Sectoral attribution of greenhouse gas and pollutant emissions using multi-species eddy covariance on a tall tower in Zurich, Switzerland

Hilland, Rainer; Hashemi, Josh; Stagakis, Stavros; Brunner, Dominik; Constantin, Lionel; Kljun, Natascha; Molinier, Betty; Hammer, Samuel; Emmenegger, Lukas; Christen, Andreas

doi:https://doi.org/10.5194/egusphere-2025-1088

Preprints

https://doi.org/10.5194/egusphere-2025-1088

Preprints

24 Mar 2025

| 24 Mar 2025

Sectoral attribution of greenhouse gas and pollutant emissions using multi-species eddy covariance on a tall tower in Zurich, Switzerland

Rainer Hilland, Josh Hashemi, Stavros Stagakis, Dominik Brunner, Lionel Constantin, Natascha Kljun, Betty Molinier, Samuel Hammer, Lukas Emmenegger, and Andreas Christen

Abstract. Eddy covariance measurement of species that are co-emitted with CO₂, such as carbon monoxide (CO) and nitrogen oxides NO and NO₂ (NO_x), provides an opportunity to attribute a total measured net flux to individual source or sink categories. This work presents eight months of continuous simultaneous measurements of fluxes (F ) of CO₂, CO, NO_x, methane (CH₄), and nitrous oxide (N₂O) from an urban tall-tower (112 m agl) in Zurich, Switzerland. Median daily fluxes of F_CO₂ were 1.47x larger in the winter (Nov–Mar) as opposed to summer (Aug–Oct) months (10.9 vs. 7.4 μmol m⁻² s⁻¹); 1.08x greater for F_CO(30 vs. 28 nmol m⁻² s⁻¹); 1.08x greater for F_{NO_x} (14 vs. 13 nmol m⁻² s⁻¹); 1.01x greater for F_CH₄ (13.5 vs. 13.3 nmol m⁻²s⁻¹); and not statistically significantly different for F_N₂O. Flux ratios of F_CO/F_CO₂ and F_{NO_x} /F_CO₂are well characterised by inventory molar emission ratios of stationary combustion and road transport in cold months. In warm months both F_CO/F_CO₂and F_{NO_x} /F_CO₂systematically exceed expected inventory ratios during the day, while no statistically significant seasonal difference is observed in F_{NO_x} /F_CO, indicating biospheric photosynthetic activity. A linear mixing model is proposed and applied to attribute half-hourly F_CO₂ , F_CO, and F_{NO_x} to stationary combustion and road transport emission categories as well as determine the biospheric F_CO₂ . Flux attribution is reasonable at certain times and from certain wind directions, but over-attributes CO and NO_x fluxes to road traffic and CO₂ fluxes to stationary combustion, and overestimates photosynthetic CO₂uptake.

Received: 07 Mar 2025 – Discussion started: 24 Mar 2025

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Rainer Hilland, Josh Hashemi, Stavros Stagakis, Dominik Brunner, Lionel Constantin, Natascha Kljun, Betty Molinier, Samuel Hammer, Lukas Emmenegger, and Andreas Christen

Status: final response (author comments only)

RC1:
'Comment on egusphere-2025-1088', Martijn Pallandt, 18 Apr 2025
General
The authors set out to create a unique multispecies dataset of urban fluxes (comprising of CO2, CO, NOx CH4 and N20) to test a new method to partition urban GHG fluxes, specifically into stationary combustion, road transport, and biosphere components. Both this dataset and the methods are novel and of importance since there is a great need for tools to properly partition fluxes in heterogeneous environments. Overall, the writing is clear, the reasoning is sound and the manuscript follows a logical order. Two problems arise during this research. One being unrealistic low values in NOx measurements which are then scaled to local measurements. The other is that the novel partition method struggles in this high flux, high heterogeneity environment, resulting in a mixture of reasonable and unrealistic partition results. The authors are clear on these limitations and suggest linking this method to more components, flux footprints, and temporal explicit emission inventories as potential improvements. Even though the results aren’t entirely as envisioned at the start, they are still valuable and serve as an important stepping stone for further research on this or similar partition methods. As an in-situ case study this research sheds a light on seasonal and daily patterns both in observed and ratios of GHG fluxes in Zurich, and is able to characterize this city as a net source of CO2, CO, NOx, CH4 and N20. It was an interesting read and I'm curious how this research further develops.
Specific

