new vertical air motion and microphysical quantity profile retrieval using a vertically pointing G-band Doppler radar. The authors provide an extensive description of the retrieval methodology, focusing on a case study demonstration using measurements collected during the CloudCube radar deployment during the EPCAPE field campaign. The retrieval itself appears pretty robust and is quite elegant. However, it does suffer from some limitations and caveats; most notably, deficient uncertainty analysis and lack of scenarios applicable for this method. On that note, it should be emphasized that the authors mention that they collected only a few instances during 2 out of 13 measurement days (case applicability < 2/13*100 = 16%). Those instances were the only times during which data suitable for the retrieval algorithm were collected – l. 108-109. Such a major limitation should be discussed in detail and acknowledged in the summary/conclusions and abstract, or otherwise, readers could develop the perception that this method is generalized and applicable to most, if not all, precipitation conditions (with matching drop sizes). Other than that, the manuscript is generally well written but has some inaccuracies and inconsistencies that should be corrected. I recommend major revisions even though I think there are straightforward solutions to most, if not all, of my major comments.
Major comments:
- Methodology: How is the minima depth dependence on the noise floor (l. 119-120) justified? For example, why should we trust more, or better say choose a much deeper minimum in case of a higher noise floor? If the minimum is well above or even slightly above the noise floor, why does it matter? If that is the actual algorithm implementation, then we potentially introduce a significant caveat here. There could be other methods to define a the notches such as a certain reduction in minima’s signal amplitude relative to the closest peaks, but my understanding is that this is not the case here.
- Methodology: Given that the raindrops follow some PSD, which is then manifested in the resolved spectra, there can be some offset in the notch location. How big is this offset? Have you examined how the retrieved PSD (in the following steps) propagates into the exact notch location or whether using different Gamma PSD input parameters modifies the exact notch location? I suspect this offset will be smaller than the mode differences illustrated in fig. 3c, yet this should be accounted for, or at least acknowledged as an uncertainty source (e.g., could this be a source for the minima inconsistencies discussed in l. 164-166?)
- Methodology: turbulent broadening calculations (l. 223-234): I could be wrong here but it seems to me that this paragraph has many inaccuracies, e.g., how was eq. 7 derived (this exact format is not in O'Connor et al. (2010)? Where are t_small and t_large referred to in the equations? L in eq. 8 only refers to the large scale (scattering volume) horizontal dimension, correct? Then what is the definition of Lsmall? I presume that t_small is simply the radar averaging time often equivalent to the dwell time, is that correct?
Perhaps most importantly, how is the dissipation rate retrieved (implicit in sigma_v_air)? This could have a critical impact on the forward-calculated variables and the conditions discussed in O’Connor et al. (2010) might bot be applicable here, for example
Are those inaccuracies mentioned above implemented in the actual retrieval? How is the resolved turbulent broadening convolved in the actual spectra as per eq. 1?
- Uncertainty quantification: given all the "moving parts" in this retrieval, I find it hard to believe such small uncertainties as the authors present and discuss (e.g., fig. 12 l. 288-290) are representative, especially given that the depicted uncertainties are merely the propagation of an arbitrary value (two bins) and nothing else. The uncertainty quantification should be revisited here and potentially discussed in more detail (e.g., could the second and third minima be incorporated, when detectable, as an uncertainty metric, e.g., as illustrated in fig. 3c? could the confidence in the Gaussian mixture model output be incorporated (or not - a possibility)? What about the turbulent broadening estimates? Does the time since radiosonde release impacts the uncertainty? etc.
- Text consistency: using radii instead of diameters is somewhat confusing, potentially deceptive, and sometimes used interchangeably. l. 61 provides an example where such a radii-diameter confusion comes into play. Drizzle is defined as drops with diameters smaller than 0.5 mm, not radii, so from this definition, even the G-band notch method applies to rain drops, not drizzle (by the self-definition of rain drop). Given that ostensibly drop diameters are more commonly used in the community and literature, I recommend converting all drop dimensions in the text and figures to diameters and check such claims as in l. 61.
Minor comments:
- Abstract l. 15 - "These bulk properties are relatively invariant to the assumptions made in the estimation of the full DSD retrieval..." - this claim is unsupported by the text and I doubt it can be supported using a single case study without a rigorous evaluation. Either rework the DSD impact on *all* retrieval components or omit this part of the sentence, e.g., start the sentence with "We suggest that large volumes of such retrievals can be useful tools ..."
- l 27 - observations --> measurements
- l 41 - Dopper moments --> higher Doppler moments (since reflectivity is the 0th moment)
- l 85 - refer to the fact that EPCAPE was a DOE ARM campaign. This is noted below on a different form but should be stated at the beginning. Also, add a reference for EPCAPE (likely the science plan - https://doi.org/10.2172/1804710)
- l 88 - This should come in the first sentence of this paragraph - also define the ARM acronym
- l 90 - data iteself is --> datasets are (data is plural of datum – change here and elsewhere where applicable, e.g., l. 141)
- l 95 - refer to the ARM radiosonde handbook - https://doi.org/10.2172/1020712.
