the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Feb 2023
Impact of the Atmospheric Boundary Layer Height on the Concentration of Chemical Species in Numerical Model Simulations
Abstract. The top of the atmospheric boundary layer (ABL) serves as a boundary separating turbulent air in the ABL and free atmosphere in the troposphere. The increase/decrease of the ABL height (ABL-H) can dilute or concentrate chemical species in the ABL. In numerical simulations the ABL-H is involved in computing the vertical diffusivity and the counter-gradient term under the unstable condition in ABL parameterization schemes. Therefore, uncertainties in the ABL-H due to the uncertainties in meteorological fields and different definitions can affect the numerical predictions of the concentration of chemical species. To understand the impacts, a 1-D diffusion model with the K-profile scheme and an 3-D air quality forecast model with the turbulent kinetic energy (TKE)-based scheme are employed to examine the sensitivities of diffusion coefficient,counter-gradient term and tracer's concentration to the ABL-H analytically and numerically. Sensitivity tests with the 1-D model show that the increase/decrease ofthe ABL-H leads to the decrease/increase of concentration of tracers under both stable and unstable conditions due to the increase/decrease of the volume of air within the ABL. Under the unstable condition, the increase/decrease of the ABL-H also enhances/weakens the vertical diffusivity, and leads to the decrease/increase of the concentration of tracers for negative/positive vertical gradient of tracer. The impact of the ABL-H through the counter-gradient term is much smaller than the impact through changing the volume of trace and vertical diffusivity in the ABL. Sensitivity tests with the 3-D numerical model with the TKE scheme show increase/decrease of the ABL-H leads to strong/weak vertical diffusivity, but hasvery small impact of the ABL-H on the concentrations of pollutants over urban centers for the ABL-Hs which is defined based on the Richardson number and temperature.
- Preprint
(674 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on acp-2023-37', Anonymous Referee #1, 06 Mar 2023
The manuscript describes some aspects of the sensitivity of simulated concentrations to the estimate of ABL height in certain boundary layer schemes. The title is misleading. The presentation is unclear and confusing. There may be some useful material here for certain specific users, but I do not see a path to acceptance for anything like the current form of this material.
General comments:
1. The title is misleading. The material is really about the errors incurred when the BL height is estimated incorrectly in certain kinds of schemes. Those errors are more or less serious depending on the scheme.
2. The authors do not seem to have a broad perspective on what is actually in use in models other than those of their own agency. Two types of mixing schemes are examined. Many of the errors noted are not significant problems in most state-of-the-art schemes. Even within this preprint, the errors in the TKE scheme are much less significant than in the K-profile scheme. This is an argument for not using the latter type of scheme, but that is never stated clearly.
3. The conceptual picture of the boundary layer presented here is oversimplified. The top of the ABL is not a material surface. It does not generally "serve as a boundary" (abstract). It is somewhat more acceptable to say that it is "a parameter describing the vertical extent of turbulent mixing" (Introduction), but then in the very next sentence the term "capping lid" is used. There are special times and places when the ABL top is well-defined and significantly impedes vertical mixing, and it happens that those are the conditions of most interest to the understanding of photochemical ozone pollution. A simple example of the more general situation: The "top" of a stable boundary layer is the point at which a more or less linear decrease of turbulence intensity reaches some selected threshold (becomes small enough). There is extensive recent literature, including review articles, describing the real boundary layer.
4. Presentation: Many aspects of the presentation are unclear. There are many equations, derived under different sets of assumptions. There is little overall concept or structure to guide the reader.
5. The key claim of the paper is buried at the end of section 2: "Under the unstable condition, the ABL-H affects tracer's concentration by changing the vertical diffusivity. This impact depends on both surface flux and the vertical gradient of tracers." This is generally true, but needs some qualification. Within a broad range, the exact value of K does not change tracer profiles significantly. Enough mixing is enough, unless there is some relatively strong competing process, for example large surface flux or large entrainment flux. Similarly, the value of K does not necessarily change the BL height if the stability gradient at the top is strong enough. In other words, the impact of a change in K is non-linear and depends on the stability profile.
Specific comments:
1. line 203: This is an instance of the conceptual problem noted in General item 3 above. The diffusivity need not be zero at the BL top; that is not a generally useful definition. It may be relatively small depending on the situation.
2. line 250: The heat flux does not jump suddenly at sunrise in reality. It is a smooth function of time. The emissions from traffic, however, are not very smooth. What is happening here is that the morning rush hour happens about the same time as the morning transition of the BL. Their relative timing depends on the season, among other things.
3. line 299: This is an example of General item 5 above. Many other factors contribute to the relationship between diffusivity and concentration. That is, after all, why we have models.
Citation: https://doi.org/10.5194/acp-2023-37-RC1 -
RC2: 'RC2', Anonymous Referee #2, 22 Mar 2023
This work presents the results of numerical simulations, which are parameterizations of more complex mechanisms at play in the atmosphere, and tries to give a general sense of how turbulent diffusivities and scalar concentrations depend on the height of the atmospheric boundary layer height (ABL-H). This is an important topic, but the results presented are merely those from a model, employed with several severe simplifications, entirely untethered from observed reality, and as such I recommend that it be rejected. Perhaps there is another journal that focuses solely on numerical modeling that might find it appropriate for their readership, but because this work makes no assessment of how it pertains to observations, I cannot see its utility to the ACP readership.
Major Concerns:
There are many other major concerns I have with this work that also indicate it is not ready for publication. First, there is an incomplete mention of the literature of scientists who have presented these kinds of relationships between ABL-H and scalar concentrations (most notably Schäfer et al., Meteorologische Zeitschrift, Vol. 15, No. 6, 647-658, December 2006; but also including, Yuval et al., Atmospheric Research 231 (2020); Wagner & Schäfer, Urban Climate 22 (2017) 64–79; Rigby & Toumi, Atmospheric Environment 42 (2008) 4932–4947.)
In my opinion there are too many figures (20 in total) and too little mechanistic understanding of the results presented. Gradient initial conditions (alpha) are compared, as are boundary layer maximum heights, as are diurnal flux patterns, as are different cities, as are ABL-H detection algorithms. The presentation seems unfiltered in terms of which variables are being altered over which range and for what reason and to what conclusion? Originally there are two flux patterns presented, one CO2-like and one H2O-like. But then later results are presented for CO and NOx (not shown) which have very different surface flux diurnal behaviors, making it very difficult to consider the model results in the second part of the paper in light of the first half.
The study is extremely specialized and specific so it is very hard to come away from it with any sense of how general it might be, or how its results might apply to any other situation. For instance, the work explicitly ignores the entrainment fluxes of all of these scalars. That omission, in and of itself, makes me weary of applying any of its conclusions to any particular circumstance in the real atmosphere.
I am unable to grasp why it is important to discern which parts of a parameterization (non-local/counter-gradient, diffusivity, dilution) are changing concentrations in a dynamic boundary layer (dzi/dt > 0). Why does that matter? Moreover, a changing boundary layer height is nearly synonymous with an entrainment flux (except in the case of purely divergent or convergent ABL flow with no turbulence at the ABL top.) That is, changes in ABL-H are mechanistically linked to entrainment fluxes (not always a diluting effect, by the way, consider CO2 over remote continental regions for example), so how can it be justified to ignore entrainment fluxes in all of this analysis? (See major concern above). In other words, the artificial isolation of using only variations in h and its impact on the counter-gradient term and diffusivity seems like it is not all that informative if changes in h are always coupled with concomitant changes induced by entrainment.
Because you are using the archetypes of CO2 (scheme 1) and H2O (scheme 2), the former being a top-down diffusion process in remote continental growing seasons, and the latter being typically bottom-up diffusion, then why not use the top-down and bottom-up diffusion schemes in this analysis? I realize that Holtslag & Moeng (1991) discuss eliminating this in some generalized cases, but they do admit, “Overall it is seen that the impact of entrainment flux cannot be neglected.” (See their Equations 25a,b).
Specific Concerns:
p.4, line 87: 500 Wm-2 is an incredibly large surface sensible heat flux. For example, summertime peak H0 in Yuma, AZ is about 250 W/m2.
p.5, line 110: Phi_M is typically reserved for the surface layer similarity dimensionless shear function, and Phi_H is the dimensionless temperature gradient function.
p.6, eq. 2-9: This vertical gradient term (alpha) is not a free variable. It is determined by the bottom-up and/or top-down gradient functions. I realize that in this work alpha represents this initial condition, but why would it matter where the model starts the value if it is different from that throughout most of the day? Moreover, Fig. 1 from Moeng & Wyngaard (1984) shows that alpha is a strong function of elevation (z) in the surface layer.
p.6, line 142: Traditional K-theory in the neutral surface layer yields K(z) is an approximately linear function of z. Even Holtslag & Moeng (1991) show this to be the case. How is a constant K with height relevant?
p.9, line 159: Fig. 3a indicates that K increases quite rapidly in early morning and evening. Maybe it is not clear exactly what times of day are being considered in this discussion.
9, line 164: How can you justify that alpha should be correctly reproduced in time in your model? Surely the gradients are established by the surface and entrainment fluxes (cf. Wyngaard & Brost, 1984), which your model has as a function of time of day. But for simplicity you say that you have neglected entrainment fluxes. The exact timing of these changes then seems somewhat arbitrary as alpha is really only relevant at t=0.
p.9, line 176: The comparison of percentage increase/decrease of mixing ratios is not very generalizable because an arbitrary constant (420 ppm) is being subtracted. If 400 ppm had been used as the offset these percentages would all change, so it is not a very meaningful result that can be generalized by a reader.
p.9, line 178: It would be much more descriptive and easier for a reader to follow if a term like CO2-like surface flux scalar, or H2O-like surface flux scalar, were used instead of the arbitrary terms ‘first’ and ‘second’.
p.9, line 179: I do not see Figure 8 resembling Fig. 2d much at all.
Figure 8: There is very little information in this figure. All four panels look indistinguishable, so I am not sure it is telling the reader anything outside of the fact that the results are insensitive to the arbitrary initial condition of scalar gradient. In any event, some discussion of this feature of the figure warrants mentioning.
p.10, Eq. 3-1: Where is the shear production (~ u*3/z) in this TKE budget equation? Under neutral to stable conditions this equation cannot sustain any turbulence whatsoever. Yet you have u* in your boundary condition for E.
p.10, line 199: Yes, but surely just above these heights the TKE, and therefore K, goes to zero. I see these values as just representing a slight uncertainty in the ABL-H.
