the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The Virga-Sniffer – a new tool to identify precipitation evaporation using ground-based remote-sensing observations
Heike Kalesse-Los
Anton Kötsche
Andreas Foth
Johannes Röttenbacher
Teresa Vogl
Jonas Witthuhn
Abstract. Combined continuous long-term ground-based remote-sensing observations with vertically pointing cloud radar and ceilometer are well-suited to identify precipitation evaporation fall streaks (so-called virga). Here we introduce the functionality and workflow of a new open-source tool, the Virga-Sniffer which was developed within the frame of RV Meteor observations during the ElUcidating the RolE of Cloud–Circulation Coupling in ClimAte (EUREC4A) field experiment in Jan–Feb 2020 in the Tropical Western Atlantic. The Virga-Sniffer Python package is highly modular and configurable and can be applied to multilayer cloud situations. In the simplest approach, it detects virga from time-height fields of cloud radar reflectivity and time series of ceilometer cloud base height. In addition, optional parameters like lifting condensation level, a surface rain flag as well as time-height fields of cloud radar mean Doppler velocity can be added to refine virga event identifications. The netcdf-output files consist of Boolean flags of virga- and cloud detection, as well as base- and top heights and depth for the detected clouds and virga. The performance of the Virga-Sniffer was assessed by comparing its results to the Cloudnet target classification resulting from using the CloudnetPy processing chain. 88 % of the pixel identified as virga by the Virga Sniffer correspond to Cloudnet classifications of precipitation. The remaining 12 % of virga pixel correspond to Cloudnet-classifications of aerosols and insects (about 7 %), cloud droplets (3 %), or clear-sky (about 1 %). Some discrepancies of the virga identification and the Cloudnet target classification can be attributed to applied temporal smoothing. Additionally, it was found that Cloudnet mostly classified aerosols and insects
at virga edges which points to a misclassification caused by CloudnetPy internal thresholds. For the RV Meteor observations during EUREC4A, about 50 % of all detected clouds with bases below the trade inversion were found to produce precipitation that evaporates before reaching the ground. The most important virga-producing clouds were either anvils of convective cells or stratocumulus clouds. 36 % of the detected virga originated from trade wind cumuli. Small virga with depths below 200 m most frequently occurred from shallow clouds with depths below 500 m, while virga depths above 1 km were mainly associated with clouds of larger depths, ranging between 500 and 1000 m. Virga depth showed no strong dependency on column-integrated liquid water path. The presented results substantiate the importance of low-level precipitation evaporation in the Atlantic lower winter trades. Possible applications of the Virga-Sniffer within the frame of EUREC4A include detailed studies of precipitation evaporation with a focus on cold pools or cloud organization, or distinguishing moist processes based on water vapor isotopic observations. However, we envision extended use of the Virga-Sniffer for other cloud regimes or scientific foci as well.
Heike Kalesse-Los et al.
Status: closed
-
RC1: 'Comment on amt-2022-252', Raphaela Vogel, 07 Nov 2022
This paper presents a new tool to identify virga below trade cumulus clouds using ground-based remote sensing observations. The tool was developed using data from the recent EUREC4A field campaign and its modular approach can be applied in multilayer cloud situations. The performance was assessed through comparison with the Cloudnet target classification, and some statistics regarding virga occurrence frequencies and depths are presented. The paper is easy to read, fits well into the scope of AMT, and should be accepted after some revisions. I have three major comments, which are summarized below, along with some more minor comments and technical corrections (in the annotated pdf).
Major comments:
1. Tool description:
I think the tool description can be improved. I couldn’t follow the explanation in Section 3.1 as I was missing some crucial information: at what temporal resolution are these analyses done? What is a CBH layer? Other confusing bits of information in Sec. 3 are:
- L152: The module numbering in Sec 3.1 is a bit counter-intuitive: why not put the smoothing as module 1 (instead of 5) and thus start with module 1?
- L154: what does 5% mean here? 5% of a given time period? Or 5% of vertical extent?
