Evolution of squall line variability and error growth in an ensemble of LES

Groot, Edward; Tost, Holger

doi:https://doi.org/10.5194/egusphere-2022-515

Preprints

https://doi.org/10.5194/egusphere-2022-515

Preprints

01 Jul 2022

| 01 Jul 2022

Evolution of squall line variability and error growth in an ensemble of LES

Edward Groot and Holger Tost

Abstract. Squall lines represent an organized form of atmospheric convection that link processes occurring at the small end of the mesoscale and processes ocurring at the large end of the mesoscale. This study analyses the initial condition sensitivity of idealized squall lines in an LES ensemble. The ensemble spread of the squall lines is evaluated using passive tracers, an ensemble sensitivity analysis, other statistical tools and an error growth metric. Analysing gravity wave dynamics, convective initiation, squall line relative motion and updraft/downdraft characteristics and transport, a chain of interacting processes is identified.

From the convective point of view ensemble spread is rooted in a secondary phase of convective initiation (30–35 min) a few km ahead of the squall line. Contrasts in the amount secondary initiation arise within the ensemble, as vertical velocity varies at the location of convective initiation within the ensemble due to differences in gravity wave amplitude and phase. Immediately after the secondary phase of initiation (30–45 min), the cold pool accelerates to velocities of 2–4 m/s (ensemble envelope).

With the spread in secondary convective initiation, upward mass transport is disturbed, which also affects downdraft mass fluxes. Furthermore, once accelerated (30–40 minutes), the cold pool nearly maintains its propagation speed in each ensemble member. It is shown that part of the errors occurring after 45–85 minutes are explained by the cold pool velocity and a correction for cold pool velocity removes a substantial fraction of the spread. A coherent anomaly of the circulation within the squall line, which is consistent with extra upward mass transport, exists during this phase of the evolution. It is proposed that the identified chain of interactions may be explained by a common mode of variability, which determines a substantial portion of the ensemble spread in the stage after 30–85 minutes in many diagnostics.

Based on a non-monotonic relation between initial conditions and local vertical velocities that cause secondary initiation, one can argue that an intrinsic limit of predictability exists, as Melhauser and Zhang (2012) do.

Received: 20 Jun 2022 – Discussion started: 01 Jul 2022

Download & links

Preprint (PDF, 2973 KB)

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 (2973 KB)

Supplement (3972 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

13 Jan 2023

Evolution of squall line variability and error growth in an ensemble of large eddy simulations

Edward Groot and Holger Tost

Atmos. Chem. Phys., 23, 565–585, https://doi.org/10.5194/acp-23-565-2023,https://doi.org/10.5194/acp-23-565-2023, 2023

Short summary

Edward Groot and Holger Tost

Interactive discussion

Status: closed

RC1: 'Comment on egusphere-2022-515', Anonymous Referee #1, 08 Jul 2022

Commendations to the authors on an interesting computation, and congratulations to the student on a PhD well earned.

However, this manuscript seems to be not well written for the professional literature, rather it is in the somewhat tedious style of a dissertation. The captive audience of a reader committee would make all their check marks: enough effort reported, quantitative analysis technique mastery, awareness of caveats, and the student made thoughtful connections to some prior papers in the literature. But earning the attention of readers and (hopefully) citation in the professional literature is a different enterprise, and a major revision should be undertaken for that purpose. However, I do not volunteer to review again - I had forgotten that I tend to dislike ACP, perhaps for its apparent incentives toward first-draft quality or length. I haven't studied the business model, but experiences seem mixed.

Main science points of the manuscript as I see them:

* Second-generation development of deep convection in a triggered, highly unstable sounding is a delicate process, which adds a burst of difference growth to an ensemble, and introduces a need for feature-centered rather than Eulerian measures.

* The main mechanisms of this delicate secondary development involve a combination of cold-pool gust front lifting plus the strength of thermally forced gravity wave displacements in the inversion layer of the lower free troposphere.

* Low-level shear enhances gust front lifting on the downshear side, adding asymmetry that is convenient for analysis purposes. It also has longer-term effects on the resulting systems, but those are not really evident here.

* Because the redevelopment process is delicate and amplifies differences, any small variation can generate an ensemble of second-generation outcomes. Here the low-level shear strength (distractingly cast as an altitude difference of the linear shear layer's top which is smaller than the grid spacing!) happens to have been used. There is no indication that the results are monotonic with this shear strength perturbation: ENS-05 has the largest perturbation, but ENS-03 has the biggest outcome difference, and the final Abstract sentence suggests so too. If the perturbation size and outcome difference magnitude are uncorrelated, then emphasizing the role of shear is misleading: lucky resonances among cold pools and internal waves aloft seem to be the main unstable difference-makers, and anything might cause ensemble divergence.

* Secondary development or non-development affects both the gust front and deep convection+wave fields at later times, of course (Fig. 3), and has clear correlates in the top height, width, volume, and thus mass flux of the squall line at any given time (however it is measured; the outlines in Fig. 5 are as informative or more so than the colored tracer difference that predominates in the attention of the viewer).

* The time scale of deep difference development is about 40 minutes, at which time about half the 80-minute ensemble spread has been established already (Fig. 6).