Fig. 10: There is no mention of what the origin of the information in this graph is, nor what the units are. I am assuming this is from a model run, but under what meteorological conditions is never mentioned or referred to. In these urban areas, I would suspect that the shear production is, all else being equal, even more important to the sustenance and character of the turbulence due to the “rough” character of urban landscapes.
p.21, line 235: It would be helpful to know what the surface flux diurnal profile looked like for these simulations (“first” or “second” type, or presumably if it is CO and NOx, it is different from either of those), and what are the different amounts for each domain? There is no mention of anything about these model runs so they are very difficult to contextualize.
p.21, line 242: Is there any information to be gleaned from the four different cities? Why does one have smaller or larger r-values? Why show the different cities if there is nothing to observe from it, especially since they are just simulated cities anyway without any knowledge of their relative emissions or geography or meteorology.
p.21, lines 249-251: Why is the mentioned positive correlation not apparent in Fig.11?
Figure 13: These CO diurnal profiles do not look like any that I have seen before. There must be a diurnal pattern to your emissions (rush hours) which normally imprints a bimodal shape on the CO diurnal pattern because the edges of rush hour tend to occur when mixing is weak (early a.m. & early evening.) See for example, Yassin et al., Environ Monit Assess (2018) 190: 372, https://doi.org/10.1007/s10661-018-6737-9.
p.21, line 257: What does this intercity variability tell us? It is one thing to report this in a set of observations, but to present it with model results without having any conjecture as to why does not seem very informative to a general readership.
p.23, line 263: Concentrations depend on ABLH and emission fluxes (assuming the chemistry is slow with respect to this small-scale mixing.) It is therefore imperative that you describe the surface flux pattern for these sensitivity studies.
p.24, line 288: I am not sure what the merit is in discussing NO2 when you do not present any of the data. It is similar to CO in some cases because its dominant source is probably the same: internal combustion. But NO2 exchanges with O3 + NO on a timescale of minutes (faster than a lot of the dynamics under consideration here), so it is not going to be conserved upon mixing, it is going to react to changes in O3 and sunlight in a way that CO will not.
Citation: https://doi.org/10.5194/acp-2023-37-RC2 -
AC2: 'Responses to the comments of reviewer #2', REN SHUZHAN, 01 May 2023
Responses to the comments of reviewer #2
We thank the reviewer for the comments and suggestions. Some suggestions have been adopted in the revised manuscript. To address the reviewer’s major concern, (1) a new paragraph has been added in section 3 in the revised version to introduce the evaluation of the high-resolution GEM-AMCH against observations, and construction of surface emissions of air pollutants for model simulation (lines 226-236); (2) details of the sensitivity test method have been added in the introduction (lines 69-73); (3) more references suggested by the reviewer have been included (lines 34-38); number of figures in section 2 and equations in section 3 has been reduced, (4) “Uncertainties” has been added in the title to reflect motivation of this work. We hope these modifications can improve the presentation of this work and make the results of our work more understandable.
The following are our point-to-point responses to the reviewer’s points (in bold font).
This work presents the results of numerical simulations, which are parameterizations of more complex mechanisms at play in the atmosphere and tries to give a general sense of how turbulent diffusivities and scalar concentrations depend on the height of the atmospheric boundary layer height (ABL-H). This is an important topic, but the results presented are merely those from a model, employed with several severe simplifications, entirely untethered from observed reality, and as such I recommend that it be rejected. Perhaps there is another journal that focuses solely on numerical modeling that might find it appropriate for their readership, but because this work makes no assessment of how it pertains to observations, I cannot see its utility to the ACP readership.
Re: Before submitting the manuscript to ACP, we checked the topics and research activities covered by ACP. It says clearly on the ACP website that “Topics include gases, aerosols, clouds, precipitation, dynamics, radiation and their role in the Earth's climate system (including the biosphere, hydrosphere, and cryosphere). Research activities include laboratory studies, field measurements, remote sensing, modelling and data analysis, and machine learning” Since the contents of this paper include both chemical species and dynamics (turbulent mixing), this paper fits the subject of ACP described on CAP website. Furthermore, this work also fits the research activity of ACP as it uses the atmospheric modelling approach.
We agree with the reviewer that it is important to use observations to evaluate model results. GEM-MACH has been evaluated extensively at different spatial scales (see Ren et al., 2020 for references). Particularly, it has been evaluated in the urban area (Great Toronto Area) with the urbanization scheme (the town energy balance model). As long as the benchmark simulation results are evaluated against observations, the impacts of model parameters described by sensitivity results should be reliable. The evaluation of GEM-MACH has been mentioned in the revised manuscript (lines 229-230). In addition, the capability of the K-profile and TKE schemes in describing the subscale effects have also been evaluated (e.g., Holtslag & Bovilie, 1883, Mailhot & Benoit, 1982).
The reviewer’s other concerns are associated with the method employed in the sensitivity studies with numerical models. Because model parameters are interconnected in a complex way, and their connections cannot be expressed analytically, sensitivity studies have been widely applied to examine the influences of model variables on numerical simulation results (one example is the sensitivity study of the climate change associated with the double CO2 in the atmosphere). By modifying the value of a specified parameter and keeping other parameters fixed, the simulated sensitivity results can well describe the impacts and help better understand the important roles of model parameters playing in model simulations. The discussion of the sensitive test method has been added in the revised version (lines 69-73).
Major Concerns:
There are many other major concerns I have with this work that also indicate it is not ready for publication. First, there is an incomplete mention of the literature of scientists who have presented these kinds of relationships between ABL-H and scalar concentrations (most notably Schäfer et al., Meteorologische Zeitschrift, Vol. 15, No. 6, 647-658, December 2006; but also including, Yuval et al., Atmospheric Research 231 (2020); Wagner & Schäfer, Urban Climate 22 (2017) 64–79; Rigby & Toumi, R 42 (2008) 4932–4947.)
Re: We thank the reviewer for drawing our attention to several research papers on the relationship between air pollutants and ABL height. Although these studies are based on observations, they can be applied for comparing model results. The first three papers have been referred to in the revised version (lines 34-38).
In my opinion there are too many figures (20 in total) and too little mechanistic understanding of the results presented. Gradient initial conditions (alpha) are compared, as are boundary layer maximum heights, as are diurnal flux patterns, as are different cities, as are ABL-H detection algorithms.
Re: The number of figures has been reduced in the revised version by combining Fig 8 and Fig. 9 together. The impacts of the ABL height on concentration through \gamma are not sensitive to \alpha in the two figures and only the differences with \alpha=0 are shown. Please also see response to p.6, eq. 2-9.
We don’t quite understand the statement about \alpha. This parameter is employed to show how the impact of the change of diffusivity (due to the change of the ABL height) is modulated by the vertical gradient of tracers numerically. The analytical form of the modulation is described by Eq. 2.10.
The presentation seems unfiltered in terms of which variables are being altered over which range and for what reason and to what conclusion? Originally there are two flux patterns presented, one CO2-like and one H2O-like. But then later results are presented for CO and NOx (not shown) which have very different surface flux diurnal behaviors, making it very difficult to consider the model results in the second part of the paper in light of the first half.
Re: Biogenic CO2 emission and CO2 emission from industrial activities are used in section 2 for sensitivity studies with the K-profile scheme. The reason that the impact on CO (and other pollutants) is discussed in section 3 is that currently we don’t have long term simulation of CO2 with GEM-MACH. Like CO2, CO can be treated as a tracer with a long lifetime and has a similar surface emission to the emission of CO2 from industrial activities.
The details of the surface emissions used in GEM-MACH simulations have been given in the revised version (lines 231-236, please also see response to p.21, line 235). The impact on NOx is mentioned just to show the chemical reactions that NOx is involved during the daytime would not change the impact of the ABL height.
Two kinds of ABL parameterization schemes (K-profile and TKE) are used to show the impacts of the ABL height. They are different schemes with different involvement of the ABL height. While their impacts on the vertical diffusivity are similar, their impacts on concentration are quite different. The K-profile scheme is implemented in a 1-D diffusion model. Because it is a simple model the modulation on the ABL height impact can be easily examined by changing vertical gradient of tracer through \alpha and by choosing different boundary conditions. Although it is difficult to do the same thing in the 3-D GEM-MACH with realistic surface emissions and 3-D meteorological and chemical fields, the modulation effects can still be applied to the 3-D case.
The study is extremely specialized and specific so it is very hard to come away from it with any sense of how general it might be, or how its results might apply to any other situation. For instance, the work explicitly ignores the entrainment fluxes of all of these scalars. That omission, in and of itself, makes me weary of applying any of its conclusions to any particular circumstance in the real atmosphere.
Re: We don’t agree with the reviewer’s statement about the usefulness of the results of this work. Because the ABL height is a parameter measuring the extent of turbulent mixing and is involved in ABL parameterization schemes, the impact of the uncertainty in the ABL height is a general issue for air quality model simulations when the ABL height is involved in the parametrization scheme. One general conclusion from the sensitivity studies with the K-profile and TKE schemes is that over-estimated/under-estimated ABL can lead to over-estimated/under-estimated vertical diffusivity.
As we pointed out in the introduction, there are large uncertainties in the ABL height. When the ABL height is involved in the parameterization schemes, the impacts of the uncertainties should be examined in order to understand the source of uncertainties of the air quality forecast. Furthermore, the results can also help understand the different model simulation results with different definition of the ABL height. Our results show that the uncertainty in the ABL height can have big impacts when the K-profile scheme is used. To our knowledge this work is the first one to examine the impacts in detail. Although only two schemes are employed in this work the results can certainly benefit not only modelers but also the air quality research community. Understanding the impacts can help identify the sources of the uncertainties in simulation results and improve forecast results.
It is incorrect to say that the entrainment fluxes are ignored in GEM-MACH simulation. These fluxes (if they exist) are included in the sensitivity studies with 3-D GEM-AMACH in section 3. For most chemical species the surface fluxes have major contributions to the concentrations. For ozone there might be entrainment fluxes when the tropopause penetrates deep into low altitude areas due to the breakdown of potential vorticity. NOx produced by lightning storms in the troposphere can also penetrate into the ABL. These are just associated with special events.
I am unable to grasp why it is important to discern which parts of a parameterization (non-local/counter-gradient, diffusivity, dilution) are changing concentrations in a dynamic boundary layer (dzi/dt > 0). Why does that matter? Moreover, a changing boundary layer height is nearly synonymous with an entrainment flux (except in the case of purely divergent or convergent ABL flow with no turbulence at the ABL top.) That is, changes in ABL-H are mechanistically linked to entrainment fluxes (not always a diluting effect, by the way, consider CO2 over remote continental regions for example), so how can it be justified to ignore entrainment fluxes in all of this analysis? (See major concern above). In other words, the artificial isolation of using only variations in h and its impact on the counter-gradient term and diffusivity seems like it is not all that informative if changes in h are always coupled with concomitant changes induced by entrainment.