- L166-168: this is a very lengthy way of saying that ‘two iterations of all 5 steps are made’.
- L168: I thought that LCL data is optional (Fig. 2), but here it seems to be necessary.
- L175: I don’t remember a definition of ‘valid radar reflectivity’.
- L200ff: These clarifications are helpful, but e.g. the minimum virga length requirement is only mentioned in Sec 3.3, and comes as a surprise here. These examples could thus be moved after Sec 3.3. Furthermore, instead of just writing ‘maximum allowed gap for virga’, the chosen default threshold could be mentioned again (I actually thought that 700m is a typo, it seemed too large for me – so repeating it would clarify this choice).
- L204: What is rg19 then? Did the ceilometer miss this second cloud layer due to the strong rain? This should be discussed.
- L210: This step of virga mask refinement is thus not optional (as suggested in Fig. 2, part 3))?
- 4: what does ‘filled cloud base’ mean?
- L269: I don’t see the multiple layers at 05:00 in Fig. 5. Is the ‘filled cloud base’ considered as a cloud? If so, I’d find this problematic, because there is obviously no cloud there.
- 4&5: Zooming into the detected virga (e.g. Fig 5c, around 04:40 or 05:45), the sub-cloud layer virga is not continuously detected, potentially due to surface rain or (for stratiform inversion cloud) positive Doppler velocity. I find that a bit problematic, as physically these rainshafts should be considered as one object, and the on-off-virga detection is a bit arbitrary. See also my major comment #3.
- To clarify the reason why some sub-cloud layer rain is not classified as virga, it would be helpful if Fig. 4 & 5 could also show the surface rain flag.
The beginning of the summary section 5 mentions that profile-by-profile information is used. I think this information should come at the beginning of Sec. 3, together with information about the temporal resolution of the analyses (e.g., it is unclear what temporal resolution the ceilometer has), and reference to the appendix, which summarizes the configuration (I only realized after the summary that there is an appendix).
I also have some issues with Fig. 2, as (i) the gray thin lines in Fig. 2 are hard to see on my print out, and (ii) the figure claims some steps to be optional, which are discussed as necessary in the text (see above). For Figure 3, the coloring is ambiguous, because detected cloud and virga should also be partly green, because they have a valid Ze. So maybe make two masks (one input and one output), or hatch the boxes with valid Ze. It would also be nice to have an example of a multi-layered cloud situation here.
2. Cloud type classification:
I have some issues with the cloud type classification here. During EUREC4A, I don’t remember any situations of stratocumulus or stratus clouds. However, deeper trade cumulus clouds with extensive stratiform cloud layers were very frequent. But these stratiform cloud layers were at some point detrained from a cumulus core rooting in the sub-cloud layer. I.e., the convection and cloud formation was surface-driven and not cloud-top driven as in stratocumulus. From a ground-based single-point perspective, this distinction is of course not easily made, because you might only capture the stratiform part of a cloud.
Although the classifications used here might be in line with the Stratocumulus Cumulogenitus (CL = 4) class of the WMO cloud atlas, I would encourage the authors to reconsider their cloud type classification. In the broader EUREC4A or trade cumulus community, we usually use different names for this ‘cloudiness aloft’ components, which are often called ‘stratiform (cloud) layers’, ‘stratiform inversion cloud’, ‘shallow anvils’, or sheared edges of deeper trade cumuli. Nuijens et al. (2014) or Vial et al. (2019, https://doi.org/10.1029/2019MS001746) are good references for how to deal with these naming issues.
3. Virga vs. evaporation from rainshafts that reach the ground:
I miss the motivation for focusing only on virga rather than all rainshafts. Although raining clouds are less frequent than clouds with virga (your Table 3), in terms of their contribution to total rain evaporation they are likely still very important. So when the main reason motivating this study is to (eventually) investigate rain evaporation, why focusing only on virga? In my eyes, the only physical reason that distinguish virga from other rainshafts is that total versus partial re-evaporation is relevant for the isotopic signal (Torri 2021, https://doi. org/10.1029/2020JD033139). But e.g. from a moisture or heat budget point of view, it doesn’t matter whether rain reaches the surface or not. It would be great if the authors could discuss their reasons for their focus on virga more explicitly.
