the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Aug 2024
Dual-tracer constraints on the Inverse-Gaussian Transit-time distribution improve the estimation of watermass ages and their temporal trends in the tropical thermocline
Abstract. Quantifying the mean state and temporal change of seawater age is crucial for understanding the role of ocean circulation and its change in the climate system. One commonly used technique to estimate the water age is the Inverse Gaussian Transit Time Distribution method (IG-TTD), which applies measurements of transient abiotic tracers like chlorofluorocarbon 12 (CFC-12). Here we use an Earth system model to evaluate how accurately the IG-TTD method infers the mean state and temporal change of true water age from 1981 to 2015 in the tropical thermocline (on isopycnal layer σ0=25.5 kg ⋅ m-3). To this end, we compared the mean age of IG-TTD (Γ) derived from simulated CFC-12 with the model "truth", the simulated ideal age. Results show that Γ underestimates the ideal age of 46.0 years by up to 50 %. We suggest that this discrepancy can be attributed to imperfect assumptions about the shapes of transit-time distribution of water parcels in the tropics and the short atmospheric history of CFC-12. As for the temporal change of seawater age, when only one transient tracer (CFC-12) is available, Γ might be an unreliable indicator and may even be of opposite sign to trends of it due to uncertainties of mixing ratio. The disparity between Γ and ideal age temporal trends can be significantly reduced by incorporating an additional abiotic tracer with a different temporal evolution, which we show by constraining Γ with sulfur hexafluoride (SF6) in addition to CFC-12.
- Preprint
(13031 KB) - Metadata XML
-
Supplement
(3830 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-2552', Rolf Sonnerup, 07 Oct 2024
This is a fine effort that provides guidance on how realistic changes in ventilation inferred from repeated CFC (and SF6) sections can be, and thus this can be an important contribution. It also focusses on the tropics, rather than the usual subtropical gyres. The work is careful and well focussed. I applaud the authors for their focus on saturation level uncertainties and trends. That said, I was disappointed that there was not much effort devoted to real-world scenarios. Researchers don't go out and measure a mean CFC-12 and a mean SF6 on a global isopycnal, they go out and measure discrete water samples. From those discrete samples, they attempt, perhaps optimistically, to infer basin-wide patterns of changes in ventilation. I suspect an 'upgrade' is easily within reach for these authors - choosing a location or two (or a section or two) and explicitly calculating the change in ventilation inferred from tracers as compared with the actual change. I would not insist on this happening - the work is fine as it is - but it may improve the impact of their paper.
Comments that the authors should consider and which (in my opinion) would improve the ms. markedly:
1). Line 153. Consider rewording this. The lessons from Waugh et al 2003 and from Stoeven et al 2015 are that CFCs and SF6 only provide meaningful constraints on Delta/Gamma when <= 1.8 or so. Those papers do not indicate that the real Delta/Gamma in the ocean is <= 1.8, merely that the tracer pair doesn’t provide useful information >= 1.8. Delta/Gamma could be 3, it could be 6, but this tracer combination wouldn’t tell us that, and wouldn't be able to distinguish 3 from 6.
2) Figs 2, 4 &5. It is very hard to see the patters described in the text. This paper focusses on the tropics. One way to improve visibility would be to remove latitudes > 40 (maybe even 35?)
Line 222-223: I cannot see this pattern in Figure 5. Perhaps re-scaling the figures will help?
3) Relatedly. Fig. 3b. Do these results vary by basin? The processes and pathways bringing tracers to the eastern tropical Pacific are quite different from, say, the Arabian Sea. This figure shows the mean.
I have the same comment about Figure 6b and 7b. Okay maybe the twin-tracer TTDs can do a good job representing changes in the global mean ventilation of this isopycnal. But people don’t go out and measure a global mean pSF6 and pCFC-12, they have section measurements from which they hope to derive larger-scale conclusions. How would these three figures look at some individual, spot locations or sections? Do these conclusions, which hold for the isopycnal mean, apply universally in individual regions?
4). Figure 4. It would be very useful and interesting to see a figure of the percent mismatch of at least one of these (your choice) from the reference, figure 2a.
5). Line 233 – This is very similar to the findings of Mecking et al 2004 based on simple pCFC ages in the early 1990s– and to the findings of Thiele and Sarmiento (1990) based on pCFC ratio ages in the 1980s. Some attribution to these pioneers, somewhere in this paper, would seem appropriate.
And some
Suggested, minor modifications, by line number
9. uncertainties in the mixing ratio
26-28. One citation will suffice – I’d recommend Jenkins 1980 and Weiss et al 1985
30. Gruber 1998
34. Doney and Bullister 1992 rather than Fine 2011
36 … individual transit times is that which results from one-dimensional transport and mixing, the Inverse Gaussian…
51 most of the studies focus on the subtropical gyres, and very few focus on the tropics.
