Exploring vertical motions in convective and stratiform precipitation using spaceborne radar observations: Insights from EarthCARE and GPM coincidence dataset

Aoki, Shunsuke; Kubota, Takuji; Turk, Francis Joseph

doi:https://doi.org/10.5194/egusphere-2025-3596

Preprints

https://doi.org/10.5194/egusphere-2025-3596

Preprints

14 Aug 2025

| 14 Aug 2025

Exploring vertical motions in convective and stratiform precipitation using spaceborne radar observations: Insights from EarthCARE and GPM coincidence dataset

Shunsuke Aoki, Takuji Kubota, and Francis Joseph Turk

Abstract. With the Doppler velocity (V_d) measurements from the Cloud Profiling Radar (CPR) onboard the Earth Cloud Aerosol and Radiation Explorer (EarthCARE), it has become possible to observe the vertical motions of hydrometeors inside cloud and precipitation globally. While W-band radar observations by CPR can capture clouds and upper-level ice hydrometeors well, Ku- and Ka-band radar observations by the Dual-frequency Precipitation Radar (DPR) onboard the Global Precipitation Measurement (GPM) Core Observatory are more effective under conditions involving rain or moderate-to-heavy ice precipitation, where attenuation and multiple scattering hinder reliable reflectivity measurements by CPR. This study constructed the EarthCARE–GPM coincidence observation dataset and investigated hydrometeor fall speeds and vertical air motion in stratiform and convective precipitation systems by integrating the complementary information from the two radars. Two case studies were conducted for stratiform and convective events, along with statistical analyses of reflectivity and V_d using nearly one year of dataset. CPR well captured ice particle growth in the upper troposphere above −10 °C, while DPR captured the properties of larger hydrometeors in the lower layers, including melting and rain layers. V_d generally increased with decreasing altitude, which is consistent with particle growth inferred from reflectivity observations from both CPR and DPR. Classification into four precipitation types based on echo top heights showed distinct differences in vertical profiles. In deep stratiform cases, V_d reveals slow downward speeds above the melting layer and faster speeds below, consistent with the bright band observed by DPR. V_d in deep convective types indicates faster-falling speed of densely rimed ice particles with high reflectivity and the presence of stronger updrafts and turbulence compared to stratiform cases. These findings indicate that V_d can provide insights into dynamical and microphysical processes inside deep clouds where the quality of reflectivity measurements in W-band deteriorates, and support future development of algorithms for precipitation retrieval and classification using V_d.

Received: 25 Jul 2025 – Discussion started: 14 Aug 2025

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 paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Shunsuke Aoki, Takuji Kubota, and Francis Joseph Turk

Status: final response (author comments only)

CC1:
'Comment on egusphere-2025-3596', Ousmane O. Sy, 24 Aug 2025
This article presents very interesting results based on a triple-frequency dataset complemented by Doppler measurements of the ESA/JAXA EarthCARE mission. It shows the high potential of coincident multi-frequency remote sensing observations with reflectivity, Doppler and passive microwave measurements. Such super-database can definitely help studies of dynamic atmospheric processes. The thermal analysis of the Doppler measurements is also eye-opening.
My minor comments are detailed below.
Comments:
Equation 1: Given the convention to represent updrafts as positive velocities, I think the equation should be

V_D= V_air – V_t + Epsilon,

Equation 2: the denominator is directly proportional to the reflectivity factor, but there is the factor (\lambda^4/\pi^5|K_W|^2) missing, unless it is implicit in the definition of \sigma_b?

Equation 1 represents all the error terms as additive (\epsilon), which is a simplification. In fact there is also a multiplicative factor (which includes the randomness of the signal) that can be mitigated only by adaptive filtering, or along-track integration. Could you please include this caveat?

Line 211: It is true that the Doppler velocity comes from the phase change of the lag-1 correlation function, which in turns is not affected by attenuation. However, the magnitude of this correlation function (module of a complex number) is important to have reliable Doppler. Otherwise, the Doppler is all salt-and-pepper.

Line 225: Does the use of Mie mean that it does not account for the shape/density variation of the frozen hydrometeors and their non-sphericity? Is that considered negligible microphysical variation?

Figs 2 and 3 are really great:

it’s great to see the various Z fields from DPR and ECPR, plus the Doppler!

The long along-track integration (10 km?) really helps to clean up aliasing in the Doppler and some of the NUBF in Fig.2;

However, doesn’t it also lead to an “over-smoothing” of the Doppler for the convective scene in Fig.3?

L.305: Isn’t it “CFED of 𝑉d shown in Fig. 4b illustrates an increase in downward velocity with decreasing increasing temperature between −20°C and 0°C,”?

The CFEDS of the W band are also great!

Fig.4a shows the stagnation of Z around [-10,0]^oC (due to competing effects of increasing unattenuated Z due to growth of particle, and, increasing attenuation) and

the slightly increasing Vd in that range shows that there is indeed a growth of particles,

The scatter-plots in Figs 5, 9 and 10 show the various datasets together.

