the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Feb 2026
Elucidating CO2 accumulation and dispersion in a semi-enclosed bay industrial park using Lidar and WRF-GHG modelling
Abstract. Industrial parks represent critical nodes in the global carbon cycle and thus require accurate monitoring and modelling to support effective carbon management. Although satellite observations, ground-based measurements, and numerical modelling frameworks have been widely utilized, these approaches inherently struggle to simultaneously achieve high temporal and spatial resolution. In this study, high-resolution Lidar is integrated with the Weather Research and Forecasting model coupled with greenhouse gas fluxes (WRF-GHG) modelling to comprehensively diagnose the CO2 accumulation–dispersion dynamics and their driving mechanisms within the Luoyuan Bay industrial park. The comprehensive analysis reveals a distinctive diurnal pattern of CO2, characterized by nighttime accumulation and daytime dispersion. Lidar observations indicate that stable atmospheric conditions and valley terrain synergistically cause CO2 to accumulate in low-lying areas at night, with concentrations exceeding 700 ppm. During daytime, strengthened southeasterly sea breezes and intensified turbulence promote its dispersion to the northwest and vertical uplift, reducing concentrations to 500-550 ppm. A significantly negative correlation between CO2 concentration and wind speed is also confirmed. While the WRF-GHG model reproduces the overall temporal variation, it systematically underestimates CO2 levels (420–460 ppm). The discrepancy is attributed to the limited spatial resolution of the emission inventory and the model’s inherent constraints in capturing terrain–wind field interactions within the bay. This study highlights the unique capabilities of coherent differential absorption Lidar, elucidates the key limitations of current modelling approaches, and provides a robust scientific basis for refining carbon verification systems and enhancing the performance of regional carbon models.
- Preprint
(3866 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-6470', Anonymous Referee #1, 16 Mar 2026
-
RC2: 'Comment on egusphere-2025-6470', Anonymous Referee #2, 09 Jun 2026
Major comments.
This paper presents an interesting demonstration of scanning lidar technology and joint CO2 and wind measurements. The interesting measurements are the most compelling element of the manuscript. The quality of the graphics and writing is good. There is potential for making this work publishable. In my opinion, however, the atmospheric modeling effort is not very useful given the limits of the modeling configuration that was used and the primary results of the paper are not highly compelling. I cannot support publication of this work in its current form.
1. The primary results of this study have been known for decades and do not warrant a new publication. 1) Pollutants accumulate near the surface when the atmospheric boundary layer is stable and they disperse during the day. 2) High winds disperse pollutants. 3) Coasts can cause sea breezes that move pollutants. 4) Mountains can trap pollutants. In my opinion these are the primary results of the study. The first two are clearly illustrated in Figure 2. The third point is demonstrated with examples and a wind rose, but the cause of the SE winds is not proven. The last point is not shown definitively but is plausible. None of these, however, is a new finding. The results obtained here are only unique in that they have been obtained with an innovative observational system. I do not think that reproducing very well established understanding of air pollution meteorology with a new instrument is worthy of publication unless the point is to demonstrate the limits of the instruments. This manuscript does not focus at all on the limits of the instruments.
2. The WRF configuration is poorly suited to the problem at hand, and the authors use this to make injustified claims about general failures of the WRF-GHG simulation system.
2A. Meteorology. Simulating the atmospheric dynamics of a small, mountainous bay is certainly challenging. Attempting to simulate this system and illustrating the limits of a model to reproduce these dynamics is a worthwhile topic of study. This WRF configuration, however, has relatively coarse vertical resolution, coarse horizontal resolution given the scale of the problem at hand, and the surface data appear fundamentally flawed with the lidar site sitting in the ocean and, according to the authors, a poor representation of the local topography. It is not surprising that this configuration is unable to simulate the complex flows observed by the lidar. Further, there is very little direct focus on model evaluation save for some general results in Table 2 and Figure 3. Figure 4-6 include both lidar data and model output, but at very different spatial resolutions and will little direct comparison. Figure 7 avoids model-data comparison. A serious attempt to simulate the bay meteorology and compared to lidar observations would be interesting. This manuscript does not include such an effort.
