the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Sep 2023
An evaluation of microphysics in a numerical model using Doppler velocity measured by ground-based radar for application to the EarthCARE satellite
Abstract. The Cloud Profiling Radar (CPR) of the Earth Cloud, Aerosol and Radiation Explorer (EarthCARE) has a new capability to observe the Doppler velocity related to the vertical air motion of the terminal velocity of hydrometeors. The new observation from space will be used to evaluate and improve the model. Before the launch of EarthCARE, we need to develop a methodology for using the CPR data for model evaluations. In this study, we evaluated simulated data by a stretched version of the global non-hydrostatic model over Japan with a ground-based CPR using an instrument design similar to the EarthCARE CPR. We chose two cases with different precipitation events in September 2019 using two cloud microphysics schemes. We introduced the categorization method for evaluating microphysics using Doppler velocity. The results show that the liquid and solid phases of hydrometeors are divided in Doppler velocity, and the model's terminal velocities of rain, snow, and graupel categories can be evaluated with the observation. The results also show that the choice of microphysics scheme has a more significant impact than the dependence on precipitation cases. We discussed the application of the EarthCARE-like simulation results using a satellite simulator.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(1746 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1746 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-1997', Anonymous Referee #1, 27 Oct 2023
Review of “An evaluation of microphysics in a numerical model using Doppler velocity measured by ground-based radar for application to the EarthCARE satellite“ by Roh et al.
The authors present a study about the potential of EarthCARE for observing Doppler velocities. In general, the topic of the paper is interesting and relevant for ACP. I recommend the paper to be accepted after major revision.
- Language: The manuscript lacks clarity due to language problems. I recommend that the authors give the paper to a native speaker to make sure they write what they intent to say. Also, I would recommend to guide the reader better to explain why a analysis was performed in a certain way. For example, in section 4 I would recommend to stress that you start with an idealized simulation without instrument effects like the Nyquist range or random errors and then make the simulation more realistic step for step. Also, it should be stressed that the hydrometeor classification is not supposed to be a universally applicable one (at least I hope this is that case) but is only used to allow for a better comparison between model and observations.
- The authors study a case related to a tropical storm with potentially high reflectivities that might lead to multiple scattering, how would that impact the results?
- What about other sources for measurement errors? E.g. cloud inhomogeneity or pointing uncertainty?
- L 160: “We assumed the contribution of vertical air velocity to Doppler velocity is relatively smaller than the terminal velocity of hydrometeors” This is a strong assumption that needs to be supported. Alternative, the authors could remove convective data points using a filter like in e.g. Mosimann, 1995.
- L 260: Without averaging, the performance of the Doppler observations is quite bad for the high mode as can be seen in Fig. 11. Is that the main message the authors want to convey with this paper? Spatial averaging would improve the results, why was it not considered?
- Did the authors correct for the effect of changing air density on hydrometeor sedimentation velocity? Wouldn’t such a correction be necessary for a threshold based classification?
Minor comments:
- L 119: Fig 1 -> Fig 2c-d
- L 166; “When we use the threshold of 2 m/s for categorising the hydrometeors, 0.2 m/s of vertical air velocity affects the 10% bias”. Does that mean that the authors expect that 0.2 m/s vertical air motion lead to a 10% error of the classification? This is only true if the data points would be equally distributed with Doppler velocity which is not the case.
- L 239: Specify latitudes for low and high modes.
- Figs 1-3, 6-11. Please add labels to all x axis, y axis and colorbars
- 6: Why is there no data above 12 km?
- 8: Not referenced in the text
- Data availability: Where are the used simulations and radar observations available?
Mosimann, L., 1995: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N.
Citation: https://doi.org/10.5194/egusphere-2023-1997-RC1 -
AC1: 'Reply on RC1', Woosub Roh, 15 Jan 2024
We appreciate the reviewer’ constructive comments.
Major comments:
- The manuscript is difficult to read because of English grammatical issues. A fairly extensive grammatical revision is required.
As an example, the first paragraph of the introduction would benefit from these changes:
(a) Remove the word "The" from the first sentence
(b) "Satellite" and "Global" should be lower case (first and second sentences)
(c) "The detailed process of hydrometeors" in fourth sentence does not make senseThe grammatical errors are too extensive to list. I would recommend that they enlist some help in the interest of ensuring that they are communicating their work effectively to their audience.
--> This draft was checked by an English-speaking researcher based on your comment.
- The hydrometeor classification scheme (section 3.2) is abruptly stated, with no underlying justification offered. Where did these categories come from (specifically the apparently arbitrary absolute values that define the cutoffs)? How much uncertainty is present in these categories?
--> We added the explanations based on some references. According to the Glossary of Meteorology of the American Meteorological Society, the diameter of a drizzle is less than 0.5 mm, and the terminal velocity is 2.068 m/s with 0.5 mm at the surface based on Foote and Toit 1969. Mosimann 1995 investigated the degree of snow crystal riming using vertical Doppler radar. He found that the degree of riming is proportional to the Doppler velocity and that there is a large fraction of graupel with the Doppler velocity greater than 2 m/s (fig. 3 in Mosimann 1995). In this classification, we did not consider the effect of air density. This classification has uncertainty from vertical air motion and air density. We think the impact of these two terms is not significant. This study does not aim for an accurate classification of hydrometeors but rather for a quantitative intercomparison of models on the same basis.
- One major lingering question I have after reading the manuscript is as follows: For the NICAM simulations, the authors obviously know the distribution of hydrometers in the model. Why not compare their suggested classification scheme to what's actually present in the model? Maybe they did in fact do this, and I simply misunderstand the meaning of the figures. For example, in Figure 6(a), is 63.1% the fraction of cloud ice/snow derived from the DV-based classification scheme? Or is the cloud ice/snow fraction that was actually present in NICAM?
--> We think the accuracy of the classification’s names is not very important in this study. Microphysics scheme has different definitions of hydrometeors, their own terminal velocity, and size distributions. We think characteristics of vertical profiles of Doppler velocity in models related to terminal velocities of hydrometeors are more important. There are several uncertainties with this categorization. Even if it's cloud ice/snow, it's possible that there are mixtures of hydrometeors like small graupel. But we can understand that the average terminal velocity in that grid is high or low, and that's expected to have an impact on clouds and precipitation. We will investigate the impact of tunings of the Doppler velocity on radiation and large circulation in a global storm-resolving model (GSRM).
- The authors later use fall-speed relationships from the model to at least somewhat clarify the hydrometeor classification system that was chosen. This leads to the larger point that the chronology of the "story" that is being told by this manuscript is somewhat muddled. The introduction should spell this out much more clearly, including the hierarchy of simulations that will be used.
--> We added the motivation of this study about the evaluations of GSRMs using the Doppler velocity in detail in the introduction part: One of the motivations of this study is to evaluate and compare the vertical distribution of hydrometeors of GSRMs using the same observational criteria. According to Roh et al. 2021, the horizontal distribution of outgoing longwave radiation of GSRMs is similar, but the simulated vertical distributions of hydrometeors of GSRMs are very different in the intercomparison data (Stevens et al. 2019). Each model used its own assumptions about the size distribution and terminal velocity of hydrometeors. We believe that the Doppler velocity is one of the criteria for understanding and constraining the vertical distributions between GSRMs using observations.
- The radar used for ground-based observations (HG-SPIDER) is polarametric, correct? Why not use the polarametric data to better quantify the actual hydrometeor classifications, and therefore enhance the classification system and associated discussion the physics?
--> HG-SPIDER is not a polarimetric radar.
- The authors state that "We found that the frequency of absolute vertical velocity above 0.2 m/s is less than 2 %, and the simulated PDF of the Doppler velocity mostly depends on the cloud microphysics." I'm not convinced that the fact that absolute vertical velocities above 0.2 m/s are a small proportion of the total (Figure 4) means that air velocity can be neglected when translating Doppler velocity to particle fall speed. Since these results are from NICAM, why not just look at this relationship directly in the model? For example, heavy rain (although it's a small proportion of cases with reflectivity exceeding -40 dBZ) is more likely to occur in regions with significant vertical ascent.
--> We investigated the impact of vertical air motion on the Doppler velocity using the Joint-Simulator. We removed the vertical air velocity for the calculation of Doppler velocity (Fig. 2). When we removed the vertical air velocity, the results were consistent with the control results (Fig. 1). However, the frequencies were concentrated in NSW6, and there was no fraction of the upward category. Most of the difference is less than 2% in the classifications.
The figure files are attached in the supplement.
Figure 1: The categorizations of the hydrometeors in NICAM simulations for NSW6 (top) and NDW6 (bottom) in case 1 (left) and case 2 (right).
Figure 2: The same as Figure 1 but for only calculations of Doppler velocity without vertical air motion.
Specific comments:
- Line 121: Do you mean that Doppler radar is free of attenuation?
--> The radar reflectivity is attenuated. The Doppler velocity is not attenuated. However, the accuracy of Doppler velocity changes because of the attenuation.
- Line 182: Looks more like about 3 m/s, doesn't it?
--> The terminal velocity of NDW6 is less than 2m/s, and the terminal velocity of NSW6 less than 3 m/s. I added a sentence like “NSW6 shows the faster terminal velocity of raindrops with less than 0.5 mm diameter.”.
- Line 189: What is the "large data sampling"?
--> The observation data is every one minute. The model output data is every one hour. So we need to have a larger sampling of data for statistical analysis.
- Line 192: Check the 0.6% number.
--> I checked the number.
- Line 215: Can you better define what is meant by "observation window"?
--> The observation window is different from the radar range. The observation window means a collected data range. The observation window depends on the PRFs, number of integration of pulse(M), and satellite altitude. The PRFs and M changes by the lookup table related to the satellite altitude. The observation window of CloudSat is 30 km (Tanelli et al. 2008).
- Line 221: Please briefly explain the -15 dBZ (in addition to the reference).
--> The errors of the Dopler velocity depend on the signal-to-noise (SNR). The lower SNR means the higher contribution of the signal noise to Doppler velocity. According to Hagihara et al. (2021), the standard deviation of random errors increases significantly when the radar reflectivity is less than −15 dBZ (SNR =6.2 dB).
- NDW6 acronym is not defined.
--> We added the explanation about NDW6 like “the NICAM Double-moment Water 6-categories (Seiki and Nakajima 2014, hereafter referred to NDW6)”.
- Figures 2 and 3 have no axes labels.
--> We added the axes labels.
- There is no reference to Figures 3 or 8 in the text.
--> We added the references.
- The inline text in Figures 6, 7, 8, and 10 is unreadable in some cases, since the text overlaps the contours. In Figure 3, the "3" and "4" are difficult to discern in places.
--> We improved all figures except Figure 1.
References
Foote, G. B., & Du Toit, P. S. 1969: Terminal velocity of raindrops aloft. Journal of Applied Meteorology (1962-1982), 249-253.
Mosimann, L.: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N, 1995.
Hagihara, Y., Ohno, Y., Horie, H., Roh, W., Satoh, M., Kubota, T., and Oki, R.: Assessments of Doppler velocity errors of EarthCARE cloud profiling radar using global cloud system resolving simulations: Effects of Doppler broadening and folding, IEEE Trans. Geosci. Remote Sens., 60, 1–9, https://doi.org/10.1109/TGRS.2021.3060828, 2021.
Kobayashi, S., Kumagai, H., & Kuroiwa, H.: A proposal of pulse-pair Doppler operation on a spaceborne cloud-profiling radar in the W band. Journal of Atmospheric and Oceanic Technology, 19(9), 1294-1306, 2002.
Tanelli, S., Durden, S. L., Im, E., Pak, K. S., Reinke, D. G., Partain, P., ... & Marchand, R. T.: CloudSat's cloud profiling radar after two years in orbit: Performance, calibration, and processing.IEEE Transactions on Geoscience and Remote Sensing, 46(11), 3560-3573, 2008.
-
AC3: 'Reply on RC1', Woosub Roh, 15 Jan 2024
I am sorry for pasting another reviewer's response.
Thank you for your constructive comments.
Major comments:
Language: The manuscript lacks clarity due to language problems. I recommend that the authors give the paper to a native speaker to make sure they write what they intent to say. Also, I would recommend to guide the reader better to explain why a analysis was performed in a certain way. For example, in section 4 I would recommend to stress that you start with an idealized simulation without instrument effects like the Nyquist range or random errors and then make the simulation more realistic step for step. Also, it should be stressed that the hydrometeor classification is not supposed to be a universally applicable one (at least I hope this is that case) but is only used to allow for a better comparison between model and observations.
The authors study a case related to a tropical storm with potentially high reflectivities that might lead to multiple scattering, how would that impact the results? What about other sources for measurement errors? E.g. cloud inhomogeneity or pointing uncertainty?
- The revised draft was checked by an English-speaking researcher based on your comment. We agree that the multiple scattering impact also affects the results related to heavy precipitation cases. We think we can clearly understand the uncertainty from the multiple scattering after the launch of the satellite. Our expectation is that the multiple scattering is not significant because of the 800 m footprint and circular polarization. We need to filter out the data related to multiple scatterings to get better results. In this paper, we assume we use calibrated Doppler velocity from the multiple scattering and the point uncertainty for evaluations of a global storm-resolving model. We will filter out the data related to multiple scattering in simulations with the same criterion as the retrieval algorithm in the future.
- The cloud inhomogeneity is not important in this resolution with less than 500m. Now, we focus on the evaluations of a km scale global model. The purpose of this study is the application of the Doppler velocity for evaluations of modeling groups.
L 160: “We assumed the contribution of vertical air velocity to Doppler velocity is relatively smaller than the terminal velocity of hydrometeors” This is a strong assumption that needs to be supported. Alternative, the authors could remove convective data points using a filter like in e.g. Mosimann, 1995.