* Outcomes (secondary develop/non-develop) appear to be coherent in the vertical at least up to midlevels (there is little curve crossing in Fig. 10... why not color all 10 lines, either arbitrarily or by the value of their initial perturbation z_top, then making reference and ENS-3 bolder for clarity?)

High-level style critique of manuscript:

Many of the key findings are quite buried behind overly detailed and sometimes defensive material about methods in a deep, murky dissertation-style text. This does not serve a professional reader’s interests optimally, since at this point the student does not need to prove amount of effort or technical mastery, but rather a sharp eye and crisp tongue about what is important. The slow resurfacing at the end into Synthesis and then Conclusion sections (also rather long) does help to pull these key points and highlights out somewhat, but those key points could be even more polished into the Abstract for instance. There is no role for suspense (historical or at readtime) in scientific writing. But those wrapup sections (4 and 5) come after a long slog of sometimes unclear prose in the late-middle (the very long sections 3.3 and 3.4).

Helpfully, section 3.2 leads it off, with simple differences between the most-different ensemble members. In this reader’s view sections 3.3 and 3.4 then add little scientifically, except a glimpse of the 10 members as spaghetti plots (why not colorize all 10 members so we can see if curves cross?). An unclear overanalysis in confusing statistical terms (terms like “source” and “error”, and “auto-correlation” for intra-ensemble rather than temporal-lag correlations) were unhelpful or confusing, and it all gave few clear insights that aren’t in the bullets above and in section 3.2. Might the paper or at least sections 3.3 and 3.4 be cut by aspirationally 50% with no loss (and a gain of clarity) on the reader’s part? Long stretches of text appear to describe figures not shown, without stating (not shown). Who needs these, who will remember them an hour after reading much less a month or year?

Detailed local comments:

Abstract: Eliminate the first sentence. Line 9: “amount —> “amount of”. disturbed —> perturbed. error —> difference. circulation —> divergent wind pattern.

Lines 19-20: The meaning of this result is that the perturbations chosen are in a non-essential field, but that even those differences grow (or explode). What does the word “intrinsic limit”, lifted from some over-realm of philosophy it seems, really add to this idea?

A general game: how many words could be trimmed or eliminated without loss of meaning? In journals with page charges there are incentives; perhaps a lack of incentive is what I have not liked about ACP (does it have length charges?)?

158: “interface height” — this is just a reference value in an analytic formula, which translates into shear strength on the 100m grid, right? This description was quite confusing.

197: “dam breaking” is never mentioned, not all readers may understand this allusion to a classic problem’s metaphor

248-250: what and where are we looking?

253: does your contouring routine treat \ sloped features different from / sloped features?

276: “difference" vs. in Fig 4 caption, sign is different (or ambiguous). Why not make it clear?

291: “will be” —> “were”

301: “reduced in the reference” seems backward to the idea of a reference vs. perturbation

302: “no less than 38.3(!)…only 23.5” — these seem like rather dramatic descriptions of excessively precise numbers that are not terribly different, and how was the second one even measured?

330: “argmax” is a useful word from the code world

Fig. 6: “autocorrelation” is often used to mean a series with time lags, perhaps say “intra-ensemble correlations at the indicated time with the minute-80 ensemble spread”

374: “Now we will continue” seems to presume the reader’s commitment, can topic phrases and topic sentences be part of a thought-thread rather than a presenter’s convenience-order?

377: “latter” —> “resulting” variable?

Fig. 7: “Correlation structure” of what field? (zonal wind, correlated with the base time series which should be given a symbol: W_ref or something like that)

389: “source” —> base time series, which could be given a symbol: W_ref or something like that for clarity?

392, 392-394, 399, 403: all not shown? This narrative timeline is perhaps too much detail and drama.

399: “trough” and “crest” — what do these mean? Is this the u field, does it even show vertical displacements at all?

Fig. 8: “Squall line structure” — what does this mean? Covariance of u with the (standardized?) W_ref or whatever is the base series?

408: “circulation” — what does it mean? We are looking at complicated multi-lobed structure of the u field.

429: the narrative gets kind of sprawling here, is this storytelling all necessary?

433: “compensating” — what does this word mean? It implies a big back-story in the authors’ minds about how things are related and constrained, makes me nervous.

442-4: huh, how does removing all buoyant gridcells remove “gravity wave contributions from saturated parcels” ?

445: “a lot vertical transports” —> “lot of” — but arguably these aren’t really transports, merely oscillations.

455-6: “cloud fraction” — in the y direction?

Fig. 9: so crude, circular blobs and various arrows and low quality, makes a poor impression

461: Paragraph break suggested, “Most of the …” is an important statement and should lead a paragraph, not be buried in a very long one.

472: “circulation describes in the earlier part of this paragraph”. First off, the paragraphs are unclear (please indent them). Also, text can always be rearranged to never have to use such a crude self-referential grammatical constructs. Repeating an earlier confusion, what is this “circulation”? It seems like a complex multi-lobed bag of features in the u field, some convective and some wave-related, is that right?

480-onward: “error” —> “difference”

Fig. 11: very useful, it shows how feature-relative differences are better than simple Eulerian differences. Why not color the lines, so we can see if they cross and interleave?

550: “differential propagation” — isn't it simply the different strength, not actually a difference in ‘propagation'? Perhaps ‘propagation of differences’?

