the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Passive Tracer Modelling at Super-Resolution with WRF-ARW to Assess Mass-Balance Schemes
Yongsheng Chen
Abstract. Accurate "super-resolution'' (Δx < 250 m) atmospheric modelling is useful for several different sectors (e.g., renewable energy, natural disaster prediction), and essential for numerous applications such as downscaling of weather and climate information to finer resolutions. It can also be used to interpret environmental observations during top-down retrieval campaigns by providing complementary data that closely correspond to real-world atmospheric pollution transport and dispersion conditions. In top-down retrievals (e.g., aircraft-based), errors in estimates can arise from assumptions about atmospheric dispersion conditions, uncertainties in measurements, and data processing. As discussed in this work and in our companion paper (Fathi and Gordon, 2022), super-resolution numerical model simulations can be utilized to investigate these sources of uncertainty and optimize the retrievals. In order to conduct a thorough model-based study of the atmospheric dynamical processes that can affect top-down retrievals, model simulations at super-resolutions on the scale of measurement frequency are required: sufficient to resolve the dynamical and turbulent processes at the scale at which measurements are conducted. Here, in the context of our modelling case studies with WRF, we demonstrate a series of best practices for improved (realistic) modelling of atmospheric pollutant dispersion at super-resolutions. These include careful considerations for grid quality over complex terrain, sub-grid TKE parameterization at the scale of large eddies, and ensuring local and global tracer mass-conservation.
For this work, super-resolution (Δx ≤ 100 m, Δt ≤ 1 s) model simulations with Large-Eddy-Simulation sub-grid scale parameterization were developed and implemented using WRF-ARW. The objective was to resolve small dynamical processes inclusive of spatio-temporal scales of high-speed (e.g., 100 m/s) airborne measurements. This was achieved by down-scaling of reanalysis data from 31.25 km to 50 m through multi-domain model nesting in the horizontal and grid-refining in the vertical. Further, WRF dynamical-solver source code was modified to simulate passive-tracer emissions within the finest resolution domain. Different meteorological case studies and several tracer emission sources were considered. Model-generated fields were evaluated against observational data and also in terms of tracer mass-conservation. Results indicated model performance within 5 % of observational data in terms of sea level pressure, temperature and humidity, and agreement within one standard deviation between modelled and observed wind fields. Model performance in terms of tracer mass conservation was within 2 % to 5 % of model input emissions.
- Preprint
(8991 KB) -
Supplement
(2899 KB) - BibTeX
- EndNote
Sepehr Fathi et al.
Status: final response (author comments only)
-
CEC1: 'Comment on egusphere-2022-1125', Astrid Kerkweg, 28 Nov 2022
Dear authors,
the GMD editorial states about Model experiment description papers: "The primary purpose of these papers is to enable modelling communities to perform the same experiments. Therefore, everything required to run the experiment must be provided, apart from the model itself." (GMD Editorial main text, see also appendix B4)
I do not see that your article is about a model experiment which is performed by a larger community with different models. Therefore, your paper fits better the "Technical and development paper" type and I will ask the editorial office to apply this manuscript type to your article.
Best regards,
Astrid Kerkweg (GMD Executive Editor)
Citation: https://doi.org/10.5194/egusphere-2022-1125-CEC1 -
RC1: 'Comment on egusphere-2022-1125', Anonymous Referee #1, 03 Jan 2023
This work features high-resolution numerical simulations with the model WRF-ARW using realistic meteorological boundary conditions and employing a multistep nesting approach. A tracer transport module is used to simulate the dispersion of various emission sources at the finest LES scale. Ground-based and airborne observations are used to evaluate meteorological parameters, which generally showed satisfactory performance, though some questions regarding the wind-speed evaluation remain. Finally, the mass-balance scheme was applied on 4-D model output to infer source emission rates and to evaluate local and global mass conservation of the tracer transport scheme. This part of the study can still be better elaborated as suggested by the comments given below. The structure of the manuscript, the quality of the figures, and the presentation of the data and results can be further improved.
Major comments
The authors evaluate the performance of the WRF tracer transport scheme regarding mass conservation and the positivity of transported concentrations. I think since this is one of the most relevant results of this study, a more detailed description of the transport schemes in the paper is necessary. What is the differential equation to be solved? How are the advection and diffusion terms discretized in space and time? What kind of flux redistribution is applied for positivity? The authors state that the turbulent diffusion step is mainly responsible for the observed negative mass creation. This seems not very plausible unless it is shown by example. Did the authors test the transport scheme with diffusion turned off? The transport scheme is supposed to be conservative (based on the flux-divergence formulation). How is it then possible that a vertically changing grid spacing results in a violation of mass conservation? This needs further clarification.