Minor comments:
1. Review of earlier approaches of virga detection or rain evaporation retrievals: In the introduction, I missed a review of earlier work focusing on virga and rain evaporation in the trades. E.g. Sarkar et al. (2020, DOI: 10.1175/MWR-D-19-0235.1) is a study that comes to my mind, but there are for sure others.
2. Results for single cloud layers: Sec. 3.5 showed that most challenges and limitations pertain to multi-layer cloud situations. To increase the robustness of the results, it would be great to see how the results (e.g. in Fig. 8 and 9) change if only single-layer clouds are considered. These results will likely be more trustworthy.
3. Comparison with Cloudnet target classification: How often does Cloudnet detect drizzle / rain when the VirgaSniffer doesn’t detect anything? I think the comparison in both directions is important.
4. Commas: I’m not an expert on commas, but I feel that some additional commas would ease the reading. I made some suggestions in the annotated pdf.
Technical corrections:
Please find some technical suggestions in the annotated pdf.
Raphaela Vogel
- AC1: 'Reply on RC1', Heike Kalesse-Los, 16 Jan 2023
-
RC2: 'Comment on amt-2022-252', Anonymous Referee #2, 14 Nov 2022
The study introduces a new tool: the virga-sniffer, this is a tool which is used to detect virga from vertically pointing radar and ceilometer observations. The authors outline the methodology of the virga-sniffer and present case studies to show the output or the virga classification and cloud base and cloud top height detection. Radar pixels that are found to contain virga are compared against Cloudnet classifications and statistics of the virga observations from the entire EUREC4A campaign are discussed. A key point is that the tool works on multiple levels of cloud and is able to detect virga between levels of cloud. This study is largely an introductory study to this new tool and outside of some preliminary results does not conduct much analysis of the virga that has been detected and its relationship with the wider atmosphere. It is my opinion that this study should be accepted after some minor revisions.
General points:
There are a large number of thresholds used within the study, how sensitive is the output of the virga-sniffer to these thresholds? Some discussion of the parameters that the tool is sensitive to is necessary. Why are they set at their current values? How does changing them effect the results?
There is some mention that the tool works without the inclusion of the LCL and the surface precipitation measurements. Some discussion of the differences in the results with and without these parameters would be useful.
Minor comments:
L98: Are roll and pitch angles allowed to be negative? If so replace this with absolute angles. If not, why is the standard deviation so much greater than the mean, this implies a very skewed distribution?
L100: Together with the previous point, if there is a sizeable inclusion of horizontal wind the pointing is relevant for the Doppler velocity. Is there any treatment or removal of Doppler velocity at large roll/pitch angles?
L196-198: In this situation it is possible to have rain from another section of cloud blown in to the column and giving the impression of rain reaching the surface. Any consideration of this situation? Use of horizontal wind e.g.?
L199: How frequently do these special cases occur and how frequently does the virga detection work with little or no complications?
L201: Is this step included when the clutter filter described earlier is also in use? Is it necessary if there is already a clutter filter?
L208-209: As previous comment about wind-blown rain detected at the surface.
L237: Include some discussion of how frequently these limitations occur and the impact they are likely to have on the overall data quality.
L252-253: Could neighbouring columns be included to mitigate this? Allowing a large vertical gap for virga seems to lead to unlikely results at times (e.g. part of the lower cloud being labelled as virga at 3.45 in Fig. 5)
L263: Due to what?
L280: If I understand this correctly the categories on the inner ring are a subset of the outer ring? If so, why do they not align for aerosols?
Fig. 6: Annotate the larger classes in the inner ring with the percentages
L313: What are the horizontal lines on Figure 8?