20: Cities leading these efforts if quite a statement, the references here need some work or the claim needs to be modified: the European commission report doi 404ed on my side, the stad Zurich website probably has no place here ->“ Informal or so-called "grey" literature may only be referred to if there is no alternative from the formal literature.”(and this is just a website not an archived report). A quick scan of the Lwasa executive summary seems to point out that GHG emissions will very likely increase through cities modernization construction, and further urbanization, though they note there is a need for decarbonization for cities.
108: Is there a particular reason why these months were chosen? A clearer distinction between winter and summer would be observable if June, July and August were chosen instead. Related to this at line 156 /table 2: the winter period is considerably longer at 5 months vs 3 and probably with worse weather, it can be informative to indicate the distribution of (successful) samples between summer and winter.
165: It is unfortunate when equipment malfunctions, time dependent in-situ measurements as these cant be redone and you have to work with what you have. However, this section goes fairly quickly from noting that observations didn’t match expectations to scaling them up. Some essential steps appear to be missing.
Was the source of the error investigated, (e.g. was it a mechanical or a calibration issue), why did it only affect NOx fluxes, and have similar problems been reported with this instrument? Without identifying the source of the issue, it is hard to justify a correction.
The NOx flux was corrected based on local measurements, but why not perform an actual recalibration/correction with calibration gasses? That would give more certainty in its accuracy than a local measurement which adds several layers of uncertainty such as transport and their measurement errors. It would also allow for additional tests on any drift, to verify that the adjusted slope is stable over time.
I assume that the throughout the paper the adjusted NOx values are used as if they were the actual measurements, and the increased uncertainty was not propagated throughout the rest of the analysis. In either case, please state clearly how this was handled.
179: Why this 4x4 square and not the entire city or a larger part of the tower footprint?
Figure 3: Not essential but a 6^th panel in this style with temperatures would be interesting since these would be the main drivers in differences between winter and summer.
301: If you can conclude based on figure 5 that the ratio can be a mixture of road transport and stationary combustion, why not a combination of respiration and road transport?
350 / figure 6: This explanation mainly fits the winter period. Summer 2-8 is largely indistinguishable from 6-7 when I look at the figure, and the afternoon rush seems to start at 12 already. Summer seems different from winter here.
375 /figure 7A: The nocturnal fluxes would be the yellow area around y0.7 x1.3 just outside the dashed lines? Without a temporal component to this figure, it is not directly clear which the nocturnal NOx fluxes are. Can you clarify.
400 and 418: What can be done about this unrealistic allocation that goes well beyond measured quantities? While the section starting around 427 discusses improvements in general, can you give specific advice on preventing this problem in future cases? Maybe adding more components or limiting allocation to know maxima?
453-455: If I followed the equations in section 2.5 correctly al equations for the final partition components include the NOx term directly or indirectly. Therefore, wouldn’t this affect all ratios?
Did you at some point test the sensitivity of this model to FNOx? Even if you are not certain of your measurements it would give an indication of the impact of such errors.
456 - 465: I am somewhat surprised to find new methods and results after the discussion of the model’s limitations. And it is not clear to me what you have done here. These two paragraphs and table 6 can benefit greatly from a rewrite to clarify the method and intent. I would advise to add a subsection to the methods section describing this sensitivity analysis in a bit more detail, and discuss these results probably before ~427
Specifically:
Table 4 nor its description make mention of a and b. what is exactly being combined here?

In the section on the linear mixing model no mention is made of the mmol mol resolution, in which way is it different here?

You are testing the sensitivity of the linear mixing model to changes in what exactly? The discussion in the second paragraph also doesn’t make clear what sensitivity has been tested here. A typical sensitivity analysis tests a range of values of certain parameters.