Re: The ABL height is involved in different components in the parameterization scheme in different ways. Thus, its uncertainty can have different impacts on the concentration of tracers. Understanding these different impacts can help identify the source of uncertainties in air quality forecasts and find ways to improve forecasts.
The effect of the entrainment flux is contained in the sensitivity results with the GEM-MACH because only the ABL height is changed. Although in some special events such as the penetration of tropopause into the low altitude areas and big storms which produce top-down ozone and NOx fluxes, respectively, the surface emissions have the major contributions to the concentrations of air pollutants and CO2. The work of Ren (2019, Journal of the Atmospheric Sciences) shows that the magnitude of entrainment flux decreases rapidly as height decreases. Because we are interested in the impacts in the lower ABL, the entrainment flux is not included in section 2 for simplicity. This has been mentioned in the revised version (line 102)
Because you are using the archetypes of CO2 (scheme 1) and H2O (scheme 2), the former being a top-down diffusion process in remote continental growing seasons, and the latter being typically bottom-up diffusion, then why not use the top-down and bottom-up diffusion schemes in this analysis? I realize that Holtslag & Moeng (1991) discuss eliminating this in some generalized cases, but they do admit, “Overall it is seen that the impact of entrainment flux cannot be neglected.” (See their Equations 25a,b).
Re: During the grown season the concentration of CO2 can increase with height. In section 2 the same bottom-up type of initial condition is used just to illustrate how the impacts of uncertainty of the ABL height can be modulated by the vertical gradient of CO2 with different surface emissions. Because the vertical gradient is weakened during the daytime due to the negative surface emission under the first kind of surface emission scheme, the top-down diffusion effect is reflected in the model integration.
The work of Ren (2019 Journal of the Atmospheric Sciences) shows that the magnitude of entrainment flux decreases rapidly as height decreases. The concentration in the lower part of the ABL is determined mainly by the surface emission and the vertical diffusivity. It is not included in section 2 for simplicity (line 102).
Specific Concerns:
p.4, line 87: 500 Wm-2 is an incredibly large surface sensible heat flux. For example, summertime peak H0 in Yuma, AZ is about 250 W/m2.
Re: The magnitude of sensible heat flux depends on many factors including the weather conditions (such clear or cloudy sky), locations (such as urban, rural), seasons etc. In the urban centers the monthly mean sensible heat flux in June can be 400W/m^2 around noon (Fig 8 in Masson 2000, Boundary-Layer Meteorology). The urban sensible heat flux contains contributions from walls, roofs, roads and traffic.
p.5, line 110: Phi_M is typically reserved for the surface layer similarity dimensionless shear function, and Phi_H is the dimensionless temperature gradient function.
Re: \phi_M has been replaced \phi_h in the revised version.
p.6, eq. 2-9: This vertical gradient term (alpha) is not a free variable. It is determined by the bottom-up and/or top-down gradient functions. I realize that in this work alpha represents this initial condition, but why would it matter where the model starts the value if it is different from that throughout most of the day? Moreover, Fig. 1 from Moeng & Wyngaard (1984) shows that alpha is a strong function of elevation (z) in the surface layer.
Re: In the real atmosphere \alpha is not a free variable. It is determined by the vertical distribution of chemical species. In the 1-D model, a specified \alpha is used to represent the background (initial) vertical gradient of tracer. The value of \alpha affects the magnitude and sign of the vertical gradient after the initial time and thus modulates the impact of the ABL height according to Eq. 2.10. Please also see response to the last major concern.
p.6, line 142: Traditional K-theory in the neutral surface layer yields K(z) is an approximately linear function of z. Even Holtslag & Moeng (1991) show this to be the case. How is a constant K with height relevant?
Re: K in Eq. 2.10 can be any function of z, and K used for sensitivity tests is also a function of z defined by Eq. 2.6. Constant K assumption was used in Ren & Stroud (2020) to examine the sign of Green’s function.
p.9, line 159: Fig. 3a indicates that K increases quite rapidly in early morning and evening. Maybe it is not clear exactly what times of day are being considered in this discussion.
Re: The rapid increase of the vertical diffusivity occurs at sunrise (6:00AM in this work). This was mentioned in line 131.
p.9, line 164: How can you justify that alpha should be correctly reproduced in time in your model? Surely the gradients are established by the surface and entrainment fluxes (cf. Wyngaard & Brost, 1984), which your model has as a function of time of day. But for simplicity you say that you have neglected entrainment fluxes. The exact timing of these changes then seems somewhat arbitrary as alpha is really only relevant at t=0.
Re: \alpha is used to define the vertical gradient of tracer at initial time. Because The value of \alpha affects the magnitude and sign of the vertical gradient after the initial time and thus modulates the impact of the ABL height. The sensitivity of the impact on \alpha illustrates the modulation described by Eq. 2.10. These changes depend on \alpha. Once \alpha is specified at initial time they are not arbitrary.
p.9, line 176: The comparison of percentage increase/decrease of mixing ratios is not very generalizable because an arbitrary constant (420 ppm) is being subtracted. If 400 ppm had been used as the offset these percentages would all change, so it is not a very meaningful result that can be generalized by a reader.
Re: The comparison in percentage has been changed to direct comparison in the revised version.
p.9, line 178: It would be much more descriptive and easier for a reader to follow if a term like CO2-like surface flux scalar, or H2O-like surface flux scalar, were used instead of the arbitrary terms ‘first’ and ‘second’.
Re: Both are CO2 emissions. One is biogenic, another is the emission from industrial activities. “biogenic” and “industrial” emissions are used in the revised version.
p.9, line 179: I do not see Figure 8 resembling Fig. 2d much at all.
Re: The similarity between the two figures means that the ABL height has similar impacts. This sentence has been removed in the revised version to avoid confusion.
Figure 8: There is very little information in this figure. All four panels look indistinguishable, so I am not sure it is telling the reader anything outside of the fact that the results are insensitive to the arbitrary initial condition of scalar gradient. In any event, some discussion of this feature of the figure warrants mentioning.
Re: Figure 8 and Fig. 9 are combined in the revised version. The impacts of the ABL height on concentration through \gamma are not sensitive to \alpha in the two figures and only the differences with \alpha=0 are shown. The insensitivity to \alpha has been mentioned.
p.10, Eq. 3-1: Where is the shear production (~ u*3/z) in this TKE budget equation? Under neutral to stable conditions this equation cannot sustain any turbulence whatsoever. Yet you have u* in your boundary condition for E.
Re: The shear production term is defined by S in (A.2) in Appendix A. In the numerical simulation with GEM-MACH the friction velocity is about 0.1(m/s) to 0.6 (m/s) in the urban center of the Great Toronto Area from which one can calculate u*^3/z at different heights.
The boundary condition under the stable condition is proposed by Mailhot & Benoit (1982).
p.10, line 199: Yes, but surely just above these heights the TKE, and therefore K, goes to zero. I see these values as just representing a slight uncertainty in the ABL-H.
Re: The definition of the ABL height is not unique as it can be based on the Richardson number, the gradient of potential temperature, the gradient of tracers etc. Although the top of the ABL is supposed to be a layer separating free atmosphere and the ABL characterized by turbulent mixing, the ABL height doesn’t impose constraint on the diffusion coefficient at the ABL top in the TKE scheme. This is because the ABL height is not present in the equation for calculating the diffusion coefficient.
Fig. 10: There is no mention of what the origin of the information in this graph is, nor what the units are. I am assuming this is from a model run, but under what meteorological conditions is never mentioned or referred to. In these urban areas, I would suspect that the shear production is, all else being equal, even more important to the sustenance and character of the turbulence due to the “rough” character of urban landscapes.
Re: The curves in the figure are the diurnal variation of diffusion coefficients at the top of the ABL defined based on the Richardson number and vertical potential temperature profile. They are monthly mean (July) GEM-MACH model simulation results. This has been mentioned in line 209 in the revised version. These figures are used to show that different definitions of the ABL height can lead to different vertical diffusivity near the top of the ABL in the TKE scheme. It also shows the different K in the K-profile and TKE schemes at the ABL top. The description of this figure can be found in lines 196-204.
Due to the existence of high buildings urban areas have larger roughness length which tends to produce larger turbulence in urban areas than in rural areas. However, the urban effect associated with the heat fluxes from roads, roofs, walls, and cars is more important in terms of generating turbulence. Please see Ren et al. (2020) for more details.
p.21, line 235: It would be helpful to know what the surface flux diurnal profile looked like for these simulations (“first” or “second” type, or presumably if it is CO and NOx, it is different from either of those), and what are the different amounts for each domain? There is no mention of anything about these model runs so they are very difficult to contextualize.
Re: A short paragraph has been added in section 3.1 in the revised version to introduce more details about the model configuration and surface emissions used in GEM-AMCH (lines 224-234). Over the Great Toronto Area the CO emission in summer increases rapidly around 7:00AM from 40 (g/s) to 134 (g/s) and continues to increase slowly to 171 (g/s) at 3:00PM before it starts to decrease. The diurnal variation of CO emission is similar to that of industrial CO2 emission in section 2.
p.21, line 242: Is there any information to be gleaned from the four different cities? Why does one have smaller or larger r-values? Why show the different cities if there is nothing to observe from it, especially since they are just simulated cities anyway without any knowledge of their relative emissions or geography or meteorology.
Re: Urban centers are big sources of air pollutants. In Ren et al., (2020) four northern American urban centers are chosen to investigate the common features of the impacts of urban effects on meteorological and chemical fields. Because the surface heat fluxes are different in the four urban areas due to different geometric parameters such as building fraction, building height, and urban canyon aspect ratio (Ren et al., 2020, 2022 (Urban Climate)), the impacts of the ABL-H are different. They are used in this work to examine the correlation between the ABL height and concentration of chemical species and the impacts of the uncertainties in ABL height. Please also see response to p.21, line 257.
p.21, lines 249-251: Why is the mentioned positive correlation not apparent in Fig.11?
Re: The positive correlations appear between 7:00AM to 8:00AM shown in Fig. 13 (not Fig. 11)
Figure 13: These CO diurnal profiles do not look like any that I have seen before. There must be a diurnal pattern to your emissions (rush hours) which normally imprints a bimodal shape on the CO diurnal pattern because the edges of rush hour tend to occur when mixing is weak (early a.m. & early evening.) See for example, Yassin et al., Environ Monit Assess (2018) 190: 372, https://doi.org/10.1007/s10661-018-6737-9.
Re: Yes, the CO and NOx emissions from traffic over GTA have a bimodal shape in summer (July). However, this shape is not clear in other urban areas due to different traffic patterns. For example, emissions are larger in the second rush hour period in Chicago and New York. This can be seen clearly in Fig. 2 in Ren et al. (2020 atmosphere). This figure shows the monthly mean diurnal variation of the heat fluxes from cars based on the emission of both CO and NOx.
