the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Tall tower eddy covariance campaign reveals complex emission patterns for carbonyl sulfide
Abstract. Carbonyl sulfide (COS) is a useful tracer of gross primary productivity (GPP), but the magnitude and variability of anthropogenic COS emissions in urban environments remain poorly constrained.
COS fluxes were investigated in Zurich, Switzerland, using tall-tower eddy covariance (EC) measurements as part of the ICOS Cities project from August 2022 to April 2023. Source-area variability was analysed using the Flux Footprint Prediction (FFP) model, and compared with biogenic COS fluxes simulated by the Simple Biosphere 4 (SiB4) model.
Zurich acted as a net sink of COS throughout most of the campaign (median −4.4 pmol m-2 s-1). The most vegetated south-western sector exhibited approximately three times the COS uptake of the south-eastern sector (−6.7 and −1.9 pmol m-2 s-1, respectively), while the south-eastern sector exhibited the weakest uptake, consistent with stronger anthropogenic fluxes. Footprint-weighted land-cover analysis showed that spatial variability in COS exchange reflected the balance between vegetation uptake and anthropogenic fluxes, with the latter masking the biospheric signal in densely built sectors. SiB4 reproduced the seasonal evolution of COS exchange but underestimated the sink following the preceding compound drought event and during an exceptionally warm winter, highlighting limitations in representing stomatal conductance and urban vegetation.
These results demonstrate both the potential and the challenges of using COS as a tracer of urban GPP. Improved anthropogenic COS emission inventories, ecosystem models, and urban flux observations are needed to improve COS-based estimates of urban carbon uptake.
- Preprint
(10362 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 11 Dec 2026)
- RC1: 'Comment on egusphere-2026-4603', Anonymous Referee #1, 07 Oct 2026 reply
-
RC2: 'Comment on egusphere-2026-4603', M.E. Whelan, 29 Oct 2026
reply
Dear Jesse Soininen and others,
I’m glad to see the effort in making carbonyl sulfide measurements in an urban setting; however, this manuscript needs to be rewritten to emphasize the useful ideas that emerged from the observations for planning future urban OCS campaigns.
This manuscript includes a substantial amount of work dedicated to interpreting a dataset that runs from the end of Northern hemisphere summer (August 2022) to spring (April 2023) over Zurich. The measurements from August to early November lack an instrument correction that allows for generating mole fraction estimates.
We expect highest OCS and CO2 uptake (and also CO2 respiration) during periods of high GPP in the summer in Zurich. Some of the authors already wrote up a convincing application of OCS observations to interpret the CO2 balance over Helsinki in 2025. But here we do not have the data to answer questions about the relative size of the summertime biosphere fluxes versus anthropogenic CO2 emission. So what is interesting/surprising/important about this study? There are many options to pursue and here is just one suggestion.
This study could be rewritten to create city-specific anthropogenic OCS estimates at the appropriate resolution for the analysis. This would include an examination of the influence of tire wear (almost zero?) and natural gas (clearly present) on observed OCS emissions.
It is clear that the Zumkehr et al., (2018) global anthropogenic dataset does not serve this study. It’s for a different year, has a course resolution compared to the sector-based analysis used here, and suggests much higher fluxes than observed. It was mentioned that there was no natural gas source in the Zumkehr database. Serendipitously, this study contains actual observations of natural gas burning. Can they be used to your advantage to figure out the influence of natural gas burning on the urban OCS observations?
I myself have tried to measure OCS emissions from tires directly and wasn’t able to see a signal with an Aerodyne instrument. It could be that tire manufacturing has changed (that initial study was in 1993) or maybe there is variability in tire composition between regions. Can we ask the question: how high are OCS emissions from tire wear likely to be based on these observations?
The treatment of SiB4 here is similarly asking too much from a model output that isn’t set up to study an urban area. The resolution of the model is coarse (0.5 degree gridcells) compared to the observation footprints and isn’t using the same land surface dataset as the rest of the analysis. One would hope that identifying model-data discrepancies can be used to improve the model or highlight surprises in the data. Here there seems to be a mismatch in how the output is being treated. While SiB4 is one of the few land surface models that generates OCS model fluxes, running a less complicated model at a more useful resolution (like pyVPRM which is involved in the ICOS effort) and examining modeled GPP might be more useful.
Even without substantial changes, this analysis has an interesting story to tell with reworking the written text. In winter, we find OCS uptake is highest when the tower observes the SW sector with the most green areas and the lowest over the densely built SE sector. This is saying that while we would expect low GPP in the winter, something continues to take up OCS. This is a useful finding. Do we know it is plants taking up OCS? The correlation and anticorrelation of CO with roads and trees, respectively, is also heartening. The sector analysis here is a useful approach, but it seems like the correlation matrices do not add substantially to the narrative of the work. With revision, this will be an excellent scientific contribution.