Continuing with table 6: The first ‘% of total’ refers to table 3 where you have 5 categories and the second ‘relative’ to only the two listed in this table 6 (sc and rt)? Please clarify.

And the model outputs should be compared to the relative column since the model only uses the sc and rt categories? If the model outputs are per wind direction, it doesn’t make much sense to compare them to the general reference inventories, wouldn’t it be more interesting then to have 5 relative inventory references: all and the 4 directions?

In line 496 it is mentioned the sensitivity to reference inputs was tested here?

468 That would be a valuable continuation of this research, though if you continue with this dataset really aim to get a proper recalibration of the NOx data.
477-489: Since not everyone might look at the methods (in detail) a disclaimer on NOx uncertainty in the second or third paragraph is appropriate.
466: “complex and heterogeneous urban environment may ultimately pose too great a challenge for the application of such a model with fixed emission factors over long periods of time and large flux footprints.” And 501: “ however in the complex and heterogeneous urban environment this information is difficult to exploit on its own.”

This is unfortunate, since these are the environments where flux partitioning is especially important. In a homogeneous environment partitioning of fluxes is of lesser importance. Hopefully next steps will improve on this method further.

Technical
~427: not essential but you could put a section break here with everything after ~427 a discussion of the assumptions and where they are met/failed.
483: tower..
Citation: https://doi.org/10.5194/egusphere-2025-1088-RC1
- AC1: 'Reply to RC1', Rainer Hilland, 10 Aug 2025
  
  The authors wish to thank Martijn Pallandt (Reviewer Comment 1) for his constructive review. Our complete response to the reviewer's comments is provided in the attached pdf file. Due to significant changes to the text, not all changes are detailed here, and we direct the reviewer to the full tracked-changes version of the resubmitted manuscript.
  
  Citation: https://doi.org/10.5194/egusphere-2025-1088-AC1
RC2:
'Comment on egusphere-2025-1088. How feasible is it to measure turbulent fluxes of greenhouse gases and pollutant species at the city scale?', Erik Velasco, 07 May 2025