2B. CO2. The model clearly entirely fails to simulate the CO2 concentrations observed at very fine resolution as observed by the lidar. The coarse horizontal resolution of the model almost guarantees this outcome. The extremely coarse resolution (in space and time) anthropogenic flux model (EDGAR) completely guarantees this failure. EDGAR fluxes are never shown. The model has the capacity to track different components of CO2, as noted by the authors, but this is never shown. The model, as configured by the authors, is certain to fail in reproducing the observed CO2 concentrations. This is not an interesting result.
3. The paper relies primarily on the lidar observations, but the lidar data are not available. This is not acceptable for publication.
I would like to suggest some directions for revision that I believe could yield a publishable manuscript.
Focusing on the limits of the lidar technology seems the most obvious to me. The authors have shown only three high concentration cases. Are there conditions where the lidar can no longer detect the industrial emissions? This could be coupled with expanding the study to describe what is observed under a wider variety of meteorological conditions. This could be helpful for illustrating the conditions where this observing system could be used to quantify emissions.
Evaluating WRF meteorological performance with the lidar in this complex environment would also be worthwhile if the authors put more effort into experimenting with the WRF configuration. It would be helpful to know if, with good vertical resolution, improved horizontal resolution and a reasonably good representation of the terrain of the bay and the temperature of the water, the atmospheric dynamics of this mountain bay could be simulated or not.
Estimating CO2 emissions from the industrial park would, of course, be very interesting. This could potentially be done with the lidar winds, a simple guess as to the regional fluxes, Lagrangian particle modeling to infer the upwind areas influencing the lidar CO2 concentrations and some simple variational experiments. I think the WRF simulation as presented is too coarse in resolution to handle the atmospheric transport in a useful way. This is just one idea.
I hope these ideas are helpful.
My detailed comments follow. Many of them will echo the points raised above.
1. Lines 12-15. “Although satellite observations, ground-based measurements, and numerical modelling frameworks have been widely utilized, these approaches inherently struggle to simultaneously achieve high temporal and spatial resolution.”
This statement is too broad to be helpful. Please make this more specific or delete it.
2. A daily cycle in CO2 accumulation and a negative correlation between CO2 concentration and wind speed are both patterns that have been understood for many decades of the study of pollutants in the atmosphere. If these are the main results of the study, I fear that this work may not yield any new understanding of atmospheric CO2.
3. Lines 22-25. “While the WRF-GHG model reproduces the overall temporal variation, it systematically underestimates CO2 levels (420-460 ppm). The discrepancy is attributed to the limited spatial resolution of the emission inventory and the model’s inherent constraints in capturing terrain–wind field interactions within the bay.”
The WRF model can be deployed in a nearly infinite array of configurations. The authors cannot argue that they have found “the model’s inherent constraints” unless they have explored the model’s potential configurations.
4. Lines 25-27. “This study highlights the unique capabilities of coherent differential absorption Lidar,
elucidates the key limitations of current modelling approaches, and provides a robust scientific basis for refining carbon verification systems and enhancing the performance of regional carbon models.”
This study highlights the capabilities of coherent DIAL, but does not provide the basis for refining carbon verification systems. I also am not really sure what a “carbon verification system” is.
5. Lines 37-39. “Currently, industrial CO2 emissions monitoring technologies mainly include in situ monitoring and remote sensing monitoring methods. In situ instrumental monitoring conducts real-time measurement of atmospheric CO2 concentrations via fixed-site online devices but struggles to effectively cover large-scale areas (Gurney et al., 2002).”
Tracking industrial emissions relies primarily upon accounting methods. Atmospheric monitoring, at this point, is used only in research settings save for a few rare exceptions. In addition, the Gurney et al., (2002) paper is archaic. It is not a good example of current research to monitor CO2 emissions using atmospheric methods.
6. Lines 78-79. “The organic combination of the two provides a new research approach.”
The use of DIAL CO2 measurements is innovative. It is an overstatement, however, to claim that this is “a new research approach.” Many previous studies have used a combination of atmospheric mole fraction and atmospheric transport observations combined with mesoscale atmospheric models to study regional atmospheric composition and pollution sources.
