the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
05 Jan 2021
05 Jan 2021
Vehicle induced turbulence and atmospheric pollution
- Air Quality Modelling and Integration Section, Air Quality Research Division, Atmospheric Science and Technology Directorate, Environment and Climate Change Canada, 4905 Dufferin Street, Toronto, Ontario, M3H 5T4, Canada
- Air Quality Modelling and Integration Section, Air Quality Research Division, Atmospheric Science and Technology Directorate, Environment and Climate Change Canada, 4905 Dufferin Street, Toronto, Ontario, M3H 5T4, Canada
Abstract. Theoretical models of the Earth's atmosphere adhere to an underlying concept of flow driven by radiative transfer and the nature of the surface over which the flow is taking place: heat from the sun and/or anthropogenic sources are the sole sources of energy driving atmospheric constituent transport. However, another source of energy is prevalent in the human environment at the very local scale – the transfer of kinetic energy from moving vehicles to the atmosphere. We show that this source of energy, due to being co-located with combustion emissions, can influence their vertical distribution to the extent of having a significant influence on lower troposphere pollutant concentrations throughout North America. The effect of vehicle-induced turbulence on freshly emitted chemicals remains notable even when taking into account more complex urban radiative transfer-driven turbulence theories at high resolution. We have designed a parameterization to account for the at-source vertical transport of freshly emitted pollutants from mobile emissions resulting from vehicle-induced turbulence, in analogy to sub-grid-scale parameterizations for plume rise emissions from large stacks. This parameterization allows vehicle-induced turbulence to be represented at the scales inherent 3D chemical transport models, allowing its impact over large regions to be represented, without the need for the computational resources and much higher resolution of large eddy simulation models. Including this sub-grid-scale parameterization for the vertical transport of emitted pollutants due to vehicle-induced turbulence into a 3D chemical transport model of the atmosphere reduces pre-existing North American nitrogen dioxide biases by a factor of eight, and improves most model performance scores for nitrogen dioxide, particulate matter and ozone (for example, reductions in root mean square errors of 20, 9 and 0.5 percent, respectively).
- Preprint
(3351 KB) -
Supplement
(5740 KB) - BibTeX
- EndNote
Paul A. Makar et al.
Status: final response (author comments only)
-
RC1: 'Comment on acp-2020-1243', Anonymous Referee #2, 16 Jan 2021
This study investigates the impact of kinetic energy from moving vehicles on the vertical distribution of combustion emissions and uses a VIT parameterization to account for the vertical transport of fresh mobile emissions in a 3D chemical transport model. This is an important topic as a better representation of vehicle emissions and mixing in atmospheric chemical transport models is crucial for an improved understanding of air pollutants. The manuscript is generally well written and aims to provide a way to improve mixing of mobile emissions in 3D regional modeling. Reasonable assumptions are made to parameterize vehicles with different sizes and running with distinct road conditions. However, the evaluation of the VIT parameterization is rather weak and there are a few major flaws in the manuscript.
- In the introduction section, it states that the LES models are typically employed at centimeter or meter level resolutions, while the mixing lengths associated with VIT are on the order of tens of meters. This incomplete review of LES studies is misleading as it indicates the vertical influence of VIT is between the scales of a LES and a 3D regional model. However, the following studies all applied LES coupled with chemistry at a horizontal resolution of tens of meters, and there are more similar LES studies not listed here.
Vinuesa and Vil.-Guerau de Arellano (2005) Atmos. Environ., 39(3), 445–461
Ouwersloot et al. (2011) Atmos. Chem. Phys., 11(20), 10681–10704
Li et al. (2016) J. Geophys. Res., 121(13), 8083-8105
Kim et al. (2016) Geophys. Res. Lett., 43(14), 7701–7708
As the VIT problem is actually on a LES scale and a LES model with chemistry has already taken into account turbulent mixing in the boundary layer, it might be more convincing to illustrate the impact of VIT on the vertical mixing of vehicle emissions if a LES model is employed.
- On line 635, it states that “An examination of all of the other possible sources of error in air-quality models is beyond the scope of this work.” This is understandable. But 3D regional models typically have difficulties representing turbulence and vertical mixing, which cause a large portion of their model-observation discrepancies. Without considering errors related to boundary layer turbulence, it is hard to evaluate the VIT parameterization developed in this study. The manuscript states that “We also emphasize that the work does not identify a deficiency in existing meteorological boundary layer turbulence models.” Does it mean the 3D model used in this study represents turbulence very well? Please clarify whether it refers to the 3D model used in this study and how “deficiency” is evaluated.