Eddy covariance is the best available method for measuring fluxes of non-reactive trace gases, including greenhouse gases and selected pollutant species, between the urban surface and overlying atmosphere. However, its application is complex and limited to urban areas relatively homogenous in terms of land cover, urban morphology, and emissions distribution. In this context, the work presented in this manuscript challenges the application of the EC method by installing an EC flux system on a tall building to measure emissions from the entire city of Zurich, Switzerland.
Fluxes of CO₂, CH₄, N₂O, CO, and NO_X were measured over eight months, three in the summer and five in the winter. The emissions source partitioning was determined using a system of equations based on the flux ratios of CO₂, CO, and NO_X. For this endeavor, emissions reported in an available bottom-up inventory were retrieved from a square area of 4 x 4 km, with the EC flux system in the center, and were considered correct in order to be used as a reference to solve such a system of equations.
The proposed approach makes sense, but there are a few issues that need to be addressed. The authors must demonstrate that the EC system was able to measure consistently turbulent fluxes, that they were representative of the targeted source area, and that the reported CO and NO_X fluxes were not affected by atmospheric chemistry.
1. Installing flux systems on very tall towers (> 100 m) may not be representative of the urban ecosystem. On one hand, tall towers may reach above the top of the collapsing boundary layer at night, causing difficulties in interpreting the flux data collected. On the other hand, if the EC instruments are installed above the upper extent of the inertial sublayer, the development of an internal boundary layer responding to regional-scale land use changes in an upwind region 100–300 times the measurement height may also affect the observed fluxes.
2. Representativeness problems may also arise when installing EC flux systems on very tall towers. First, the system would collect fluxes associated with eddies of different sizes and ages, with some originating before the beginning of the averaging flux period and possibly from beyond the city boundaries. Second, as shown by the footprint analysis, the system collected flux data beyond the square area of 4 x 4 km used as a study case.
3. Wouldn't it have been more appropriate, both in terms of theory and operation, to have installed the flux system on a much smaller platform if only an area of 4 x 4 km was to be evaluated?
4. It is necessary to prove the good performance of the closed-path system. First, the capacity of the MGA7 analyzer to measure the targeted species must be shown. Information on the calibration procedure is required. Second, the spectra and cospectra of the measured variables must be examined to ensure that the flux system was fully capable of measuring turbulent fluxes.
5. The EC method is limited to non-reactive species or species with low chemical reactivity, particularly in highly reactive atmospheres like those in urban areas. Measured fluxes of NO_x and CO, especially the former, may not represent emissions at surface level, since chemical reactions might drastically lower their abundance before reaching the height at which the EC instruments were installed. Consider the lifetime of these species by calculating their reactivity with the hydroxyl (OH) radical and the time it takes an air parcel to travel from the surface to the top of the tower. If OH data for Zurich are not available, use data reported for other cities to get a rough estimate (e.g., Price et al., 2025; https://doi.org/10.1038/s43247-025-02308-y). The time it takes an air parcel to reach the top of the tower can be determined using the standard deviation of the vertical wind speed fluctuations.
6. Eddy covariance flux towers are used to evaluate the accuracy of emissions estimated using bottom-up approaches, not the other way around. If the emissions reported in the inventory are presumed to be correct, what is the study’s purpose? This reviewer has used eddy covariance flux measurements to evaluate the accuracy of gridded emission inventories of CO₂ (Velasco et al., 2014; https://doi.org/10.1016/j.atmosenv.2014.08.018) and selected volatile organic compounds (Velasco et al., 2005, https://doi.org/10.1029/2005GL023356; Velasco et al., 2009, https://doi.org/10.5194/acp-9-7325-2009). These studies could provide insight on the use of flux towers for such a purpose.
Specific comments
P1, L2. Define all abbreviations and acronyms in the abstract and main text when they appear for the first time.
P1, L5. Does ‘1.47x’ mean 1.47 times the variable ‘x’? Write ‘1.5 times larger …’ Review the entire text and correct this spelling error.
P1, L5-8. The main text indicates that the differences observed between summer and winter were statistically significant; please also indicate this here. Include a variability metric.
P1, L19. … 70% of global energy-related GHG emissions … directly and/or indirectly?
P2, L22 Define bottom-up approach.
P2, L23. What do you mean by temporally coarse (yearly, monthly, daily, hourly)? Consider that many cities estimate gridded emissions for air quality forecasting on an hourly basis.
P2, L28. To avoid possible useless discussions, replace ‘the only method for’ with ‘the best available method’.
P2, L47. Provide more examples, as this reference only reports EC flux measurements in one city. Alternatively, refer to review papers on the EC application over urban surfaces: Feigenwinter et al., 2012 (https://doi.org/10.1007/978-94-007-2351-1_16) or Velasco et al., 2010 (https://doi.org/10.1111/j.1749-8198.2010.00384.x).
P2, L47. The EC flux method has also been used to measure fluxes of selected volatile organic compounds and aerosols to evaluate the accuracy of bottom-up emission inventories.
P2, L50. Carbon monoxide is a criteria pollutant for regulatory purposes, and therefore environmental agencies must report its ambient concentrations on an hourly basis.
P3, L67-74. You are listing what was done in this study, but you are missing the reason. State clearly the hypothesis to be tested and the questions to be answered.
P3, L75. A more detailed description of the monitored footprint is required, in this case, the entire city of Zurich. Provide information on building characteristics and energy consumption, indicate the types of industries and vehicle fleet characteristics, include population density, describe the city's vegetation, and provide tree density. Also, describe the climate throughout the year.