- We checked the upward motion using the observation data. The frequency of the upward motion is very rare. The time interval of our observation data is less than a second, and we used one-minute integrated data for the analysis. So, we think convective data points are reduced by the integration. Additionally, we investigated the impact of vertical air motion on the Doppler velocity using our satellite simulator. We removed the vertical air velocity for the calculation of Doppler velocity. When we removed the vertical air velocity, the results were consistent with the control results. However, the frequencies were concentrated, and there was no fraction of the upward category. Most of the difference is less than 2% in the diagrams. We added these results in the revised draft.
L 260: Without averaging, the performance of the Doppler observations is quite bad for the high mode as can be seen in Fig. 11. Is that the main message the authors want to convey with this paper? Spatial averaging would improve the results, why was it not considered?
- We checked the resolution dependency of the simulation results using NICAM. We found the impact of resolution dependency is not larger than the choice of microphysics schemes. So, we expect the 10km integration data to be useful for the evaluation or intercomparison of global storm-resolving models (GSRMs).
Did the authors correct for the effect of changing air density on hydrometeor sedimentation velocity? Wouldn’t such a correction be necessary for a threshold based classification?
- For the Doppler velocity, we considered the effect of changing air density. However, we did not consider the classification method. We think the air density affects the classification of the hydrometeors. We think the impact is not significant for the classification. Our purpose is a simple evaluation method for intercomparison or evaluation of GSRMs.
Specific comments:
L 119: Fig 1 -> Fig 2c-d
- We modified it based on your comment.
L 166; “When we use the threshold of 2 m/s for categorising the hydrometeors, 0.2 m/s of vertical air velocity affects the 10% bias”. Does that mean that the authors expect that 0.2 m/s vertical air motion lead to a 10% error of the classification? This is only true if the data points would be equally distributed with Doppler velocity which is not the case.
- We removed the explanation about that.
L 239: Specify latitudes for low and high modes.
- We specify the latitudes like “the high and low modes will be used depending on latitudes: low mode (−1 to 16 km) at latitudes of 60°–90° and high mode (−1 to 20 km) at latitudes of 0°–60° (Hagihara et al., 2022). ”
Figs 1-3, 6-11. Please add labels to all x axis, y axis and colorbars
- We added the labels to all x-axis, y-axis, and color bars.
6: Why is there no data above 12 km?
- The vertical range of the observation data is until 12 km.
8: Not referenced in the text
Data availability: Where are the used simulations and radar observations available?
- Simulation data will be available, but we need to discuss to open our observation data to the public. We will upload the available data before the publication.
Mosimann, L., 1995: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N.
Citation: https://doi.org/10.5194/egusphere-2023-1997-AC3
-
RC2: 'Comment on egusphere-2023-1997', Anonymous Referee #2, 04 Nov 2023
SUMMARY AND OVERALL ASSESSMENT
The manuscript deals with several aspects of how new EarthCare Doppler velocity (hereafter DV) measurements might address problems of hydrometeor identification using a variety of observations and simulations.
There are a number of interesting results in the manuscript, but ultimately I don't think the methodology or goals were explained well enough for me to have a clear understanding of what this study is really trying to communicate. For example, is the focus of this paper to define a DV-based hydrometeor classification system, or is that just a tool? I'm really unclear on the underlying "story" this work is telling. More clearly defining the goals of the study, and steps to reach that goal, will go a long way toward bringing this study into a publishable state.
MAJOR COMMENTS
1. The manuscript is difficult to read because of English grammatical issues. A fairly extensive grammatical revision is required.
As an example, the first paragraph of the introduction would benefit from these changes:
(a) Remove the word "The" from the first sentence
(b) "Satellite" and "Global" should be lower case (first and second sentences)
(c) "The detailed process of hydrometeors" in fourth sentence does not make senseThe grammatical errors are too extensive to list. I would recommend that they enlist some help in the interest of ensuring that they are communicating their work effectively to their audience.
2. The hydrometeor classification scheme (section 3.2) is abruptly stated, with no underlying justification offered. Where did these categories come from (specifically the apparently arbitrary absolute values that define the cutoffs)? How much uncertainty is present in these categories?
3. One major lingering question I have after reading the manuscript is as follows: For the NICAM simulations, the authors obviously know the distribution of hydrometers in the model. Why not compare their suggested classification scheme to what's actually present in the model? Maybe they did in fact do this, and I simply misunderstand the meaning of the figures. For example, in Figure 6(a), is 63.1% the fraction of cloud ice/snow derived from the DV-based classification scheme? Or is the cloud ice/snow fraction that was actually present in NICAM?
4. The authors later use fall-speed relationships from the model to at least somewhat clarify the hydrometeor classification system that was chosen. This leads to the larger point that the chronology of the "story" that is being told by this manuscript is somewhat muddled. The introduction should spell this out much more clearly, including the hierarchy of simulations that will be used.
5. The radar used for ground-based observations (HG-SPIDER) is polarametric, correct? Why not use the polarametric data to better quantify the actual hydrometeor classifications, and therefore enhance the classification system and associated discussion the physics?
6. The authors state that "We found that the frequency of absolute vertical velocity above 0.2 m/s is less than 2 %, and the simulated PDF of the Doppler velocity mostly depends on the cloud microphysics." I'm not convinced that the fact that absolute vertical velocities above 0.2 m/s are a small proportion of the total (Figure 4) means that air velocity can be neglected when translating Doppler velocity to particle fall speed. Since these results are from NICAM, why not just look at this relationship directly in the model? For example, heavy rain (although it's a small proportion of cases with reflectivity exceeding -40 dBZ) is more likely to occur in regions with significant vertical ascent.
7. What procedure is used to produce Figure 6, and how does it differ from the procedure used to produce Figure 7? Figure 6 apparently does not use the Joint simulator, so what does it use? Why does Figure 6 cut off at about 12 km when NICAM clearly produces results above 12 km?
The rather significant changes between Figures 6 and 7 are dealt with in a cursory way, but the changes are significant and what causes them need to be better explained.
MINOR COMMENTS
- Line 121: Do you mean that Doppler radar is free of attenuation?
- Line 182: Looks more like about 3 m/s, doesn't it?
- Line 189: What is the "large data sampling"?
- Line 192: Check the 0.6% number.
- Line 215: Can you better define what is meant by "observation window"?
- Line 221: Please briefly explain the -15 dBZ (in addition to the reference).
- NDW6 acronym is not defined.
- Figures 2 and 3 have no axes labels.
- There is no reference to Figures 3 or 8 in the text.
- The inline text in Figures 6, 7, 8, and 10 is unreadable in some cases, since the text overlaps the contours. In Figure 3, the "3" and "4" are difficult to discern in places.
Citation: https://doi.org/10.5194/egusphere-2023-1997-RC2 -
AC4: 'Reply on RC2', Woosub Roh, 15 Jan 2024
We appreciate the reviewer’s constructive comments.
Major comments:
1. The manuscript is difficult to read because of English grammatical issues. A fairly extensive grammatical revision is required.
As an example, the first paragraph of the introduction would benefit from these changes:
(a) Remove the word "The" from the first sentence
(b) "Satellite" and "Global" should be lower case (first and second sentences)
(c) "The detailed process of hydrometeors" in fourth sentence does not make senseThe grammatical errors are too extensive to list. I would recommend that they enlist some help in the interest of ensuring that they are communicating their work effectively to their audience.
- The revised draft was checked by an English-speaking researcher based on your comment.
2. The hydrometeor classification scheme (section 3.2) is abruptly stated, with no underlying justification offered. Where did these categories come from (specifically the apparently arbitrary absolute values that define the cutoffs)? How much uncertainty is present in these categories?
- We added the explanations based on some references. According to the Glossary of Meteorology of the American Meteorological Society, the diameter of a drizzle is less than 0.5 mm, and the terminal velocity is 2.068 m/s with 0.5 mm at the surface based on Foote and Toit 1969. Mosimann 1995 investigated the degree of snow crystal riming using vertical Doppler radar. He found that the degree of riming is proportional to the Doppler velocity and that there is a large fraction of graupel with the Doppler velocity greater than 2 m/s (fig. 3 in Mosimann 1995). In this classification, we did not consider the effect of air density. This classification has uncertainty from vertical air motion and air density. We think the impact of these two terms is not significant. This study does not aim for an accurate classification of hydrometeors but rather for a quantitative intercomparison of models on the same basis.
3. One major lingering question I have after reading the manuscript is as follows: For the NICAM simulations, the authors obviously know the distribution of hydrometers in the model. Why not compare their suggested classification scheme to what's actually present in the model? Maybe they did in fact do this, and I simply misunderstand the meaning of the figures. For example, in Figure 6(a), is 63.1% the fraction of cloud ice/snow derived from the DV-based classification scheme? Or is the cloud ice/snow fraction that was actually present in NICAM?
- We think the accuracy of the classification’s names is not very important in this study. Microphysics scheme has different definitions of hydrometeors, their own terminal velocity, and size distributions. We think characteristics of vertical profiles of Doppler velocity in models related to terminal velocities of hydrometeors are more important. There are several uncertainties with this categorization. Even if it's cloud ice/snow, it's possible that there are mixtures of hydrometeors like small graupel. But we can understand that the average terminal velocity in that grid is high or low, and that's expected to have an impact on clouds and precipitation. We will investigate the impact of tunings of the Doppler velocity on radiation and large circulation in a global storm-resolving model (GSRM).
4. The authors later use fall-speed relationships from the model to at least somewhat clarify the hydrometeor classification system that was chosen. This leads to the larger point that the chronology of the "story" that is being told by this manuscript is somewhat muddled. The introduction should spell this out much more clearly, including the hierarchy of simulations that will be used.
- We added the motivation of this study about the evaluations of GSRMs using the Doppler velocity in detail in the introduction part: One of the motivations of this study is to evaluate and compare the vertical distribution of hydrometeors of GSRMs using the same observational criteria. According to Roh et al. 2021, the horizontal distribution of outgoing longwave radiation of GSRMs is similar, but the simulated vertical distributions of hydrometeors of GSRMs are very different in the intercomparison data (Stevens et al. 2019). Each model used its own assumptions about the size distribution and terminal velocity of hydrometeors. We believe that the Doppler velocity is one of the criteria for understanding and constraining the vertical distributions between GSRMs using observations.
5. The radar used for ground-based observations (HG-SPIDER) is polarametric, correct? Why not use the polarametric data to better quantify the actual hydrometeor classifications, and therefore enhance the classification system and associated discussion the physics?
- HG-SPIDER is not a polarimetric radar.
6. The authors state that "We found that the frequency of absolute vertical velocity above 0.2 m/s is less than 2 %, and the simulated PDF of the Doppler velocity mostly depends on the cloud microphysics." I'm not convinced that the fact that absolute vertical velocities above 0.2 m/s are a small proportion of the total (Figure 4) means that air velocity can be neglected when translating Doppler velocity to particle fall speed. Since these results are from NICAM, why not just look at this relationship directly in the model? For example, heavy rain (although it's a small proportion of cases with reflectivity exceeding -40 dBZ) is more likely to occur in regions with significant vertical ascent.
- We investigated the impact of vertical air motion on the Doppler velocity using the Joint-Simulator. We removed the vertical air velocity for the calculation of Doppler velocity (Fig. 2). When we removed the vertical air velocity, the results were consistent with the control results (Fig. 1). However, the frequencies were concentrated in NSW6, and there was no fraction of the upward category. Most of the difference is less than 2% in the classifications.
The figure files are attached in the supplement.
Figure 1: The categorizations of the hydrometeors in NICAM simulations for NSW6 (top) and NDW6 (bottom) in case 1 (left) and case 2 (right).
Figure 2: The same as Figure 1 but for only calculations of Doppler velocity without vertical air motion.
7. What procedure is used to produce Figure 6, and how does it differ from the procedure used to produce Figure 7? Figure 6 apparently does not use the Joint simulator, so what does it use? Why does Figure 6 cut off at about 12 km when NICAM clearly produces results above 12 km?
The rather significant changes between Figures 6 and 7 are dealt with in a cursory way, but the changes are significant and what causes them need to be better explained.
- We used the Joint Simulator for Figure 6 and Figure 7. The differences between Fig. 6 and Fig. 7 are the setting of ground observation and the EarthCARE satellite. The observation range of the ground observation is up to 12 km, and the vertical resolution is different from EarthCARE. The CFADs of radar reflectivity are different because of the attenuation of rain. However, the results of Doppler velocity are very similar to ground observation. We expect there is an impact on data of Doppler velocity larger than -15 dBZ because of attenuation. Before we introduce the impact of random errors based on the observation window, we need to introduce the simulated Doppler velocity like EarthCARE.
Specific comments:
- Line 121: Do you mean that Doppler radar is free of attenuation?
- The radar reflectivity is attenuated. The Doppler velocity is not attenuated. However, the accuracy of Doppler velocity changes because of the attenuation.
- Line 182: Looks more like about 3 m/s, doesn't it?
- The terminal velocity of NDW6 is less than 2m/s, and the terminal velocity of NSW6 less than 3 m/s. I added a sentence like “NSW6 shows the faster terminal velocity of raindrops with less than 0.5 mm diameter.”.
- Line 189: What is the "large data sampling"?
- The observation data is every one minute. The model output data is every one hour. So we need to have a larger sampling of data for statistical analysis
- Line 192: Check the 0.6% number.
- I checked the number.
- Line 215: Can you better define what is meant by "observation window"?
- The observation window is different from the radar range. The observation window means a collected data range. The observation window depends on the PRFs, number of integration of pulse(M), and satellite altitude. The PRFs and M changes by the lookup table related to the satellite altitude. The observation window of CloudSat is 30 km (Tanelli et al. 2008).
- Line 221: Please briefly explain the -15 dBZ (in addition to the reference).