Would it help to show the concentration in log(counts) instead of counts? It may be that there aren’t enough points to consider a log scale…

L.429 and 501: rimmed or rimed? (please correct various instances in the article).

I was wondering if it would be worth showing plots/statistics of the DWR (ZKu-ZKa) as well?

it is directly related to Dm, which plays a central role in the PSD used to estimate Vt in Eqs 3 and 4, and

it is a clear indicator when the attenuation of Ka is excessive.

In the interpretation of Fig.8, the larger standard deviation of the Doppler is attributed to turbulence. Couldn’t it be caused also/instead by 1) shear or 2) the microphysical variability of Vt? Or are these implied in the term “turbulence”?
Citation: https://doi.org/10.5194/egusphere-2025-3596-CC1
RC1:
'Comment on egusphere-2025-3596', Anonymous Referee #1, 13 Sep 2025
General Comments: To demonstrate the utility of the newly available EarthCARE data, the authors in this study investigate the EarthCARE CPR radar measured Doppler velocity in convective and stratiform precipitating clouds and interpret it in relation to precipitation growth processes. The uniqueness of this study is the combined use of both EarthCARE CPR and GPM DPR data, which allows the authors to explore the difference of particle growth processes in stratiform and convective clouds. The paper seems to serve 2 purposes: First, it demonstrates the quality and usefulness of the first-ever space-borne cloud radar Doppler velocity measurements, and second, using the Doppler velocity measurements, it confirms some of the understandings on microphysical processes in convective and stratiform clouds. The paper is well structured, and the messages are well presented. It will be a good contribution to this special collection of papers on EarthCARE. I suggest accepting after addressing some minor concerns.
Specific Comments:
Doppler velocity Vd in the EarthCARE product. In the manuscript, the authors stated that the Doppler velocity in the EarthCARE product is derived from “the phase shift of the radar signal and is therefore less affected by attenuation”. I understand that the retrieval of Vd is not the focus of this paper, but I do like to see a brief explanation on this topic in the Data and Methods section. Because many of the cases involved in this study are related to moderate to heavy rainfall, CPR should suffer significant attenuation particularly for the rain portion in the vertical profiles. Are there any studies on the impact on Vd retrieval accuracy by attenuation when using phase shift method?

10 km horizontal integration. Please explain the reason for 10 km integration for EarthCARE data. The DPR footprint size is about 5 km, then you have to average 2 DPR pixels to match 13 (=10./0.75) EarthCARE pixels? Is there a reason not using ~7 EarthCARE pixels to match 1 DPR pixel? Will the results be different if do so? In short, the decision to use 10 km seems to be somewhat arbitrary, may need couple sentence to justify.

The use of temperature as vertical coordinate. The use of temperature as vertical coordinate is an interesting way to investigate microphysical processes. However, there is a shortcoming when global data are mixed into one figure such as Figure 4. I suspect that most of data near 20C are from tropics or warm season mid-latitudes. In the meantime, data near 0C are from almost all the places. When we put all data into one figure, explaining the features in a way that particles are falling from aloft to lower part is somewhat misleading. I’d like the authors mention this shortcoming, and remind readers that future studies should separate data into groups with similar temperature range in the vertical.

Just a comment. It is great to see in Figure 5 that Vd in the cold range (-10C) is around -1 m/s and the derived Vt is matching well with measured Vd. This gives us great confidence that the Vd quality is high.

Misc.

Line 55. “The Tropical … (TRMM) was launched in 1997, and the TRMM carried …”. I think it is better to say: “The Tropical … (TRMM) satellite was launched in 1997, and it carried …”
Line 105. I don’t see “CSATGPM” appearing in any place before this point. Please define it.
Line 137. The exclusion of 5 and 10 range bins are somewhat arbitrary. Are they about 0.5 and 1.0 km, respectively? Please add a couple of sentences to explain why excluding these many bins is enough.
Line 164-165. Earlier in the text, it is mentioned that EarthCARE data is integrated to a 10 km “pixel”. Here it sounds like the matching is between 1 EarthCARE original pixel (750 m size) with 1 DPR pixel (5 km size). Please clarify.
Line 276-277. Do you have a rough number of profiles (in percent) that is detected by CPR but not DPR?
Figure 6. An interesting feature is that most stratiform precipitation tops (by DPR) are around -15C although their cloud tops (by CPR) are all over the place. Any explanations?
Line 447. “theoretical W-band terminal velocity” -> “theoretical terminal velocity”. Terminal velocity should not be band-dependent.
Citation: https://doi.org/10.5194/egusphere-2025-3596-RC1
RC2: 'Comment on egusphere-2025-3596', Anonymous Referee #2, 29 Sep 2025