P3, L83-84. Was 17 m above roof level sufficiently high to prevent wind flows from obstructing turbulent flux measurements? Demonstrate that the building in which the system was mounted did not create disturbing flows.
P3, L84-87. But the footprint analysis indicates that the EC system measured fluxes from a much larger radius. In the best scenario, a 1.5 km radius will cover 30-40% of the observed source area. This reviewer does not believe that the characteristics of the district where the EC system was installed are representative of the overall footprint that was observed.
P4, L88. Include the instruments' accuracies and uncertainties as provided by manufacturers in Table 1.
P4, L88. How do the CO₂ concentrations and fluxes measured by the open-path and closed-path flux systems compare?
P4, L90-100. Were both instruments periodically calibrated? Failure to do so may result in a serious omission in the study.
P4, L103. FFP flux footprint?
P4, L103. Which source did you use to get the boundary layer height that the footprint model requests as input data?
P5, Fig. 1. The 30% footprint contour looks suspicious. There are two lines in the west sector, but none in the east sector.
P6, L110-112. Are there regulations in place about domestic heating according to the time of year?
P7, L126 Open-path instruments and sonic anemometers do not work during rain events. Thus, periods affected by rain must be excluded for further analysis.
P7, 131-134. How do you define a spike? See Schmid et al. 2000 (https://doi.org/10.1016/S0168-1923(00)00140-4).
P7, L135-137. Did the sampling tube and response time of the closed-path analyzer dampen the turbulent signal? Did you examine the spectra and cospectra of the measured variables to determine the influence of potential attenuations. You need to do it to demonstrate that the flux system is capable of measuring turbulence fluxes of trace gases via the EC method.
P7, L140. Explain what each flag indicates.
P7, L146-147. This is unfortunate. According to Figure 1, this sector covers the most urbanized area of Zurich.
P7, L154-155. How large was the storage flux?
P8, Table 2. Instead of listing the number of averaging periods, show the percentage of averaging periods used for further analysis for each climatological season analyzed. You may add the percentage of periods removed due to instrument problems and calibration, rain events, or failure to reach the flag 0 and u* thresholds. Note, periods with low u* usually do not meet the stationarity test.
P8, L156-158. The study covered three months in late summer and early autumn and five months during the winter. Furthermore, 40-50% of the averaging periods were excluded for further analysis due to quality issues. Was the amount of data collected sufficient to draw conclusive results? Could the difference between data collected in summer and winter be a determining factor in drawing conclusions?
P8, L167-170. Check the method used by the NO₂ monitor used as a reference. Measurements of NO₂ from standard chemiluminescence monitors equipped with molybdenum oxide converters may suffer from interferences and end up reporting concentrations exceeding up to 50%. See Dunlea et al. 2007 (https://doi.org/10.5194/acp-7-2691-2007).
P8, L175. What is the temporal resolution of the emissions inventory (hours, days, months, years)?
P9, Table 3. Use Mg as a unit. Tonnes are not accepted by the International System of Units.
P9, L179. I would say the 4 x 4 km grid cell designated to investigate flux partitioning covers no more than 60% of the observed source area. Now, if the study was limited to that area of the city, was it necessary to install such a tall tower? A much smaller flux tower would have been more effective and would have covered the target area much better.
P9, L181. Perhaps ‘anthropogenic CO₂ emissions’ instead of ‘non-respiration CO₂ emissions’.
P10, L200. Do these ratios imply that the emissions estimates in the inventory are essentially correct?
P10, L205-210. You need to account for NO_X lifetime and reactivity throughout the diurnal course.
P12, Fig. 3. Were there differences between weekdays and weekends? The analysis must account for emission variations based on the day of the week. Studies in other cities have found substantial differences.
P12, Section 3. This section is lengthy and tough to read. Many of the results might be better presented in a table or figure. Avoid filling the text with numbers.
P12, Section 3.1. Add a variability metric to all figures.
P13, L247-249. Why?
P13, L260-261. Why?
P14, Table 5. Add a variability metric. You may additionally provide observed fluxes based on wind sector.
P14, Fig. 4. Nice figure!
P15, L290-291. This is not true for the case of reactive species such as NO_X and CO.
P15, L299-302. This reviewer does not find such a strong agreement when considering Fig. 5 as a reference. He observes an important variability in the measured ratios and a significant difference against those derived from the emissions inventory. To arrive at this statement, it is necessary to provide the hourly ratios obtained from the emissions inventory.
P16, L327-328. Similar to the previous comment. Comparing hourly ratios from field observations with daily ratios from the emissions inventory does not lead to this conclusion.
P17, L355. Perhaps ‘seasonal changes in FCO₂’.
P18, Fig 7b. A panel may be presented for each case. It is a bit confusing to identify the histogram corresponding to each ratio and season.
P19, L387-389. The type and characteristics of the vegetation determine this. Is Zurich's vegetation similar to that in Indianapolis, USA?
P19, L394-396. Are these differences significant compared to those reported in the emissions inventory? If there is a difference, might it be due to differences in the accounted footprint or modeled area?
P20, Fig. 8b. For NO_X, the stationary combustion sector's contribution looks suspiciously low compared to the total flux of CO and CO₂.
P20, Fig. 8c. The biogenic CO₂ uptake during winter looks suspiciously intense. Is Zurich's vegetation evergreen?
P21, L421-426. Could this be related to differences between the flux tower’s footprint and the 4 x 4 km area extracted from the emissions inventory?
P21, L431. This assumption was clearly not met using such a tall EC flux tower.
P21, L432. Please provide examples of 'biospheric' emission sources of CO and NO_x within Zurich's built-up area.
P22, Section 4. Conclusion should not be a results summary. You may add a section summarizing main findings.
P23, Table 6. Are these results for the entire 8-month measurement period, or just the summer or winter periods?
P23, Table 6. Explain the potential reasons for the negative CO and NOx fluxes reported in this table.
P24, L499-506. This is the type of statement that should be in the conclusion section.