587-592: “compensation…opening up space…” This makes me nervous about possible misconceptions regarding causality and how mass continuity works and is maintained.

From line 500 or so: the paper is getting rather long and verbose… can it be streamlined? If all the key results have been shown, why not gather them crisply and close?

Congratulations again on an interesting study and well-deserved student degree.

Citation: https://doi.org/10.5194/egusphere-2022-515-RC1
RC2:
'Comment on egusphere-2022-515', Anonymous Referee #2, 01 Aug 2022
Review of “Evolution of squall line variability and error growth in an ensemble of LES“ by Edward Groot and Holger Tost

The study investigates the evolution of an ensemble of simulated squall lines. The simulations are performed with a comparatively fine computational mesh, and an ensemble consisting of 10 members is run. The characteristics of the simulated squall lines are discussed, and the evolution of the spread of the ensemble is investigated. Thereby, the focus is on points of the simulation, that are decisive for the generation of spread between individual members. The points in time when the convection starts to interact with itself through cold-pool dynamics and gravity waves turn out to be the points where the distance between ensemble members increases most. It is very nice to see simulations performed at such a high resolution, and very good to see that an ensemble was run. Yet, I do have some questions on the setup of the model. The quality of the manuscript is good, the text is well-written, but tends to be lengthy, descriptive and repetitive in some parts, with important information lacking at other instances. In these parts a short but concise statement on essential information to reproduce the study is needed. Overall, I suggest major revisions before the paper can be published.

Major Comments:

vertical grid spacing: while the authors invest a lot of computational resources into running an ensemble at a high horizontal resolution, the vertical grid spacing is relatively poor. The equidistant spacing of 100 m in the vertical is in my view inadequate to resolve the cold-pool dynamics and maybe also the melting layer properly. Especially as the cold pool plays a vital role for the further spread between ensemble members, a fine grid seems to be critical to resolve differences in the evolution of the cold pool, its interaction with the flow and the feedback onto the squall line dynamics. I suggest to rerun the control simulation with a vertically stretched grid and to document the differences in cold-pool dynamics with the equidistant grid.

Please provide more information about the employed tracers, for example as a further subsection in section 2. How are they transported by the flow, in which way is the coupling to microphysics (sedimentation) and turbulence realized, what is the treatment of the tracers in the surface layer? The last point my be especially important as the lowest atmospheric layer is very deep, and without a careful treatment tracers may be stuck in the surface layer.

The Weisman and Klemp (1982) sounding is established and popular. Yet, it has been criticized for being very unstable and favourable for convection. For the case at hand this means that in all ensemble members a squall line develops. A profile that is less favourable for convective development the ensemble spread may be much larger, as some members may not be able to produce a vivid squall line.

The reference simulation appears at one end of the spectrum, while I would have expected it somewhere in the middle of the ensemble. Do you have an explanation for this behaviour?

Minor Comments:

subsection 2.1: please give more details about the formulation of microphysics, especially the treatment of the condensation process (via saturation adjustment ? ) and the evaporation process, as they will be crucial for the development of up- and downdrafts and thereby ensemble spread.

Please provide more information on the perturbation of the initial conditions, especially on the perturbation of z_i. In which way is the perturbation of z_i transferred to the atmospheric profile.

The top of the model domain is at 20 km, with a sponge extending down to 15 km. Taking a look at e.g. Figure 8 some of the convection seems to interact with the sponge already. Did you see any signals of interaction of the convection with the sponge?

Figure 1: Please specify the computation of the parcel ascent. The red line seems to start at some elevated point. Please also remove the “(left)” statement.

Section 3.1: please provide some detail about the computation of the radar reflectivity

section 3.3.1: there is some directional shear in the simulations given by the increasing v velocity component. The averaging over the y direction ignores this directional shear. In which way did you account for this?

Section 3.3.2: please give more detail on the ensemble sensitivity analysis. The section is impossible to understand without first taking a look into the cited papers. The 4th and 5th paragraph of the subsection is hard to follow, there is no figure supporting the statement “during the first 15 minutes of simulation time ...” and “After 15-20 minutes”

Section 3.3.3 downdraft selection: by selecting only grid points that contain hydrometeors, the downdrafts where all rain has been evaporated will be disregarded. A better choice could be to increase the magnitude and/or to check for hydrometeors above.

Line 461-467: I cannot follow the argumentation here. Judging from Figure 10 the downdraft mass flux at a height of ~1.5-2km seems to show the largest variability. Please be more specific or rephrase.

Line 590-593: I disagree with this statement. Downdrafts will carve their space, irrespective if there is space available or not, thereby killing updrafts.

Technical Points:

please add labels to figure panels for better readability

Line 9: insert “of” after “amount”

Line 37: “can be” → has been

Line 68: “of error” → “the error”

Line 100: “an comprehensive” → “a comprehensive”

Line 165: insert “km” after “40”

Figure 3: the units at the lower right corner of the figure are very hard to read, please also repeat in the figure caption

Line 270: I do not understand the statement “so the upward transport of mid-level entrainment...”. Did you mean the upward transport of diluted air masses that have undergone entrainment processes before ?

caption Figure 4: insert “of” before “a new line”

caption Figure 5: “salmon color”: I interpret this as white ?