60 … represent ideal age and temporal changes in ideal age…. (cumbersome but clearer)
147 …from measurement of a single transient tracer
159 - Tricky – the saturation level mis-assumption causes the derived age to be older than it should, however it is still younger than the ideal age (or, water age)
Maybe try wording like
…causes the tracer-derived age to be older than with a more realistic, undersaturated, boundary condition.
160 Shao et al 2013 should be cited here as well.
162 …rise in saturation levels, TTD-derived water ages range from being…
192 …calculation, so fixed spatially homogeneous…
205 More precisely, the spatial variance of Gamma increases with higher Delta/Gamma, with 1.2 being most representative of the spatial variation in ideal age.
(If I’m interpreting the Fig. 3a correctly, ignore this remark if I’m not)
251-252. …along outcrops and in the Atlantic, Pacific and Indian Ocean ranges from 80% to 100% (Fig. S3).
262 This is an important result. I would start the sentence ‘Notably, the assumption…’
278 An even more important result. Possibly clarify a bit …our results suggest that Gamma constrained by CFC-12 alone is not able…
325 Does this also mean that ‘ages’ from those concentrations would be about half the ideal age? That result is very similar to yours, and similar to those of Mecking et al 2004.
333. Rather than Shao et al I’d choose Waugh et al 2003 as the citation
333. ‘only began to be measurable in the atmosphere after 1936 and 1953’
These weren’t really measured back then, they were invented and presumably released into the atmosphere around those dates. Bullister’s reconstruction is a model using CFC and SF6 production and assuming some release function that is tuned to CFC SF6 measurements when those started to be readily available – 1970s?
Rather than muddy the waters, you could just write ‘These gases were released into the atmosphere after 1936 and 1953, respectively…’
337 – A set of 39-Ar simulations would be nice, but what would be nicer still is a set of 39-Ar measurements. For that study, when you do it, you may need to focus your efforts on waters that are a bit older. Good luck!
Fig. 8 – I can’t help but notice that all of these ‘data’ treatments capture the brief slowdown events of 1990 and 2006. Whether real-world sampling would capture this is another question, but you might note in the text that these events are reflected in the tracer concentrations (and in their derived ages).
Citations provided by reviewer
Doney, S. C., and J. L. Bullister (1992) A chlorofluorocarbon section in the eastern North Atlantic, Deep Sea Research 39, 11-12, 1857-1883.
Gruber, N., 1998 Anthropogenic CO2 in the Atlantic Ocean, Glob. Biogeochem. Cycles 12, 1, doi.org/10.1029/97GB03658
Jenkins, William J.. 1980. "Tritium and 3He in the Sargasso Sea." Journal of Marine Research 38, (3). https://elischolar.library.yale.edu/journal_of_marine_research/1518
Mecking, S., M. J. Warner, C. E. Greene, S. L. Hautala, and R. E. Sonnerup 2004 Influence of mixing on CFC uptake and CFC ages in the North Pacific thermocline, J. Geophys. Res. Oceans 109, C7, https://doi.org/10.1029/2003JC001988
Weiss, R., Bullister, J., Gammon, R. et al. Atmospheric chlorofluoromethanes in the deep equatorial Atlantic. Nature 314, 608–610 (1985). https://doi.org/10.1038/314608a0
Citation: https://doi.org/10.5194/egusphere-2024-2552-RC1 -
RC2: 'Comment on egusphere-2024-2552', Anonymous Referee #2, 13 Jan 2025
Review on 'Dual-tracer constraints on the Inverse-Gaussian Transit-time distribution improve the estimation of watermass ages and their temporal trends in the tropical thermocline' by Haichao Guo et al.
The aim of this study is to compare the 'real' mean (or ideal) ages with the mean ages inferred from Inverse Gaussian (IG) functions for the isopycnal sigma_theta=25.5 (including thermocline and intermediate waters) over the period 1981-2015. The ideal age cannot be observed, but the IG functions can be inferred form the observations of anthropogenic tracers like CFCs and SF6. Hence, it is of interest, in how far these observational inferred mean ages agree with the 'theoretical' ideal age. This can only be tested in a model study. The authors use the FOCI model to simulate mean age, CFC-12 and SF6. After a short model evaluation, the IG functions are inferred for different cases: from the modeled CFC-12 data alone, assuming fixed Delta/Gamma ratios, and by inferring both IG parameters Delta and Gamma from the modeled SF6 and CFC-12 fields. The IG parameter Gamma (mean age) and its temporal change between 1981 and 2015 is compared with the modeled ideal age.