This work analyses Ku-, Ka-, and W-band radar measurements of precipitating clouds using a one-year dataset of matchup EarthCARE and GPM observations. Following a quick demonstration of two case studies, statistical properties of radar reflectivity and Doppler velocity profiles are investigated for different particle phases (liquid/solid) and different precipitation types (shallow/convective/stratiform). A new method is devised to evaluate the vertical air motion in raining layers by subtracting the terminal velocity deduced by DPR-derived DSD from the EarthCARE Doppler velocity.
This is a well written paper presenting robust analysis results that align with theoretical expectations and physical intuition. The authors' effort to construct a matchup EarthCARE and GPM dataset should be applauded and will be welcomed by the cloud/precipitation science community. I suggest a few revisions that are mostly minor in nature with the possible exception of the first point. Otherwise I would recommend that the paper be published in AMT.
Main comments --------------------

1. JAXA's EarthCARE CPR Level-2 cloud product (L2a CPR_CLP) contains its own vertical air motion estimated from Doppler velocity. The CPR reflectivitiy (and hence DSD estimates crucial for V_t and V_air as well) is subject to heavy attenuation for intense rain as the authors pointed out (ll. 210-211). That being said, a substantial number of CPR reflectivities would be still usable, being not entirely washed out by attenuation even beneath the 0-degree level as far as I can tell from Fig. 2c and 3c. This means that there would be plenty of simultaneous measurements available for both EarthCARE-provided V_air and GPM Dm estimates.

I am curious how consistent the V_t estimates are between the CPR standard product and the current method using DPR-derived DSD. You would find discrepancies because the CPR_CLP relies on its own built-in DSD assumption which is not guaranteed to accord with DPR Dm. An additional plot or two comparing the CPR-only and DPR-based V_t and V_air estimates would tell us how reliable the CPR products are, offering useful information for EarthCARE algorithm developers and interested users.
2. In the paper, V_t and V_air are shown only for rain layers (Fig. 10). Why not add V_t and V_air for solid precipitation too (Fig. 9)? The DPR Dm might not be as reliable for snow as for rain because the KuPR is not sensitive enough to small frozen hydrometeors, but a comparison with the CPR_CLP product would be worth studying for solid precipitation as well.
Specific points -----------------

3. l. 55: and The TRMM -> and the TRMM
4. l. 73: weaker hydrometeors -> smaller hydrometeors
5. l.135: are more frequently appear -> more frequently appear
6. ll. 144-146: I am puzzled by the claim that "only data from the HS mode are used". The KaHS mode has been reassigned to match the outer KuPR swath to complement the inner KaMS swath since the scan pattern was changed in May 2018. As far as I can tell from Figs. 2 and 3, the whole DPR swath (that is, both MS and HS modes) seems to be analysed in this work.
7. l. 153: which calculated -> which were calculated
8. l.156: "equatorial" region may be better replaced by low- and mid-latitude regions. 65S-N is much wider than the equatorial region.
9. l. 170: samplings -> samples
10. Figure 4c/d: Why are the KuPR and KaPR CFEDs sharply (presumably artificially) cut off above a certain level, with a temperature threshold around -42C for Ku and -35C for Ka?
11. l. 303: "However" does not really fit the context here. Please try "On the other hand" or "In contrast" instead.
12. Figure 6: The joint histogram of Ku- and W-band Z reminds me of Fig. 8 of Stephens and Wood (2007, DOI: 10.1175/MWR3321.1). They showed CFADs separated for different cloud types, which also bears resemblance to the present work. I just thought this might be worth a brief discussion.
13. l.375: "Intense" deep convection (DC-I) could be misleading, given that higher echo tops do not necessarily guarantee more intense convection (e.g., Hamada et al., 2015, DOI: 10.1038/ncomms7213). Something like Tall deep convection (DC-T) may be a safer alternative.
14. l. 394: is generally low -> are generally low

15. l. 415: Here again, "However" may be better rephrased by "By contrast" etc.

Citation: https://doi.org/10.5194/egusphere-2025-3596-RC2

Shunsuke Aoki, Takuji Kubota, and Francis Joseph Turk

Viewed

Total article views: 1,682 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
1,590	79	13	1,682	18	38

HTML: 1,590
PDF: 79
XML: 13
Total: 1,682
BibTeX: 18
EndNote: 38

Views and downloads (calculated since 14 Aug 2025)

Month	HTML	PDF	XML	Total
Aug 2025	642	31	6	679
Sep 2025	922	41	7	970
Oct 2025	26	7	0	33

Cumulative views and downloads (calculated since 14 Aug 2025)

Month	HTML	PDF	XML	Total
Aug 2025	642	31	6	679
Sep 2025	922	41	7	970
Oct 2025	26	7	0	33

Viewed (geographical distribution)

Total article views: 1,643 (including HTML, PDF, and XML) Thereof 1,643 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 05 Oct 2025

Short summary

The EarthCARE/CPR provides the first spaceborne Doppler velocity measurements, while the GPM/DPR excels at observing rain and heavy snow, which are more attenuated in CPR. Using coincident observations from both radars, we examined vertical motions in stratiform and convective precipitation systems. The synergy between the radars enables a more comprehensive understanding of hydrometeor fall speeds and vertical air motions across different precipitation types.


Total:	0
HTML:	0
PDF:	0
XML:	0