- l 97 – Add a reference to the KAZR handbook - https://doi.org/10.2172/1035855
- l 99 - that is the wrong reference (LD handbook) - change to VDIS handbook - https://doi.org/10.2172/1251384 - note that the title on osti is wrong but the document is correct).
- l 108-110 - What is "sufficient span of elevations and times"? This should be elaborated as per the first paragraph of my review.
- l 108 - campaign --> deployment
- l 117 - Add reference to scipy
- l 125-130 – The example case should be described in detail, and a panel showing the mean Doppler velocity should be added to Fig. 2. For non-experts, the melting layer echoes are not clear, and questions could arise concerning the cloud base height vs. melting layer indications. Without further elaborating on this scenario, one might think that we have a ringing effect rather than big ice particles melting and generating the interesting observed spectra consisting of a wide range of rain drop sizes and therefore large spectral widths, evolving *below cloud base*.
- l 128-130 – quick-looking at other ARM data from this case, those ceilometer variations looks physical to me. Recommend removing this sentence.
- l 131 - by "single 2-D spectrum" do you refer to a spectrogram (Doppler array values vs. height) or something else?
- l 138 - provide reference to SciKit-Learn
- l 147-149 - the stark outlier is discussed but how is the cyan pint at ~250 m next to the purple points justified?
- l 159 - define vt and k
- l 159-160 - that is incorrect. Air density is not measured by the radiosonde, but can be calculated using radiosonde measurements.
- l 178 -|K^2| (the dielectric factor) is not the square magnitude of the refraction index, but it depends on it
- l 183-184 - "In the Stokes regime, ..." - this sentence is unclear - recommend rewording.
- l 185-186 - can you explain the local minima in fig. 4b?
- l 195 - what does "previous" refer to?
- l 204 - this is the two-way optical depth so either define it as such or remove the factor of 2 here and add it in eq. 6.
- l 244 - recommend using an alternative nomenclature for lambda because it is immediately associated with the wavelength rather than the regularization weight.
- l 228 - t is the radar averaging time, not the turbulence time scale
- l 271-277 - I am not sure I'd consider the depicted PSDs as showing good agreement. The reason here is that one would expect the drops to evaporate between the lowest (out-of-cloud) range gate and the surface; hence, one would expect a PSD offset to the left (decreasing sizes) in the case of the VDIS, whereas in the plotted spectra (fig 11b), the shift is to the right (apparently increasing sizes). I recommend (a) revisiting the text here and (b) I am less familiar with the CloudCube radar and do not know how susceptible it is for near-surface clutter effects, but it might be good to also examine the retrieval 50 or 100 m above the first range to reduce that probability - even if we are further away from the surface, the signal might be cleaner and hence, might generate better agreement with surface observations.
- l 307 - provide reference for that deployment
- l 285 - define rho_w
- l 287 - define the volume in the equation
- l 308 - "to provide unprecedented observations of profiles of ..." such microphysical quantity retrievals are not unprecedented. Perhaps you meant something like: "to robustly provide profiles of ..."?
- Data availability statement: The KAZR and VDIS datasets require references including their DOI, which I believe can be retrieved from ARM.
- Table 1 - define FMCW. Also, many of those parameters (at least for KAZR) are configurable so the caption should specifically refer to EPCAPE.
- fig 3 onward - provide the units for each plotted spectrograph as in fig. 2b
- fig 6 and elsewhere - add units to for N(r) within or next to the logarithm.
- fig 8 - turbulence velocity? Do you mean turbulent broadening?
- fig 9: caption - this seems like a spectrum (1D) rather than 2D. To which example spectrum (altitude and time) do you refer here?
Also, is the reflectivity here really in linear units? This appears inconsistent with the dBZ units in fig. 2b. Same goes for N(r), which seems more log-scale than linear.
- fig 10 - specify RML
- fig 11 - nice results - I'd consider 2 dBZ difference (at least up to 1.1 km) to be well within the KAZR measurement uncertainty (3 dBZ; see handbook). To support your case, adding such error bars/uncertainty envelope around the measured profile would be helpful to demonstrate that the forward calculated values are within the uncertainty range.
- fig 13 - specify the times each spectrograph corresponds to.
Also, I think that a large fraction of readers would be interested in the vertical air motion aspect of this retrieval (as the manuscript title suggests) - recommend adding profiles to this figure.
Citation: https://doi.org/10.5194/egusphere-2025-618-RC1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 101 | 20 | 4 | 125 | 2 | 2 |
- HTML: 101
- PDF: 20
- XML: 4
- Total: 125
- BibTeX: 2
- EndNote: 2
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 63 | 48 |
| China | 2 | 22 | 17 |
| France | 3 | 8 | 6 |
| Germany | 4 | 5 | 3 |
| Japan | 5 | 5 | 3 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 63