7. Line 98. “This further worsens local air pollution.”
This statement implies a comparison to an earlier state of polluted air. Perhaps you mean to say, “The terrain can trap local emissions causing poor local air quality.”
8. Lines 103-106. “a range resolution of 120 m, and a time resolution of 1 min. It can simultaneously measure CO2 concentrations and wind fields. The CO2 measurement accuracy has been confirmed in prior
validation, measured at a background concentration of 463 ppm, showing a mean error of 2.05 ppm and a standard deviation of 7.18 ppm.”
Concerns: 1) If there is “prior validation” then please include a citation for this work. 2) Does the CO2 measurement performance apply to the range and time resolution noted - 1 minute and 120 meters? This is implied but not stated clearly. 3) Please explain what a “mean error” represents. If the measurement is always biased, it can be corrected. Is this the case?
9. Lines 106-107. “When compared to an optical cavity ring-down spectrometer, the correlation coefficient reaches 0.91 and the root-mean-square error (RMSE) is 5.24 ppm (Yu et al., 2024).”
Apologies, but this is confusing. How is this related to the previous performance metrics, especially the quoted standard deviation of the CO2 measurement? Is this the same work? If so, why are these quantities different? Is the time and range resolution for this comparison also 1 minute and 120 meters? Please be precise when stating instrument performance specifications.
10. Lines 107-109. “For wind field measurements, the RMSE of horizontal wind speed is less than 0.6 m s-1, the RMSE of wind direction is less than 11°, and the vertical wind speed measurement accuracy is
0.2 m s-1 (Yuan et al., 2022)”
As above, do these metrics refer to 1 minute and 120m temporal and spatial resolution data? How have these metrics been determined? How is vertical wind speed accuracy determined? Finally, this sentence is missing a period.
11. Figure 1. Please explain the color legend in 1(b). Is this elevation of the terrain?
12. Figure 1. It would help to have a better view of the innermost domain and the bay. At present it is difficult to see how the lidar observational domain sits within the bay, and the size of the bay compared to the innermost WRF domain.
13. Line 138-139. “The Copernicus Atmosphere Monitoring Service (CAMS)”
GHG boundary conditions are a critical element of the simulation and a new topic. This should be part of a new paragraph.
14. Lines 142-143. “We selected the three-hourly interval products from the first day of each daily forecast, with a spatial resolution of 0.1° and 137 vertical model levels.”
I don’t understand. What are these forecasts used for? The WRF simulation only needs one set of initial conditions but it needs continuous boundary conditions for the entire period of study. Does “each daily forecast” suggest that WRF is being run independently for each day of the study, with new boundary and initial conditions? Is any spin up time allowed? CAMS and WRF GHG fields could be quite different in equilibrium depending on both fluxes and transport. No spin up time could result in a simulation that is dominated by the modeled GHG fields adjusting to the differences between these two modeling systems.
15. Line 150. There is no discussion of anthropogenic emissions. Is WRF run without any anthropogenic CO2 emissions estimate? Why? How is this comparable to the observations over this industrial area?
16. Lines 158-161. “To accurately assess the impact of industrial park emissions on the surrounding environment, a forested area within the Lidar’s azimuth range of 315°–320° and radial distance of 2.7–3 km was selected as the downwind site. Dominantly composed of evergreen broad leaved forests, this area maintains a high photosynthetic rate year-round. With a vertical elevation difference of approximately
450 m from the Lidar installation, it effectively reflects the concentration level of pollutants after dispersion.”
I am quite confused by this text. Why does a location 3km away and at a higher elevation “reflect the concentration level of pollutants after dispersion.” This is not nearly far enough downwind (if it is downwind) for emissions to be well mixed in the atmospheric boundary layer. Please explain what is meant by “after dispersion.” The altitude is also confusing. Is this ‘downwind site’ 450m above sea level? Or is 450m the elevation in the atmosphere / above ground? as measured by the lidar?