Line 295: insert comma after “first”

Line 312: insert a space between the full-stop after “LFC”

Line 370: insert “the” before “y-averaged” and “x-z plane”

Line 392: unit “km” should be text font.

Line 420: “similar” → “similarly”

Line 471: remove parenthesis

Line 509: insert “in” after “differences”

Line 632: something is missing here

Line 729: “which is varies” → “which is varied” or “which varies”
Citation: https://doi.org/10.5194/egusphere-2022-515-RC2
AC1: 'AC: response to RC1 and RC2', Edward Groot, 20 Sep 2022

Please find the comments attached in the .pdf.

Citation: https://doi.org/10.5194/egusphere-2022-515-AC1
AC2: 'AC: response to RC1 and RC2', Edward Groot, 20 Sep 2022

The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-515/egusphere-2022-515-AC2-supplement.pdf

Citation: https://doi.org/10.5194/egusphere-2022-515-AC2

Interactive discussion

Status: closed

RC1: 'Comment on egusphere-2022-515', Anonymous Referee #1, 08 Jul 2022

Commendations to the authors on an interesting computation, and congratulations to the student on a PhD well earned.

However, this manuscript seems to be not well written for the professional literature, rather it is in the somewhat tedious style of a dissertation. The captive audience of a reader committee would make all their check marks: enough effort reported, quantitative analysis technique mastery, awareness of caveats, and the student made thoughtful connections to some prior papers in the literature. But earning the attention of readers and (hopefully) citation in the professional literature is a different enterprise, and a major revision should be undertaken for that purpose. However, I do not volunteer to review again - I had forgotten that I tend to dislike ACP, perhaps for its apparent incentives toward first-draft quality or length. I haven't studied the business model, but experiences seem mixed.

Main science points of the manuscript as I see them:

* Second-generation development of deep convection in a triggered, highly unstable sounding is a delicate process, which adds a burst of difference growth to an ensemble, and introduces a need for feature-centered rather than Eulerian measures.

* The main mechanisms of this delicate secondary development involve a combination of cold-pool gust front lifting plus the strength of thermally forced gravity wave displacements in the inversion layer of the lower free troposphere.

* Low-level shear enhances gust front lifting on the downshear side, adding asymmetry that is convenient for analysis purposes. It also has longer-term effects on the resulting systems, but those are not really evident here.

* Because the redevelopment process is delicate and amplifies differences, any small variation can generate an ensemble of second-generation outcomes. Here the low-level shear strength (distractingly cast as an altitude difference of the linear shear layer's top which is smaller than the grid spacing!) happens to have been used. There is no indication that the results are monotonic with this shear strength perturbation: ENS-05 has the largest perturbation, but ENS-03 has the biggest outcome difference, and the final Abstract sentence suggests so too. If the perturbation size and outcome difference magnitude are uncorrelated, then emphasizing the role of shear is misleading: lucky resonances among cold pools and internal waves aloft seem to be the main unstable difference-makers, and anything might cause ensemble divergence.

* Secondary development or non-development affects both the gust front and deep convection+wave fields at later times, of course (Fig. 3), and has clear correlates in the top height, width, volume, and thus mass flux of the squall line at any given time (however it is measured; the outlines in Fig. 5 are as informative or more so than the colored tracer difference that predominates in the attention of the viewer).

* The time scale of deep difference development is about 40 minutes, at which time about half the 80-minute ensemble spread has been established already (Fig. 6).

* Outcomes (secondary develop/non-develop) appear to be coherent in the vertical at least up to midlevels (there is little curve crossing in Fig. 10... why not color all 10 lines, either arbitrarily or by the value of their initial perturbation z_top, then making reference and ENS-3 bolder for clarity?)

High-level style critique of manuscript:

Many of the key findings are quite buried behind overly detailed and sometimes defensive material about methods in a deep, murky dissertation-style text. This does not serve a professional reader’s interests optimally, since at this point the student does not need to prove amount of effort or technical mastery, but rather a sharp eye and crisp tongue about what is important. The slow resurfacing at the end into Synthesis and then Conclusion sections (also rather long) does help to pull these key points and highlights out somewhat, but those key points could be even more polished into the Abstract for instance. There is no role for suspense (historical or at readtime) in scientific writing. But those wrapup sections (4 and 5) come after a long slog of sometimes unclear prose in the late-middle (the very long sections 3.3 and 3.4).

Helpfully, section 3.2 leads it off, with simple differences between the most-different ensemble members. In this reader’s view sections 3.3 and 3.4 then add little scientifically, except a glimpse of the 10 members as spaghetti plots (why not colorize all 10 members so we can see if curves cross?). An unclear overanalysis in confusing statistical terms (terms like “source” and “error”, and “auto-correlation” for intra-ensemble rather than temporal-lag correlations) were unhelpful or confusing, and it all gave few clear insights that aren’t in the bullets above and in section 3.2. Might the paper or at least sections 3.3 and 3.4 be cut by aspirationally 50% with no loss (and a gain of clarity) on the reader’s part? Long stretches of text appear to describe figures not shown, without stating (not shown). Who needs these, who will remember them an hour after reading much less a month or year?