- Although the manuscript states that “The use of the VIT parameterization has been demonstrated to result in decreases in air-quality model error,” this is not convincing as the changes in the metrics used to evaluate model performance are inconsistent and the differences in these metrics between the VIT simulation and No VIT simulation are quite small. It is necessary to show whether adding VIT actually leads to statistically significant differences. Statistical significance can be calculated based on the differences (VIT simulation – No VIT simulation) in daily averaged NO2, PM2.5 and O3 at each site. Alternatively, estimates of vehicle km travelled can be used as a criterion to select sites, then significance could be calculated based on the selected sites with similar traffic conditions and background meteorological conditions.
- To evaluate the VIT parameterization, using observations from surface monitoring sites only is not sufficient. Due to the limitations and uncertainties acknowledged in this study, it is actually better to develop and test the VIT parameterization based on a small domain, maybe city size, which has relatively simply traffic and meteorological conditions as well as observational vertical profiles of chemical species for evaluation. The manuscript shows vertical cross-sections in Figure 10. Without the observed vertical distributions of NO2, PM2.5 and O3, it is hard to determine whether using the VIT parameterization leads to an improvement.
- On line 108, the manuscript states “Here we make use of both the observational and LES modelling studies to devise a parameterization for VIT.” This is the last place in the manuscript that LES is referred to, so it is rather confusing how LES modeling studies are used in this study. As discussed above, please acknowledge other LES studies, and also careful elaborate how “LES modeling studies” are used here. If not used, please also clarify.
-
RC2: 'Comment on acp-2020-1243', Anonymous Referee #1, 02 Feb 2021
This study develops/proposes a parameterization for an additional source of atmospheric dispersion due to moving vehicles on roads (vehicle-induced turbulence, VIT) for use in 3-D chemical transport models. The topic is relevant as mobile sources are often dominant contributors to air pollution in many urban areas while grid scales typically employed by 3-D CTMs are not sufficiently fine to adequately represent those roadway sources. The manuscript is well organized, and the methodology and study outcomes are effectively presented.
What's missing in the current manuscript is a proper evaluation of the proposed VIT scheme. The authors evaluated performance of a 3-D CTM with and without the VIT scheme against observations and showed that the model performance was generally better with the VIT scheme. As the authors also noted, however, model performance of a 3-D CTM is affected by a number of factors, and good model performance doesn't necessarily mean that the model is right for the right reasons. For example, improved model performance could be resulted from model biases due to over-estimated vehicle emissions being reduced by increased mixing by the VIT scheme. While the 3-D CTM simulations serve well as a sensitivity analysis (it's clearly shown that the proposed VIT scheme implemented in a 3-D CTM has significant impacts on the model results), a better evaluation of the VIT scheme may be to directly compare the scheme in a simplified version of the CTM (e.g., with a single horizontal grid cell with multiple vertical layers) with a finer-resolution LES or CFD model, using a small test case with a more controlled setup (e.g., a hypothetical roadway with a predefined vehicle configuration). At least, more discussion on this should be added.
Minor technical issues are listed below.
Line 64: "a vehicle11" ?
Line 282: "added via the “F” terms in (6)" -> "added via the “F” terms in (8)"
Line 302: "All six panels also show a trend of ∂K⁄∂z becoming more negative" - revise this sentence; many of the panels actually show positive ∂K⁄∂z.
Line 358: "with different values for the input coefficients of thermal turbulent transfer coefficient (K)" -> "with different values for the input coefficients of thermal turbulent transfer coefficient (K) and for the lower boundary conditions (E)"
Line 479: "metrics used to here (see Methods)" -> "metrics used here"
Line 497: "Figure 7(a,b)" -> "Figure 7(a,c)"; "Figure 7(c)" -> "Figure 7(e)"
Line 587: "S11" -> "S10"
Line 830: Figure 1 caption says "the length scale of turbulence immediately behind the leading vehicle, a large transport truck is only 3m, while the length scale immediately behind the trailing vehicle in the ensemble (an identical transport truck) is 12.73m", but Table 1 shows that the mixing length for an isolated lead diesel cargo truck is 5.13m and that for the 2nd diesel cargo truck in an ensemble is 14.64m. Explain the discrepancies.
Line 836: In Figure 2 caption, "at low (a,c) and high (c,d) resolution" -> "at low (b,d) and high (a,c) resolution"
Line 845: Figure 4(b) caption says "equation (8)", but the legend says "eqn (6)".
Paul A. Makar et al.
Paul A. Makar et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 189 | 80 | 5 | 274 | 18 | 2 | 5 |
- HTML: 189
- PDF: 80
- XML: 5
- Total: 274
- Supplement: 18
- BibTeX: 2
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1