17. Lines 161-163. “Emission source data were obtained from the park’s emission area within the Lidar’s detection range, i.e., the industrial production zone with an azimuth of 315°–4° and a radial distance of 500–1500 m.”
What is “emission source data”?
18. Lines 165-167. “During the entire study period, the average CO2 concentration at the emission source was 578 ppm, while the average concentration at the downwind site was 487 ppm, with a difference of 91 ppm (Fig. 2a). Meanwhile, the mean wind speed was 2.7 m s-1.”
I don’t think averages over several weeks with variable meteorology and emissions are very useful.
19. Lines 168-169. “the daily maximum concentration of the emission source exceeded 900 ppm on each day.”
What is “the emission source”? Is this a location that the authors claim is the primary source of CO2 emissions?
20. Given the complex spatial setting, a view of the “emission source” and the “downwind site” in a horizontal / vertical cross section view along the rough direction of the lidar beam might help to explain how the authors have chosen to subselect their observations.
21. Figure 2. What is the time and space resolution of the data shown in Figure 2(a)? Is that the data resolution used to compute the statistics in Figure 2(b), and shown in the data points of Figure 2(c)?
22. Lines 175-176. “with afternoon concentrations slightly higher than nighttime levels. This difference is closely related to changes in land-sea breeze circulation and turbulence.”
This might be true, but it is not at all evident from the observations shown. Please either show this relationship or revise this statement.
23. Lines 178-183. “Meanwhile, as a semi-enclosed bay, Luoyuan Bay exhibits a prominent land-sea breeze effect: prevailing southeasterly winds transport high concentration of CO2 from industrial emissions to the
downwind site in the north-west. However, the surrounding mountainous terrain impedes horizontal dispersion, forcing part of the CO2 to lift vertically, which further promotes pollutant dilution. In addition, the photosynthesis of forests enters an active period during the daytime, and the absorption of CO2 partially offsets the increment of industrial emissions, preventing a sharp rise in concentrations at the downwind site.”
This all might be true, but nothing in this manuscript demonstrates these hypothesized relationships. Speculative reasoning without a basis in observation might be suitable for the discussion, but these are not results.
24. Figure 2(d) is nice to see, but this relationship has been shown many, many times for a wide variety of pollutants.
25. Table 2 lacks units, it does not describe the domain studied and it does not describe the resolution of the data. The sources of data are also not clear. This is not a useful table. This certainly does not present “validation” of the WRF simulation. Figure 3 is more useful.
26. Lines 200-202. “The WRF-GHG simulation results were compared and validated against Lidar detection data and sensor-collected data, as shown in Table 2. Figure 3d presents a comparison between the WRF simulated CO2 concentrations and the Lidar measurements at the park’s emission source.”
Sentences like these do not belong in the text of a manuscript. These should be in the captions for the table and figure. Please remove these from the text. Start each paragraph of the results with a topic sentence that presents the primary result that you wish to discuss in the following paragraph. Please follow this approach throughout the manuscript.
27. Lines 202-203. “While the two datasets exhibit good consistency in temporal variability, substantial discrepancies remain in their absolute values.”
I’m sorry, these two CO2 values look utterly unrelated. There is not at all “good consistency in temporal variability.”
28. Lines 205-206. “The EDGAR_2024_GHG dataset used in the simulation has a latitudinal and longitudinal resolution of only 0.1° and a temporal resolution of monthly data.”
This should be described in the methods.
29. Lines 203-205. “Additionally, the northern Luoyuan Bay where the park is situated is surrounded by mountains on three sides and adjacent to the sea on one side; this unique terrain poses a significant
challenge for the terrain data inherently used by WRF.”
As noted previously, it would be very helpful to see a distance / altitude cross section of the data locations and the model grids and altitude levels that are being evaluated in this study. It would help a great deal to show the actual terrain and the terrain data used in WRF. Finally, a visualization of the CO2 fluxes from EDGAR would be exceptionally helpful. Without this information this text is not very interpretable. What, for example, is “the terrain data inherently used by WRF” and why does this cause a problem?