Detailed local comments:

Abstract: Eliminate the first sentence. Line 9: “amount —> “amount of”. disturbed —> perturbed. error —> difference. circulation —> divergent wind pattern.

Lines 19-20: The meaning of this result is that the perturbations chosen are in a non-essential field, but that even those differences grow (or explode). What does the word “intrinsic limit”, lifted from some over-realm of philosophy it seems, really add to this idea?

A general game: how many words could be trimmed or eliminated without loss of meaning? In journals with page charges there are incentives; perhaps a lack of incentive is what I have not liked about ACP (does it have length charges?)?

158: “interface height” — this is just a reference value in an analytic formula, which translates into shear strength on the 100m grid, right? This description was quite confusing.

197: “dam breaking” is never mentioned, not all readers may understand this allusion to a classic problem’s metaphor

248-250: what and where are we looking?

253: does your contouring routine treat \ sloped features different from / sloped features?

276: “difference" vs. in Fig 4 caption, sign is different (or ambiguous). Why not make it clear?

291: “will be” —> “were”

301: “reduced in the reference” seems backward to the idea of a reference vs. perturbation

302: “no less than 38.3(!)…only 23.5” — these seem like rather dramatic descriptions of excessively precise numbers that are not terribly different, and how was the second one even measured?

330: “argmax” is a useful word from the code world

Fig. 6: “autocorrelation” is often used to mean a series with time lags, perhaps say “intra-ensemble correlations at the indicated time with the minute-80 ensemble spread”

374: “Now we will continue” seems to presume the reader’s commitment, can topic phrases and topic sentences be part of a thought-thread rather than a presenter’s convenience-order?

377: “latter” —> “resulting” variable?

Fig. 7: “Correlation structure” of what field? (zonal wind, correlated with the base time series which should be given a symbol: W_ref or something like that)

389: “source” —> base time series, which could be given a symbol: W_ref or something like that for clarity?

392, 392-394, 399, 403: all not shown? This narrative timeline is perhaps too much detail and drama.

399: “trough” and “crest” — what do these mean? Is this the u field, does it even show vertical displacements at all?

Fig. 8: “Squall line structure” — what does this mean? Covariance of u with the (standardized?) W_ref or whatever is the base series?

408: “circulation” — what does it mean? We are looking at complicated multi-lobed structure of the u field.

429: the narrative gets kind of sprawling here, is this storytelling all necessary?

433: “compensating” — what does this word mean? It implies a big back-story in the authors’ minds about how things are related and constrained, makes me nervous.

442-4: huh, how does removing all buoyant gridcells remove “gravity wave contributions from saturated parcels” ?

445: “a lot vertical transports” —> “lot of” — but arguably these aren’t really transports, merely oscillations.

455-6: “cloud fraction” — in the y direction?

Fig. 9: so crude, circular blobs and various arrows and low quality, makes a poor impression

461: Paragraph break suggested, “Most of the …” is an important statement and should lead a paragraph, not be buried in a very long one.

472: “circulation describes in the earlier part of this paragraph”. First off, the paragraphs are unclear (please indent them). Also, text can always be rearranged to never have to use such a crude self-referential grammatical constructs. Repeating an earlier confusion, what is this “circulation”? It seems like a complex multi-lobed bag of features in the u field, some convective and some wave-related, is that right?

480-onward: “error” —> “difference”

Fig. 11: very useful, it shows how feature-relative differences are better than simple Eulerian differences. Why not color the lines, so we can see if they cross and interleave?

550: “differential propagation” — isn't it simply the different strength, not actually a difference in ‘propagation'? Perhaps ‘propagation of differences’?

587-592: “compensation…opening up space…” This makes me nervous about possible misconceptions regarding causality and how mass continuity works and is maintained.

From line 500 or so: the paper is getting rather long and verbose… can it be streamlined? If all the key results have been shown, why not gather them crisply and close?

Congratulations again on an interesting study and well-deserved student degree.

Citation: https://doi.org/10.5194/egusphere-2022-515-RC1
RC2:
'Comment on egusphere-2022-515', Anonymous Referee #2, 01 Aug 2022
Review of “Evolution of squall line variability and error growth in an ensemble of LES“ by Edward Groot and Holger Tost

The study investigates the evolution of an ensemble of simulated squall lines. The simulations are performed with a comparatively fine computational mesh, and an ensemble consisting of 10 members is run. The characteristics of the simulated squall lines are discussed, and the evolution of the spread of the ensemble is investigated. Thereby, the focus is on points of the simulation, that are decisive for the generation of spread between individual members. The points in time when the convection starts to interact with itself through cold-pool dynamics and gravity waves turn out to be the points where the distance between ensemble members increases most. It is very nice to see simulations performed at such a high resolution, and very good to see that an ensemble was run. Yet, I do have some questions on the setup of the model. The quality of the manuscript is good, the text is well-written, but tends to be lengthy, descriptive and repetitive in some parts, with important information lacking at other instances. In these parts a short but concise statement on essential information to reproduce the study is needed. Overall, I suggest major revisions before the paper can be published.