Due to the contribution from other sources the bimodal shape of the total CO and NOx emissions are not obvious. Over the Great Toronto Area the CO emission in summer increases rapidly around 7:00AM from 40 (g/s) to 134 (g/s) and continues to increase slowly to 171 (g/s) at 3:00PM before it starts to decrease.
p.21, line 257: What does this intercity variability tell us? It is one thing to report this in a set of observations, but to present it with model results without having any conjecture as to why does not seem very informative to a general readership.
Re: Urban areas are a big source of air pollutants. Each urban area has different geometric parameters such as building fraction, building height, and urban canyon aspect ratio (Ren et al., 2020, 2022 (Urban Climate)). The work of Ren et al. (2020) shows that the surface heat fluxes in the four urban centers are different due to different geometric parameters. Because the surface flux can modulate the impact of the ABL height (in convective velocity), the correlation between concentration and the ABL height reflects the modulation effect. This has been mentioned in the revised version (lines 239-242)
p.23, line 263: Concentrations depend on ABLH and emission fluxes (assuming the chemistry is slow with respect to this small-scale mixing.) It is therefore imperative that you describe the surface flux pattern for these sensitivity studies.
Re: More details of the surface emissions are given in the revised version (lines 234-235). Please also see response to p.21, line 235
p.24, line 288: I am not sure what the merit is in discussing NO2 when you do not present any of the data. It is similar to CO in some cases because its dominant source is probably the same: internal combustion. But NO2 exchanges with O3 + NO on a timescale of minutes (faster than a lot of the dynamics under consideration here), so it is not going to be conserved upon mixing, it is going to react to changes in O3 and sunlight in a way that CO will not.
Re: NOx is a major air pollutant and is involved in photochemical reactions during the daytime. Because the uncertainty of the ABL height affects all the chemical species involved in the chemical reactions, the additional impact from other species on NOx needs to be examined by comparing the impact on CO. This has been mentioned in the revised version (lines 308-309).
Citation: https://doi.org/10.5194/acp-2023-37-AC2
-
AC2: 'Responses to the comments of reviewer #2', REN SHUZHAN, 01 May 2023
-
AC1: 'Responses to the comments of reviewer #1', REN SHUZHAN, 01 May 2023
Responses to the comments of reviewer #1
The reviewer’s comments and suggestions are appreciated. Some suggestions have been adopted in the revised manuscript. The following are our point-to-point responses to the reviewer’s points (in bold font).
The manuscript describes some aspects of the sensitivity of simulated concentrations to the estimate of ABL height in certain boundary layer schemes. The title is misleading.
Re: Following the suggestion the title of the manuscript has been modified to reflect the contents of this work. Please see the response to general comment 1 below.
The presentation is unclear and confusing. There may be some useful material here for certain specific users, but I do not see a path to acceptance for anything like the current form of this material.
Re: We don’t agree with the reviewer’s statement about the usefulness of the results of this work. Because the ABL height is a parameter measuring the extent of turbulent mixing and it is involved in ABL parameterization schemes, the impact of the uncertainty in the ABL height is a general issue for air quality model simulations when the ABL height is involved in the parametrization scheme.
As we pointed out in the introduction, there are large uncertainties in the ABL height. When the ABL height is involved in the parameterization schemes, the impacts of the uncertainties should be examined in order to understand the source of uncertainties of the air quality forecast. Furthermore, the results can also help understand the different model simulation results with different definition of the ABL height. Our results show that the uncertainty in the ABL height can have big impacts when the K-profile scheme is used. To our knowledge this work is the first one to examine the impacts in detail. Although only two schemes are employed in this work the results can certainly benefit not only modelers but also the air quality research community. Understanding the impacts can help identify the sources of the uncertainties in simulation results and improve forecast results.
The manuscript has been revised by including the following modifications: (1) adding “uncertainties in the title reflect the motivation of this work, (2) adding a new paragraph in section 3 to introduce the evaluation of the high-resolution GEM-AMCH against observations, and construction of surface emissions of air pollutants for model simulation (lines 225-235), (3) adding the details of the sensitivity test method in the introduction (lines 69-73), (4) adding more references about the studies on the impacts of the ABL height on chemical species (lines 34-38), (5) reducing number of figures in section 2 and equations in section 3. We hope that these changes can clarify confusions in the previous version and help readers better understand the results of this work.
General comments:
- The title is misleading. The material is really about the errors incurred when the BL height is estimated incorrectly in certain kinds of schemes. Those errors are more or less serious depending on the scheme.
Re: The reason that the title was chosen for this paper is that our work discusses not only the impact of uncertainties in the ABL height on the simulation of chemical species but also the impact of different definitions of the ABL height used in simulations. Following the suggestion of the reviewer, “uncertainties” has been added in the title in the revised version to make the motivation of this work clearer.
- The authors do not seem to have a broad perspective on what is actually in use in models other than those of their own agency. Two types of mixing schemes are examined. Many of the errors noted are not significant problems in most state-of-the-art schemes. Even within this preprint, the errors in the TKE scheme are much less significant than in the K-profile scheme. This is an argument for not using the latter type of scheme, but that is never stated clearly.
Re: We don’t quite understand the first sentence. Yes, there are many kinds of parameterization schemes, and we cannot examine the impacts with all the schemes. TKE and K-profile schemes are two schemes with the involvement of the ABL height and are widely employed in numerical models. The reason that they are employed in this work is not because the ABL height is involved explicitly in the schemes (under unstable condition) but also because it is involved in completely different ways. While the ABL height affects diffusivity in the lower vertical boundary condition through the convective velocity in the TKE scheme, it affects diffusivity within the ABL not only through the convective velocity but also affects the calculation of diffusion coefficient directly in the K-profile scheme. Therefore, examining the impacts of the two schemes can give us a comprehensive understanding of the impacts of the ABL height.
We also don’t understand the statement “Many of the errors noted are not significant problems in most state-of-the-art schemes”. Whether the impacts are significant or not should be based on careful examination. To the best of our knowledge our work is the first one to examine the impacts in detail. We would appreciate it if the reviewer could point out references of other similar studies based on the state-of-the-art schemes.
Both analytical and numerical results of this work show that the ABL height has very different impacts on tracers. However, these results don’t suggest which scheme should be used. They just help modelers to identify the source of uncertainties in air quality forecast and improve the performance of the forecast.
- The conceptual picture of the boundary layer presented here is oversimplified. The top of the ABL is not a material surface. It does not generally "serve as a boundary" (abstract). It is somewhat more acceptable to say that it is "a parameter describing the vertical extent of turbulent mixing" (Introduction), but then in the very next sentence the term "capping lid" is used. There are special times and places when the ABL top is well-defined and significantly impedes vertical mixing, and it happens that those are the conditions of most interest to the understanding of photochemical ozone pollution. A simple example of the more general situation: The "top" of a stable boundary layer is the point at which a more or less linear decrease of turbulence intensity reaches some selected threshold (becomes small enough). There is extensive recent literature, including review articles, describing the real boundary layer.
Re: We agree with the reviewer that the top of the ABL is not a material surface. It is not an observed variable but a derived one based on different variables such as the Richardson number, vertical gradient of potential, vertical gradient of chemical tracers etc. Therefore, there is no unique boundary layer top as its definition is not unique.
The ABL is characterized by turbulent mixing and the troposphere is turbulence free. “Boundary” and “capping lid” are used to describe the extent to which the vertical mixing and its effects on chemical species can reach vertically. To avoid confusion these two terms have been removed in the revised version.
- Presentation: Many aspects of the presentation are unclear. There are many equations, derived under different sets of assumptions. There is little overall concept or structure to guide the reader.
Re: In the revised version, paragraphs have been added to explain the details of the sensitivity test method and introduce more details of configuration of GEM-MACH and construction of surface emissions of air pollutants for GEM-MACH. Hope these modifications can make the presentation clearer.
Results of sensitivity studies are produced by numerical models. Due to complex processes and interconnections of variables, analytical methods are often employed to interpret the numerical results. Analytical results can provide independent evaluations of numerical results.
Equations in section 2 are basic equations to define variables associated with the ABL height (convective velocity, counter gradient term, diffusion coefficient etc.) in the parameterization schemes. They are simple and straightforward and can be found easily in the textbooks of boundary meteorology and published papers. Reference is given in this section (Holtslag and Moeng 1991). The impacts estimated based on the equations agree with numerical results suggesting the results can indeed provide independent evaluations of numerical results.
Readers may not be familiar with the equations in section 3. In the revised version only three equations (Eqs. 3.2, 3.3 and 3.7) are kept and the rest equations have been moved to appendix.
- The key claim of the paper is buried at the end of section 2: "Under the unstable condition, the ABL-H affects tracer's concentration by changing the vertical diffusivity. This impact depends on both surface flux and the vertical gradient of tracers." This is generally true but needs some qualification. Within a broad range, the exact value of K does not change tracer profiles significantly. Enough mixing is enough, unless there is some relatively strong competing process, for example large surface flux or large entrainment flux. Similarly, the value of K does not necessarily change the BL height if the stability gradient at the top is strong enough. In other words, the impact of a change in K is non-linear and depends on the stability profile.
Re: This claim is not only mentioned at the end of section 2 but also mentioned in the abstract (lines 13-24) and emphasized in summary and discussion (lines 312-314). To emphasize the claim a new sentence has been added to the abstract in the revised version. The dependence of the impact of K on the vertical gradient is discussed quantitively in section 2 based on different values of \alpha which is a measure of the vertical gradient of tracer.
In this work only the impact of the ABL height on K is discussed. If the stability gradient is strong at the top, turbulent mixing would be suppressed. The impact of K on the ABL height would be weak.
Specific comments:
- line 203: This is an instance of the conceptual problem noted in General item 3 above. The diffusivity need not be zero at the BL top; that is not a generally useful definition. It may be relatively small depending on the situation.
Re: Specific values are often used in the ABL height definitions. When the Richardson number is used the critical value is 0.25. If the vertical gradient of potential temperature is used, the critical value is 0. One may not use these values in practice, but these theoretical values provide guidance for the definition of the ABL height. For the same reason zero value of diffusion coefficient is mentioned. Another reason to mention this is for comparing the value of K at the top of the ABL in the K-profile and TKE schemes. It is constrained to be zero in the former scheme and no similar constraint in the latter scheme.
- line 250: The heat flux does not jump suddenly at sunrise in reality. It is a smooth function of time. The emissions from traffic, however, are not very smooth. What is happening here is that the morning rush hour happens about the same time as the morning transition of the BL. Their relative timing depends on the season, among other things.
