|Review of Wenger et al., Atmospheric radiocarbon measurements to quantify CO2 emissions in the UK from 2014 to 2015|
This paper describes a set of radiocarbon (14C) measurements from two sites in Ireland and the UK, and uses these measurements to determine recently added fossil fuel CO2 (ffCO2). Carbon monoxide (CO) from the same sites is also measured, and the potential and challenges of using CO as a ffCO2 tracer are considered. Transport model simulations are performed for all 3 C isotopes in CO2 and these model results are used to (a) diagnose the influence of nuclear industry 14C production and heterotrophic respiration fluxes on 14C and hence the calculated ffCO2 values for this region and (b) compare the modelled and calculated ffCO2 and CO values. The results demonstrate that although previous research has shown that nuclear industry 14C emissions can be problematic for ffCO2 studies in the UK, many parts of the UK are not overly influenced by this problem.
I reviewed an earlier version of this paper. Thank you for the exceptional improvement over the previous version. This revision is clear and easy to follow. It is now apparent that the overall concept and strategy are strong, and the paper is appropriate for publication in ACP. I have a number of comments, but overall only minor revisions are needed before publication.
Line 27. Core Writing Team reference – I think this is a reference to the most recent IPCC Working Group I document, but the reference is incomplete.
Lines 44-46. There are many more studies where 14CO2 has been used to estimate ffCO2, many of them in places other than just those listed, including in Asia.
Line 63. “as each source emits with a different CO:ffCO2 ratio. Please reference this statement.
Lines 78-81 and lines 91-95. So the flasks are collected from the 185 m height, but the in situ CO observations are from a lower height. It seems possible that the difference in CO mole fraction between flasks and in situ could simply be due to the difference in sampling height rather than a scale issue. Please comment on this.
Is there any reason to suspect that the in situ – flask offset could indicate a real problem with CO in the flasks, or are you confident that the offset is only a scale issue? If the former, then is there any possibility that other species from the same flasks could also have a problem? Please clarify.
In general, I would have thought that it would be better to use the CO from the same flasks/height as the 14C measurement, since they are being used together. How would using the flask CO data (rather than the in situ data) change the results and interpretation?
Line 83. Miller et al 2012 used free tropospheric measurements from the same aircraft flights as background, not upwind sites. Be careful about calling sites “unpolluted”, as all sites will be influenced by local or regional pollution to some extent.
Line 85. I believe you mean Turnbull et al 2015, not 2014.
Lines 134-135. Please edit for grammar.
Lines 153-155. It is not strictly correct that fractionation discriminates 14C twice as much as 13C . Farnhi et al (2017) discuss this in detail and show that the approximation used in S&P 1977 is sufficient given the current uncertainties in the 14C measurement.
Fahrni, S. M.; Southon, J. R.; Santos, G. M.; Palstra, S. W. L.; Meijer, H. A. J.; Xu, X., Reassessment of the 13 C/ 12 C and 14 C/ 12 C isotopic fractionation ratio and its impact on high-precision radiocarbon dating. Geochimica et Cosmochimica Acta 2017, 213, 330-345.
Line 154. I think you mean 14C, not 14C!
Normalising “should” remove reservoir specific differences? Are you suggesting that it might not be effective? If so, please explain.
Line 173. Please discuss a little more about how the background was calculated. Later you indicate the the background uncertainty for 14C is ~4‰, which is much higher than the 14C measurement uncertainty which is quoted for the TAC site measurements. Were the MHD measurements done to lower precision, or is uncertainty calculated in some other way?
Please add a figure (perhaps in the supplementary material) that shows the MHD 14C data and the median 14C values that were used as background.
Line 178. Suggest adding a brief note about why only nuclear and heterotrophic respiration are considered, and not other sources such as the ocean.
Line 205 and line 217. Why is 10^15 ‰ first mentioned, then 7.3x 10^14 then used? And please explain where this value comes from.
Line 231. Please include (in the supplement) a table of the 14C values and sampling info (lab numbers, time/date sampled, etc)
Line 231. “Daily mean values” are these 24 hour means, or are they only daytime? If only part of the day, please indicate which hours of the day are used.
What time of day were the flask samples taken, and does that match with the modelled time of day used?
Lines 240-246. How does the uncertainty in 13C play out in terms of the 14C measurements that are the focus of this paper? Does it matter?
Line 249. See previous comment about uncertainty in background 14C values.
Line 266. Clarify that “ffCO2 equivalent” is the correction terms in equation 4.
Lines 276 – 279. From figure 6, it appears that the nuclear correction is 100-200% larger than the ffCO2 value at the TAC site, yet the text implies (but doesn’t explicitly say) that it is a much smaller correction. Please add some more detail about the relative contribution of the nuclear (and biosphere) corrections to the ffCO2 calculation, and the implications for the reliability of ffCO2 for this site and the UK.