Minor comments are included below in case they prove useful after broader scale revisions. I look forward to the publication of this study.
Solidarity,
Mary Whelan
Minor Comments
L85 Describe the white lines and the white cross in the caption for Fig 1.
L94 This mention of things in the West is a little confusing in terms of interpreting data. The West is included in both the SW and NW sectors.
L153 One would expect there to be a change in storage for a tall tower over time except for high turbulence conditions. One way to assess the suitability of the method would be to compare to the method that relies on concentration measurements at multiple heights below the EC instrumentation - but we don’t have that here. So what is there to do? It would be interesting to see a plot of Fsto with u*.
L154 This makes it sound like you only have inaccurate mole fraction measurements before November. If this is the case, we have no measurements before November. This is fine, we don’t have to mention them in this analysis.
L157 Why do we need the data before the background was run? It seems like if we don’t have a good handle on Ftot, we should leave these observations out of the analysis. It’s not like this data does not have an interesting story if we leave out data from before the calibration or “background” system was up and running.
L166 Well done dealing with this inconvenient chimney.
L181 Not having used FFP before, I’m not sure what the ABL flag is. Can you please describe it?
L203 SiB4 was run at 0.5 degrees, which I think is around 40 km at the latitude of Zurich. This seems mismatched compared to the resolution of your landuse map and footprints.
L237 Given that we’re only looking at a single 0.5 by 0.5 degree cell of SiB4, it seems hard to make a comment on the performance of the tower measurement procedure versus the model. The tower doesn’t typically see the entire cell of SiB4 and so we’re typically always experiencing the (c) condition. Unless I am missing something. Also, why are we using Fmeas?
L259 It is noted much later that the Zumkehr et al (2018) dataset has missing sources. It is important to note that this particular dataset will not be updated beyond 2012. Instead of comparing a database we know to have limitations, can we somehow build on the critique of Belviso et al., 2023?
L284 Say where the measurements were taken (can refer to Fig 1). “Compound event” could mean many things - we can say “compound soil and atmospheric drought event” to be clear.
L297 It is unfortunate that we don’t have storage values for summer time observations, when we expect OCS and CO2 uptake to be highest.
Including the measured CO2 fluxes to the SiB4 fluxes is just a little confusing because you expect anthropogenic fluxes of CO2 that SiB4 does not model.
Finally, the gray lines don’t look dashed?
L306-311 For Figure A4, it’s ambiguous whether we are seeing only the chimney flag in orange, or if we are seeing the “influence of combustion on the fluxes” -which would be the flag 3 data incorporated into the average diurnal cycle. The latter would be a little problematic because the influence of the chimney would be mixed up with the frequency with which flag 3 occurs.
L324 The fluxes should be seasonal. Are these diurnal fluxes averaged over all months?
L394-395 This conclusion could be the title of this section, or the first sentence. It helps understand where the analysis is going at the beginning.
L397 Comparing the SiB4 flux to the observed OCS flux here is confusing. The SiB4 flux covers the entire area, whereas we have already seen there is variability in the observations based on which way the wind is blowing. In other words, the observations never actually cover the whole area that the SiB4 model is attempting to capture.
L437-439 Why were the measurements declared trustworthy? I have used these QCLs - I am not doubting that the measurements are okay. More words probably need to be used here.
L455-456 This needs its own paragraph. Looking at the difference between weekdays and weekends has been a handy tool for CO2 analysis in cities. If these observations cannot see that, the hypothesis here is that the surrounding forest is dominating the fluxes. This means that the tower height cannot see the influence of the urban biosphere flux? Can you see the expected weekday/weekend difference in CO2?
L464-465 Could it be that the OCS uptake was suppressed during winter and there wasn’t enough OCS uptake to make a GPP estimate?
L495 Except for deserts, any soil when hot and dry will emit OCS. Agricultural soil will emit OCS that rivals plant OCS uptake. Who’s to say if urban soil is more like agricultural soil? For interpreting the August emissions, there is a straightforward empirical soil OCS model that could be run with only soil moisture and soil temperature as inputs (https://doi.org/10.1029/2022JG006858). Full disclosure: this is my model. It provides a first estimate of soil fluxes.
L506 I’m not sure how the snow conditions of the Finnish forest explain the Zurich fluxes.
L529-L534 The way this is presented, it sounds like we are saying it is either uptake by soils or uptake by vegetation. It’s likely both. The last sentence here is stating the motivation of the entire exercise.
L540 Lowest COS emissions? Or lowest net COS fluxes. There is likely uptake and emission happening simultaneously.
L569 Here is where a conversation about tire wear might be appropriate.