Re: The observed diurnal variation of surface heat flux shown in Fig. 1.9 in Oke’s book (boundary layer climate) suggests the heat flux can jump rapidly at sunrise. Although the variation of sensible heat flux is not as rapid as latent flux and flux from soil, the sum of the three components has a big jump at sunrise. Corresponding to the rapid change of heat flux, the vertical diffusivity also experiences a rapid change. Of course, the jump of heat flux depends on weather, season and other conditions as suggested by the reviewer.
The concentration of tracers near the ground is affected by both vertical mixing and surface emissions. The quick change of diffusivity at sunrise can lead to a rapid drop of concentration even when the surface emission is strong.
- line 299: This is an example of General item 5 above. Many other factors contribute to the relationship between diffusivity and concentration. That is, after all, why we have models.
Re: We agree with the reviewer’s point.
Citation: https://doi.org/10.5194/acp-2023-37-AC1
-
RC2: 'RC2', Anonymous Referee #2, 22 Mar 2023
Status: closed
-
RC1: 'Comment on acp-2023-37', Anonymous Referee #1, 06 Mar 2023
The manuscript describes some aspects of the sensitivity of simulated concentrations to the estimate of ABL height in certain boundary layer schemes. The title is misleading. The presentation is unclear and confusing. There may be some useful material here for certain specific users, but I do not see a path to acceptance for anything like the current form of this material.
General comments:
1. The title is misleading. The material is really about the errors incurred when the BL height is estimated incorrectly in certain kinds of schemes. Those errors are more or less serious depending on the scheme.
2. The authors do not seem to have a broad perspective on what is actually in use in models other than those of their own agency. Two types of mixing schemes are examined. Many of the errors noted are not significant problems in most state-of-the-art schemes. Even within this preprint, the errors in the TKE scheme are much less significant than in the K-profile scheme. This is an argument for not using the latter type of scheme, but that is never stated clearly.
3. The conceptual picture of the boundary layer presented here is oversimplified. The top of the ABL is not a material surface. It does not generally "serve as a boundary" (abstract). It is somewhat more acceptable to say that it is "a parameter describing the vertical extent of turbulent mixing" (Introduction), but then in the very next sentence the term "capping lid" is used. There are special times and places when the ABL top is well-defined and significantly impedes vertical mixing, and it happens that those are the conditions of most interest to the understanding of photochemical ozone pollution. A simple example of the more general situation: The "top" of a stable boundary layer is the point at which a more or less linear decrease of turbulence intensity reaches some selected threshold (becomes small enough). There is extensive recent literature, including review articles, describing the real boundary layer.
4. Presentation: Many aspects of the presentation are unclear. There are many equations, derived under different sets of assumptions. There is little overall concept or structure to guide the reader.
5. The key claim of the paper is buried at the end of section 2: "Under the unstable condition, the ABL-H affects tracer's concentration by changing the vertical diffusivity. This impact depends on both surface flux and the vertical gradient of tracers." This is generally true, but needs some qualification. Within a broad range, the exact value of K does not change tracer profiles significantly. Enough mixing is enough, unless there is some relatively strong competing process, for example large surface flux or large entrainment flux. Similarly, the value of K does not necessarily change the BL height if the stability gradient at the top is strong enough. In other words, the impact of a change in K is non-linear and depends on the stability profile.
Specific comments:
1. line 203: This is an instance of the conceptual problem noted in General item 3 above. The diffusivity need not be zero at the BL top; that is not a generally useful definition. It may be relatively small depending on the situation.
2. line 250: The heat flux does not jump suddenly at sunrise in reality. It is a smooth function of time. The emissions from traffic, however, are not very smooth. What is happening here is that the morning rush hour happens about the same time as the morning transition of the BL. Their relative timing depends on the season, among other things.
3. line 299: This is an example of General item 5 above. Many other factors contribute to the relationship between diffusivity and concentration. That is, after all, why we have models.
Citation: https://doi.org/10.5194/acp-2023-37-RC1 -
RC2: 'RC2', Anonymous Referee #2, 22 Mar 2023
This work presents the results of numerical simulations, which are parameterizations of more complex mechanisms at play in the atmosphere, and tries to give a general sense of how turbulent diffusivities and scalar concentrations depend on the height of the atmospheric boundary layer height (ABL-H). This is an important topic, but the results presented are merely those from a model, employed with several severe simplifications, entirely untethered from observed reality, and as such I recommend that it be rejected. Perhaps there is another journal that focuses solely on numerical modeling that might find it appropriate for their readership, but because this work makes no assessment of how it pertains to observations, I cannot see its utility to the ACP readership.
Major Concerns:
There are many other major concerns I have with this work that also indicate it is not ready for publication. First, there is an incomplete mention of the literature of scientists who have presented these kinds of relationships between ABL-H and scalar concentrations (most notably Schäfer et al., Meteorologische Zeitschrift, Vol. 15, No. 6, 647-658, December 2006; but also including, Yuval et al., Atmospheric Research 231 (2020); Wagner & Schäfer, Urban Climate 22 (2017) 64–79; Rigby & Toumi, Atmospheric Environment 42 (2008) 4932–4947.)
In my opinion there are too many figures (20 in total) and too little mechanistic understanding of the results presented. Gradient initial conditions (alpha) are compared, as are boundary layer maximum heights, as are diurnal flux patterns, as are different cities, as are ABL-H detection algorithms. The presentation seems unfiltered in terms of which variables are being altered over which range and for what reason and to what conclusion? Originally there are two flux patterns presented, one CO2-like and one H2O-like. But then later results are presented for CO and NOx (not shown) which have very different surface flux diurnal behaviors, making it very difficult to consider the model results in the second part of the paper in light of the first half.
The study is extremely specialized and specific so it is very hard to come away from it with any sense of how general it might be, or how its results might apply to any other situation. For instance, the work explicitly ignores the entrainment fluxes of all of these scalars. That omission, in and of itself, makes me weary of applying any of its conclusions to any particular circumstance in the real atmosphere.
I am unable to grasp why it is important to discern which parts of a parameterization (non-local/counter-gradient, diffusivity, dilution) are changing concentrations in a dynamic boundary layer (dzi/dt > 0). Why does that matter? Moreover, a changing boundary layer height is nearly synonymous with an entrainment flux (except in the case of purely divergent or convergent ABL flow with no turbulence at the ABL top.) That is, changes in ABL-H are mechanistically linked to entrainment fluxes (not always a diluting effect, by the way, consider CO2 over remote continental regions for example), so how can it be justified to ignore entrainment fluxes in all of this analysis? (See major concern above). In other words, the artificial isolation of using only variations in h and its impact on the counter-gradient term and diffusivity seems like it is not all that informative if changes in h are always coupled with concomitant changes induced by entrainment.
Because you are using the archetypes of CO2 (scheme 1) and H2O (scheme 2), the former being a top-down diffusion process in remote continental growing seasons, and the latter being typically bottom-up diffusion, then why not use the top-down and bottom-up diffusion schemes in this analysis? I realize that Holtslag & Moeng (1991) discuss eliminating this in some generalized cases, but they do admit, “Overall it is seen that the impact of entrainment flux cannot be neglected.” (See their Equations 25a,b).
Specific Concerns:
p.4, line 87: 500 Wm-2 is an incredibly large surface sensible heat flux. For example, summertime peak H0 in Yuma, AZ is about 250 W/m2.
p.5, line 110: Phi_M is typically reserved for the surface layer similarity dimensionless shear function, and Phi_H is the dimensionless temperature gradient function.
p.6, eq. 2-9: This vertical gradient term (alpha) is not a free variable. It is determined by the bottom-up and/or top-down gradient functions. I realize that in this work alpha represents this initial condition, but why would it matter where the model starts the value if it is different from that throughout most of the day? Moreover, Fig. 1 from Moeng & Wyngaard (1984) shows that alpha is a strong function of elevation (z) in the surface layer.
p.6, line 142: Traditional K-theory in the neutral surface layer yields K(z) is an approximately linear function of z. Even Holtslag & Moeng (1991) show this to be the case. How is a constant K with height relevant?
p.9, line 159: Fig. 3a indicates that K increases quite rapidly in early morning and evening. Maybe it is not clear exactly what times of day are being considered in this discussion.
9, line 164: How can you justify that alpha should be correctly reproduced in time in your model? Surely the gradients are established by the surface and entrainment fluxes (cf. Wyngaard & Brost, 1984), which your model has as a function of time of day. But for simplicity you say that you have neglected entrainment fluxes. The exact timing of these changes then seems somewhat arbitrary as alpha is really only relevant at t=0.
p.9, line 176: The comparison of percentage increase/decrease of mixing ratios is not very generalizable because an arbitrary constant (420 ppm) is being subtracted. If 400 ppm had been used as the offset these percentages would all change, so it is not a very meaningful result that can be generalized by a reader.
p.9, line 178: It would be much more descriptive and easier for a reader to follow if a term like CO2-like surface flux scalar, or H2O-like surface flux scalar, were used instead of the arbitrary terms ‘first’ and ‘second’.
p.9, line 179: I do not see Figure 8 resembling Fig. 2d much at all.
Figure 8: There is very little information in this figure. All four panels look indistinguishable, so I am not sure it is telling the reader anything outside of the fact that the results are insensitive to the arbitrary initial condition of scalar gradient. In any event, some discussion of this feature of the figure warrants mentioning.
p.10, Eq. 3-1: Where is the shear production (~ u*3/z) in this TKE budget equation? Under neutral to stable conditions this equation cannot sustain any turbulence whatsoever. Yet you have u* in your boundary condition for E.
p.10, line 199: Yes, but surely just above these heights the TKE, and therefore K, goes to zero. I see these values as just representing a slight uncertainty in the ABL-H.
Fig. 10: There is no mention of what the origin of the information in this graph is, nor what the units are. I am assuming this is from a model run, but under what meteorological conditions is never mentioned or referred to. In these urban areas, I would suspect that the shear production is, all else being equal, even more important to the sustenance and character of the turbulence due to the “rough” character of urban landscapes.
p.21, line 235: It would be helpful to know what the surface flux diurnal profile looked like for these simulations (“first” or “second” type, or presumably if it is CO and NOx, it is different from either of those), and what are the different amounts for each domain? There is no mention of anything about these model runs so they are very difficult to contextualize.
p.21, line 242: Is there any information to be gleaned from the four different cities? Why does one have smaller or larger r-values? Why show the different cities if there is nothing to observe from it, especially since they are just simulated cities anyway without any knowledge of their relative emissions or geography or meteorology.
p.21, lines 249-251: Why is the mentioned positive correlation not apparent in Fig.11?