L313: Given the large number of virga reaching 300m it would be interesting to see any meteorological observations both surface based or radio/dropsondes to look at profiles of humidity and temperature.
L325: By eye there appears to be a loose trend along a line from approx. (0, 0.2) to (1, 1.5). Have you looked at any statistics for these data?
Fig. 8, 9b: The y-axis scale is irregular, I assume it should be 250 m per label. Add the extra sig fig to make this clearer
Fig. A1: needs colorbar
Spelling/Grammar/Typos:
L19(x2), 20, 31, 197: Using above/below is ambiguous when talking about the atmosphere, especially in relation to temperature which changes with height. Use greater than, less than etc.
L112: Define MPI before use
L154: less -> fewer
L261: remove the comma
L334: 1.5 m -> 1.5 km
L357: pixel -> pixels
L363: “As application”, I’m not sure what was intended here
L403: suses -> uses
L404: remove comma
L457: remove paragraph
Citation: https://doi.org/10.5194/amt-2022-252-RC2 - AC2: 'Reply on RC2', Heike Kalesse-Los, 16 Jan 2023
-
RC3: 'Comment on amt-2022-252', Anonymous Referee #3, 17 Nov 2022
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-252/amt-2022-252-RC3-supplement.pdf
- AC3: 'Reply on RC3', Heike Kalesse-Los, 16 Jan 2023
Status: closed
-
RC1: 'Comment on amt-2022-252', Raphaela Vogel, 07 Nov 2022
This paper presents a new tool to identify virga below trade cumulus clouds using ground-based remote sensing observations. The tool was developed using data from the recent EUREC4A field campaign and its modular approach can be applied in multilayer cloud situations. The performance was assessed through comparison with the Cloudnet target classification, and some statistics regarding virga occurrence frequencies and depths are presented. The paper is easy to read, fits well into the scope of AMT, and should be accepted after some revisions. I have three major comments, which are summarized below, along with some more minor comments and technical corrections (in the annotated pdf).
Major comments:
1. Tool description:
I think the tool description can be improved. I couldn’t follow the explanation in Section 3.1 as I was missing some crucial information: at what temporal resolution are these analyses done? What is a CBH layer? Other confusing bits of information in Sec. 3 are:
- L152: The module numbering in Sec 3.1 is a bit counter-intuitive: why not put the smoothing as module 1 (instead of 5) and thus start with module 1?
- L154: what does 5% mean here? 5% of a given time period? Or 5% of vertical extent?
- L166-168: this is a very lengthy way of saying that ‘two iterations of all 5 steps are made’.
- L168: I thought that LCL data is optional (Fig. 2), but here it seems to be necessary.
- L175: I don’t remember a definition of ‘valid radar reflectivity’.
- L200ff: These clarifications are helpful, but e.g. the minimum virga length requirement is only mentioned in Sec 3.3, and comes as a surprise here. These examples could thus be moved after Sec 3.3. Furthermore, instead of just writing ‘maximum allowed gap for virga’, the chosen default threshold could be mentioned again (I actually thought that 700m is a typo, it seemed too large for me – so repeating it would clarify this choice).
- L204: What is rg19 then? Did the ceilometer miss this second cloud layer due to the strong rain? This should be discussed.
- L210: This step of virga mask refinement is thus not optional (as suggested in Fig. 2, part 3))?
- 4: what does ‘filled cloud base’ mean?
- L269: I don’t see the multiple layers at 05:00 in Fig. 5. Is the ‘filled cloud base’ considered as a cloud? If so, I’d find this problematic, because there is obviously no cloud there.
- 4&5: Zooming into the detected virga (e.g. Fig 5c, around 04:40 or 05:45), the sub-cloud layer virga is not continuously detected, potentially due to surface rain or (for stratiform inversion cloud) positive Doppler velocity. I find that a bit problematic, as physically these rainshafts should be considered as one object, and the on-off-virga detection is a bit arbitrary. See also my major comment #3.