- The errors of the Dopler velocity depend on the signal-to-noise (SNR). The lower SNR means the higher contribution of the signal noise to Doppler velocity. According to Hagihara et al. (2021), the standard deviation of random errors increases significantly when the radar reflectivity is less than −15 dBZ (SNR =6.2 dB).
- NDW6 acronym is not defined.
- We added the explanation about NDW6 like “the NICAM Double-moment Water 6-categories (Seiki and Nakajima 2014, hereafter referred to NDW6)”.
- Figures 2 and 3 have no axes labels.
- We added the axes labels.
- There is no reference to Figures 3 or 8 in the text.
- We added the references.
- The inline text in Figures 6, 7, 8, and 10 is unreadable in some cases, since the text overlaps the contours. In Figure 3, the "3" and "4" are difficult to discern in places.
- We improved all figures except Figure 1.
References
Foote, G. B., & Du Toit, P. S. 1969: Terminal velocity of raindrops aloft. Journal of Applied Meteorology (1962-1982), 249-253.
Mosimann, L.: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N, 1995.
Hagihara, Y., Ohno, Y., Horie, H., Roh, W., Satoh, M., Kubota, T., and Oki, R.: Assessments of Doppler velocity errors of EarthCARE cloud profiling radar using global cloud system resolving simulations: Effects of Doppler broadening and folding, IEEE Trans. Geosci. Remote Sens., 60, 1–9, https://doi.org/10.1109/TGRS.2021.3060828, 2021.
Kobayashi, S., Kumagai, H., & Kuroiwa, H.: A proposal of pulse-pair Doppler operation on a spaceborne cloud-profiling radar in the W band. Journal of Atmospheric and Oceanic Technology, 19(9), 1294-1306, 2002.
Tanelli, S., Durden, S. L., Im, E., Pak, K. S., Reinke, D. G., Partain, P., ... & Marchand, R. T.: CloudSat's cloud profiling radar after two years in orbit: Performance, calibration, and processing.IEEE Transactions on Geoscience and Remote Sensing, 46(11), 3560-3573, 2008.
-
AC4: 'Reply on RC2', Woosub Roh, 15 Jan 2024
-
RC3: 'Comment on egusphere-2023-1997', Anonymous Referee #3, 08 Nov 2023
This study aims to use model simulations of two precipitation events by NICAM and a satellite simulator to investigate how spaceborne Doppler velocity (Vdop) measurement could be useful to evaluate microphysics, with an emphasis on categorization of hydrometeor types using thresholds in Doppler velocities. The topic is relevant to the satellite mission, EarthCARE. The language in the manuscript, , the justification of the scientific methods, and the presentation of the study needs to be significant improved before it can be considered for publication.
It is not clear why the authors chose to use threshold of the Vdop to categorize hydrometeor types. From a storm resolving model, all the hydrometeors are known. However, the authors chose not to use that information, but applying thresholds directly to the Vdop from a forward model (or satellite simulator). Why does this manuscript do it this way? Is it because the EarthCARE retrieval algorithm have such a component? Even if it is trying to mimic a component in the retrieval algorithms, storm resolving model could be better used to provide context for retrievals. For example, snow and graupel (other hydrometeors too) must be mixed in a large number of model grid points. How could the categorization of Vdop (such as those shown in Fig. 6) could be used to single out the mass of a certain hydrometeor type? Do you need to consider a ratio of mixing of hydrometeor types? Is that the retrieval would provide, or not?
The presentation of the study and figures need to be improved. I suggest adding a figure showing the observed reflectivity and Doppler velocities. as well as the simulated reflectivity and Vdop. Figure 1 is only the precipitation rate. The text in Figs. 3, 6, 7, 8, 10, 11 are not illegible. A table could be necessary for comparisons. The 2% in Fig. 4 is not accurate (should it be 0.3% for grid points with absolute air motion > 0.5 m/s?). The magnitude of the model-resolved vertical air motion and the model horizontal resolution also needs to be elaborated a bit more. Ignoring vertical air motion, directly relating Fall velocity to Vdop is not a scientific sound method for interpretation. Much more justification needs to be made.
Below are some editing I made while read the manuscript.
Line 27: “... sampling of ...”, this sentence needs to be reworded. You mean footprint of space-borne radar? Spatial sampling scales?
Line 34: “... in the same body ...” needs to be reworded. You mean same space craft?
Line 38: “synergetic”? Maybe using “synergistic”?
Line 90: I think the variable in Figure 1 is “precipitation rate”.
Line 102 to 107: You did not mention how the simulations do for Case 2. The simulations missed the part of precipitation over the ocean to the east near 36 and 37 north latitudes. This should be added.
Fig. 2 caption: What is the unit of the color scale? Percentage? Please be specific. Also note this figure is for observation.
Line 119: It should be Figure 2 c and d.
Line 120: Why “Doppler velocity is free to attenuation”?
Line 123: “Two high-frequency modes are near the melting layer...” This description is rather vague. It is just the difference above and below the melting layer.
Line 126 “less than -2 m/s”? The figure shows “greater than -2 m/s”. Do you mean the absolute value?
Line 128: if you talk about aliasing, I think you should give the CPR’s measurement range. When the velocity is out of the range, aliasing would happen.
Line 129: Not using radar reflectivity is due to strong attenuation, right? You’d better mention this when you talk about the reflectivity attenuation earlier.
Last paragraph in Page 6: It looks like it is about Figure 3. Please refer to Figure 3.
Comparing Fig. 3 and 2: why you use CFADs vs. joint histogram? Are they different in your paper? Give units to your plots.
Comparing Fig. 3a to Fig. 2a: Are their differences solely due to unfolding? Why there is so much difference from 8 km and above?
Line 162: reword it to “produces a large bias and makes the results unreliable.”
Line 164: Is it 2%? From the black solid line, I read 0.997, from the black dash and orange lines, I read 0.9985 to 0.999. So, it should be 0.3% or less. Why do you think it's 2%?
Line 164-165: are you talking about the simulated PDF of the Doppler velocity or the vertical air velocity?
Line 169: About the results are affected by the horizontal resolution of the model. The issue of the dependence of the vertical air motion on model resolution needs more details. Please restate the resolution of the model you are using and the coarse resolution you are referring to. Please also refer to previous studies about the dependence of vertical air motion on model horizontal resolution (e.g., Lebo and Morrison 2015 Monthly Weather Review page 4355-4375 or other study that you find appropriate.)
Line 178: It is vague for rain vs cloud water vs drizzle. Please explicitly state how rain, cloud water, drizzle is separated in your categorization method. In Fig. 5 there is no cloud water or drizzle shown.
Figure 6: text on this figure is not readable. You need to adjust the location of the text for hydrometeor percentages, and the color of the text that labels the regions.
Line 193: You show the choice of microphysics scheme has more influence as compared to the case dependence. So, how this method could be used to provide useful information to the EarthCARE retrieval methods?
Line 200: The cloud ice has 0 m/s terminal velocity as shown in Fig. 5 in the single moment scheme. How do you interpret the growth of snow from cloud ice from Fig. 6?
Line218: “It decrease the sampling of liquid hydrometeors and upward motion (Fig. 7)” – Do you mean the attenuation from the spaceborne CPR in the liquid hydrometeor layer is more severe?
Line 222: “we found an increase ....” no text is illegible in your figures, same here in Fig. 7 and 8. You need to use a table to show the comparisons.
Line 310: correct “thee” to “three”
References:
Lebo, Z. J., H. Morrison, 2015, Effects of Horizontal and Vertical Grid Spacing on Mixing in Simulated Squall Lines and Implications for Convective Strength and Structure. Monthly Weather Review, 4355-4375.
Citation: https://doi.org/10.5194/egusphere-2023-1997-RC3 -
AC2: 'Reply on RC3', Woosub Roh, 15 Jan 2024
We appreciate the reviewer’s constructive comments.
Major comments:
It is not clear why the authors chose to use threshold of the Vdop to categorize hydrometeor types. From a storm resolving model, all the hydrometeors are known. However, the authors chose not to use that information, but applying thresholds directly to the Vdop from a forward model (or satellite simulator). Why does this manuscript do it this way? Is it because the EarthCARE retrieval algorithm have such a component? Even if it is trying to mimic a component in the retrieval algorithms, storm resolving model could be better used to provide context for retrievals. For example, snow and graupel (other hydrometeors too) must be mixed in a large number of model grid points. How could the categorization of Vdop (such as those shown in Fig. 6) could be used to single out the mass of a certain hydrometeor type? Do you need to consider a ratio of mixing of hydrometeor types? Is that the retrieval would provide, or not?
The presentation of the study and figures need to be improved. I suggest adding a figure showing the observed reflectivity and Doppler velocities. as well as the simulated reflectivity and Vdop. Figure 1 is only the precipitation rate. The text in Figs. 3, 6, 7, 8, 10, 11 are not illegible. A table could be necessary for comparisons. The 2% in Fig. 4 is not accurate (should it be 0.3% for grid points with absolute air motion > 0.5 m/s?). The magnitude of the model-resolved vertical air motion and the model horizontal resolution also needs to be elaborated a bit more. Ignoring vertical air motion, directly relating Fall velocity to Vdop is not a scientific sound method for interpretation. Much more justification needs to be made.
- We improved all figures except Figure 1.
- We added the explanations for the thresholds of Doppler velocity in the draft. For drizzle, according to the Glossary of Meteorology of the American Meteorological Society, the diameter of a drizzle is less than 0.5 mm, and the terminal velocity is 2.068 m/s with 0.5 mm at the surface based on Foote and Toit 1969. Mosimann 1995 investigated the degree of snow crystal riming using vertical Doppler radar. He found that the degree of riming is proportional to the Doppler velocity and that there is a large fraction of graupel with the Doppler velocity greater than 2 m/s (fig. 3 in Mosimann 1995). In this classification, we did not consider the effect of air density. This classification has uncertainty from vertical air motion and air density. We think the impact of these two terms is not significant. This study does not aim for an accurate classification of hydrometeors but rather for a quantitative intercomparison of models on the same basis. We also think the name of categorization is not perfect, because the Doppler velocity has information on mixtures of different hydrometeors. The naming of this classification is related to the average Doppler velocity to understand the model’s performance. Each model has their own categories of hydrometeors and characteristics like terminal velocity, density, and size distribution. The motivation of this study is that we need to compare hydrometeors with the same criterion as the Doppler velocity.We want to use the same categorization for the understanding of microphysics schemes or intercomparisons of GSRMs. We added the paragraph in the introduction part: One of the motivations of this study is to evaluate and compare the vertical distribution of hydrometeors of GSRMs using the same observational criteria. According to Roh et al. 2021, the horizontal distribution of outgoing longwave radiation of GSRMs is similar, but the simulated vertical distributions of hydrometeors of GSRMs are very different in the intercomparison data (Stevens et al. 2019). Each model used its own assumptions about the size distribution and terminal velocity of hydrometeors. We believe that the Doppler velocity is one of the criteria for understanding and constraining the vertical distributions of hydrometeors between GSRMs using observations.
- We investigated the impact of vertical air motion on the Doppler velocity using our satellite simulator. We removed the vertical air velocity about the calculation of Doppler velocity (Fig. 2 in the supplement). When we removed the vertical air velocity, the results were consistent with the control results (Fig. 1 in the supplement). However, the frequencies were concentrated in NSW6, and there was no fraction of the upward category. Most of the differences are less than 2% in the classifications.
The figures are attached in the supplement.
Figure 1: The categorizations of the hydrometeors in NICAM simulations for NSW6 (top) and NDW6 (bottom) in case 1 (left) and case 2 (right).
Figure 2: The same as Figure 1 but for only calculations of Doppler velocity without vertical air motion.
Specific comments:
Line 27: “... sampling of ...”, this sentence needs to be reworded. You mean footprint of space-borne radar? Spatial sampling scales?
- It is different from the footprint. For example, the horizontal sampling of CPR is approximately 500m, and the footprint is approximately 800m (e.g. Kollias et al. 2014). I changed “the along-track sampling”.
Line 34: “... in the same body ...” needs to be reworded. You mean same space craft?
- We changed it based on your comment.
Line 38: “synergetic”? Maybe using “synergistic”?
- We changed it based on your comment.
Line 90: I think the variable in Figure 1 is “precipitation rate”.
- We changed it based on your comment.
Line 102 to 107: You did not mention how the simulations do for Case 2. The simulations missed the part of precipitation over the ocean to the east near 36 and 37 north latitudes. This should be added.
- We added the explanation about your notice about precipitation.
Fig. 2 caption: What is the unit of the color scale? Percentage? Please be specific. Also note this figure is for observation.
- We added the unit of the color scale.
Line 119: It should be Figure 2 c and d.
- We changed it based on your comment.
Line 120: Why “Doppler velocity is free to attenuation”?
- For the precipitation area, the observed radar reflectivity is not reliable because of the attenuation. The Doppler velocity is not attenuated, but the data quality is not good in the highly attenuated areas.
Line 123: “Two high-frequency modes are near the melting layer...” This description is rather vague. It is just the difference above and below the melting layer.
- We modified to “there are two different modes above and below the melting layer.”
Line 126 “less than -2 m/s”? The figure shows “greater than -2 m/s”. Do you mean the absolute value?
- It is not an absolute value. The rimmed ice particle has a Doppler velocity of less than -2m/s, like -4 or -5m/s.
Line 128: if you talk about aliasing, I think you should give the CPR’s measurement range. When the velocity is out of the range, aliasing would happen.
- I explained the range of the Doppler velocity in the next sentence.
Line 129: Not using radar reflectivity is due to strong attenuation, right? You’d better mention this when you talk about the reflectivity attenuation earlier.
- We moved the sentence to the paragraph describing attenuation.
Last paragraph in Page 6: It looks like it is about Figure 3. Please refer to Figure 3.
- We referred to Figure 3 in the first sentence.
Comparing Fig. 3 and 2: why you use CFADs vs. joint histogram? Are they different in your paper? Give units to your plots.