Figure 13: These CO diurnal profiles do not look like any that I have seen before. There must be a diurnal pattern to your emissions (rush hours) which normally imprints a bimodal shape on the CO diurnal pattern because the edges of rush hour tend to occur when mixing is weak (early a.m. & early evening.) See for example, Yassin et al., Environ Monit Assess (2018) 190: 372, https://doi.org/10.1007/s10661-018-6737-9.
p.21, line 257: What does this intercity variability tell us? It is one thing to report this in a set of observations, but to present it with model results without having any conjecture as to why does not seem very informative to a general readership.
p.23, line 263: Concentrations depend on ABLH and emission fluxes (assuming the chemistry is slow with respect to this small-scale mixing.) It is therefore imperative that you describe the surface flux pattern for these sensitivity studies.
p.24, line 288: I am not sure what the merit is in discussing NO2 when you do not present any of the data. It is similar to CO in some cases because its dominant source is probably the same: internal combustion. But NO2 exchanges with O3 + NO on a timescale of minutes (faster than a lot of the dynamics under consideration here), so it is not going to be conserved upon mixing, it is going to react to changes in O3 and sunlight in a way that CO will not.
Citation: https://doi.org/10.5194/acp-2023-37-RC2 -
AC2: 'Responses to the comments of reviewer #2', REN SHUZHAN, 01 May 2023
Responses to the comments of reviewer #2
We thank the reviewer for the comments and suggestions. Some suggestions have been adopted in the revised manuscript. To address the reviewer’s major concern, (1) a new paragraph has been added in section 3 in the revised version to introduce the evaluation of the high-resolution GEM-AMCH against observations, and construction of surface emissions of air pollutants for model simulation (lines 226-236); (2) details of the sensitivity test method have been added in the introduction (lines 69-73); (3) more references suggested by the reviewer have been included (lines 34-38); number of figures in section 2 and equations in section 3 has been reduced, (4) “Uncertainties” has been added in the title to reflect motivation of this work. We hope these modifications can improve the presentation of this work and make the results of our work more understandable.
The following are our point-to-point responses to the reviewer’s points (in bold font).
This work presents the results of numerical simulations, which are parameterizations of more complex mechanisms at play in the atmosphere and tries to give a general sense of how turbulent diffusivities and scalar concentrations depend on the height of the atmospheric boundary layer height (ABL-H). This is an important topic, but the results presented are merely those from a model, employed with several severe simplifications, entirely untethered from observed reality, and as such I recommend that it be rejected. Perhaps there is another journal that focuses solely on numerical modeling that might find it appropriate for their readership, but because this work makes no assessment of how it pertains to observations, I cannot see its utility to the ACP readership.
Re: Before submitting the manuscript to ACP, we checked the topics and research activities covered by ACP. It says clearly on the ACP website that “Topics include gases, aerosols, clouds, precipitation, dynamics, radiation and their role in the Earth's climate system (including the biosphere, hydrosphere, and cryosphere). Research activities include laboratory studies, field measurements, remote sensing, modelling and data analysis, and machine learning” Since the contents of this paper include both chemical species and dynamics (turbulent mixing), this paper fits the subject of ACP described on CAP website. Furthermore, this work also fits the research activity of ACP as it uses the atmospheric modelling approach.
We agree with the reviewer that it is important to use observations to evaluate model results. GEM-MACH has been evaluated extensively at different spatial scales (see Ren et al., 2020 for references). Particularly, it has been evaluated in the urban area (Great Toronto Area) with the urbanization scheme (the town energy balance model). As long as the benchmark simulation results are evaluated against observations, the impacts of model parameters described by sensitivity results should be reliable. The evaluation of GEM-MACH has been mentioned in the revised manuscript (lines 229-230). In addition, the capability of the K-profile and TKE schemes in describing the subscale effects have also been evaluated (e.g., Holtslag & Bovilie, 1883, Mailhot & Benoit, 1982).
The reviewer’s other concerns are associated with the method employed in the sensitivity studies with numerical models. Because model parameters are interconnected in a complex way, and their connections cannot be expressed analytically, sensitivity studies have been widely applied to examine the influences of model variables on numerical simulation results (one example is the sensitivity study of the climate change associated with the double CO2 in the atmosphere). By modifying the value of a specified parameter and keeping other parameters fixed, the simulated sensitivity results can well describe the impacts and help better understand the important roles of model parameters playing in model simulations. The discussion of the sensitive test method has been added in the revised version (lines 69-73).
Major Concerns:
There are many other major concerns I have with this work that also indicate it is not ready for publication. First, there is an incomplete mention of the literature of scientists who have presented these kinds of relationships between ABL-H and scalar concentrations (most notably Schäfer et al., Meteorologische Zeitschrift, Vol. 15, No. 6, 647-658, December 2006; but also including, Yuval et al., Atmospheric Research 231 (2020); Wagner & Schäfer, Urban Climate 22 (2017) 64–79; Rigby & Toumi, R 42 (2008) 4932–4947.)
Re: We thank the reviewer for drawing our attention to several research papers on the relationship between air pollutants and ABL height. Although these studies are based on observations, they can be applied for comparing model results. The first three papers have been referred to in the revised version (lines 34-38).
In my opinion there are too many figures (20 in total) and too little mechanistic understanding of the results presented. Gradient initial conditions (alpha) are compared, as are boundary layer maximum heights, as are diurnal flux patterns, as are different cities, as are ABL-H detection algorithms.
Re: The number of figures has been reduced in the revised version by combining Fig 8 and Fig. 9 together. The impacts of the ABL height on concentration through \gamma are not sensitive to \alpha in the two figures and only the differences with \alpha=0 are shown. Please also see response to p.6, eq. 2-9.
We don’t quite understand the statement about \alpha. This parameter is employed to show how the impact of the change of diffusivity (due to the change of the ABL height) is modulated by the vertical gradient of tracers numerically. The analytical form of the modulation is described by Eq. 2.10.
The presentation seems unfiltered in terms of which variables are being altered over which range and for what reason and to what conclusion? Originally there are two flux patterns presented, one CO2-like and one H2O-like. But then later results are presented for CO and NOx (not shown) which have very different surface flux diurnal behaviors, making it very difficult to consider the model results in the second part of the paper in light of the first half.
Re: Biogenic CO2 emission and CO2 emission from industrial activities are used in section 2 for sensitivity studies with the K-profile scheme. The reason that the impact on CO (and other pollutants) is discussed in section 3 is that currently we don’t have long term simulation of CO2 with GEM-MACH. Like CO2, CO can be treated as a tracer with a long lifetime and has a similar surface emission to the emission of CO2 from industrial activities.
The details of the surface emissions used in GEM-MACH simulations have been given in the revised version (lines 231-236, please also see response to p.21, line 235). The impact on NOx is mentioned just to show the chemical reactions that NOx is involved during the daytime would not change the impact of the ABL height.
Two kinds of ABL parameterization schemes (K-profile and TKE) are used to show the impacts of the ABL height. They are different schemes with different involvement of the ABL height. While their impacts on the vertical diffusivity are similar, their impacts on concentration are quite different. The K-profile scheme is implemented in a 1-D diffusion model. Because it is a simple model the modulation on the ABL height impact can be easily examined by changing vertical gradient of tracer through \alpha and by choosing different boundary conditions. Although it is difficult to do the same thing in the 3-D GEM-MACH with realistic surface emissions and 3-D meteorological and chemical fields, the modulation effects can still be applied to the 3-D case.
The study is extremely specialized and specific so it is very hard to come away from it with any sense of how general it might be, or how its results might apply to any other situation. For instance, the work explicitly ignores the entrainment fluxes of all of these scalars. That omission, in and of itself, makes me weary of applying any of its conclusions to any particular circumstance in the real atmosphere.
Re: We don’t agree with the reviewer’s statement about the usefulness of the results of this work. Because the ABL height is a parameter measuring the extent of turbulent mixing and is involved in ABL parameterization schemes, the impact of the uncertainty in the ABL height is a general issue for air quality model simulations when the ABL height is involved in the parametrization scheme. One general conclusion from the sensitivity studies with the K-profile and TKE schemes is that over-estimated/under-estimated ABL can lead to over-estimated/under-estimated vertical diffusivity.
As we pointed out in the introduction, there are large uncertainties in the ABL height. When the ABL height is involved in the parameterization schemes, the impacts of the uncertainties should be examined in order to understand the source of uncertainties of the air quality forecast. Furthermore, the results can also help understand the different model simulation results with different definition of the ABL height. Our results show that the uncertainty in the ABL height can have big impacts when the K-profile scheme is used. To our knowledge this work is the first one to examine the impacts in detail. Although only two schemes are employed in this work the results can certainly benefit not only modelers but also the air quality research community. Understanding the impacts can help identify the sources of the uncertainties in simulation results and improve forecast results.
It is incorrect to say that the entrainment fluxes are ignored in GEM-MACH simulation. These fluxes (if they exist) are included in the sensitivity studies with 3-D GEM-AMACH in section 3. For most chemical species the surface fluxes have major contributions to the concentrations. For ozone there might be entrainment fluxes when the tropopause penetrates deep into low altitude areas due to the breakdown of potential vorticity. NOx produced by lightning storms in the troposphere can also penetrate into the ABL. These are just associated with special events.
I am unable to grasp why it is important to discern which parts of a parameterization (non-local/counter-gradient, diffusivity, dilution) are changing concentrations in a dynamic boundary layer (dzi/dt > 0). Why does that matter? Moreover, a changing boundary layer height is nearly synonymous with an entrainment flux (except in the case of purely divergent or convergent ABL flow with no turbulence at the ABL top.) That is, changes in ABL-H are mechanistically linked to entrainment fluxes (not always a diluting effect, by the way, consider CO2 over remote continental regions for example), so how can it be justified to ignore entrainment fluxes in all of this analysis? (See major concern above). In other words, the artificial isolation of using only variations in h and its impact on the counter-gradient term and diffusivity seems like it is not all that informative if changes in h are always coupled with concomitant changes induced by entrainment.
Re: The ABL height is involved in different components in the parameterization scheme in different ways. Thus, its uncertainty can have different impacts on the concentration of tracers. Understanding these different impacts can help identify the source of uncertainties in air quality forecasts and find ways to improve forecasts.
The effect of the entrainment flux is contained in the sensitivity results with the GEM-MACH because only the ABL height is changed. Although in some special events such as the penetration of tropopause into the low altitude areas and big storms which produce top-down ozone and NOx fluxes, respectively, the surface emissions have the major contributions to the concentrations of air pollutants and CO2. The work of Ren (2019, Journal of the Atmospheric Sciences) shows that the magnitude of entrainment flux decreases rapidly as height decreases. Because we are interested in the impacts in the lower ABL, the entrainment flux is not included in section 2 for simplicity. This has been mentioned in the revised version (line 102)
Because you are using the archetypes of CO2 (scheme 1) and H2O (scheme 2), the former being a top-down diffusion process in remote continental growing seasons, and the latter being typically bottom-up diffusion, then why not use the top-down and bottom-up diffusion schemes in this analysis? I realize that Holtslag & Moeng (1991) discuss eliminating this in some generalized cases, but they do admit, “Overall it is seen that the impact of entrainment flux cannot be neglected.” (See their Equations 25a,b).