It would be helpful to show the ffCO2 values calculated for each sample – this could be an additional panel in either figure 2 or figure 3 or a separate figure (it would be nice to show the time series of ffCO2 from the model as well). Currently the only place ffCO2 values are shown is in figure 4, which is useful but we can’t tell which point relates to which correction in figure 3.
Line 288. Is there a typo in the boiler inspection date? Seems like it should be 10th June 2014, not 10th July 2014 – the July date couldn’t have caused a problem at TAC on June 13th!
Lines 295-300. Does the model include emissions from continental Europe, and if so, what is the quality of those emission estimates? I’m wondering if the problem is model transport during this low wind speed period, or if there is an issue with the emissions from this region as well?
Line 305. Explain and/or reference the 1 ppm to 2.5 ‰ relationship.
Lines 308-309. You say that the observed ffCO2 uncertainty is relatively large while the UK ffCO2 emissions are relatively low. I think you mean that the ffCO2 mole fractions (not emissions themselves) are relatively low – ie signal to noise is poor.
Line 313-314. Please reference the previous work on CO:ffCO2 ratios and their variability.
Line 315. Do diesel cars really have that low an emission ratio? And what about petrol cars? There are a number of recent studies that show onroad CO:ffCO2 ratios of around 5-15 ppb/ppm (depending on the country, emissions controls, etc).
Lines 321-327. The fit in figure 5 looks to be strongly constrained by the two high points, and if they were excluded it looks like you would get a much higher ratio. Maybe those two points are samples where there was a strong local influence in the sample (e.g. a car idling nearby)? Further, curve fits to noisy data, and data with uncertainty in both axes is tricky. What kind of linear regression was used; were the ffCO2 uncertainties accounted for in the regression; how would excluding the two high points change the slope and it’s interpretation?
Lines 328-329. Reference this statement.
Lines 336-340. This argument that variability in the CO:ffCO2 ratio is due to variability in traffic CO:ffCO2 ratios needs some more justification. There are quite a few studies of onroad emission ratios that could be referred to. Those studies show that indeed, individual vehicles do vary considerably in their CO:ffCO2 emission ratio, but it is not clear that the variability in individual vehicles translates to variability in tower measurements where the traffic signal is a mix of many, many vehicles.
Lines 340 – 344. If I understand correctly, the CO:ffCO2 ratio was derived from the 14CO2 measurements and CO measurements taken at the same time at the TAC site (figure 5). Then that CO:ffCO2 ratio is applied to the full time series of in situ CO measurements. The agreement between the model and observed ffCO2 is pretty decent (figure 4). So the results in figure S8 showing that the CO derived ffCO2 time series doesn’t agree with the model seems at odds with everything else. This needs some more discussion and explanation. Some thoughts:
(a) the CO:ffCO2 ratio shown in figure 5 is skewed way too low by those two high points that might be locally influenced. If they were excluded, you’d get a higher ratio and therefore the CO derived ffCO2 time series would have a smaller magnitude and match the model better.
(b) Does figure S8 show just the daytime when (presumably) flasks were sampled, or does it include nighttime data? I’d be surprised if the model does a good job at night, so if nighttime data is shown it could be confusing things.
Liens 348-350. As in a previous comment, figure 6 seems to show a very high contribution from the nuclear industry, whereas the text asserts that the influence is small. This needs to be clarified!
Line 358. In this paper, the flask samples were very short grab samples, and Turnbull et el 2012 used a 1 hour integrated sample. So where does the 3 hour integration come from? What’s the justification for choosing 3 hours vs 1 hour or some other time period? This paper doesn’t make any comparison of integrated samples vs grab samples, so I don’t see how the outcomes of THIS study can lead to a recommendation that integrated samples are better – although it is probably true that integrated samples have advantages in many situations.
Lines 360-362. What is the justification for suggesting conditional sampling? As in the previous comment, it’s probably useful to do that, but THIS study doesn’t add any new information about the usefulness of doing so. It’s worth noting that conditional sampling might be a great idea, but it is also likely to be much more difficult to do!
Lines 364-369. This should first be presented in the methods and results, not just in the discussion section.
Figure 6 only shows the ratio of nuclear correction to ffCO2. Whereas the magnitude of ffCO2 will also be important in considering site locations, so that the signal is large enough to be measurable.
Lines 370-374. This study doesn’t seem to provide any evidence that the method would work better at the city scale. Further, if discussing London, please indicate London’s location in figure 6.
Comment on grammar and language. I noted several very minor typos and grammar issues as I read. I have not commented on them all since they are likely to change in revision, but I suggest a good check through for such issues before resubmission.
Comment on authorship. Have the current authors considered whether the NOAA and INSTAAR scientists who contributed the 14C and other flask measurements should be included as authors on this paper?