- For understanding the vertical structure of the radar reflectivity and Doppler velocity, the CFADs are better because of the different sampling numbers per height. For the quantitative analysis, we thought the joint histogram was better than CFADs.
Comparing Fig. 3a to Fig. 2a: Are their differences solely due to unfolding? Why there is so much difference from 8 km and above?
- The difference is the normalization by each height or normalization by total height. The difference is the number of data samples at each height. The number of sampling data is not so many in case 1 above 8 km. So, the distribution is different between the CFADs and the joint histogram.
Line 162: reword it to “produces a large bias and makes the results unreliable.”
- We changed it based on your comment.
Line 164: Is it 2%? From the black solid line, I read 0.997, from the black dash and orange lines, I read 0.9985 to 0.999. So, it should be 0.3% or less. Why do you think it's 2%?
- We changed it by 0.2% based on your comment.
Line 164-165: are you talking about the simulated PDF of the Doppler velocity or the vertical air velocity?
- We talked about the simulated PDFs of the vertical air velocity.
Line 169: About the results are affected by the horizontal resolution of the model. The issue of the dependence of the vertical air motion on model resolution needs more details. Please restate the resolution of the model you are using and the coarse resolution you are referring to. Please also refer to previous studies about the dependence of vertical air motion on model horizontal resolution (e.g., Lebo and Morrison 2015 Monthly Weather Review page 4355-4375 or other study that you find appropriate.)
- Thank you for your interesting comments. We think the impact of horizontal resolution on the Doppler velocity and the contributions of vertical air velocity on the Doppler velocity is different between large-eddy simulation (LES) and cloud system–resolving model (CRM). When we reduce the horizontal resolution in our model (CRM), the contribution of vertical air velocity decreases (Fig. 3 in the supplement). We have also done LES with a different regional model and a different single-moment microphysics. We found that the LES with a 100-m resolution reproduced the fifth upward regime in the joint histogram well. We think the upward regime is mainly related to the turbulences in clouds (observation also shows a similar upward Doppler velocity related to turbulences). We will prepare a paper about this issue. The LES with a coarser resolution (250m resolution) has a lower fraction of the upward regime.
The figures are attached in the supplement.
Figure 3. The resolution dependency of the cumulative PDF of the absolute vertical air velocity for the sampling data with larger than -40 dBZ in NDW6 for case 1 (left) and case 2 (right).
Line 178: It is vague for rain vs cloud water vs drizzle. Please explicitly state how rain, cloud water, drizzle is separated in your categorization method. In Fig. 5 there is no cloud water or drizzle shown.
- The separation among cloud water, drizzle, and rain in this categorization method is the size of the liquid hydrometeors. The cloud water and drizzle have less than 0.5 mm diameter, and the terminal velocity is less than 2 m/s. In Fig. 5, the rain terminal velocity with less than 0.5 mm consists of the drizzle and cloud water categorization. However, two microphysics schemes have only a rain category in the scheme, which consists of rain and drizzle in this classification.
Figure 6: text on this figure is not readable. You need to adjust the location of the text for hydrometeor percentages, and the color of the text that labels the regions.
- We improved the figure based on your comment.
Line 193: You show the choice of microphysics scheme has more influence as compared to the case dependence. So, how this method could be used to provide useful information to the EarthCARE retrieval methods?
- We think these results are useful for understanding the uncertainty of the retrieval method. The retrieval method has its own assumptions about size distributions and terminal velocity. We think these results show the importance of the microphysical setting in developing the retrieval method.
Line 200: The cloud ice has 0 m/s terminal velocity as shown in Fig. 5 in the single moment scheme. How do you interpret the growth of snow from cloud ice from Fig. 6?
- In Fig. 6, the growth of snow from cloud ice is not clear. The maximum and minimum Doppler velocity show the increase of Doppler velocity from the top to the melting layer. The main reason is the autoconversion process from cloud ice to snow occurs at higher than 12 km (Fig. 7).
Line 218: “It decreases the sampling of liquid hydrometeors and upward motion (Fig. 7)” – Do you mean the attenuation from the spaceborne CPR in the liquid hydrometeor layer is more severe?
- It is not related to attenuation. The sampling areas of the satellite related to ice hydrometeors increased than the ground observation from 12 km to 20 km. There are more fractions of ice hydrometeors with 8 km, and it decreases the fraction of liquid hydrometeors in the total fraction.
Line 222: “we found an increase ....” no text is illegible in your figures, same here in Fig. 7 and 8. You need to use a table to show the comparisons.
- We improved the figures and added the tables from the figures.
Line 310: correct “thee” to “three”
- We changed it based on your comment.
References:
Foote, G. B., & Du Toit, P. S.: Terminal velocity of raindrops aloft. Journal of Applied Meteorology (1962-1982), 249-253.1969.
Mosimann, L.: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N, 1995.
Lebo, Z. J., H. Morrison, 2015, Effects of Horizontal and Vertical Grid Spacing on Mixing in Simulated Squall Lines and Implications for Convective Strength and Structure. Monthly Weather Review, 4355-4375.
Kollias, Pavlos, et al. "Evaluation of EarthCARE cloud profiling radar Doppler velocity measurements in particle sedimentation regimes." Journal of Atmospheric and Oceanic Technology 31.2 (2014): 366-386.
Roh, W., Satoh, M., & Hohenegger, C.: Intercomparison of cloud properties in DYAMOND simulations over the Atlantic Ocean. Journal of the Meteorological Society of Japan. Ser. II, 99(6), 1439-1451, 2021.
-
AC2: 'Reply on RC3', Woosub Roh, 15 Jan 2024
-
EC1: 'Comment on egusphere-2023-1997', Robin Hogan, 22 Jan 2024
Dear authors, thank you for your revision. The manuscript still needs a further round of reviews before it can be considered for publishing in AMT. In addition (and also for the benefit of the reviewers), please also take note the following comments, where the line and figure numbers refer to your revised version of the manuscript. Where I am referring to a comment from a reviewer, I indicate with R1, R2 or R3.1. Your response to the reviewers has muddled up their numbers: Reviewers 1, 2 and 3 you have called reviewers 3, 1 and 2 in your response. [I note that your response to reviewer 2 (whom you call reviewer 1) misquotes from a previous "reviewer 1" from your earlier submission to the AMT special issue.] In future responses, please use the same numbering as on the Copernicus web page for your manuscript.2. (R1) Regarding multiple scattering, it is reasonable to neglect it in this manuscript, but you need to at least discuss the issue and outline a plan for dealing with it in future when you have real EarthCARE data. For example, cite Battaglia et al. (JGR 2011) who proposed a way of identifying data likely to be affected by it.3. (R1) It is good that at least the simulated data will be released, but any revised version of the manuscript will need to cite the data in a public repository.4. (R2) It is not true that SPIDER is non-polarimetric - that is what the "P" stands for! However, the potential to improve the classification using linear depolarisation measurements may be limited.5. The word "data" is the plural of "datum". Therefore, strictly, you should say "Satellite data *have*" and similarly throughout the manuscript.6. New line 31 (34 in tracked-changes version): You say the EarthCARE satellite is "challenging", which implies "problematic"... surely this is not what you mean. Isn't this paragraph explaining the opportunities from EarthCARE rather than any problems?7. (R3) New line 131 (141 with tracked-changes) and elsewhere: The presence of riming, graupel and hail seems to be asserted much too readily based on a simple velocity threshold. You can only get riming if supercooled water is present, so the process cannot occur at temperatures colder than around -40 deg C yet the "riming region" seems to extend up to all heights, which is not realistic. Moreover, when interpreting the new Figs. 8a and 8c, and in section 3.3, you say that NDW6 has much more graupel than NSW6 just because the particles are falling faster. But this is the dominant mode in the Doppler velocity distribution between 5 and 10 km from a 24-hour sample: is graupel really dominating the distribution for such a long period? It seems much more likely that you are observing the ice/snow in the model that has been assigned too high a fall velocity. In any case, you have the model data so can check whether or not these values are really due to graupel rather than over-relying on a simple velocity threshold.8. New line 133 (145 with tracked-changes): You say that values greater than 2 m/s occur with high frequency - I can see some values here but they look quite rare, so not really "high frequency".9. Figure 4: I am puzzled by the y axis. I understand this is the fraction of absolute velocities between 0 m/s and the value on the x axis, in which case the value on the y axis at x=0 m/s indicates the fraction of absolute velocities between 0 and 0 m/s, which is surely y=0. So why is the value instead between 0.985 and 0.997?10. I can find no reference in the new text to Fig. 6.11. The captions to your new tables 1-3 need to clarify that the numbers are percentages, and you have written "faction" when you mean "fraction".Citation: https://doi.org/
10.5194/egusphere-2023-1997-EC1 -
AC5: 'Reply on EC1', Woosub Roh, 14 Mar 2024
Dear Editor,
We sincerely appreciate your insightful comments and the considerable effort you have invested in reviewing our manuscript.
- Your response to the reviewers has muddled up their numbers: Reviewers 1, 2 and 3 you have called reviewers 3, 1 and 2 in your response. [I note that your response to reviewer 2 (whom you call reviewer 1) misquotes from a previous "reviewer 1" from your earlier submission to the AMT special issue.] In future responses, please use the same numbering as on the Copernicus web page for your manuscript.
→ I am sorry for my mistake. I will use the same numbering.
- (R1) Regarding multiple scattering, it is reasonable to neglect it in this manuscript, but you need to at least discuss the issue and outline a plan for dealing with it in future when you have real EarthCARE data. For example, cite Battaglia et al. (JGR 2011) who proposed a way of identifying data likely to be affected by it.
→Thank you for your comments. I added this issue in the summary like “This study did not account for the complexities of multiple scattering effects (Battaglia et al., 2011) and pointing uncertainties (Tanelli et al., 2005) in simulating Doppler velocities. These aspects are crucial for accurately assessing the Doppler velocity capabilities on the EarthCARE, including the impact of Pulse Repetition Frequencies (PRFs). Understanding these influences is challenging prior to the satellite's launch. We will investigate these effects on the evaluation result in the future.”
- (R1) It is good that at least the simulated data will be released, but any revised version of the manuscript will need to cite the data in a public repository.
→ We uploaded the data and added “The snapshot data of simulated radar reflectivity and Doppler velocity of 94 GHz Cloud Profiling Radar (CPR) are available from https://doi.org/10.5281/zenodo.10813626 (Roh et al., 2024). We made Fig. 10 using these data.” in the data availability part.
- (R2) It is not true that SPIDER is non-polarimetric - that is what the "P" stands for! However, the potential to improve the classification using linear depolarisation measurements may be limited.
→ Thank you for your comment. I misunderstood the coauthors' comments. The HG-SPIDER was developed as a polarimetric radar. HG-SPIDER did not observe the depolarization ratio in these cases.
- The word "data" is the plural of "datum". Therefore, strictly, you should say "Satellite data *have*" and similarly throughout the manuscript.
→ We changed the sentence based on your comment.
- New line 31 (34 in tracked-changes version): You say the EarthCARE satellite is "challenging", which implies "problematic"... surely this is not what you mean. Isn't this paragraph explaining the opportunities from EarthCARE rather than any problems?
→ Thank you for your comment. I changed to “innovative” from “challenging”.
- (R3) New line 131 (141 with tracked-changes) and elsewhere: The presence of riming, graupel and hail seems to be asserted much too readily based on a simple velocity threshold. You can only get riming if supercooled water is present, so the process cannot occur at temperatures colder than around -40 deg C yet the "riming region" seems to extend up to all heights, which is not realistic. Moreover, when interpreting the new Figs. 8a and 8c, and in section 3.3, you say that NDW6 has much more graupel than NSW6 just because the particles are falling faster. But this is the dominant mode in the Doppler velocity distribution between 5 and 10 km from a 24-hour sample: is graupel really dominating the distribution for such a long period? It seems much more likely that you are observing the ice/snow in the model that has been assigned too high a fall velocity. In any case, you have the model data so can check whether or not these values are really due to graupel rather than over-relying on a simple velocity threshold.
→ It has uncertainty to characterize graupel or hail based on only Doppler velocity, but I think it's a better basis than radar reflectivity. The observation is consistent with the location of the riming area below 10 km. It is only dominant in the strong convection case. However, the simulations are different.
I agree with your concern about the misinterpretation of the simulation. We notice that simulated graupel and classified graupel by Doppler velocity are different. We explained, “NDW6 overestimates the graupel/hail regime associated with large snow or ice crystals. This result indicates that the terminal velocity of the snow is overestimated compared to the observation.” in new L252-254 (old L226-227).
We also checked the impact of the simulated graupel and snow on the joint histogram. The simulated large snow in NDW6 and NSW6 was classified as graupel/hail by the Doppler velocity threshold. NDW6 especially overestimates the size or terminal velocity of snow compared to observation. We think we need to improve the size distribution or terminal velocity of snow in NDW6 based on observation on the ground or the EarthCARE CPR.
- New line 133 (145 with tracked-changes): You say that values greater than 2 m/s occur with high frequency - I can see some values here but they look quite rare, so not really "high frequency".
→ Thank you for your comment. We removed “high’ in the text.
- Figure 4: I am puzzled by the y axis. I understand this is the fraction of absolute velocities between 0 m/s and the value on the x axis, in which case the value on the y axis at x=0 m/s indicates the fraction of absolute velocities between 0 and 0 m/s, which is surely y=0. So why is the value instead between 0.985 and 0.997?
→ The x=0 means the vertical air velocity is less than 0.2 m/s
- I can find no reference in the new text to Fig. 6.
→ There is the reference of Fig. 6 in Line 213.
- The captions to your new tables 1-3 need to clarify that the numbers are percentages, and you have written "faction" when you mean "fraction".
→ We added “%” in the tables.