Re: During the grown season the concentration of CO2 can increase with height. In section 2 the same bottom-up type of initial condition is used just to illustrate how the impacts of uncertainty of the ABL height can be modulated by the vertical gradient of CO2 with different surface emissions. Because the vertical gradient is weakened during the daytime due to the negative surface emission under the first kind of surface emission scheme, the top-down diffusion effect is reflected in the model integration.
The work of Ren (2019 Journal of the Atmospheric Sciences) shows that the magnitude of entrainment flux decreases rapidly as height decreases. The concentration in the lower part of the ABL is determined mainly by the surface emission and the vertical diffusivity. It is not included in section 2 for simplicity (line 102).
Specific Concerns:
p.4, line 87: 500 Wm-2 is an incredibly large surface sensible heat flux. For example, summertime peak H0 in Yuma, AZ is about 250 W/m2.
Re: The magnitude of sensible heat flux depends on many factors including the weather conditions (such clear or cloudy sky), locations (such as urban, rural), seasons etc. In the urban centers the monthly mean sensible heat flux in June can be 400W/m^2 around noon (Fig 8 in Masson 2000, Boundary-Layer Meteorology). The urban sensible heat flux contains contributions from walls, roofs, roads and traffic.
p.5, line 110: Phi_M is typically reserved for the surface layer similarity dimensionless shear function, and Phi_H is the dimensionless temperature gradient function.
Re: \phi_M has been replaced \phi_h in the revised version.
p.6, eq. 2-9: This vertical gradient term (alpha) is not a free variable. It is determined by the bottom-up and/or top-down gradient functions. I realize that in this work alpha represents this initial condition, but why would it matter where the model starts the value if it is different from that throughout most of the day? Moreover, Fig. 1 from Moeng & Wyngaard (1984) shows that alpha is a strong function of elevation (z) in the surface layer.
Re: In the real atmosphere \alpha is not a free variable. It is determined by the vertical distribution of chemical species. In the 1-D model, a specified \alpha is used to represent the background (initial) vertical gradient of tracer. The value of \alpha affects the magnitude and sign of the vertical gradient after the initial time and thus modulates the impact of the ABL height according to Eq. 2.10. Please also see response to the last major concern.
p.6, line 142: Traditional K-theory in the neutral surface layer yields K(z) is an approximately linear function of z. Even Holtslag & Moeng (1991) show this to be the case. How is a constant K with height relevant?
Re: K in Eq. 2.10 can be any function of z, and K used for sensitivity tests is also a function of z defined by Eq. 2.6. Constant K assumption was used in Ren & Stroud (2020) to examine the sign of Green’s function.
p.9, line 159: Fig. 3a indicates that K increases quite rapidly in early morning and evening. Maybe it is not clear exactly what times of day are being considered in this discussion.
Re: The rapid increase of the vertical diffusivity occurs at sunrise (6:00AM in this work). This was mentioned in line 131.
p.9, line 164: How can you justify that alpha should be correctly reproduced in time in your model? Surely the gradients are established by the surface and entrainment fluxes (cf. Wyngaard & Brost, 1984), which your model has as a function of time of day. But for simplicity you say that you have neglected entrainment fluxes. The exact timing of these changes then seems somewhat arbitrary as alpha is really only relevant at t=0.
Re: \alpha is used to define the vertical gradient of tracer at initial time. Because The value of \alpha affects the magnitude and sign of the vertical gradient after the initial time and thus modulates the impact of the ABL height. The sensitivity of the impact on \alpha illustrates the modulation described by Eq. 2.10. These changes depend on \alpha. Once \alpha is specified at initial time they are not arbitrary.
p.9, line 176: The comparison of percentage increase/decrease of mixing ratios is not very generalizable because an arbitrary constant (420 ppm) is being subtracted. If 400 ppm had been used as the offset these percentages would all change, so it is not a very meaningful result that can be generalized by a reader.
Re: The comparison in percentage has been changed to direct comparison in the revised version.
p.9, line 178: It would be much more descriptive and easier for a reader to follow if a term like CO2-like surface flux scalar, or H2O-like surface flux scalar, were used instead of the arbitrary terms ‘first’ and ‘second’.
Re: Both are CO2 emissions. One is biogenic, another is the emission from industrial activities. “biogenic” and “industrial” emissions are used in the revised version.
p.9, line 179: I do not see Figure 8 resembling Fig. 2d much at all.
Re: The similarity between the two figures means that the ABL height has similar impacts. This sentence has been removed in the revised version to avoid confusion.
Figure 8: There is very little information in this figure. All four panels look indistinguishable, so I am not sure it is telling the reader anything outside of the fact that the results are insensitive to the arbitrary initial condition of scalar gradient. In any event, some discussion of this feature of the figure warrants mentioning.
Re: Figure 8 and Fig. 9 are combined in the revised version. The impacts of the ABL height on concentration through \gamma are not sensitive to \alpha in the two figures and only the differences with \alpha=0 are shown. The insensitivity to \alpha has been mentioned.
p.10, Eq. 3-1: Where is the shear production (~ u*3/z) in this TKE budget equation? Under neutral to stable conditions this equation cannot sustain any turbulence whatsoever. Yet you have u* in your boundary condition for E.
Re: The shear production term is defined by S in (A.2) in Appendix A. In the numerical simulation with GEM-MACH the friction velocity is about 0.1(m/s) to 0.6 (m/s) in the urban center of the Great Toronto Area from which one can calculate u*^3/z at different heights.
The boundary condition under the stable condition is proposed by Mailhot & Benoit (1982).
p.10, line 199: Yes, but surely just above these heights the TKE, and therefore K, goes to zero. I see these values as just representing a slight uncertainty in the ABL-H.
Re: The definition of the ABL height is not unique as it can be based on the Richardson number, the gradient of potential temperature, the gradient of tracers etc. Although the top of the ABL is supposed to be a layer separating free atmosphere and the ABL characterized by turbulent mixing, the ABL height doesn’t impose constraint on the diffusion coefficient at the ABL top in the TKE scheme. This is because the ABL height is not present in the equation for calculating the diffusion coefficient.
Fig. 10: There is no mention of what the origin of the information in this graph is, nor what the units are. I am assuming this is from a model run, but under what meteorological conditions is never mentioned or referred to. In these urban areas, I would suspect that the shear production is, all else being equal, even more important to the sustenance and character of the turbulence due to the “rough” character of urban landscapes.
Re: The curves in the figure are the diurnal variation of diffusion coefficients at the top of the ABL defined based on the Richardson number and vertical potential temperature profile. They are monthly mean (July) GEM-MACH model simulation results. This has been mentioned in line 209 in the revised version. These figures are used to show that different definitions of the ABL height can lead to different vertical diffusivity near the top of the ABL in the TKE scheme. It also shows the different K in the K-profile and TKE schemes at the ABL top. The description of this figure can be found in lines 196-204.
Due to the existence of high buildings urban areas have larger roughness length which tends to produce larger turbulence in urban areas than in rural areas. However, the urban effect associated with the heat fluxes from roads, roofs, walls, and cars is more important in terms of generating turbulence. Please see Ren et al. (2020) for more details.
p.21, line 235: It would be helpful to know what the surface flux diurnal profile looked like for these simulations (“first” or “second” type, or presumably if it is CO and NOx, it is different from either of those), and what are the different amounts for each domain? There is no mention of anything about these model runs so they are very difficult to contextualize.
Re: A short paragraph has been added in section 3.1 in the revised version to introduce more details about the model configuration and surface emissions used in GEM-AMCH (lines 224-234). Over the Great Toronto Area the CO emission in summer increases rapidly around 7:00AM from 40 (g/s) to 134 (g/s) and continues to increase slowly to 171 (g/s) at 3:00PM before it starts to decrease. The diurnal variation of CO emission is similar to that of industrial CO2 emission in section 2.
p.21, line 242: Is there any information to be gleaned from the four different cities? Why does one have smaller or larger r-values? Why show the different cities if there is nothing to observe from it, especially since they are just simulated cities anyway without any knowledge of their relative emissions or geography or meteorology.
Re: Urban centers are big sources of air pollutants. In Ren et al., (2020) four northern American urban centers are chosen to investigate the common features of the impacts of urban effects on meteorological and chemical fields. Because the surface heat fluxes are different in the four urban areas due to different geometric parameters such as building fraction, building height, and urban canyon aspect ratio (Ren et al., 2020, 2022 (Urban Climate)), the impacts of the ABL-H are different. They are used in this work to examine the correlation between the ABL height and concentration of chemical species and the impacts of the uncertainties in ABL height. Please also see response to p.21, line 257.
p.21, lines 249-251: Why is the mentioned positive correlation not apparent in Fig.11?
Re: The positive correlations appear between 7:00AM to 8:00AM shown in Fig. 13 (not Fig. 11)
Figure 13: These CO diurnal profiles do not look like any that I have seen before. There must be a diurnal pattern to your emissions (rush hours) which normally imprints a bimodal shape on the CO diurnal pattern because the edges of rush hour tend to occur when mixing is weak (early a.m. & early evening.) See for example, Yassin et al., Environ Monit Assess (2018) 190: 372, https://doi.org/10.1007/s10661-018-6737-9.
Re: Yes, the CO and NOx emissions from traffic over GTA have a bimodal shape in summer (July). However, this shape is not clear in other urban areas due to different traffic patterns. For example, emissions are larger in the second rush hour period in Chicago and New York. This can be seen clearly in Fig. 2 in Ren et al. (2020 atmosphere). This figure shows the monthly mean diurnal variation of the heat fluxes from cars based on the emission of both CO and NOx.
Due to the contribution from other sources the bimodal shape of the total CO and NOx emissions are not obvious. Over the Great Toronto Area the CO emission in summer increases rapidly around 7:00AM from 40 (g/s) to 134 (g/s) and continues to increase slowly to 171 (g/s) at 3:00PM before it starts to decrease.
p.21, line 257: What does this intercity variability tell us? It is one thing to report this in a set of observations, but to present it with model results without having any conjecture as to why does not seem very informative to a general readership.
Re: Urban areas are a big source of air pollutants. Each urban area has different geometric parameters such as building fraction, building height, and urban canyon aspect ratio (Ren et al., 2020, 2022 (Urban Climate)). The work of Ren et al. (2020) shows that the surface heat fluxes in the four urban centers are different due to different geometric parameters. Because the surface flux can modulate the impact of the ABL height (in convective velocity), the correlation between concentration and the ABL height reflects the modulation effect. This has been mentioned in the revised version (lines 239-242)
p.23, line 263: Concentrations depend on ABLH and emission fluxes (assuming the chemistry is slow with respect to this small-scale mixing.) It is therefore imperative that you describe the surface flux pattern for these sensitivity studies.
