the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The potential of drone observations to improve air quality predictions by 4D-var
Abstract. Vertical profiles of atmospheric pollutants, acquired by unmanned aerial vehicles (UAVs, known as drones), represent a new type of observation that can help to fill the existing observation gap in the planetary boundary layer. In this article, the first study of assimilating air pollutant observations from drones is presented to evaluate the impact on local air quality analysis. The study uses the high-resolution air quality model EURAD-IM (EURopean Air pollution Dispersion – Inverse Model), including the four-dimensional variational data assimilation system (4D-var), to perform the assimilation of ozone (O3) and nitrogen oxide (NO) vertical profiles. 4D-var takes advantage of the inverse technique and allows for simultaneous adjustments of initial values and emissions rates. The drone data was collected during the MesSBAR (Automatisierte luftgestützte Messung der SchadstoffBelastung in der erdnahen Atmosphäre in urbanen Räumen / Automated airborne measurement of air pollution levels in the near earth atmosphere in urban areas) field campaign, which was conducted on 22–23 September 2021 in Wesseling, Germany. The two-day analyses reveal that the 4D-var assimilation of high-resolution drone measurements has a beneficial impact on the representation of regional air quality in the model. On both days, a significant improvement in the vertical distribution of O3 and NO is noticed in the analysis compared to the reference simulation without data assimilation. Moreover, the validation against independent observations shows an overall improvement in the bias, root-mean-square error, and correlation for O3, NO, and NO2 (nitrogen dioxide) ground concentrations at the measurement site as well as in the surrounding region. Furthermore, the assimilation allows for the deduction of emission correction factors in the grid cells surrounding the measurement site, which significantly contribute to the observed improvement in the analysis.
- Preprint
(10998 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2024-517', Anonymous Referee #1, 07 Apr 2024
Drones equipped with low-cost sensors is a new approach for measuring trace gases within the planetary boundary layer. They enable vertical measurements of such gases, complimenting gaps in our current observational system. This manuscript explores a new application of drone observations. The authors use 2-day drone observations of O3 and NO at a polluted site in Germany to perform data assimilation. They found improved performance in simulated levels of O3 and NOx after assimilation, showing enhancements not only in local vertical profiles but also across a broader spatial region at ground level. This topic is interesting, and the results show some potential for future applications in improving environmental forecasts. The manuscript is well organized and easy to follow.
My major concern is about the representation error and the correction of emissions. The authors use observations from a single location near the emission sources, to correct a model grid cell averaging 5km x 5km, which introduces representation error. Indeed, it seems logical for the model to initially underestimate drone-observed NO (due to the model’s coarse resolution being unable to resolve fine-scale mobile emissions) and overestimate O3 (because the model’s resolution cannot resolve the close-to-source NO titration of O3). In such cases, directly aligning the model with single-spot drone observations and attributing this difference to emissions might result in a significant upward correction of NO emissions, as shown in the authors’ results (with correction factors up to 15 on the second day, and the locations of these corrections are suspicious). However, these corrections could be false. Therefore, their findings regarding emission corrections need cautious interpretation and thorough discussion.
Other comments are outlined below.
Major comments:
- In Section 2 and 3, could the authors consider adding an introduction on the accuracy and uncertainty of drone observations, and on how the observation error is specifically treated in their 4D-var system?
- Line 234-240, what is the authors’ definition of “optimal” conditions for efficient emission optimization? How do these conditions affect emission optimization, and are the two cases in this study considered optimal? Clarifications are needed.
- Table 6, what is the definition of “partial cost” and how is it calculated? Could the authors consider adding this information?
- Please change “NOx” to “NOx” throughout the text.
Citation: https://doi.org/10.5194/egusphere-2024-517-RC1 -
AC1: 'Reply on RC1', Hassnae Erraji, 30 Jul 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-517/egusphere-2024-517-AC1-supplement.pdf
-
RC2: 'Comment on egusphere-2024-517', Anonymous Referee #2, 22 Apr 2024
Drones bring new opportunities to improve air pollution monitoring vertically within the Planetary Boundary Layer due to their portability, flexibility, and affordability. The authors apply drone profile measurement of O3 and NO to optimize the anthropogenic emissions using the 4D-var data assimilation system of EURAD-IM. As the first application of drone data assimilation within a CTM, research is interesting, which also offers new insights and implications for future studies on emission assimilation.