Major Comments:

vertical grid spacing: while the authors invest a lot of computational resources into running an ensemble at a high horizontal resolution, the vertical grid spacing is relatively poor. The equidistant spacing of 100 m in the vertical is in my view inadequate to resolve the cold-pool dynamics and maybe also the melting layer properly. Especially as the cold pool plays a vital role for the further spread between ensemble members, a fine grid seems to be critical to resolve differences in the evolution of the cold pool, its interaction with the flow and the feedback onto the squall line dynamics. I suggest to rerun the control simulation with a vertically stretched grid and to document the differences in cold-pool dynamics with the equidistant grid.

Please provide more information about the employed tracers, for example as a further subsection in section 2. How are they transported by the flow, in which way is the coupling to microphysics (sedimentation) and turbulence realized, what is the treatment of the tracers in the surface layer? The last point my be especially important as the lowest atmospheric layer is very deep, and without a careful treatment tracers may be stuck in the surface layer.

The Weisman and Klemp (1982) sounding is established and popular. Yet, it has been criticized for being very unstable and favourable for convection. For the case at hand this means that in all ensemble members a squall line develops. A profile that is less favourable for convective development the ensemble spread may be much larger, as some members may not be able to produce a vivid squall line.

The reference simulation appears at one end of the spectrum, while I would have expected it somewhere in the middle of the ensemble. Do you have an explanation for this behaviour?

Minor Comments:

subsection 2.1: please give more details about the formulation of microphysics, especially the treatment of the condensation process (via saturation adjustment ? ) and the evaporation process, as they will be crucial for the development of up- and downdrafts and thereby ensemble spread.

Please provide more information on the perturbation of the initial conditions, especially on the perturbation of z_i. In which way is the perturbation of z_i transferred to the atmospheric profile.

The top of the model domain is at 20 km, with a sponge extending down to 15 km. Taking a look at e.g. Figure 8 some of the convection seems to interact with the sponge already. Did you see any signals of interaction of the convection with the sponge?

Figure 1: Please specify the computation of the parcel ascent. The red line seems to start at some elevated point. Please also remove the “(left)” statement.

Section 3.1: please provide some detail about the computation of the radar reflectivity

section 3.3.1: there is some directional shear in the simulations given by the increasing v velocity component. The averaging over the y direction ignores this directional shear. In which way did you account for this?

Section 3.3.2: please give more detail on the ensemble sensitivity analysis. The section is impossible to understand without first taking a look into the cited papers. The 4th and 5th paragraph of the subsection is hard to follow, there is no figure supporting the statement “during the first 15 minutes of simulation time ...” and “After 15-20 minutes”

Section 3.3.3 downdraft selection: by selecting only grid points that contain hydrometeors, the downdrafts where all rain has been evaporated will be disregarded. A better choice could be to increase the magnitude and/or to check for hydrometeors above.

Line 461-467: I cannot follow the argumentation here. Judging from Figure 10 the downdraft mass flux at a height of ~1.5-2km seems to show the largest variability. Please be more specific or rephrase.

Line 590-593: I disagree with this statement. Downdrafts will carve their space, irrespective if there is space available or not, thereby killing updrafts.

Technical Points:

please add labels to figure panels for better readability

Line 9: insert “of” after “amount”

Line 37: “can be” → has been

Line 68: “of error” → “the error”

Line 100: “an comprehensive” → “a comprehensive”

Line 165: insert “km” after “40”

Figure 3: the units at the lower right corner of the figure are very hard to read, please also repeat in the figure caption

Line 270: I do not understand the statement “so the upward transport of mid-level entrainment...”. Did you mean the upward transport of diluted air masses that have undergone entrainment processes before ?

caption Figure 4: insert “of” before “a new line”

caption Figure 5: “salmon color”: I interpret this as white ?

Line 295: insert comma after “first”

Line 312: insert a space between the full-stop after “LFC”

Line 370: insert “the” before “y-averaged” and “x-z plane”

Line 392: unit “km” should be text font.

Line 420: “similar” → “similarly”

Line 471: remove parenthesis

Line 509: insert “in” after “differences”

Line 632: something is missing here

Line 729: “which is varies” → “which is varied” or “which varies”
Citation: https://doi.org/10.5194/egusphere-2022-515-RC2
AC1: 'AC: response to RC1 and RC2', Edward Groot, 20 Sep 2022

Please find the comments attached in the .pdf.

Citation: https://doi.org/10.5194/egusphere-2022-515-AC1
AC2: 'AC: response to RC1 and RC2', Edward Groot, 20 Sep 2022

The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2022/egusphere-2022-515/egusphere-2022-515-AC2-supplement.pdf

Citation: https://doi.org/10.5194/egusphere-2022-515-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

AR by Edward Groot on behalf of the Authors (11 Oct 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (24 Oct 2022) by Thijs Heus

RR by Anonymous Referee #3 (05 Dec 2022)

Suggestions for revision or reasons for rejection

This manuscript discusses the growth of differences that arise in an ensemble of Large Eddy Simulations of an idealised squall line. The authors identify and discuss different stages of growth in error (assuming simulation differences represent error growth). They identify two stages of error growth in their simulations, and argue the first of these is associated with differences in gravity waves generated by the initial convection. These gravity waves lead to the initiation of secondary cells ahead of the squall line in some of the simulations, which impact on cold pool development later on. The second stage of error growth is associated with differences in cold pool propagation.