Citation: https://doi.org/10.5194/egusphere-2025-1088-RC2
- AC2: 'Reply to RC2', Rainer Hilland, 10 Aug 2025
  
  The authors wish to thank Erik Velasco (reviewer comment 2) for this thorough and critical review. Our complete response to all comments is provided in the attached pdf file. Due to significant changes to the text, not all changes are detailed here, and we direct the reviewer to the full tracked-changes version of the resubmitted manuscript.
  
  Citation: https://doi.org/10.5194/egusphere-2025-1088-AC2

Rainer Hilland, Josh Hashemi, Stavros Stagakis, Dominik Brunner, Lionel Constantin, Natascha Kljun, Betty Molinier, Samuel Hammer, Lukas Emmenegger, and Andreas Christen

Viewed

Total article views: 838 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
727	93	18	838	15	36

HTML: 727
PDF: 93
XML: 18
Total: 838
BibTeX: 15
EndNote: 36

Views and downloads (calculated since 24 Mar 2025)

Month	HTML	PDF	XML	Total
Mar 2025	155	28	3	186
Apr 2025	115	28	5	148
May 2025	68	5	2	75
Jun 2025	63	11	4	78
Jul 2025	53	9	2	64
Aug 2025	171	7	1	179
Sep 2025	102	5	1	108

Cumulative views and downloads (calculated since 24 Mar 2025)

Month	HTML	PDF	XML	Total
Mar 2025	155	28	3	186
Apr 2025	115	28	5	148
May 2025	68	5	2	75
Jun 2025	63	11	4	78
Jul 2025	53	9	2	64
Aug 2025	171	7	1	179
Sep 2025	102	5	1	108

Viewed (geographical distribution)

Total article views: 826 (including HTML, PDF, and XML) Thereof 826 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 14 Sep 2025

Short summary

We present a study of simultaneously measured fluxes of carbon dioxide (CO₂) and co-emitted species in the city of Zurich. Flux measurements of CO₂alone can’t be attributed to specific emission sectors, such as road transport or residential heating. We present a model which uses the measured ratios of CO₂ to carbon monoxide (CO) and nitrogen oxides (NO_x) as well as sector-specific reference ratios, to attribute measured fluxes to their emission sectors.


Total:	0
HTML:	0
PDF:	0
XML:	0