The authors’ effort in conducting drone measurements, analyzing the data, and presenting their works are greatly appreciated. However, I still have some concerns in terms of their methodology.
The authors use of drone measurements at a single point to infer NO emissions across a 5 km × 5 km grid box, which raises concerns regarding data adequacy, potentially introducing significant bias in estimating emission correction factors. Particularly, the NO and NO2 emission correction factors derived from DA_23SEP exhibit a 4 to 5-fold increase compared to DA_22SEP. The results seem to be counterintuitive as anthropogenic emissions typically exhibit small changes over two consecutive days, unless extreme events occur that lead to significant changes in NO and NO2 emissions. Even though the authors get ‘improved’ simulations after the assimilation, I assume the observed large differences in NOx concentrations would be more related to the daily variations in transport, either horizontally or vertically. In such cases, improving the representation of the meteorology of the model would likely be more beneficial than merely adjusting pollutant emissions. The authors might consider comparing meteorological parameters in their simulations with observations, particularly focusing on variables like winds, to identify potential discrepancies.
The calculated emission correction factors for NO and NO2 can be even larger than 15, suggesting the regional emission inventory they use in their simulation have an uncertainty of over 1400%, which might not be a reasonable value for emission correction. I suggest the authors checking their anthropogenic emission inventory and previous emission assimilation studies to identify a scientifically reasonable range for their emission correction.
Additionally, I am not sure whether the authors’ observational results would affect by the wind conditions. Both horizontal and vertical wind speeds can exceed the ascent rate (i.e., 1m/s) of their instrument, potentially affecting data collection. How did they keep their instrument ascending at a constant rate in a fixed location? NO is a highly reactive air pollutant which can be converted to NO2 quickly upon emission. Consequently, NO usually presents significant decreasing gradient in the vertical direction. However, the observed NO concentration at 350 m can double the surface NO concentration in the manuscript (e.g., F10, F11). The result is confusing. The authors may want to check the credibility of their data and present relevant explanations.
I recommend that the authors incorporate more observations, if possible, rather than relying solely on data from a single location, in their assimilation process. By doing so, they can enhance the robustness and credibility of their results, mitigating any potential suspicions regarding the validity of their findings.
Specific comments:
L162: I assume the authors may want to say ‘with data assimilation’ here or above (i.e., L158)? Otherwise, all the four experiments are conducted ‘without’ data assimilation, which is not consistent with what is shown in Table 2.
L218-219 I am not sure whether anthropogenic emissions can have such large differences in two consecutive days.
L227-228 As this is a model study, more quantitative analysis is expected.
Citation: https://doi.org/10.5194/egusphere-2024-517-RC2 -
AC2: 'Reply on RC2', Hassnae Erraji, 30 Jul 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-517/egusphere-2024-517-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Hassnae Erraji, 30 Jul 2024
Status: closed
-
RC1: 'Comment on egusphere-2024-517', Anonymous Referee #1, 07 Apr 2024
Drones equipped with low-cost sensors is a new approach for measuring trace gases within the planetary boundary layer. They enable vertical measurements of such gases, complimenting gaps in our current observational system. This manuscript explores a new application of drone observations. The authors use 2-day drone observations of O3 and NO at a polluted site in Germany to perform data assimilation. They found improved performance in simulated levels of O3 and NOx after assimilation, showing enhancements not only in local vertical profiles but also across a broader spatial region at ground level. This topic is interesting, and the results show some potential for future applications in improving environmental forecasts. The manuscript is well organized and easy to follow.
My major concern is about the representation error and the correction of emissions. The authors use observations from a single location near the emission sources, to correct a model grid cell averaging 5km x 5km, which introduces representation error. Indeed, it seems logical for the model to initially underestimate drone-observed NO (due to the model’s coarse resolution being unable to resolve fine-scale mobile emissions) and overestimate O3 (because the model’s resolution cannot resolve the close-to-source NO titration of O3). In such cases, directly aligning the model with single-spot drone observations and attributing this difference to emissions might result in a significant upward correction of NO emissions, as shown in the authors’ results (with correction factors up to 15 on the second day, and the locations of these corrections are suspicious). However, these corrections could be false. Therefore, their findings regarding emission corrections need cautious interpretation and thorough discussion.