30. Lines 210-213. Correlation coefficients spanning a few weeks show that WRF simulates the daily cycle and weather patterns reasonably well. That is not surprising.
31. Lines 214-216. “Among these, the correlation coefficient (R) for 10m wind speed is 0.7 with RMSE of 1.45 m s-1, indicating that WRF underestimates wind speed to a certain extent;”
RMSE and R values say nothing about biases. These data do not indicate that WRF underestimates wind speed.
32. Figure 4. The comparisons between the CO2 lidar and the WRF CO2 simulations are not useful. The observations are on an entirely different spatial scale than the model and represent a complex set of distances above ground. The large area WRF simulations are the surface layer where, especially in stable conditions but truly at any time of day close to a strong source, will be quite different than concentrations aloft. The authors have already noted that EDGAR has roughly 10km resolution hence cannot possibly represent emissions from this microscale domain (lidar) effectively. And if I understand the WRF maps, the model terrain puts the lidar in the water! I do not think anything useful can be learned from this simulation.
33. Figure 4. What is the time and space resolution of the lidar data displayed. This is never mentioned but is quite important. If these are snapshots in time, why have these been selected?
34. Lines 243-246. “Only low-intensity emissions were detected at the sources, and CO2 was transported
from the southeast to the northwest. At 18:37 LT, the sea breeze weakened, and the wind field in the park became turbulent (wind speed: 2–4 m s-1). The emission intensity increased at the source located at 1 km radial distance with 330°–350° azimuth, with CO2 concentration reaching around 750 ppm and dispersing to the far field.”
One minor concern. Winds can be turbulent regardless of the mean wind speed. Mean winds becoming lower(?) do not mean that atmospheric turbulence has increased. The opposite is often true. A more significant concern: The text appears to associate the strength of CO2 sources with the observed CO2 concentrations. Concentrations are the result of both the emissions strength and the ventilation. Source strength cannot be judged solely via atmospheric concentrations.
35. Lines 252-253. ‘“Combined with mountain obstruction and a decrease in Boundary Layer Height (BLH),”
I have not seen any measurements of BLH. these would be useful and could be obtained by a lidar. If this is an assumption, that should be clearly stated.
36. I am afraid that I do not see much scientific content in Figure 4 and the associated discussion, save for a demonstration of the lidar technology. I have the same concerns for Figures 5 and 6.
37. Lines 299-304. “Joint analysis of the high concentration period (24–26 March) reveals that both the Lidar and WRF model captured the core dynamic of CO2: "nighttime accumulation and daytime dispersion". The dispersion and accumulation processes are jointly influenced by wind fields (sea-land breeze transition, wind speed) and terrain (mountain obstruction). The Lidar accurately identifies local emission plumes and small scale dispersion details, while the model reproduces the overall regional variation trend. These two approaches complement each other to verify the complex dispersion mechanism of CO2 in the semi-enclosed bay.”
The day/night cycle of CO2 accumulation close to the surface where sources are located is not a significant or publishable finding. The variations in wind related to the bay location are interesting but it is not clear that the WRF simulation reproduces these patterns. I’m afraid that I do not agree that these are complementary approaches for this set of observations.
38. Lines 329-334 should be in the figure caption, not the text.
39. Figure 7(f). What is purpose of “the inversion layer”?
40. Why does Figure 7 show mostly model output, and mostly at altitude beyond the range of the lidar system? What is the significance of showing these modeled fields? A comparison to the lidar winds and ABL depth estimate with this type of resolution and view might be valuable. It would be helpful to know if WRF was simulating similar BLH mixing and winds in this complex region. The CO2 fields from the model are not useful.
41. Paragraph starting at line 335. This is reasonable but not at all new or unique. And this discussion neglects the impact of wind speed. The accumulation of pollutants close to the ground in stable conditions is not a new finding worthy of publication.