The implementation details of the mass-balance technique are still only poorly described. The authors should give some details on how the advective and diffusive fluxes (FC,H , FC,H T, FC,V, FC,V,T) at the flux-box boundaries are computed in discrete form. How are the scalar values, which are defined at mass points, interpolated/reconstructed at the cell faces, where the velocity components are defined? Does the offline computation of the fluxes based on model output match the flux computation of the WRF advection/diffusion scheme?
The horizontal grid-refinement ratio of 1:5 in the nesting process is quite large (typically 1:2 or 1:3 is used). Are there any specific reasons for using such a large grid-refinement ratio? Are there any tests available from which nesting errors could be estimated? How large are the vertical grid refinement ratios near the ground? âz for d03 and d04 are missing in the paper.
Regarding the wind speed evaluation with WBEA observations: The sensitivity of wind speed to model resolution was not satisfactorily explained. It is not convincing that the nesting error would result in a systematic increase in wind speed at each nesting step. Such a conclusion certainly cannot be drawn from Daniels et al. (2016). This brings me to the question of whether the vertical interpolation of model data to the measurement height of 10m was based on a proper logarithmic law? A comparison of vertical profiles of wind speed between domains d03, d04, and d05 would be helpful to further investigate this large apparent sensitivity.
Regarding Figure 5: Did the authors really show the full vertical extent of model data from the first model layer above ground? If so, it is strange that the wind speed in the model does not decrease towards the ground. The WEBA data shows consistently lower wind speeds than the aircraft data, which obviously reflects the logarithmic wind law within the surface layer. Your model results give no hint of that.
Figure 7 alone provides a very weak basis for the discussion of the vertical mixing characteristics of the plumes. Some features the authors describe (like the effect of the Athabasca river basin) may be just attributable to random turbulent fluctuations. I suggest providing at least an additional plot for temporal mean turbulent statistics, like the average vertical turbulent flux of tracer mass or turbulent stresses.
The structure of the manuscript needs to be improved:
- Section 2.2: I suggest separating the model description from the technical setup more clearly. The paragraphs do not seem to follow any clear logic. I would suggest the following paragraphs:
- A concise description of the WRF model
- The tracer module should be described in more detail (governing equation + spatial discretization), as it is the main focus of the study.
- Introduction of the simulation domains d0-d5 with the static information (horizontal + vertical resolution, horizontal extent, coverage) along with the nesting technique
- Simulation settings of domains d0-d5 + driving data of coarsest domain d0
- Implementation of the point and area emission sources in d05
- Section 3.1: The authors should first provide a complete and purely descriptive comparison of model data (domain d05) with WEBA and aircraft observations before they draw any further conclusions regarding model performance. The content from line 310 until the end of the section describes in fact only model sensitivity and does not provide a model evaluation at the location of the oil-sand facility. From line 321 onwards the authors just mostly repeat the content of Table 4, which is quite confusing to read. I would suggest dedicating this paragraph to model sensitivity, labeling it as such, and presenting the data in a more structured way.
- Section 3.3: It is started here again with a model description. The authors should move this first paragraph to Section 2.2, where the tracer module is introduced.
Many figures need revision:
- The subplots of multipanel figures should be labeled in alphabetic order and referenced as such in the caption. The caption should only contain precise and concise descriptive information and no evaluative comments. The caption of Fig. S2 is incomplete and contains an evaluative comment. What exactly is “level tracer count sum” in Figure 7, S4, and S5? The authors should only use terms that are clearly defined.
- The colorbars of Figures 11 and S1 are still missing. The colorbars of Figures 7, S3, S4, S5, S6, and S7 can be improved (logarithmic scaling of ticks and larger tick labels).
- It is difficult to see the wind barbs in Figures S3, S6, and S7, and the thin contour lines (I assume this is terrain height?) further interfere and do not provide further useful information.
- The labeling of the time axis in Figures 4 and S2 needs to be revised (what is the “20” in front of the time stamp?). Please also include bias values in each subplot of Figure S2, like done for rsme.
- For a better comparison of data in Figure 3, I suggest using the same percentage axis (e.g., 0, 10, 20, …, 60) in the wind-rose plots.
- The tick labels of the x- and y-axis need to be increased in Figures C1-C3. What simulation domain is presented in Figures C1-C3, and is the depicted data horizontally averaged?
Minor comments
Line 29: missing comma before “which”
Line 51: “were successful”
Line 65: Please provide also an estimate for the sampling frequency besides the flying speed. How one gets dx ≤ 100m and dt ≤ 1s?