Reference
Roh, W., & Satoh, M.: 94 GHz cloud radar simulation data using NICAM and Joint simulator for evaluation of ground-based radar and application to the EarthCARE satellite [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10813626, 2024.
Citation: https://doi.org/10.5194/egusphere-2023-1997-AC5
-
AC5: 'Reply on EC1', Woosub Roh, 14 Mar 2024
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-1997', Anonymous Referee #1, 27 Oct 2023
Review of “An evaluation of microphysics in a numerical model using Doppler velocity measured by ground-based radar for application to the EarthCARE satellite“ by Roh et al.
The authors present a study about the potential of EarthCARE for observing Doppler velocities. In general, the topic of the paper is interesting and relevant for ACP. I recommend the paper to be accepted after major revision.
- Language: The manuscript lacks clarity due to language problems. I recommend that the authors give the paper to a native speaker to make sure they write what they intent to say. Also, I would recommend to guide the reader better to explain why a analysis was performed in a certain way. For example, in section 4 I would recommend to stress that you start with an idealized simulation without instrument effects like the Nyquist range or random errors and then make the simulation more realistic step for step. Also, it should be stressed that the hydrometeor classification is not supposed to be a universally applicable one (at least I hope this is that case) but is only used to allow for a better comparison between model and observations.
- The authors study a case related to a tropical storm with potentially high reflectivities that might lead to multiple scattering, how would that impact the results?
- What about other sources for measurement errors? E.g. cloud inhomogeneity or pointing uncertainty?
- L 160: “We assumed the contribution of vertical air velocity to Doppler velocity is relatively smaller than the terminal velocity of hydrometeors” This is a strong assumption that needs to be supported. Alternative, the authors could remove convective data points using a filter like in e.g. Mosimann, 1995.
- L 260: Without averaging, the performance of the Doppler observations is quite bad for the high mode as can be seen in Fig. 11. Is that the main message the authors want to convey with this paper? Spatial averaging would improve the results, why was it not considered?
- Did the authors correct for the effect of changing air density on hydrometeor sedimentation velocity? Wouldn’t such a correction be necessary for a threshold based classification?
Minor comments:
- L 119: Fig 1 -> Fig 2c-d
- L 166; “When we use the threshold of 2 m/s for categorising the hydrometeors, 0.2 m/s of vertical air velocity affects the 10% bias”. Does that mean that the authors expect that 0.2 m/s vertical air motion lead to a 10% error of the classification? This is only true if the data points would be equally distributed with Doppler velocity which is not the case.
- L 239: Specify latitudes for low and high modes.
- Figs 1-3, 6-11. Please add labels to all x axis, y axis and colorbars
- 6: Why is there no data above 12 km?
- 8: Not referenced in the text
- Data availability: Where are the used simulations and radar observations available?
Mosimann, L., 1995: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N.
Citation: https://doi.org/10.5194/egusphere-2023-1997-RC1 -
AC1: 'Reply on RC1', Woosub Roh, 15 Jan 2024
We appreciate the reviewer’ constructive comments.
Major comments:
- The manuscript is difficult to read because of English grammatical issues. A fairly extensive grammatical revision is required.
As an example, the first paragraph of the introduction would benefit from these changes:
(a) Remove the word "The" from the first sentence
(b) "Satellite" and "Global" should be lower case (first and second sentences)
(c) "The detailed process of hydrometeors" in fourth sentence does not make senseThe grammatical errors are too extensive to list. I would recommend that they enlist some help in the interest of ensuring that they are communicating their work effectively to their audience.
--> This draft was checked by an English-speaking researcher based on your comment.
- The hydrometeor classification scheme (section 3.2) is abruptly stated, with no underlying justification offered. Where did these categories come from (specifically the apparently arbitrary absolute values that define the cutoffs)? How much uncertainty is present in these categories?
--> We added the explanations based on some references. According to the Glossary of Meteorology of the American Meteorological Society, the diameter of a drizzle is less than 0.5 mm, and the terminal velocity is 2.068 m/s with 0.5 mm at the surface based on Foote and Toit 1969. Mosimann 1995 investigated the degree of snow crystal riming using vertical Doppler radar. He found that the degree of riming is proportional to the Doppler velocity and that there is a large fraction of graupel with the Doppler velocity greater than 2 m/s (fig. 3 in Mosimann 1995). In this classification, we did not consider the effect of air density. This classification has uncertainty from vertical air motion and air density. We think the impact of these two terms is not significant. This study does not aim for an accurate classification of hydrometeors but rather for a quantitative intercomparison of models on the same basis.
- One major lingering question I have after reading the manuscript is as follows: For the NICAM simulations, the authors obviously know the distribution of hydrometers in the model. Why not compare their suggested classification scheme to what's actually present in the model? Maybe they did in fact do this, and I simply misunderstand the meaning of the figures. For example, in Figure 6(a), is 63.1% the fraction of cloud ice/snow derived from the DV-based classification scheme? Or is the cloud ice/snow fraction that was actually present in NICAM?
--> We think the accuracy of the classification’s names is not very important in this study. Microphysics scheme has different definitions of hydrometeors, their own terminal velocity, and size distributions. We think characteristics of vertical profiles of Doppler velocity in models related to terminal velocities of hydrometeors are more important. There are several uncertainties with this categorization. Even if it's cloud ice/snow, it's possible that there are mixtures of hydrometeors like small graupel. But we can understand that the average terminal velocity in that grid is high or low, and that's expected to have an impact on clouds and precipitation. We will investigate the impact of tunings of the Doppler velocity on radiation and large circulation in a global storm-resolving model (GSRM).
- The authors later use fall-speed relationships from the model to at least somewhat clarify the hydrometeor classification system that was chosen. This leads to the larger point that the chronology of the "story" that is being told by this manuscript is somewhat muddled. The introduction should spell this out much more clearly, including the hierarchy of simulations that will be used.
--> We added the motivation of this study about the evaluations of GSRMs using the Doppler velocity in detail in the introduction part: One of the motivations of this study is to evaluate and compare the vertical distribution of hydrometeors of GSRMs using the same observational criteria. According to Roh et al. 2021, the horizontal distribution of outgoing longwave radiation of GSRMs is similar, but the simulated vertical distributions of hydrometeors of GSRMs are very different in the intercomparison data (Stevens et al. 2019). Each model used its own assumptions about the size distribution and terminal velocity of hydrometeors. We believe that the Doppler velocity is one of the criteria for understanding and constraining the vertical distributions between GSRMs using observations.
- The radar used for ground-based observations (HG-SPIDER) is polarametric, correct? Why not use the polarametric data to better quantify the actual hydrometeor classifications, and therefore enhance the classification system and associated discussion the physics?
--> HG-SPIDER is not a polarimetric radar.
- The authors state that "We found that the frequency of absolute vertical velocity above 0.2 m/s is less than 2 %, and the simulated PDF of the Doppler velocity mostly depends on the cloud microphysics." I'm not convinced that the fact that absolute vertical velocities above 0.2 m/s are a small proportion of the total (Figure 4) means that air velocity can be neglected when translating Doppler velocity to particle fall speed. Since these results are from NICAM, why not just look at this relationship directly in the model? For example, heavy rain (although it's a small proportion of cases with reflectivity exceeding -40 dBZ) is more likely to occur in regions with significant vertical ascent.
--> We investigated the impact of vertical air motion on the Doppler velocity using the Joint-Simulator. We removed the vertical air velocity for the calculation of Doppler velocity (Fig. 2). When we removed the vertical air velocity, the results were consistent with the control results (Fig. 1). However, the frequencies were concentrated in NSW6, and there was no fraction of the upward category. Most of the difference is less than 2% in the classifications.
The figure files are attached in the supplement.
Figure 1: The categorizations of the hydrometeors in NICAM simulations for NSW6 (top) and NDW6 (bottom) in case 1 (left) and case 2 (right).
Figure 2: The same as Figure 1 but for only calculations of Doppler velocity without vertical air motion.
Specific comments:
- Line 121: Do you mean that Doppler radar is free of attenuation?
--> The radar reflectivity is attenuated. The Doppler velocity is not attenuated. However, the accuracy of Doppler velocity changes because of the attenuation.
- Line 182: Looks more like about 3 m/s, doesn't it?
--> The terminal velocity of NDW6 is less than 2m/s, and the terminal velocity of NSW6 less than 3 m/s. I added a sentence like “NSW6 shows the faster terminal velocity of raindrops with less than 0.5 mm diameter.”.
- Line 189: What is the "large data sampling"?
--> The observation data is every one minute. The model output data is every one hour. So we need to have a larger sampling of data for statistical analysis.
- Line 192: Check the 0.6% number.
--> I checked the number.
- Line 215: Can you better define what is meant by "observation window"?
--> The observation window is different from the radar range. The observation window means a collected data range. The observation window depends on the PRFs, number of integration of pulse(M), and satellite altitude. The PRFs and M changes by the lookup table related to the satellite altitude. The observation window of CloudSat is 30 km (Tanelli et al. 2008).
- Line 221: Please briefly explain the -15 dBZ (in addition to the reference).
--> The errors of the Dopler velocity depend on the signal-to-noise (SNR). The lower SNR means the higher contribution of the signal noise to Doppler velocity. According to Hagihara et al. (2021), the standard deviation of random errors increases significantly when the radar reflectivity is less than −15 dBZ (SNR =6.2 dB).
- NDW6 acronym is not defined.
--> We added the explanation about NDW6 like “the NICAM Double-moment Water 6-categories (Seiki and Nakajima 2014, hereafter referred to NDW6)”.
- Figures 2 and 3 have no axes labels.
--> We added the axes labels.
- There is no reference to Figures 3 or 8 in the text.
--> We added the references.
- The inline text in Figures 6, 7, 8, and 10 is unreadable in some cases, since the text overlaps the contours. In Figure 3, the "3" and "4" are difficult to discern in places.
--> We improved all figures except Figure 1.
References
Foote, G. B., & Du Toit, P. S. 1969: Terminal velocity of raindrops aloft. Journal of Applied Meteorology (1962-1982), 249-253.
Mosimann, L.: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N, 1995.
Hagihara, Y., Ohno, Y., Horie, H., Roh, W., Satoh, M., Kubota, T., and Oki, R.: Assessments of Doppler velocity errors of EarthCARE cloud profiling radar using global cloud system resolving simulations: Effects of Doppler broadening and folding, IEEE Trans. Geosci. Remote Sens., 60, 1–9, https://doi.org/10.1109/TGRS.2021.3060828, 2021.
Kobayashi, S., Kumagai, H., & Kuroiwa, H.: A proposal of pulse-pair Doppler operation on a spaceborne cloud-profiling radar in the W band. Journal of Atmospheric and Oceanic Technology, 19(9), 1294-1306, 2002.
Tanelli, S., Durden, S. L., Im, E., Pak, K. S., Reinke, D. G., Partain, P., ... & Marchand, R. T.: CloudSat's cloud profiling radar after two years in orbit: Performance, calibration, and processing.IEEE Transactions on Geoscience and Remote Sensing, 46(11), 3560-3573, 2008.
-
AC3: 'Reply on RC1', Woosub Roh, 15 Jan 2024
I am sorry for pasting another reviewer's response.
Thank you for your constructive comments.
Major comments:
Language: The manuscript lacks clarity due to language problems. I recommend that the authors give the paper to a native speaker to make sure they write what they intent to say. Also, I would recommend to guide the reader better to explain why a analysis was performed in a certain way. For example, in section 4 I would recommend to stress that you start with an idealized simulation without instrument effects like the Nyquist range or random errors and then make the simulation more realistic step for step. Also, it should be stressed that the hydrometeor classification is not supposed to be a universally applicable one (at least I hope this is that case) but is only used to allow for a better comparison between model and observations.
The authors study a case related to a tropical storm with potentially high reflectivities that might lead to multiple scattering, how would that impact the results? What about other sources for measurement errors? E.g. cloud inhomogeneity or pointing uncertainty?
- The revised draft was checked by an English-speaking researcher based on your comment. We agree that the multiple scattering impact also affects the results related to heavy precipitation cases. We think we can clearly understand the uncertainty from the multiple scattering after the launch of the satellite. Our expectation is that the multiple scattering is not significant because of the 800 m footprint and circular polarization. We need to filter out the data related to multiple scatterings to get better results. In this paper, we assume we use calibrated Doppler velocity from the multiple scattering and the point uncertainty for evaluations of a global storm-resolving model. We will filter out the data related to multiple scattering in simulations with the same criterion as the retrieval algorithm in the future.
- The cloud inhomogeneity is not important in this resolution with less than 500m. Now, we focus on the evaluations of a km scale global model. The purpose of this study is the application of the Doppler velocity for evaluations of modeling groups.
L 160: “We assumed the contribution of vertical air velocity to Doppler velocity is relatively smaller than the terminal velocity of hydrometeors” This is a strong assumption that needs to be supported. Alternative, the authors could remove convective data points using a filter like in e.g. Mosimann, 1995.
- We checked the upward motion using the observation data. The frequency of the upward motion is very rare. The time interval of our observation data is less than a second, and we used one-minute integrated data for the analysis. So, we think convective data points are reduced by the integration. Additionally, we investigated the impact of vertical air motion on the Doppler velocity using our satellite simulator. We removed the vertical air velocity for the calculation of Doppler velocity. When we removed the vertical air velocity, the results were consistent with the control results. However, the frequencies were concentrated, and there was no fraction of the upward category. Most of the difference is less than 2% in the diagrams. We added these results in the revised draft.
L 260: Without averaging, the performance of the Doppler observations is quite bad for the high mode as can be seen in Fig. 11. Is that the main message the authors want to convey with this paper? Spatial averaging would improve the results, why was it not considered?
- We checked the resolution dependency of the simulation results using NICAM. We found the impact of resolution dependency is not larger than the choice of microphysics schemes. So, we expect the 10km integration data to be useful for the evaluation or intercomparison of global storm-resolving models (GSRMs).