Re: More details of the surface emissions are given in the revised version (lines 234-235). Please also see response to p.21, line 235
p.24, line 288: I am not sure what the merit is in discussing NO2 when you do not present any of the data. It is similar to CO in some cases because its dominant source is probably the same: internal combustion. But NO2 exchanges with O3 + NO on a timescale of minutes (faster than a lot of the dynamics under consideration here), so it is not going to be conserved upon mixing, it is going to react to changes in O3 and sunlight in a way that CO will not.
Re: NOx is a major air pollutant and is involved in photochemical reactions during the daytime. Because the uncertainty of the ABL height affects all the chemical species involved in the chemical reactions, the additional impact from other species on NOx needs to be examined by comparing the impact on CO. This has been mentioned in the revised version (lines 308-309).
Citation: https://doi.org/10.5194/acp-2023-37-AC2
-
AC2: 'Responses to the comments of reviewer #2', REN SHUZHAN, 01 May 2023
-
AC1: 'Responses to the comments of reviewer #1', REN SHUZHAN, 01 May 2023
Responses to the comments of reviewer #1
The reviewer’s comments and suggestions are appreciated. Some suggestions have been adopted in the revised manuscript. The following are our point-to-point responses to the reviewer’s points (in bold font).
The manuscript describes some aspects of the sensitivity of simulated concentrations to the estimate of ABL height in certain boundary layer schemes. The title is misleading.
Re: Following the suggestion the title of the manuscript has been modified to reflect the contents of this work. Please see the response to general comment 1 below.
The presentation is unclear and confusing. There may be some useful material here for certain specific users, but I do not see a path to acceptance for anything like the current form of this material.
Re: We don’t agree with the reviewer’s statement about the usefulness of the results of this work. Because the ABL height is a parameter measuring the extent of turbulent mixing and it is involved in ABL parameterization schemes, the impact of the uncertainty in the ABL height is a general issue for air quality model simulations when the ABL height is involved in the parametrization scheme.
As we pointed out in the introduction, there are large uncertainties in the ABL height. When the ABL height is involved in the parameterization schemes, the impacts of the uncertainties should be examined in order to understand the source of uncertainties of the air quality forecast. Furthermore, the results can also help understand the different model simulation results with different definition of the ABL height. Our results show that the uncertainty in the ABL height can have big impacts when the K-profile scheme is used. To our knowledge this work is the first one to examine the impacts in detail. Although only two schemes are employed in this work the results can certainly benefit not only modelers but also the air quality research community. Understanding the impacts can help identify the sources of the uncertainties in simulation results and improve forecast results.
The manuscript has been revised by including the following modifications: (1) adding “uncertainties in the title reflect the motivation of this work, (2) adding a new paragraph in section 3 to introduce the evaluation of the high-resolution GEM-AMCH against observations, and construction of surface emissions of air pollutants for model simulation (lines 225-235), (3) adding the details of the sensitivity test method in the introduction (lines 69-73), (4) adding more references about the studies on the impacts of the ABL height on chemical species (lines 34-38), (5) reducing number of figures in section 2 and equations in section 3. We hope that these changes can clarify confusions in the previous version and help readers better understand the results of this work.
General comments:
- The title is misleading. The material is really about the errors incurred when the BL height is estimated incorrectly in certain kinds of schemes. Those errors are more or less serious depending on the scheme.
Re: The reason that the title was chosen for this paper is that our work discusses not only the impact of uncertainties in the ABL height on the simulation of chemical species but also the impact of different definitions of the ABL height used in simulations. Following the suggestion of the reviewer, “uncertainties” has been added in the title in the revised version to make the motivation of this work clearer.
- The authors do not seem to have a broad perspective on what is actually in use in models other than those of their own agency. Two types of mixing schemes are examined. Many of the errors noted are not significant problems in most state-of-the-art schemes. Even within this preprint, the errors in the TKE scheme are much less significant than in the K-profile scheme. This is an argument for not using the latter type of scheme, but that is never stated clearly.
Re: We don’t quite understand the first sentence. Yes, there are many kinds of parameterization schemes, and we cannot examine the impacts with all the schemes. TKE and K-profile schemes are two schemes with the involvement of the ABL height and are widely employed in numerical models. The reason that they are employed in this work is not because the ABL height is involved explicitly in the schemes (under unstable condition) but also because it is involved in completely different ways. While the ABL height affects diffusivity in the lower vertical boundary condition through the convective velocity in the TKE scheme, it affects diffusivity within the ABL not only through the convective velocity but also affects the calculation of diffusion coefficient directly in the K-profile scheme. Therefore, examining the impacts of the two schemes can give us a comprehensive understanding of the impacts of the ABL height.
We also don’t understand the statement “Many of the errors noted are not significant problems in most state-of-the-art schemes”. Whether the impacts are significant or not should be based on careful examination. To the best of our knowledge our work is the first one to examine the impacts in detail. We would appreciate it if the reviewer could point out references of other similar studies based on the state-of-the-art schemes.
Both analytical and numerical results of this work show that the ABL height has very different impacts on tracers. However, these results don’t suggest which scheme should be used. They just help modelers to identify the source of uncertainties in air quality forecast and improve the performance of the forecast.
- The conceptual picture of the boundary layer presented here is oversimplified. The top of the ABL is not a material surface. It does not generally "serve as a boundary" (abstract). It is somewhat more acceptable to say that it is "a parameter describing the vertical extent of turbulent mixing" (Introduction), but then in the very next sentence the term "capping lid" is used. There are special times and places when the ABL top is well-defined and significantly impedes vertical mixing, and it happens that those are the conditions of most interest to the understanding of photochemical ozone pollution. A simple example of the more general situation: The "top" of a stable boundary layer is the point at which a more or less linear decrease of turbulence intensity reaches some selected threshold (becomes small enough). There is extensive recent literature, including review articles, describing the real boundary layer.
Re: We agree with the reviewer that the top of the ABL is not a material surface. It is not an observed variable but a derived one based on different variables such as the Richardson number, vertical gradient of potential, vertical gradient of chemical tracers etc. Therefore, there is no unique boundary layer top as its definition is not unique.
The ABL is characterized by turbulent mixing and the troposphere is turbulence free. “Boundary” and “capping lid” are used to describe the extent to which the vertical mixing and its effects on chemical species can reach vertically. To avoid confusion these two terms have been removed in the revised version.
- Presentation: Many aspects of the presentation are unclear. There are many equations, derived under different sets of assumptions. There is little overall concept or structure to guide the reader.
Re: In the revised version, paragraphs have been added to explain the details of the sensitivity test method and introduce more details of configuration of GEM-MACH and construction of surface emissions of air pollutants for GEM-MACH. Hope these modifications can make the presentation clearer.
Results of sensitivity studies are produced by numerical models. Due to complex processes and interconnections of variables, analytical methods are often employed to interpret the numerical results. Analytical results can provide independent evaluations of numerical results.
Equations in section 2 are basic equations to define variables associated with the ABL height (convective velocity, counter gradient term, diffusion coefficient etc.) in the parameterization schemes. They are simple and straightforward and can be found easily in the textbooks of boundary meteorology and published papers. Reference is given in this section (Holtslag and Moeng 1991). The impacts estimated based on the equations agree with numerical results suggesting the results can indeed provide independent evaluations of numerical results.
Readers may not be familiar with the equations in section 3. In the revised version only three equations (Eqs. 3.2, 3.3 and 3.7) are kept and the rest equations have been moved to appendix.
- The key claim of the paper is buried at the end of section 2: "Under the unstable condition, the ABL-H affects tracer's concentration by changing the vertical diffusivity. This impact depends on both surface flux and the vertical gradient of tracers." This is generally true but needs some qualification. Within a broad range, the exact value of K does not change tracer profiles significantly. Enough mixing is enough, unless there is some relatively strong competing process, for example large surface flux or large entrainment flux. Similarly, the value of K does not necessarily change the BL height if the stability gradient at the top is strong enough. In other words, the impact of a change in K is non-linear and depends on the stability profile.
Re: This claim is not only mentioned at the end of section 2 but also mentioned in the abstract (lines 13-24) and emphasized in summary and discussion (lines 312-314). To emphasize the claim a new sentence has been added to the abstract in the revised version. The dependence of the impact of K on the vertical gradient is discussed quantitively in section 2 based on different values of \alpha which is a measure of the vertical gradient of tracer.
In this work only the impact of the ABL height on K is discussed. If the stability gradient is strong at the top, turbulent mixing would be suppressed. The impact of K on the ABL height would be weak.
Specific comments:
- line 203: This is an instance of the conceptual problem noted in General item 3 above. The diffusivity need not be zero at the BL top; that is not a generally useful definition. It may be relatively small depending on the situation.
Re: Specific values are often used in the ABL height definitions. When the Richardson number is used the critical value is 0.25. If the vertical gradient of potential temperature is used, the critical value is 0. One may not use these values in practice, but these theoretical values provide guidance for the definition of the ABL height. For the same reason zero value of diffusion coefficient is mentioned. Another reason to mention this is for comparing the value of K at the top of the ABL in the K-profile and TKE schemes. It is constrained to be zero in the former scheme and no similar constraint in the latter scheme.
- line 250: The heat flux does not jump suddenly at sunrise in reality. It is a smooth function of time. The emissions from traffic, however, are not very smooth. What is happening here is that the morning rush hour happens about the same time as the morning transition of the BL. Their relative timing depends on the season, among other things.
Re: The observed diurnal variation of surface heat flux shown in Fig. 1.9 in Oke’s book (boundary layer climate) suggests the heat flux can jump rapidly at sunrise. Although the variation of sensible heat flux is not as rapid as latent flux and flux from soil, the sum of the three components has a big jump at sunrise. Corresponding to the rapid change of heat flux, the vertical diffusivity also experiences a rapid change. Of course, the jump of heat flux depends on weather, season and other conditions as suggested by the reviewer.
The concentration of tracers near the ground is affected by both vertical mixing and surface emissions. The quick change of diffusivity at sunrise can lead to a rapid drop of concentration even when the surface emission is strong.
- line 299: This is an example of General item 5 above. Many other factors contribute to the relationship between diffusivity and concentration. That is, after all, why we have models.
Re: We agree with the reviewer’s point.
Citation: https://doi.org/10.5194/acp-2023-37-AC1
-
RC2: 'RC2', Anonymous Referee #2, 22 Mar 2023
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 925 | 198 | 49 | 1,172 | 43 | 60 |
- HTML: 925
- PDF: 198
- XML: 49
- Total: 1,172
- BibTeX: 43
- EndNote: 60
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1