Lines 85-86: “from the Joint Canada-Alberta Implementation Plan on Oil Sands Monitoring airborne campaign (JOSM 2013).”
Line 114: “are considered”.
Line 123: “large area rectangular surface source” is a bit strange. It should be clear to the reader that it is a surface source. Maybe “large rectangular area source” is better? Please check also other instances in the manuscript, e.g., “multi-section line surface source“ (line 124).
Table 2: Please provide an extra column for “Size” for the numbers ∼20 km, ∼50 km2, etc.
Caption of Figure 1: Is Hwy an abbreviation for highway?
Line 142: This is not a complete sentence. Suggested correction: “…simultaneously. The process where the coarse “parent” domain’s output is interpolated to provide initial and lateral boundary conditions for the fine “child” domain is referred to as one-way nesting.”
Line 149: “and this is”
Line 153: Please either consistently use “JOSM 2013” or “2013 JOSM” throughout the manuscript.
Caption of Table3: What exactly are “Coarse” and “Fine” for the vertical grid? Please provide a clear explanation with numbers here or in the main text. According to the main text, “Fine” is not exactly the same for d04 and d05? What is Ztop? Please use precise language and introduce every variable properly.
Line 165: “in the figure Figure 2b.”
Line 167: “complex flow conditions”
Line 169: Please provide a reference with section numbers for “In the following sections”.
Lines 170-175: This paragraph deals again with vertical model resolution (already introduced before line 160) and contains repeated information (see line 158).
Line 183: Is this really part of the subgrid-scale parameterization or just numerical diffusion to remove spurious small-scale noise of the advection scheme? What exactly is the default option?
Line 185: What exactly are these modifications? I checked Blaylock et al. (2017) and could not find any WRF modifications related to tracer transport.
Line 205: Please provide the section number here (“…are discussed in Section xx”).
Line 308: “were more severe is larger”
Line 308: “We discuss later“ Where exactly?
Line 311: “due to the fact that”
Line 311: “(see above)“ Where exactly this is shown?
Line 339: “below” Where exactly (Section xx)?
Line 361: What is an “emission flight”?
Line 361: “in relevant publications” Which publications? Only Fathi et al. (2021)?
Line 391: If I am not wrong, mass conservation is determined by the spatial discretization (flux-divergence formulation) and is not directly related to the Runge-Kutta scheme, which is just a time integration scheme.
Line 397: “The turbulent diffusion step in the model is also prone to creating negative mass, but to a lesser degree.” Can the authors please provide a reference for this? Standard second-order diffusion does not create negative mass (see e.g., Fig. 2a in Xue (2000): High-Order Monotonic Numerical Diffusion and Smoothing).
Lines 463 and 473: use singular “Table 6”
Line 465: “with model input emission rates (MIE)”
Line 481: Why negligible estimated emission rates? The estimated emission rates for case 2 were within 5% of MIE (except case CNRL0).
Lines 485-490: Based on the results, I would be cautious to draw such a conclusion. Higher resolution does not always mean more accurate wind fields (see case 1, where d03 seems to be more accurate than d05). Where can I find the observation of (FC,H < 20%)?
Line 502: “as indicated in Fig. 11”
Line 524: Where are these modifications described in the manuscript? Is this related to the hard coding of emissions?
Lines 536-538: This is quite speculative unless the authors provide more details on the numerical schemes.
Lines 543-544: What happens if the tracer plume is vertically advected into layers with an irregular grid spacing? Then there is the same problem with losing mass.
Lines 552-553: See the previous comment.
Citation: https://doi.org/10.5194/egusphere-2022-1125-RC1 -
RC2: 'Comment on egusphere-2022-1125', Anonymous Referee #2, 09 Jan 2023
“Passive Tracer Modelling at Super-Resolution with WRF-ARW to Assess Mass-Balance Schemes” by Sepehr Fathi, Mark Gordon, and Yongsheng Chen
Recommendation: Major revisions
General comments:
This manuscript introduced a turbulence-resolving (or super-resolution) passive tracer modeling system and assessed its performance in terms of 1) meteorological parameters in the atmospheric boundary layer (ABL), 2) characteristics of small-scale passive tracer structures in the ABL, and 3) conservation of passive tracers. The super-resolution tracer modeling system was developed based on a one-way nesting capability of the WRF model. Specifically, a gradual downscaling from reanalysis scale (31.25 km) to an LES scale (50 m) was implemented in horizontal, and grid-refining in vertical was applied for the two innermost sub-kilometer grid-spacing domains. Meteorological evaluations were performed for three cases in comparison to station and aircraft observations of surface pressure, temperature, humidity, and wind. The first major concern I have is about the methodology used in meteorological evaluations, especially the inconsistency of the reference data (base case), i.e., the use of d03 results for evaluation of d04 results vs the use of d04 results for evaluation of d05 results etc. The characteristics of plume dispersion were then investigated focusing on differences caused by emission scenarios — e.g., emission source height, and emission type (point, line, and surface) — and by meteorological conditions — i.e., three different weather cases. My second major comment is that this manuscript lacks discussions on role of the ABL turbulent mixing in the plume characteristics, even though the results presented in the manuscript indicate that the ABL turbulence plays a significant role in determining the tracer dispersion that appears differently according to the emission source height and meteorological conditions.