Other comments are outlined below.
Major comments:
- In Section 2 and 3, could the authors consider adding an introduction on the accuracy and uncertainty of drone observations, and on how the observation error is specifically treated in their 4D-var system?
- Line 234-240, what is the authors’ definition of “optimal” conditions for efficient emission optimization? How do these conditions affect emission optimization, and are the two cases in this study considered optimal? Clarifications are needed.
- Table 6, what is the definition of “partial cost” and how is it calculated? Could the authors consider adding this information?
- Please change “NOx” to “NOx” throughout the text.
Citation: https://doi.org/10.5194/egusphere-2024-517-RC1 -
AC1: 'Reply on RC1', Hassnae Erraji, 30 Jul 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-517/egusphere-2024-517-AC1-supplement.pdf
-
RC2: 'Comment on egusphere-2024-517', Anonymous Referee #2, 22 Apr 2024
Drones bring new opportunities to improve air pollution monitoring vertically within the Planetary Boundary Layer due to their portability, flexibility, and affordability. The authors apply drone profile measurement of O3 and NO to optimize the anthropogenic emissions using the 4D-var data assimilation system of EURAD-IM. As the first application of drone data assimilation within a CTM, research is interesting, which also offers new insights and implications for future studies on emission assimilation.
The authors’ effort in conducting drone measurements, analyzing the data, and presenting their works are greatly appreciated. However, I still have some concerns in terms of their methodology.
The authors use of drone measurements at a single point to infer NO emissions across a 5 km × 5 km grid box, which raises concerns regarding data adequacy, potentially introducing significant bias in estimating emission correction factors. Particularly, the NO and NO2 emission correction factors derived from DA_23SEP exhibit a 4 to 5-fold increase compared to DA_22SEP. The results seem to be counterintuitive as anthropogenic emissions typically exhibit small changes over two consecutive days, unless extreme events occur that lead to significant changes in NO and NO2 emissions. Even though the authors get ‘improved’ simulations after the assimilation, I assume the observed large differences in NOx concentrations would be more related to the daily variations in transport, either horizontally or vertically. In such cases, improving the representation of the meteorology of the model would likely be more beneficial than merely adjusting pollutant emissions. The authors might consider comparing meteorological parameters in their simulations with observations, particularly focusing on variables like winds, to identify potential discrepancies.
The calculated emission correction factors for NO and NO2 can be even larger than 15, suggesting the regional emission inventory they use in their simulation have an uncertainty of over 1400%, which might not be a reasonable value for emission correction. I suggest the authors checking their anthropogenic emission inventory and previous emission assimilation studies to identify a scientifically reasonable range for their emission correction.
Additionally, I am not sure whether the authors’ observational results would affect by the wind conditions. Both horizontal and vertical wind speeds can exceed the ascent rate (i.e., 1m/s) of their instrument, potentially affecting data collection. How did they keep their instrument ascending at a constant rate in a fixed location? NO is a highly reactive air pollutant which can be converted to NO2 quickly upon emission. Consequently, NO usually presents significant decreasing gradient in the vertical direction. However, the observed NO concentration at 350 m can double the surface NO concentration in the manuscript (e.g., F10, F11). The result is confusing. The authors may want to check the credibility of their data and present relevant explanations.
I recommend that the authors incorporate more observations, if possible, rather than relying solely on data from a single location, in their assimilation process. By doing so, they can enhance the robustness and credibility of their results, mitigating any potential suspicions regarding the validity of their findings.
Specific comments:
L162: I assume the authors may want to say ‘with data assimilation’ here or above (i.e., L158)? Otherwise, all the four experiments are conducted ‘without’ data assimilation, which is not consistent with what is shown in Table 2.
L218-219 I am not sure whether anthropogenic emissions can have such large differences in two consecutive days.
L227-228 As this is a model study, more quantitative analysis is expected.
Citation: https://doi.org/10.5194/egusphere-2024-517-RC2 -
AC2: 'Reply on RC2', Hassnae Erraji, 30 Jul 2024
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-517/egusphere-2024-517-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Hassnae Erraji, 30 Jul 2024
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
441 | 138 | 26 | 605 | 22 | 21 |
- HTML: 441
- PDF: 138
- XML: 26
- Total: 605
- BibTeX: 22
- EndNote: 21
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1