Although the findings of this study may not be groundbreaking, the topic is certainly of interest to the atmospheric dynamics community. The authors clearly show a good knowledge of the literature on squall lines and error propagation in numerical models. However, in its current form the manuscript has a number of issues that need to be addressed before it can be published. Most of these are only textual, but resolving them will require restructuring the results section.

1) The key messages of the study and the motivation for it aren’t made sufficiently clear at the onset.
Although there is a discussion of the mechanisms that will be studied, it is not clear at the onset what motivated the study (what questions were left open by previous studies). A clearer focus and more guidance would also help to streamline the results section.

2) Several sections are difficult to read/interpret. Some sections could be reduced as they contain too much detail. A previous reviewer has already given some very helpful feedback here, which could be exploited further.
- Section 2.4 is hard to interpret without more context.
- Section 3 as a whole contains many numbers and details, which make it hard to read. The key messages are given in section 4. It would be better to only retain what is needed to support section 4. It is probably best to integrate these sections, so that results and their interpretation are presented together.
- The supplementary material could likewise be reduced further (at this point, I have focused on the main text).

3) The figures need some improvement (both in terms of presentation and clarity)
It would be clearer to plot the tracer in each simulation (and only keep the difference plot if the differences are not clear from a direct comparison) in figures 4 and 5. In figure 5, I wonder if both tracers are needed to tell the story.
For figure 6, why not simply show the trajectory over time for each of the simulations and use that as a starting point for discussion? The lag-correlation is harder to interpret.
For figure 8 and 9, using equations rather than words in the text would make it clearer what precisely is plotted. In terms of presentation style, some axis labels are missing, and some text overlaps. Time units switch between minutes and hours.

4) Details about the differences between the simulations that form the ensemble are unclear.
As I understand it, the simulations differ by the height of the zonal shear layer, but it is not clear which member corresponds to which interface height. Though the authors argue the interface height does not monotonically relate to e.g. “w_loc”, it would still be good to order the simulations by it.

5) In section 3.3.2, the precursors may not always be driving the target.
For example, a higher precipitation flux may cause higher evaporation and then faster cold pool propagation. That said, faster propagation could indeed also lead to more intense convection. In this context, it may be worth looking at a paper by Alfaro (2017) in JAS “Low-Tropospheric Shear in the Structure of Squall Lines: Impacts on Latent Heating under Layer-Lifting Ascent”.

DETAILED COMMENTS

General
- Check the text for compound (multi-word) adjectives, and hyphenate these: e.g. “three dimensional” → “three-dimensional”; “high resolution simulations” → “high-resolution simulations”.
- Remove/replace words that can be left out with no loss of information, e.g: “Presented diagnostics” → “diagnostics” ; “used scheme” → “microphysics scheme”; “The applied initial conditions” → “The initial conditions”.
- The subject “One” is overused in the text. I realise some authors try to avoid “we”, but the use of the first person makes it clearer whether the authors agree with a line of thought or not.
- Where two references are given outside parentheses, replace “;” by “and” (e.g. line 605).

In several places in the introduction, the text is vague/general/unclear, for example:
- Line 17: “Given the increasing computational resources”
- Line 21: “It also includes the aspect of representation”
- Line 25: “How squall lines depend on microphysics, shear and instability has been investigated rigorously by now, (e.g. Morrison et al., 2009; Grant et al., 2018; Adams-Selin, 2020a, b).” (also note the comma here)
- Line 35: “This was the core feature of both sensitivity studies.”
- Line 69: “A sensitivity of these discrete convective cell was identified, which lead to a dependence of initiation on the active treatment of radiation.”