Did the authors correct for the effect of changing air density on hydrometeor sedimentation velocity? Wouldn’t such a correction be necessary for a threshold based classification?
- For the Doppler velocity, we considered the effect of changing air density. However, we did not consider the classification method. We think the air density affects the classification of the hydrometeors. We think the impact is not significant for the classification. Our purpose is a simple evaluation method for intercomparison or evaluation of GSRMs.
Specific comments:
L 119: Fig 1 -> Fig 2c-d
- We modified it based on your comment.
L 166; “When we use the threshold of 2 m/s for categorising the hydrometeors, 0.2 m/s of vertical air velocity affects the 10% bias”. Does that mean that the authors expect that 0.2 m/s vertical air motion lead to a 10% error of the classification? This is only true if the data points would be equally distributed with Doppler velocity which is not the case.
- We removed the explanation about that.
L 239: Specify latitudes for low and high modes.
- We specify the latitudes like “the high and low modes will be used depending on latitudes: low mode (−1 to 16 km) at latitudes of 60°–90° and high mode (−1 to 20 km) at latitudes of 0°–60° (Hagihara et al., 2022). ”
Figs 1-3, 6-11. Please add labels to all x axis, y axis and colorbars
- We added the labels to all x-axis, y-axis, and color bars.
6: Why is there no data above 12 km?
- The vertical range of the observation data is until 12 km.
8: Not referenced in the text
Data availability: Where are the used simulations and radar observations available?
- Simulation data will be available, but we need to discuss to open our observation data to the public. We will upload the available data before the publication.
Mosimann, L., 1995: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N.
Citation: https://doi.org/10.5194/egusphere-2023-1997-AC3
-
RC2: 'Comment on egusphere-2023-1997', Anonymous Referee #2, 04 Nov 2023
SUMMARY AND OVERALL ASSESSMENT
The manuscript deals with several aspects of how new EarthCare Doppler velocity (hereafter DV) measurements might address problems of hydrometeor identification using a variety of observations and simulations.
There are a number of interesting results in the manuscript, but ultimately I don't think the methodology or goals were explained well enough for me to have a clear understanding of what this study is really trying to communicate. For example, is the focus of this paper to define a DV-based hydrometeor classification system, or is that just a tool? I'm really unclear on the underlying "story" this work is telling. More clearly defining the goals of the study, and steps to reach that goal, will go a long way toward bringing this study into a publishable state.
MAJOR COMMENTS
1. The manuscript is difficult to read because of English grammatical issues. A fairly extensive grammatical revision is required.
As an example, the first paragraph of the introduction would benefit from these changes:
(a) Remove the word "The" from the first sentence
(b) "Satellite" and "Global" should be lower case (first and second sentences)
(c) "The detailed process of hydrometeors" in fourth sentence does not make senseThe grammatical errors are too extensive to list. I would recommend that they enlist some help in the interest of ensuring that they are communicating their work effectively to their audience.
2. The hydrometeor classification scheme (section 3.2) is abruptly stated, with no underlying justification offered. Where did these categories come from (specifically the apparently arbitrary absolute values that define the cutoffs)? How much uncertainty is present in these categories?
3. One major lingering question I have after reading the manuscript is as follows: For the NICAM simulations, the authors obviously know the distribution of hydrometers in the model. Why not compare their suggested classification scheme to what's actually present in the model? Maybe they did in fact do this, and I simply misunderstand the meaning of the figures. For example, in Figure 6(a), is 63.1% the fraction of cloud ice/snow derived from the DV-based classification scheme? Or is the cloud ice/snow fraction that was actually present in NICAM?
4. The authors later use fall-speed relationships from the model to at least somewhat clarify the hydrometeor classification system that was chosen. This leads to the larger point that the chronology of the "story" that is being told by this manuscript is somewhat muddled. The introduction should spell this out much more clearly, including the hierarchy of simulations that will be used.
5. The radar used for ground-based observations (HG-SPIDER) is polarametric, correct? Why not use the polarametric data to better quantify the actual hydrometeor classifications, and therefore enhance the classification system and associated discussion the physics?
6. The authors state that "We found that the frequency of absolute vertical velocity above 0.2 m/s is less than 2 %, and the simulated PDF of the Doppler velocity mostly depends on the cloud microphysics." I'm not convinced that the fact that absolute vertical velocities above 0.2 m/s are a small proportion of the total (Figure 4) means that air velocity can be neglected when translating Doppler velocity to particle fall speed. Since these results are from NICAM, why not just look at this relationship directly in the model? For example, heavy rain (although it's a small proportion of cases with reflectivity exceeding -40 dBZ) is more likely to occur in regions with significant vertical ascent.
7. What procedure is used to produce Figure 6, and how does it differ from the procedure used to produce Figure 7? Figure 6 apparently does not use the Joint simulator, so what does it use? Why does Figure 6 cut off at about 12 km when NICAM clearly produces results above 12 km?
The rather significant changes between Figures 6 and 7 are dealt with in a cursory way, but the changes are significant and what causes them need to be better explained.
MINOR COMMENTS
- Line 121: Do you mean that Doppler radar is free of attenuation?
- Line 182: Looks more like about 3 m/s, doesn't it?
- Line 189: What is the "large data sampling"?
- Line 192: Check the 0.6% number.
- Line 215: Can you better define what is meant by "observation window"?
- Line 221: Please briefly explain the -15 dBZ (in addition to the reference).
- NDW6 acronym is not defined.
- Figures 2 and 3 have no axes labels.
- There is no reference to Figures 3 or 8 in the text.
- The inline text in Figures 6, 7, 8, and 10 is unreadable in some cases, since the text overlaps the contours. In Figure 3, the "3" and "4" are difficult to discern in places.
Citation: https://doi.org/10.5194/egusphere-2023-1997-RC2 -
AC4: 'Reply on RC2', Woosub Roh, 15 Jan 2024
We appreciate the reviewer’s constructive comments.
Major comments:
1. The manuscript is difficult to read because of English grammatical issues. A fairly extensive grammatical revision is required.
As an example, the first paragraph of the introduction would benefit from these changes:
(a) Remove the word "The" from the first sentence
(b) "Satellite" and "Global" should be lower case (first and second sentences)
(c) "The detailed process of hydrometeors" in fourth sentence does not make senseThe grammatical errors are too extensive to list. I would recommend that they enlist some help in the interest of ensuring that they are communicating their work effectively to their audience.
- The revised draft was checked by an English-speaking researcher based on your comment.
2. The hydrometeor classification scheme (section 3.2) is abruptly stated, with no underlying justification offered. Where did these categories come from (specifically the apparently arbitrary absolute values that define the cutoffs)? How much uncertainty is present in these categories?
- We added the explanations based on some references. According to the Glossary of Meteorology of the American Meteorological Society, the diameter of a drizzle is less than 0.5 mm, and the terminal velocity is 2.068 m/s with 0.5 mm at the surface based on Foote and Toit 1969. Mosimann 1995 investigated the degree of snow crystal riming using vertical Doppler radar. He found that the degree of riming is proportional to the Doppler velocity and that there is a large fraction of graupel with the Doppler velocity greater than 2 m/s (fig. 3 in Mosimann 1995). In this classification, we did not consider the effect of air density. This classification has uncertainty from vertical air motion and air density. We think the impact of these two terms is not significant. This study does not aim for an accurate classification of hydrometeors but rather for a quantitative intercomparison of models on the same basis.
3. One major lingering question I have after reading the manuscript is as follows: For the NICAM simulations, the authors obviously know the distribution of hydrometers in the model. Why not compare their suggested classification scheme to what's actually present in the model? Maybe they did in fact do this, and I simply misunderstand the meaning of the figures. For example, in Figure 6(a), is 63.1% the fraction of cloud ice/snow derived from the DV-based classification scheme? Or is the cloud ice/snow fraction that was actually present in NICAM?
- We think the accuracy of the classification’s names is not very important in this study. Microphysics scheme has different definitions of hydrometeors, their own terminal velocity, and size distributions. We think characteristics of vertical profiles of Doppler velocity in models related to terminal velocities of hydrometeors are more important. There are several uncertainties with this categorization. Even if it's cloud ice/snow, it's possible that there are mixtures of hydrometeors like small graupel. But we can understand that the average terminal velocity in that grid is high or low, and that's expected to have an impact on clouds and precipitation. We will investigate the impact of tunings of the Doppler velocity on radiation and large circulation in a global storm-resolving model (GSRM).
4. The authors later use fall-speed relationships from the model to at least somewhat clarify the hydrometeor classification system that was chosen. This leads to the larger point that the chronology of the "story" that is being told by this manuscript is somewhat muddled. The introduction should spell this out much more clearly, including the hierarchy of simulations that will be used.
- We added the motivation of this study about the evaluations of GSRMs using the Doppler velocity in detail in the introduction part: One of the motivations of this study is to evaluate and compare the vertical distribution of hydrometeors of GSRMs using the same observational criteria. According to Roh et al. 2021, the horizontal distribution of outgoing longwave radiation of GSRMs is similar, but the simulated vertical distributions of hydrometeors of GSRMs are very different in the intercomparison data (Stevens et al. 2019). Each model used its own assumptions about the size distribution and terminal velocity of hydrometeors. We believe that the Doppler velocity is one of the criteria for understanding and constraining the vertical distributions between GSRMs using observations.
5. The radar used for ground-based observations (HG-SPIDER) is polarametric, correct? Why not use the polarametric data to better quantify the actual hydrometeor classifications, and therefore enhance the classification system and associated discussion the physics?
- HG-SPIDER is not a polarimetric radar.
6. The authors state that "We found that the frequency of absolute vertical velocity above 0.2 m/s is less than 2 %, and the simulated PDF of the Doppler velocity mostly depends on the cloud microphysics." I'm not convinced that the fact that absolute vertical velocities above 0.2 m/s are a small proportion of the total (Figure 4) means that air velocity can be neglected when translating Doppler velocity to particle fall speed. Since these results are from NICAM, why not just look at this relationship directly in the model? For example, heavy rain (although it's a small proportion of cases with reflectivity exceeding -40 dBZ) is more likely to occur in regions with significant vertical ascent.
- We investigated the impact of vertical air motion on the Doppler velocity using the Joint-Simulator. We removed the vertical air velocity for the calculation of Doppler velocity (Fig. 2). When we removed the vertical air velocity, the results were consistent with the control results (Fig. 1). However, the frequencies were concentrated in NSW6, and there was no fraction of the upward category. Most of the difference is less than 2% in the classifications.
The figure files are attached in the supplement.
Figure 1: The categorizations of the hydrometeors in NICAM simulations for NSW6 (top) and NDW6 (bottom) in case 1 (left) and case 2 (right).
Figure 2: The same as Figure 1 but for only calculations of Doppler velocity without vertical air motion.
7. What procedure is used to produce Figure 6, and how does it differ from the procedure used to produce Figure 7? Figure 6 apparently does not use the Joint simulator, so what does it use? Why does Figure 6 cut off at about 12 km when NICAM clearly produces results above 12 km?
The rather significant changes between Figures 6 and 7 are dealt with in a cursory way, but the changes are significant and what causes them need to be better explained.
- We used the Joint Simulator for Figure 6 and Figure 7. The differences between Fig. 6 and Fig. 7 are the setting of ground observation and the EarthCARE satellite. The observation range of the ground observation is up to 12 km, and the vertical resolution is different from EarthCARE. The CFADs of radar reflectivity are different because of the attenuation of rain. However, the results of Doppler velocity are very similar to ground observation. We expect there is an impact on data of Doppler velocity larger than -15 dBZ because of attenuation. Before we introduce the impact of random errors based on the observation window, we need to introduce the simulated Doppler velocity like EarthCARE.
Specific comments:
- Line 121: Do you mean that Doppler radar is free of attenuation?
- The radar reflectivity is attenuated. The Doppler velocity is not attenuated. However, the accuracy of Doppler velocity changes because of the attenuation.
- Line 182: Looks more like about 3 m/s, doesn't it?
- The terminal velocity of NDW6 is less than 2m/s, and the terminal velocity of NSW6 less than 3 m/s. I added a sentence like “NSW6 shows the faster terminal velocity of raindrops with less than 0.5 mm diameter.”.
- Line 189: What is the "large data sampling"?
- The observation data is every one minute. The model output data is every one hour. So we need to have a larger sampling of data for statistical analysis
- Line 192: Check the 0.6% number.
- I checked the number.
- Line 215: Can you better define what is meant by "observation window"?
- The observation window is different from the radar range. The observation window means a collected data range. The observation window depends on the PRFs, number of integration of pulse(M), and satellite altitude. The PRFs and M changes by the lookup table related to the satellite altitude. The observation window of CloudSat is 30 km (Tanelli et al. 2008).
- Line 221: Please briefly explain the -15 dBZ (in addition to the reference).
- The errors of the Dopler velocity depend on the signal-to-noise (SNR). The lower SNR means the higher contribution of the signal noise to Doppler velocity. According to Hagihara et al. (2021), the standard deviation of random errors increases significantly when the radar reflectivity is less than −15 dBZ (SNR =6.2 dB).
- NDW6 acronym is not defined.
- We added the explanation about NDW6 like “the NICAM Double-moment Water 6-categories (Seiki and Nakajima 2014, hereafter referred to NDW6)”.
- Figures 2 and 3 have no axes labels.
- We added the axes labels.
- There is no reference to Figures 3 or 8 in the text.
- We added the references.
- The inline text in Figures 6, 7, 8, and 10 is unreadable in some cases, since the text overlaps the contours. In Figure 3, the "3" and "4" are difficult to discern in places.
- We improved all figures except Figure 1.
References
Foote, G. B., & Du Toit, P. S. 1969: Terminal velocity of raindrops aloft. Journal of Applied Meteorology (1962-1982), 249-253.