42. Paragraph starting with line 342. Mountainous terrain trapping local pollution sources is also not a new discovery.
43. Lines 353-360. The source for this text is not clear. More importantly, the purpose of this text is not clear. I presume this is a discussion of the simulated meteorology and CO2 transport, but this is not at all clear. The text speaks of CO2 emitted by the industrial park accumulating in the boundary layer, but the authors have shown that the simulation does not at all realistically capture the local emissions. A comparison of simulated and observed meteorology and distributions of CO2 in space and time would be worthwhile, but that is not the content of this text. A simple description of the environment in this bay is interesting but not surprising or new in terms of general meteorological properties.
44. Lines 373-380 are a summary of what has already been presented in the manuscript. This is not suitable text for discussion or conclusions. The rest of the paragraph introduces a new figure, but it is a summary figure rather than conclusions. Conclusions should focus on the message that readers should take away from the new findings. That is not a summary.
45. If the manuscript were focused on proving the existence of the day/night pattern shown in figure 9 using both the WRF simulation and the lidar data, and comparing these when appropriate (e.g. atmospheric winds, BLH) this would be a better manuscript. Bringing up Figure 9 in the conclusions isn’t appropriate. Further, the manuscript does not bring together lidar and WRF throughout to illustrate the validity of the pattern of events shown in Figure 9.
46. Line 390. This paragraph could be a conclusion about the limitations of the authors’ implementation of WRF-GHG, but none of these issues would be new understanding for the research community. Groups that work with numerical models to simulate atmospheric flow in complex terrain would not expect this configuration to be successful.
47. Lines 395-399. This discussion of the boundary layer parameterization in WRF is speculation. The authors have not proven anything about the WRF boundary layer parameterization and its response to complex terrain.
48. The paragraph starting with line 400 is a valid summary of the observational findings of the manuscript. The demonstration of the lidar system is worthwhile. These are results, not conclusions.
49. Line 412 is the start of conclusions that I believe are suitable for this work.
50. Line 412-413. “This study demonstrates the advantages of Lidar in monitoring industrial CO2 emissions with high spatiotemporal resolution and highlights the limitations of the WRF-GHG model in characterizing complex terrain and refined emission sources.”
I disagree with this statement in two ways. First, this manuscript does not monitoring industrial CO2 emissions. It detects CO2 from industrial emissions. That is quite different from monitoring CO2 emissions. If these data were used to quantify CO2 emissions from the industrial part, this would be a reasonable conclusion. Second, this manuscript does not highlight any general limitations of WRF-GHG in simulating this environment. It does show that the authors have chosen a configuration of WRF-GHG that is not suited to simulating this environment. More work on the model is needed before any broad conclusions can be drawn about the limits of this modeling system in general. It is not clear to me that showing that this configuration of WRF cannot reproduce the meteorology of the system is helpful. It is not useful to show that this implementation of WRF-GHG fails to simulated atmospheric CO2 in this environment given the extremely inappropriate use of EDGAR to try to simulate CO2 emissions in this very local environment.
51. This final paragraph contains some useful thoughts about research needs but it is quite long.
52. Code and data availability. There is nothing in this statement about the observations save that it can be obtained by contacting the corresponding author. That is not sufficient data access for a peer-reviewed publication relying primarily on new observations.
Citation: https://doi.org/10.5194/egusphere-2025-6470-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 971 | 448 | 90 | 1,509 | 75 | 80 |
- HTML: 971
- PDF: 448
- XML: 90
- Total: 1,509
- BibTeX: 75
- EndNote: 80
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Summary
This manuscript integrates high-resolution coherent differential absorption Lidar observations with WRF-GHG modelling to comprehensively diagnose the CO2 accumulation and dispersion dynamics within the semi-enclosed Luoyuan Bay industrial park. The comprehensive analysis successfully reveals a distinctive diurnal pattern characterized by nighttime accumulation and daytime dispersion. The Lidar accurately identifies local emission plumes and small-scale dispersion details, while the WRF model reproduces the overall regional temporal variation trend. This study highlights the unique capabilities of Lidar technology in complex coastal terrains and provides a robust scientific basis for enhancing the performance of regional carbon models.
Recommendation
Overall, the work is innovative, well-structured, and provides valuable observational and modeling insights. I recommend that the manuscript be considered suitable for publication after the following minor points are addressed.
Specific Comments