Line-by-line
- Line 1: Remove the first occurrence of “ensemble”.
- Line 4: “cold pool acceleration within the ensemble envelope” → “differences in cold pool propagation within the ensemble”
- Line 18: “augmented” → Probably this is not the right word. Do you mean that these systems have been studied in CRM, and more recently also in LES?
- Line 19: “high resolutions”
- Line 24: “true convergence” → note that there may be convergence of bulk properties (see e.g. work by Wolfgang Langhans and others), even if there is no numerical convergence.
- Line 39: Mentioning the work of Lorenz (1969) here already would be beneficial.
- Line 43: “Despite this focus they have compared the error growth of divergent and rotational wind components and found that divergent winds are mostly affecting larger scale errors.” → No need to start this sentence with “despite”.
- Line 51: These 4 points could be shortened, especially the ones that are not of interest here.
- Line 98: “, but by targeting at an inflow layer” → “, and by targeting at an inflow layer”
- Line 107: “are pointed out” → rephrase
- Line 119: “using the model version of Bryan (2019)” → is there a version number?
- Line 120: “All grid cells have a 200 m horizontal grid spacing and 100 m in the vertical.” → “All grid cells have a 200 m horizontal grid spacing and 100 m vertical grid spacing.”
- Line 123: “2 moment” “two-moment”
- Line 125: “Radiation is not actively resolved.” → are any tendencies prescribed?
- Line 141: “criticality” → “importance” (or explain what is meant)?
- Line 146: Even local noise would generate some 3D behaviour, though this may be more confined to small scales.
- Line 166: “Furthermore, the ensemble members all have slightly different boundary conditions, as controlled by their own evolution nearby/at the boundaries. The boundary conditions are solely based on their conditions, with the first derivatives set to zero right at the boundary.” → This is unclear, possibly what is meant is that the values at the boundary are different, even though the same type of boundary conditions is applied.
- Line 184: “valid” → “valid for”
- Line 209: y-variations may not be “smoother”, but on smaller scales?
- Line 234: “analysed in Section 3.2.2 for further analysis” → rephrase
- Figure 3: the upper levels could be left out.
- Line 247: “ocurs” → “occurs”
- Line 255: “about a couple of km” → “several km”
- Line 268: “time of installation until given output time” → rephrase
- Line 279: “peculiar patterns” → rephrase
- Line 289: “massively different” → rephrase
- Line 310: “as it has been” → “as has been”
- Line 312: “in the former section” → “in the previous section”
- Line 342: The use of the word “uncertainty” here is confusing, as it refers to simulations with small differences.
- Section 3.2.2: w_loc is always the value after 30 minutes, is this correct?
- L 384: “many more faster gravity wave signals” → the three comparatives make it hard to interpret this sentence.
- L 385: “would definitely pass the statistical significance test” → why not simply check it passes.
- Line 396: “The signal of the this stage” → ??
- Line 416: “(positive for updraft detection)” → remove
- Line 439: remove second occurrence of “causes”
- Line 466: “a process leading to highly unbalanced tendencies dominates error tendencies.” → explain
- Line 468: the word “climatological” is confusing in the context of the present paper.
- Line 471: “The zonal wind approach is differs” → differs
- Line 493: remove “do”
- Line 567: “doing” → “making”
- Line 587: “halfway of” → “half of”
- Section 4.2: Start the discussion with outcomes, rather than limitations (e.g. “This study identifies a highly relevant window for squall line error growth in the first 80 minutes of the
development, in which systematic variability in the squall line relative flow has fully developed and also decays.”)
- Line 599-630: This section can be shortened to focus on the (new) results presented in the manuscript.
- Line 614: “In spite of their analysis mostly carried out in spectral space” (“being” missing)
- Line 633: “apart from the depth of the shear layer” → put at the end of the sentence.
- Line 631: The idealisation certainly has the advantage that it simplifies the analysis, but also means that the spin-up of convection happens everywhere across the y-axis at the same time, which may not be representative of the way differences develop in an NWP ensemble.
- Line 634: “The high degree of 2D” → rephrase
- Line 640: the correlations look plausible, but the strength of the evidence seems overstated here. It can be questioned if the ensemble members can be treated as fully independent realisations given the small differences between them. Here, it would be good to show that w_loc is indeed poorly correlated with the interface height.
- Line 690: “Noteworthy” → “Notably”? Can you clarify this sentence?

Hide

ED: Publish subject to minor revisions (review by editor) (13 Dec 2022) by Thijs Heus

AR by Edward Groot on behalf of the Authors (22 Dec 2022) Author's response Author's tracked changes Manuscript

ED: Publish as is (23 Dec 2022) by Thijs Heus

AR by Edward Groot on behalf of the Authors (23 Dec 2022) Author's response

Journal article(s) based on this preprint

13 Jan 2023

Evolution of squall line variability and error growth in an ensemble of large eddy simulations

Edward Groot and Holger Tost

Atmos. Chem. Phys., 23, 565–585, https://doi.org/10.5194/acp-23-565-2023,https://doi.org/10.5194/acp-23-565-2023, 2023

Short summary

Edward Groot and Holger Tost

Supplement

https://doi.org/10.5194/egusphere-2022-515-supplement

Data sets

Output data and namelist - README file ’Evolution of squall line variability and error growth in an ensemble of LES’ Groot, Edward https://doi.org/10.5281/zenodo.6619313

Edward Groot and Holger Tost

Viewed

Total article views: 446 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
342	88	16	446	36	11	5

HTML: 342
PDF: 88
XML: 16
Total: 446
Supplement: 36
BibTeX: 11
EndNote: 5

Views and downloads (calculated since 01 Jul 2022)

Month	HTML	PDF	XML	Total
Jul 2022	163	36	8	207
Aug 2022	60	17	3	80
Sep 2022	45	12	4	61
Oct 2022	19	6	0	25
Nov 2022	27	7	0	34
Dec 2022	19	7	0	26
Jan 2023	9	3	1	13
Feb 2023	0
Mar 2023	0

Cumulative views and downloads (calculated since 01 Jul 2022)

Month	HTML	PDF	XML	Total
Jul 2022	163	36	8	207
Aug 2022	60	17	3	80
Sep 2022	45	12	4	61
Oct 2022	19	6	0	25
Nov 2022	27	7	0	34
Dec 2022	19	7	0	26
Jan 2023	9	3	1	13
Feb 2023	0
Mar 2023	0

Viewed (geographical distribution)

Total article views: 414 (including HTML, PDF, and XML) Thereof 414 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 24 Mar 2023

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (2973 KB)
Metadata XML

Short summary

Thunderstorm systems play an important role in the dynamics of the Earth’s atmosphere and some of them form a well organised line: squall lines. Simulations of such squall lines with very small initial perturbations are compared to an unperturbed control simulation. The evolution of perturbations and processes amplifying them are analysed. It is shown that the formation of new secondary thunderstorm cells (after the initial primary cells) directly ahead of the line affects the spread strongly.


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