Mosimann, L.: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N, 1995.
Hagihara, Y., Ohno, Y., Horie, H., Roh, W., Satoh, M., Kubota, T., and Oki, R.: Assessments of Doppler velocity errors of EarthCARE cloud profiling radar using global cloud system resolving simulations: Effects of Doppler broadening and folding, IEEE Trans. Geosci. Remote Sens., 60, 1–9, https://doi.org/10.1109/TGRS.2021.3060828, 2021.
Kobayashi, S., Kumagai, H., & Kuroiwa, H.: A proposal of pulse-pair Doppler operation on a spaceborne cloud-profiling radar in the W band. Journal of Atmospheric and Oceanic Technology, 19(9), 1294-1306, 2002.
Tanelli, S., Durden, S. L., Im, E., Pak, K. S., Reinke, D. G., Partain, P., ... & Marchand, R. T.: CloudSat's cloud profiling radar after two years in orbit: Performance, calibration, and processing.IEEE Transactions on Geoscience and Remote Sensing, 46(11), 3560-3573, 2008.
-
AC4: 'Reply on RC2', Woosub Roh, 15 Jan 2024
-
RC3: 'Comment on egusphere-2023-1997', Anonymous Referee #3, 08 Nov 2023
This study aims to use model simulations of two precipitation events by NICAM and a satellite simulator to investigate how spaceborne Doppler velocity (Vdop) measurement could be useful to evaluate microphysics, with an emphasis on categorization of hydrometeor types using thresholds in Doppler velocities. The topic is relevant to the satellite mission, EarthCARE. The language in the manuscript, , the justification of the scientific methods, and the presentation of the study needs to be significant improved before it can be considered for publication.
It is not clear why the authors chose to use threshold of the Vdop to categorize hydrometeor types. From a storm resolving model, all the hydrometeors are known. However, the authors chose not to use that information, but applying thresholds directly to the Vdop from a forward model (or satellite simulator). Why does this manuscript do it this way? Is it because the EarthCARE retrieval algorithm have such a component? Even if it is trying to mimic a component in the retrieval algorithms, storm resolving model could be better used to provide context for retrievals. For example, snow and graupel (other hydrometeors too) must be mixed in a large number of model grid points. How could the categorization of Vdop (such as those shown in Fig. 6) could be used to single out the mass of a certain hydrometeor type? Do you need to consider a ratio of mixing of hydrometeor types? Is that the retrieval would provide, or not?
The presentation of the study and figures need to be improved. I suggest adding a figure showing the observed reflectivity and Doppler velocities. as well as the simulated reflectivity and Vdop. Figure 1 is only the precipitation rate. The text in Figs. 3, 6, 7, 8, 10, 11 are not illegible. A table could be necessary for comparisons. The 2% in Fig. 4 is not accurate (should it be 0.3% for grid points with absolute air motion > 0.5 m/s?). The magnitude of the model-resolved vertical air motion and the model horizontal resolution also needs to be elaborated a bit more. Ignoring vertical air motion, directly relating Fall velocity to Vdop is not a scientific sound method for interpretation. Much more justification needs to be made.
Below are some editing I made while read the manuscript.
Line 27: “... sampling of ...”, this sentence needs to be reworded. You mean footprint of space-borne radar? Spatial sampling scales?
Line 34: “... in the same body ...” needs to be reworded. You mean same space craft?
Line 38: “synergetic”? Maybe using “synergistic”?
Line 90: I think the variable in Figure 1 is “precipitation rate”.
Line 102 to 107: You did not mention how the simulations do for Case 2. The simulations missed the part of precipitation over the ocean to the east near 36 and 37 north latitudes. This should be added.
Fig. 2 caption: What is the unit of the color scale? Percentage? Please be specific. Also note this figure is for observation.
Line 119: It should be Figure 2 c and d.
Line 120: Why “Doppler velocity is free to attenuation”?
Line 123: “Two high-frequency modes are near the melting layer...” This description is rather vague. It is just the difference above and below the melting layer.
Line 126 “less than -2 m/s”? The figure shows “greater than -2 m/s”. Do you mean the absolute value?
Line 128: if you talk about aliasing, I think you should give the CPR’s measurement range. When the velocity is out of the range, aliasing would happen.
Line 129: Not using radar reflectivity is due to strong attenuation, right? You’d better mention this when you talk about the reflectivity attenuation earlier.
Last paragraph in Page 6: It looks like it is about Figure 3. Please refer to Figure 3.
Comparing Fig. 3 and 2: why you use CFADs vs. joint histogram? Are they different in your paper? Give units to your plots.
Comparing Fig. 3a to Fig. 2a: Are their differences solely due to unfolding? Why there is so much difference from 8 km and above?
Line 162: reword it to “produces a large bias and makes the results unreliable.”
Line 164: Is it 2%? From the black solid line, I read 0.997, from the black dash and orange lines, I read 0.9985 to 0.999. So, it should be 0.3% or less. Why do you think it's 2%?
Line 164-165: are you talking about the simulated PDF of the Doppler velocity or the vertical air velocity?
Line 169: About the results are affected by the horizontal resolution of the model. The issue of the dependence of the vertical air motion on model resolution needs more details. Please restate the resolution of the model you are using and the coarse resolution you are referring to. Please also refer to previous studies about the dependence of vertical air motion on model horizontal resolution (e.g., Lebo and Morrison 2015 Monthly Weather Review page 4355-4375 or other study that you find appropriate.)
Line 178: It is vague for rain vs cloud water vs drizzle. Please explicitly state how rain, cloud water, drizzle is separated in your categorization method. In Fig. 5 there is no cloud water or drizzle shown.
Figure 6: text on this figure is not readable. You need to adjust the location of the text for hydrometeor percentages, and the color of the text that labels the regions.
Line 193: You show the choice of microphysics scheme has more influence as compared to the case dependence. So, how this method could be used to provide useful information to the EarthCARE retrieval methods?
Line 200: The cloud ice has 0 m/s terminal velocity as shown in Fig. 5 in the single moment scheme. How do you interpret the growth of snow from cloud ice from Fig. 6?
Line218: “It decrease the sampling of liquid hydrometeors and upward motion (Fig. 7)” – Do you mean the attenuation from the spaceborne CPR in the liquid hydrometeor layer is more severe?
Line 222: “we found an increase ....” no text is illegible in your figures, same here in Fig. 7 and 8. You need to use a table to show the comparisons.
Line 310: correct “thee” to “three”
References:
Lebo, Z. J., H. Morrison, 2015, Effects of Horizontal and Vertical Grid Spacing on Mixing in Simulated Squall Lines and Implications for Convective Strength and Structure. Monthly Weather Review, 4355-4375.
Citation: https://doi.org/10.5194/egusphere-2023-1997-RC3 -
AC2: 'Reply on RC3', Woosub Roh, 15 Jan 2024
We appreciate the reviewer’s constructive comments.
Major comments:
It is not clear why the authors chose to use threshold of the Vdop to categorize hydrometeor types. From a storm resolving model, all the hydrometeors are known. However, the authors chose not to use that information, but applying thresholds directly to the Vdop from a forward model (or satellite simulator). Why does this manuscript do it this way? Is it because the EarthCARE retrieval algorithm have such a component? Even if it is trying to mimic a component in the retrieval algorithms, storm resolving model could be better used to provide context for retrievals. For example, snow and graupel (other hydrometeors too) must be mixed in a large number of model grid points. How could the categorization of Vdop (such as those shown in Fig. 6) could be used to single out the mass of a certain hydrometeor type? Do you need to consider a ratio of mixing of hydrometeor types? Is that the retrieval would provide, or not?
The presentation of the study and figures need to be improved. I suggest adding a figure showing the observed reflectivity and Doppler velocities. as well as the simulated reflectivity and Vdop. Figure 1 is only the precipitation rate. The text in Figs. 3, 6, 7, 8, 10, 11 are not illegible. A table could be necessary for comparisons. The 2% in Fig. 4 is not accurate (should it be 0.3% for grid points with absolute air motion > 0.5 m/s?). The magnitude of the model-resolved vertical air motion and the model horizontal resolution also needs to be elaborated a bit more. Ignoring vertical air motion, directly relating Fall velocity to Vdop is not a scientific sound method for interpretation. Much more justification needs to be made.
- We improved all figures except Figure 1.
- We added the explanations for the thresholds of Doppler velocity in the draft. For drizzle, according to the Glossary of Meteorology of the American Meteorological Society, the diameter of a drizzle is less than 0.5 mm, and the terminal velocity is 2.068 m/s with 0.5 mm at the surface based on Foote and Toit 1969. Mosimann 1995 investigated the degree of snow crystal riming using vertical Doppler radar. He found that the degree of riming is proportional to the Doppler velocity and that there is a large fraction of graupel with the Doppler velocity greater than 2 m/s (fig. 3 in Mosimann 1995). In this classification, we did not consider the effect of air density. This classification has uncertainty from vertical air motion and air density. We think the impact of these two terms is not significant. This study does not aim for an accurate classification of hydrometeors but rather for a quantitative intercomparison of models on the same basis. We also think the name of categorization is not perfect, because the Doppler velocity has information on mixtures of different hydrometeors. The naming of this classification is related to the average Doppler velocity to understand the model’s performance. Each model has their own categories of hydrometeors and characteristics like terminal velocity, density, and size distribution. The motivation of this study is that we need to compare hydrometeors with the same criterion as the Doppler velocity.We want to use the same categorization for the understanding of microphysics schemes or intercomparisons of GSRMs. We added the paragraph in the introduction part: One of the motivations of this study is to evaluate and compare the vertical distribution of hydrometeors of GSRMs using the same observational criteria. According to Roh et al. 2021, the horizontal distribution of outgoing longwave radiation of GSRMs is similar, but the simulated vertical distributions of hydrometeors of GSRMs are very different in the intercomparison data (Stevens et al. 2019). Each model used its own assumptions about the size distribution and terminal velocity of hydrometeors. We believe that the Doppler velocity is one of the criteria for understanding and constraining the vertical distributions of hydrometeors between GSRMs using observations.
- We investigated the impact of vertical air motion on the Doppler velocity using our satellite simulator. We removed the vertical air velocity about the calculation of Doppler velocity (Fig. 2 in the supplement). When we removed the vertical air velocity, the results were consistent with the control results (Fig. 1 in the supplement). However, the frequencies were concentrated in NSW6, and there was no fraction of the upward category. Most of the differences are less than 2% in the classifications.
The figures are attached in the supplement.
Figure 1: The categorizations of the hydrometeors in NICAM simulations for NSW6 (top) and NDW6 (bottom) in case 1 (left) and case 2 (right).
Figure 2: The same as Figure 1 but for only calculations of Doppler velocity without vertical air motion.
Specific comments:
Line 27: “... sampling of ...”, this sentence needs to be reworded. You mean footprint of space-borne radar? Spatial sampling scales?
- It is different from the footprint. For example, the horizontal sampling of CPR is approximately 500m, and the footprint is approximately 800m (e.g. Kollias et al. 2014). I changed “the along-track sampling”.
Line 34: “... in the same body ...” needs to be reworded. You mean same space craft?
- We changed it based on your comment.
Line 38: “synergetic”? Maybe using “synergistic”?
- We changed it based on your comment.
Line 90: I think the variable in Figure 1 is “precipitation rate”.
- We changed it based on your comment.
Line 102 to 107: You did not mention how the simulations do for Case 2. The simulations missed the part of precipitation over the ocean to the east near 36 and 37 north latitudes. This should be added.
- We added the explanation about your notice about precipitation.
Fig. 2 caption: What is the unit of the color scale? Percentage? Please be specific. Also note this figure is for observation.
- We added the unit of the color scale.
Line 119: It should be Figure 2 c and d.
- We changed it based on your comment.
Line 120: Why “Doppler velocity is free to attenuation”?
- For the precipitation area, the observed radar reflectivity is not reliable because of the attenuation. The Doppler velocity is not attenuated, but the data quality is not good in the highly attenuated areas.
Line 123: “Two high-frequency modes are near the melting layer...” This description is rather vague. It is just the difference above and below the melting layer.
- We modified to “there are two different modes above and below the melting layer.”
Line 126 “less than -2 m/s”? The figure shows “greater than -2 m/s”. Do you mean the absolute value?
- It is not an absolute value. The rimmed ice particle has a Doppler velocity of less than -2m/s, like -4 or -5m/s.
Line 128: if you talk about aliasing, I think you should give the CPR’s measurement range. When the velocity is out of the range, aliasing would happen.
- I explained the range of the Doppler velocity in the next sentence.
Line 129: Not using radar reflectivity is due to strong attenuation, right? You’d better mention this when you talk about the reflectivity attenuation earlier.
- We moved the sentence to the paragraph describing attenuation.
Last paragraph in Page 6: It looks like it is about Figure 3. Please refer to Figure 3.
- We referred to Figure 3 in the first sentence.
Comparing Fig. 3 and 2: why you use CFADs vs. joint histogram? Are they different in your paper? Give units to your plots.
- For understanding the vertical structure of the radar reflectivity and Doppler velocity, the CFADs are better because of the different sampling numbers per height. For the quantitative analysis, we thought the joint histogram was better than CFADs.
Comparing Fig. 3a to Fig. 2a: Are their differences solely due to unfolding? Why there is so much difference from 8 km and above?
- The difference is the normalization by each height or normalization by total height. The difference is the number of data samples at each height. The number of sampling data is not so many in case 1 above 8 km. So, the distribution is different between the CFADs and the joint histogram.
Line 162: reword it to “produces a large bias and makes the results unreliable.”
- We changed it based on your comment.
Line 164: Is it 2%? From the black solid line, I read 0.997, from the black dash and orange lines, I read 0.9985 to 0.999. So, it should be 0.3% or less. Why do you think it's 2%?
- We changed it by 0.2% based on your comment.