- To clarify the reason why some sub-cloud layer rain is not classified as virga, it would be helpful if Fig. 4 & 5 could also show the surface rain flag.
The beginning of the summary section 5 mentions that profile-by-profile information is used. I think this information should come at the beginning of Sec. 3, together with information about the temporal resolution of the analyses (e.g., it is unclear what temporal resolution the ceilometer has), and reference to the appendix, which summarizes the configuration (I only realized after the summary that there is an appendix).
I also have some issues with Fig. 2, as (i) the gray thin lines in Fig. 2 are hard to see on my print out, and (ii) the figure claims some steps to be optional, which are discussed as necessary in the text (see above). For Figure 3, the coloring is ambiguous, because detected cloud and virga should also be partly green, because they have a valid Ze. So maybe make two masks (one input and one output), or hatch the boxes with valid Ze. It would also be nice to have an example of a multi-layered cloud situation here.
2. Cloud type classification:
I have some issues with the cloud type classification here. During EUREC4A, I don’t remember any situations of stratocumulus or stratus clouds. However, deeper trade cumulus clouds with extensive stratiform cloud layers were very frequent. But these stratiform cloud layers were at some point detrained from a cumulus core rooting in the sub-cloud layer. I.e., the convection and cloud formation was surface-driven and not cloud-top driven as in stratocumulus. From a ground-based single-point perspective, this distinction is of course not easily made, because you might only capture the stratiform part of a cloud.
Although the classifications used here might be in line with the Stratocumulus Cumulogenitus (CL = 4) class of the WMO cloud atlas, I would encourage the authors to reconsider their cloud type classification. In the broader EUREC4A or trade cumulus community, we usually use different names for this ‘cloudiness aloft’ components, which are often called ‘stratiform (cloud) layers’, ‘stratiform inversion cloud’, ‘shallow anvils’, or sheared edges of deeper trade cumuli. Nuijens et al. (2014) or Vial et al. (2019, https://doi.org/10.1029/2019MS001746) are good references for how to deal with these naming issues.
3. Virga vs. evaporation from rainshafts that reach the ground:
I miss the motivation for focusing only on virga rather than all rainshafts. Although raining clouds are less frequent than clouds with virga (your Table 3), in terms of their contribution to total rain evaporation they are likely still very important. So when the main reason motivating this study is to (eventually) investigate rain evaporation, why focusing only on virga? In my eyes, the only physical reason that distinguish virga from other rainshafts is that total versus partial re-evaporation is relevant for the isotopic signal (Torri 2021, https://doi. org/10.1029/2020JD033139). But e.g. from a moisture or heat budget point of view, it doesn’t matter whether rain reaches the surface or not. It would be great if the authors could discuss their reasons for their focus on virga more explicitly.
Minor comments:
1. Review of earlier approaches of virga detection or rain evaporation retrievals: In the introduction, I missed a review of earlier work focusing on virga and rain evaporation in the trades. E.g. Sarkar et al. (2020, DOI: 10.1175/MWR-D-19-0235.1) is a study that comes to my mind, but there are for sure others.
2. Results for single cloud layers: Sec. 3.5 showed that most challenges and limitations pertain to multi-layer cloud situations. To increase the robustness of the results, it would be great to see how the results (e.g. in Fig. 8 and 9) change if only single-layer clouds are considered. These results will likely be more trustworthy.
3. Comparison with Cloudnet target classification: How often does Cloudnet detect drizzle / rain when the VirgaSniffer doesn’t detect anything? I think the comparison in both directions is important.
4. Commas: I’m not an expert on commas, but I feel that some additional commas would ease the reading. I made some suggestions in the annotated pdf.
Technical corrections:
Please find some technical suggestions in the annotated pdf.