Major comments:
1. Revisions are needed in the methodology used in meteorological evaluations.
1.1. Wind speed, which is the most critical meteorological variable in accurate prediction of tracer dispersion modeling, shows large biases in comparison to observations. The authors suggested possible reasons of the large bias, including the NARR reanalysis data that have positive biases. To take account of the impacts of this reanalysis error carried over to the simulations, the authors used coarser resolution simulations to evaluate nested-domain simulations, instead of observations. I agree to the impact of reanalysis errors on the nested domains, but I think evaluation results in comparison to both observations and coarser-domain results need to be presented together, together with tables summarizing bias and RMSE scores, in the main text.
1.2. This study used percentage error as an evaluation metric. Most meteorological variables, except for wind speed, have large absolute values (e.g., pressure, RH, temperature, and wind direction), therefore using the percentage error as an evaluation metric could mislead about the performance of the modeling results. I suggest adding a table that summarizes bias and RMSE scores of the meteorological variables in comparison to observations, similar to Table 4.
1.3. There are a number of places that plots and main texts are inconsistent. Wind-rose diagrams in Figure 3 show that winds are from north-east and east, while the main text mentions that the wind directions are from west and west south west (Line 265, Page 11).
2. This manuscript lacks discussions on the role of turbulent mixing in determining plume characteristics, while the results presented in the manuscript indicate it is critical to understand the differences of the plume characteristics by emission source height and also between cases.
2.1. Plume characteristics from the emission scenario CNRL0 indicate the different role of turbulent mixing across the atmospheric boundary layer (ABL) top: i.e., within the ABL where turbulent mixing plays a dominant role in vertical structure of the ABL, including tracer concentration, vs. above the ABL where turbulent motions are suppressed by negative buoyance of the stably stratified inversion layer overlying the ABL.
In Case 1, the plume dispersion resulting from the emission scenario CNRL0 shows very different behaviors (e.g., Figure 7) from other stack scenarios. The source height of the CNRL0 scenario is located at 483 m, which is around the ABL top where the vertical mixing by turbulence ceases due to the capping temperature inversion. The ABL height can be inferred from the potential temperature profile in Figure C1, which shows a stably stratified layer with small spatial variability (standard deviation) around 500 m. The tracers emitted from other sources are quickly (within a few minutes) mixed in vertical within the ABL from the surface to the ABL top, leading to a similar vertical structure at ~ 10 km downstream regardless of the source height (Figure 7). On the other hand, the vertical dispersion of tracers in the CNRL0 scenario is confined to a smaller vertical domain, due to the relatively weaker turbulence mixing above the ABL top than within the ABL.
2.2. Differences in plume characteristics between Case 2 vs. other two cases can also be explained by the role of turbulence mixing. The meteorological profiles shown in Figure C2 indicate stronger turbulence activities in Case 2 than in other two cases, with the ABL top at around 1300~1500 m. In Case 2, I guess the plume dispersion of the CNRL0 would be very similar to other CNRL scenarios (though not shown nor mentioned in the manuscript), because all sources are located within the ABL in this case. Based on the meteorological profiles of Case 3 presented in Figure C3, I think Case 1 and Case 3 would show very similar results; the ABL top is located around 500 m in both cases, resulting in only the CNRL0 emission source being above the ABL top. More in-depth comparison of the plume characteristics between Case 1 and Case 2 could be made, based on the different role of turbulence mixing between the two scenarios. The manuscript did not provide any of these important points.
Citation: https://doi.org/10.5194/egusphere-2022-1125-RC2
Sepehr Fathi et al.
Sepehr Fathi et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 400 | 110 | 12 | 522 | 40 | 6 | 3 |
- HTML: 400
- PDF: 110
- XML: 12
- Total: 522
- Supplement: 40
- BibTeX: 6
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1