Line 164-165: are you talking about the simulated PDF of the Doppler velocity or the vertical air velocity?
- We talked about the simulated PDFs of the vertical air velocity.
Line 169: About the results are affected by the horizontal resolution of the model. The issue of the dependence of the vertical air motion on model resolution needs more details. Please restate the resolution of the model you are using and the coarse resolution you are referring to. Please also refer to previous studies about the dependence of vertical air motion on model horizontal resolution (e.g., Lebo and Morrison 2015 Monthly Weather Review page 4355-4375 or other study that you find appropriate.)
- Thank you for your interesting comments. We think the impact of horizontal resolution on the Doppler velocity and the contributions of vertical air velocity on the Doppler velocity is different between large-eddy simulation (LES) and cloud system–resolving model (CRM). When we reduce the horizontal resolution in our model (CRM), the contribution of vertical air velocity decreases (Fig. 3 in the supplement). We have also done LES with a different regional model and a different single-moment microphysics. We found that the LES with a 100-m resolution reproduced the fifth upward regime in the joint histogram well. We think the upward regime is mainly related to the turbulences in clouds (observation also shows a similar upward Doppler velocity related to turbulences). We will prepare a paper about this issue. The LES with a coarser resolution (250m resolution) has a lower fraction of the upward regime.
The figures are attached in the supplement.
Figure 3. The resolution dependency of the cumulative PDF of the absolute vertical air velocity for the sampling data with larger than -40 dBZ in NDW6 for case 1 (left) and case 2 (right).
Line 178: It is vague for rain vs cloud water vs drizzle. Please explicitly state how rain, cloud water, drizzle is separated in your categorization method. In Fig. 5 there is no cloud water or drizzle shown.
- The separation among cloud water, drizzle, and rain in this categorization method is the size of the liquid hydrometeors. The cloud water and drizzle have less than 0.5 mm diameter, and the terminal velocity is less than 2 m/s. In Fig. 5, the rain terminal velocity with less than 0.5 mm consists of the drizzle and cloud water categorization. However, two microphysics schemes have only a rain category in the scheme, which consists of rain and drizzle in this classification.
Figure 6: text on this figure is not readable. You need to adjust the location of the text for hydrometeor percentages, and the color of the text that labels the regions.
- We improved the figure based on your comment.
Line 193: You show the choice of microphysics scheme has more influence as compared to the case dependence. So, how this method could be used to provide useful information to the EarthCARE retrieval methods?
- We think these results are useful for understanding the uncertainty of the retrieval method. The retrieval method has its own assumptions about size distributions and terminal velocity. We think these results show the importance of the microphysical setting in developing the retrieval method.
Line 200: The cloud ice has 0 m/s terminal velocity as shown in Fig. 5 in the single moment scheme. How do you interpret the growth of snow from cloud ice from Fig. 6?
- In Fig. 6, the growth of snow from cloud ice is not clear. The maximum and minimum Doppler velocity show the increase of Doppler velocity from the top to the melting layer. The main reason is the autoconversion process from cloud ice to snow occurs at higher than 12 km (Fig. 7).
Line 218: “It decreases the sampling of liquid hydrometeors and upward motion (Fig. 7)” – Do you mean the attenuation from the spaceborne CPR in the liquid hydrometeor layer is more severe?
- It is not related to attenuation. The sampling areas of the satellite related to ice hydrometeors increased than the ground observation from 12 km to 20 km. There are more fractions of ice hydrometeors with 8 km, and it decreases the fraction of liquid hydrometeors in the total fraction.
Line 222: “we found an increase ....” no text is illegible in your figures, same here in Fig. 7 and 8. You need to use a table to show the comparisons.
- We improved the figures and added the tables from the figures.
Line 310: correct “thee” to “three”
- We changed it based on your comment.
References:
Foote, G. B., & Du Toit, P. S.: Terminal velocity of raindrops aloft. Journal of Applied Meteorology (1962-1982), 249-253.1969.
Mosimann, L.: An improved method for determining the degree of snow crystal riming by vertical Doppler radar. Atmos. Res., 37, 305–323, doi:10.1016/0169-8095(94)00050-N, 1995.
Lebo, Z. J., H. Morrison, 2015, Effects of Horizontal and Vertical Grid Spacing on Mixing in Simulated Squall Lines and Implications for Convective Strength and Structure. Monthly Weather Review, 4355-4375.
Kollias, Pavlos, et al. "Evaluation of EarthCARE cloud profiling radar Doppler velocity measurements in particle sedimentation regimes." Journal of Atmospheric and Oceanic Technology 31.2 (2014): 366-386.
Roh, W., Satoh, M., & Hohenegger, C.: Intercomparison of cloud properties in DYAMOND simulations over the Atlantic Ocean. Journal of the Meteorological Society of Japan. Ser. II, 99(6), 1439-1451, 2021.
-
AC2: 'Reply on RC3', Woosub Roh, 15 Jan 2024
-
EC1: 'Comment on egusphere-2023-1997', Robin Hogan, 22 Jan 2024
Dear authors, thank you for your revision. The manuscript still needs a further round of reviews before it can be considered for publishing in AMT. In addition (and also for the benefit of the reviewers), please also take note the following comments, where the line and figure numbers refer to your revised version of the manuscript. Where I am referring to a comment from a reviewer, I indicate with R1, R2 or R3.1. Your response to the reviewers has muddled up their numbers: Reviewers 1, 2 and 3 you have called reviewers 3, 1 and 2 in your response. [I note that your response to reviewer 2 (whom you call reviewer 1) misquotes from a previous "reviewer 1" from your earlier submission to the AMT special issue.] In future responses, please use the same numbering as on the Copernicus web page for your manuscript.2. (R1) Regarding multiple scattering, it is reasonable to neglect it in this manuscript, but you need to at least discuss the issue and outline a plan for dealing with it in future when you have real EarthCARE data. For example, cite Battaglia et al. (JGR 2011) who proposed a way of identifying data likely to be affected by it.3. (R1) It is good that at least the simulated data will be released, but any revised version of the manuscript will need to cite the data in a public repository.4. (R2) It is not true that SPIDER is non-polarimetric - that is what the "P" stands for! However, the potential to improve the classification using linear depolarisation measurements may be limited.5. The word "data" is the plural of "datum". Therefore, strictly, you should say "Satellite data *have*" and similarly throughout the manuscript.6. New line 31 (34 in tracked-changes version): You say the EarthCARE satellite is "challenging", which implies "problematic"... surely this is not what you mean. Isn't this paragraph explaining the opportunities from EarthCARE rather than any problems?7. (R3) New line 131 (141 with tracked-changes) and elsewhere: The presence of riming, graupel and hail seems to be asserted much too readily based on a simple velocity threshold. You can only get riming if supercooled water is present, so the process cannot occur at temperatures colder than around -40 deg C yet the "riming region" seems to extend up to all heights, which is not realistic. Moreover, when interpreting the new Figs. 8a and 8c, and in section 3.3, you say that NDW6 has much more graupel than NSW6 just because the particles are falling faster. But this is the dominant mode in the Doppler velocity distribution between 5 and 10 km from a 24-hour sample: is graupel really dominating the distribution for such a long period? It seems much more likely that you are observing the ice/snow in the model that has been assigned too high a fall velocity. In any case, you have the model data so can check whether or not these values are really due to graupel rather than over-relying on a simple velocity threshold.8. New line 133 (145 with tracked-changes): You say that values greater than 2 m/s occur with high frequency - I can see some values here but they look quite rare, so not really "high frequency".9. Figure 4: I am puzzled by the y axis. I understand this is the fraction of absolute velocities between 0 m/s and the value on the x axis, in which case the value on the y axis at x=0 m/s indicates the fraction of absolute velocities between 0 and 0 m/s, which is surely y=0. So why is the value instead between 0.985 and 0.997?10. I can find no reference in the new text to Fig. 6.11. The captions to your new tables 1-3 need to clarify that the numbers are percentages, and you have written "faction" when you mean "fraction".Citation: https://doi.org/
10.5194/egusphere-2023-1997-EC1 -
AC5: 'Reply on EC1', Woosub Roh, 14 Mar 2024
Dear Editor,
We sincerely appreciate your insightful comments and the considerable effort you have invested in reviewing our manuscript.
- Your response to the reviewers has muddled up their numbers: Reviewers 1, 2 and 3 you have called reviewers 3, 1 and 2 in your response. [I note that your response to reviewer 2 (whom you call reviewer 1) misquotes from a previous "reviewer 1" from your earlier submission to the AMT special issue.] In future responses, please use the same numbering as on the Copernicus web page for your manuscript.
→ I am sorry for my mistake. I will use the same numbering.
- (R1) Regarding multiple scattering, it is reasonable to neglect it in this manuscript, but you need to at least discuss the issue and outline a plan for dealing with it in future when you have real EarthCARE data. For example, cite Battaglia et al. (JGR 2011) who proposed a way of identifying data likely to be affected by it.
→Thank you for your comments. I added this issue in the summary like “This study did not account for the complexities of multiple scattering effects (Battaglia et al., 2011) and pointing uncertainties (Tanelli et al., 2005) in simulating Doppler velocities. These aspects are crucial for accurately assessing the Doppler velocity capabilities on the EarthCARE, including the impact of Pulse Repetition Frequencies (PRFs). Understanding these influences is challenging prior to the satellite's launch. We will investigate these effects on the evaluation result in the future.”
- (R1) It is good that at least the simulated data will be released, but any revised version of the manuscript will need to cite the data in a public repository.
→ We uploaded the data and added “The snapshot data of simulated radar reflectivity and Doppler velocity of 94 GHz Cloud Profiling Radar (CPR) are available from https://doi.org/10.5281/zenodo.10813626 (Roh et al., 2024). We made Fig. 10 using these data.” in the data availability part.
- (R2) It is not true that SPIDER is non-polarimetric - that is what the "P" stands for! However, the potential to improve the classification using linear depolarisation measurements may be limited.
→ Thank you for your comment. I misunderstood the coauthors' comments. The HG-SPIDER was developed as a polarimetric radar. HG-SPIDER did not observe the depolarization ratio in these cases.
- The word "data" is the plural of "datum". Therefore, strictly, you should say "Satellite data *have*" and similarly throughout the manuscript.
→ We changed the sentence based on your comment.
- New line 31 (34 in tracked-changes version): You say the EarthCARE satellite is "challenging", which implies "problematic"... surely this is not what you mean. Isn't this paragraph explaining the opportunities from EarthCARE rather than any problems?
→ Thank you for your comment. I changed to “innovative” from “challenging”.
- (R3) New line 131 (141 with tracked-changes) and elsewhere: The presence of riming, graupel and hail seems to be asserted much too readily based on a simple velocity threshold. You can only get riming if supercooled water is present, so the process cannot occur at temperatures colder than around -40 deg C yet the "riming region" seems to extend up to all heights, which is not realistic. Moreover, when interpreting the new Figs. 8a and 8c, and in section 3.3, you say that NDW6 has much more graupel than NSW6 just because the particles are falling faster. But this is the dominant mode in the Doppler velocity distribution between 5 and 10 km from a 24-hour sample: is graupel really dominating the distribution for such a long period? It seems much more likely that you are observing the ice/snow in the model that has been assigned too high a fall velocity. In any case, you have the model data so can check whether or not these values are really due to graupel rather than over-relying on a simple velocity threshold.
→ It has uncertainty to characterize graupel or hail based on only Doppler velocity, but I think it's a better basis than radar reflectivity. The observation is consistent with the location of the riming area below 10 km. It is only dominant in the strong convection case. However, the simulations are different.
I agree with your concern about the misinterpretation of the simulation. We notice that simulated graupel and classified graupel by Doppler velocity are different. We explained, “NDW6 overestimates the graupel/hail regime associated with large snow or ice crystals. This result indicates that the terminal velocity of the snow is overestimated compared to the observation.” in new L252-254 (old L226-227).
We also checked the impact of the simulated graupel and snow on the joint histogram. The simulated large snow in NDW6 and NSW6 was classified as graupel/hail by the Doppler velocity threshold. NDW6 especially overestimates the size or terminal velocity of snow compared to observation. We think we need to improve the size distribution or terminal velocity of snow in NDW6 based on observation on the ground or the EarthCARE CPR.
- New line 133 (145 with tracked-changes): You say that values greater than 2 m/s occur with high frequency - I can see some values here but they look quite rare, so not really "high frequency".
→ Thank you for your comment. We removed “high’ in the text.
- Figure 4: I am puzzled by the y axis. I understand this is the fraction of absolute velocities between 0 m/s and the value on the x axis, in which case the value on the y axis at x=0 m/s indicates the fraction of absolute velocities between 0 and 0 m/s, which is surely y=0. So why is the value instead between 0.985 and 0.997?
→ The x=0 means the vertical air velocity is less than 0.2 m/s
- I can find no reference in the new text to Fig. 6.
→ There is the reference of Fig. 6 in Line 213.
- The captions to your new tables 1-3 need to clarify that the numbers are percentages, and you have written "faction" when you mean "fraction".
→ We added “%” in the tables.
Reference
Roh, W., & Satoh, M.: 94 GHz cloud radar simulation data using NICAM and Joint simulator for evaluation of ground-based radar and application to the EarthCARE satellite [Data set]. Zenodo. https://doi.org/10.5281/zenodo.10813626, 2024.
Citation: https://doi.org/10.5194/egusphere-2023-1997-AC5
-
AC5: 'Reply on EC1', Woosub Roh, 14 Mar 2024
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 470 | 133 | 58 | 661 | 21 | 17 |
- HTML: 470
- PDF: 133
- XML: 58
- Total: 661
- BibTeX: 21
- EndNote: 17
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Woosub Roh
Masaki Satoh
Yuichiro Hagihara
Hiroaki Horie
Yuichi Ohno
Takuji Kubuta
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1746 KB) - Metadata XML