L639-645 We want to use the OCS tracer here to tease apart CO2 uptake. There isn’t a great deal of CO2 uptake at night when the atmospheric dynamics make the measurement more difficult. From my perspective, the reason to do COS measurements in Zurich at night is to see possible anthropogenic emissions.
L652 Would a recommendation that multiple towers are necessary be appropriate here? Also, there are many ways to calculate tower footprints. Perhaps a different approach (with higher resolution meteorological data that takes into account topography) would be more effective here.
L660-669 See my comments about SiB4 above.
Fig A9 needs more of a caption.
Citation: https://doi.org/10.5194/egusphere-2026-4603-RC2 -
RC3: 'Comment on egusphere-2026-4603', Anonymous Referee #3, 10 Nov 2026
reply
This manuscript presents tall-tower observations of COS, CO₂, and CO exchange in Zurich, combining them with footprint analysis and SiB4 simulations. Urban COS observations are valuable for understanding anthropogenic influences on the COS budget and for assessing the potential of COS as a tracer for photosynthetic uptake. Therefore, this dataset has the potential to contribute to the journal of Atmospheric Chemistry and Physics.
However, the manuscript currently contains extensive descriptions of the observations but lacks sufficiently detailed quantitative analysis to establish the processes responsible for the reported patterns. The manuscript documents numerous seasonal, diurnal, and sectoral patterns, but the scientific advancement beyond simply describing these patterns is not sufficiently established. Much of the discussion proposes explanations based on analogies with previous studies, rather than testing those explanations against the current dataset. Consequently, several sections read more like a literature-based discussion of possible mechanisms than an analysis demonstrating which mechanisms actually operate at this site. The main concerns are the unclear connection between the research questions and the analyses, the insufficient treatment of measurement and sampling uncertainty, and the interpretation of model–observation differences as anthropogenic emissions despite significant spatial and temporal mismatches. Therefore, I recommend a major revision.
Specific Comments
The introduction spends considerable space on listing partitioning methods, but it does not clearly explain which unresolved limitation motivates this study or how the proposed analysis addresses it. The methodological details tend to obscure the scientific rationale.The title mentions only COS, whereas substantial portions of the manuscript describe CO₂ and CO. A COS-focused title is reasonable if the other gases provide supporting evidence, but their roles need to be made explicit.
Lines 22–23: The assertion that Eddy Covariance (EC) is the only observational method for studying ecosystem-scale CO₂ exchange is too categorical. Other micrometeorological approaches, including flux-gradient methods, can estimate exchange under appropriate assumptions. Please describe the specific advantage of EC more precisely. Also, distinguish the turbulent flux measured by EC from the storage term estimated from concentration changes; the latter is a separate component of the control-volume budget, not a turbulent EC measurement.
Line ~260: It is unclear how the footprint domain overlaps with the four inventory cells.
Figure A4: The caption mentions CH₄ and panels (g) and (h), although neither appears in the figure. Please rewrite the caption to match the six displayed panels and three gases.
Citation: https://doi.org/10.5194/egusphere-2026-4603-RC3
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 360 | 141 | 84 | 585 | 82 | 68 |
- HTML: 360
- PDF: 141
- XML: 84
- Total: 585
- BibTeX: 82
- EndNote: 68
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
Thank you for investigating the potential use of COS as a proxy for GPP in urban areas, where anthropogenic sources can play a significant role. This is a highly important topic, as only a limited number of measurements have been conducted to date that cover and attempt to disentangle the different sources and sinks present in urban environments.
The manuscript is very well written and detailed. The advantages and disadvantages of using COS as a proxy in urban areas are well discussed. However, in my view, the implications and potential applications of using COS as a proxy for GPP estimation could be highlighted more clearly.
Furthermore, I only have a few minor comments, which are listed below.
Title: “tall tower”, how important is this? would Eddy covariance be enough. now the fact that you used a tall tower looks crucial since it is in the title and at the front.
L23-24: please add a reference
L37: “The potential of COS to track CO2 uptake is based on the similar pathway to the leaf through the leaf stomata.” This is a difficult sentence, could you rephrase it?
L40: “destroyed”, hydrolysed would be more precise I believe
L98: were there no condensation problems using an unheated line?
L100: if the measurements are not used, no need to write it down unless you want to specify they are available for downloading
L113: also here please focus on what was used for the specific experiment, could become unnecessary complicated otherwise.
L175: a natural question comes up when reading this section, which flag was used in the analysis? This will be answered later on but maybe also stating here could be helpful.
L312: Fcos_meas is discussed after the model results but in the figure legend and description it is labelled first. Please keep same orders.
Fig 6. Could significance be added?
L533: this connection to GPP is very nice, if possible, more of those implications connections between your findings and GPP could be made.