Raphaela Vogel
- AC1: 'Reply on RC1', Heike Kalesse-Los, 16 Jan 2023
-
RC2: 'Comment on amt-2022-252', Anonymous Referee #2, 14 Nov 2022
The study introduces a new tool: the virga-sniffer, this is a tool which is used to detect virga from vertically pointing radar and ceilometer observations. The authors outline the methodology of the virga-sniffer and present case studies to show the output or the virga classification and cloud base and cloud top height detection. Radar pixels that are found to contain virga are compared against Cloudnet classifications and statistics of the virga observations from the entire EUREC4A campaign are discussed. A key point is that the tool works on multiple levels of cloud and is able to detect virga between levels of cloud. This study is largely an introductory study to this new tool and outside of some preliminary results does not conduct much analysis of the virga that has been detected and its relationship with the wider atmosphere. It is my opinion that this study should be accepted after some minor revisions.
General points:
There are a large number of thresholds used within the study, how sensitive is the output of the virga-sniffer to these thresholds? Some discussion of the parameters that the tool is sensitive to is necessary. Why are they set at their current values? How does changing them effect the results?
There is some mention that the tool works without the inclusion of the LCL and the surface precipitation measurements. Some discussion of the differences in the results with and without these parameters would be useful.
Minor comments:
L98: Are roll and pitch angles allowed to be negative? If so replace this with absolute angles. If not, why is the standard deviation so much greater than the mean, this implies a very skewed distribution?
L100: Together with the previous point, if there is a sizeable inclusion of horizontal wind the pointing is relevant for the Doppler velocity. Is there any treatment or removal of Doppler velocity at large roll/pitch angles?
L196-198: In this situation it is possible to have rain from another section of cloud blown in to the column and giving the impression of rain reaching the surface. Any consideration of this situation? Use of horizontal wind e.g.?
L199: How frequently do these special cases occur and how frequently does the virga detection work with little or no complications?
L201: Is this step included when the clutter filter described earlier is also in use? Is it necessary if there is already a clutter filter?
L208-209: As previous comment about wind-blown rain detected at the surface.
L237: Include some discussion of how frequently these limitations occur and the impact they are likely to have on the overall data quality.
L252-253: Could neighbouring columns be included to mitigate this? Allowing a large vertical gap for virga seems to lead to unlikely results at times (e.g. part of the lower cloud being labelled as virga at 3.45 in Fig. 5)
L263: Due to what?
L280: If I understand this correctly the categories on the inner ring are a subset of the outer ring? If so, why do they not align for aerosols?
Fig. 6: Annotate the larger classes in the inner ring with the percentages
L313: What are the horizontal lines on Figure 8?
L313: Given the large number of virga reaching 300m it would be interesting to see any meteorological observations both surface based or radio/dropsondes to look at profiles of humidity and temperature.
L325: By eye there appears to be a loose trend along a line from approx. (0, 0.2) to (1, 1.5). Have you looked at any statistics for these data?
Fig. 8, 9b: The y-axis scale is irregular, I assume it should be 250 m per label. Add the extra sig fig to make this clearer
Fig. A1: needs colorbar
Spelling/Grammar/Typos:
L19(x2), 20, 31, 197: Using above/below is ambiguous when talking about the atmosphere, especially in relation to temperature which changes with height. Use greater than, less than etc.
L112: Define MPI before use
L154: less -> fewer
L261: remove the comma
L334: 1.5 m -> 1.5 km
L357: pixel -> pixels
L363: “As application”, I’m not sure what was intended here
L403: suses -> uses
L404: remove comma
L457: remove paragraph
Citation: https://doi.org/10.5194/amt-2022-252-RC2 - AC2: 'Reply on RC2', Heike Kalesse-Los, 16 Jan 2023
-
RC3: 'Comment on amt-2022-252', Anonymous Referee #3, 17 Nov 2022
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-252/amt-2022-252-RC3-supplement.pdf
- AC3: 'Reply on RC3', Heike Kalesse-Los, 16 Jan 2023
Heike Kalesse-Los et al.
Heike Kalesse-Los et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 330 | 137 | 17 | 484 | 5 | 4 |
- HTML: 330
- PDF: 137
- XML: 17
- Total: 484
- BibTeX: 5
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1