This comparison of the mean age inferred from tracer data with the 'real' mean (ideal) age is important for the understanding and interpretation of tracer derived ages. A correct understanding of them helps to detect changes in ocean ventilation and, e.g. to infer anthropogenic carbon or ocean utilization rates from transient tracer data. This study provides a significant contribution to this topic, although the model analyses is restricted to the isopycnal sigma_theta=25.5.
The text is clear and well written, whereas the figures could need some improvement.
General comments:
For the case of constant Delta/Gamma ratios, the values 0f 0.8, 1.0, 1.2 and 1.4 are chosen. When inferring Delta/Gamma from CFC-12 and SF6, the color bar reaches from 0.2 to 1.8 (the same range has been used in He et al. 2018 to infer anthropogenic carbon from IG functions). Why is the range of the assumed Delta/Gamma ratios so much smaller (one could choose e.g. 0.2, 0.6, 1.0. 1.4 and 1.8)? (For the case Delta/Gamma=1.8 I would expect that the IG derived mean age is larger than the ideal age at least for the earlier years.)In this study, only absolute values for the differences and temporal changes in age are presented. This implies, that difference between ideal and tracer derived age values and a temporal change in the ages is weighted equally, independent from the age value itself. I wonder whether this is appropriate. For young waters (Gamma~5 yr), an age change of +- 2 years over the considered time period or a difference between ideal and tracer derived age of ~2 yr is significant, whereas for old waters (Gamma~100 yr), such changes/differences would be negligible. I would thus suggest to also calculate relative age differences between ideal and tracer derived ages and also relative changes of age over time. If the results for the relative age changes/differences do not substantially differ from the absolute changes/differences presented here, this should be mentioned in the text. Otherwise, the relative age changes/differences need be discussed in addition to the absolute changes.
The analyses focuses on the globally integrated/averaged mean age of the tracer inferred IG functions, i.e. the global distribution is 'condensed' to one number. This implies that regional differences might cancel out (e.g. the trend in age and the difference between tracer derived mean age and ideal age could differ between regions and even have opposite signs). In reality, also age changes for specific regions (e. g. upwelling, or water mass formation regions) are of interest, not only globally averaged values. Therefor, I would suggest to show at least one map with the differences between tracer derived and ideal age (for the 'best' tracer derived mean age), and also one map with the differences in the temporal trend between tracer derived ('best' result) and ideal age.
Why is the mean age not inferred by calculating the Delta/Gamma ratio from SF6 and CFC-12 at every grid point for every year?
The results presented here are based on spatially variable Delta/Gamma ratios, but without temporal change (the Delta/Gamma ratios from the years 2000, 2005, 2010 and 2015 are applied to the whole time series). Maybe the age calculation with the actual, time varying Delta/Gamma ratios could even replace the four differnt age calculations presented here.
Specific comments:
l. 119-120 and Fig. 1
Why has the isopycnal 26.0+-0.5 been chosen? The whole analyses is restricted to the isopycnal 25.5,
wouldn't it be more reasonable to show the data for this isopycnal (25.5 +-0.5) here?l.323-327
The results from the study in Peacock and Maltrud (2006) are interpretated wrongly.
First, the IG derived CFC values are smaller (half of) than the CFC values derived from the 'real' TTD (actual value). This would imply that the IG derived TTD is too old compared to the real TTD, in l. 326-327 the opposite is stated.
Second, the mean age of the 'real' and of the IG TTD are identical, because the parameters Delta and Gamma for the IG function are derived by calculating mean age and width of the 'real' TTD. Hence, it is wrong to say the mean age of the IG TTD differs from the 'real' water age. The difference is that the IG function contains a smaller fraction of young water, thus the inferred CFC values are smaller and the water 'appears' older compared to the 'real' TTD. One could also conclude that the shape of the 'real' TTD in this case differs significantly from the shape of an IG function.
l.326
'directly simulated CFC-like tracer'
This is misleading, as the CFC-like tracer in Peacock and Maltrud (2006) is inferred from the modeled TTD (convolution integral of TTD and assumed tracer surface concentration). This is not what I understand as 'directly simulated'.Regarding the difference between tracer derived and 'real' TTDs also the study from Chouksey et al. (2022) could be mentioned. There, tracer inferred IG-TTDs are compared with TTDs inferred from modeled (numerical) floats for the AAIW range. In some cases, the tracer derived TTDs are younger, in some cases older than the float based TTDs. Also, the shape of the float derived TTDs sometimes differs from the shape of an IG function.
l.327-328
Here, the study from Steinfeldt et al. (2024) could be cited. These authors found an increase of age (and hence a negative anomaly of anthropogenic carbon) with time for the old deep waters of the Atlantic when parameterizing the TTD as a single IG function. Assuming a contribution of an additional 'old' TTD leads to smaller age changes over time.l.351-352
'...and also from the cut-off of the long-tail of old ages in the spectrum due to the limited
atmospheric history length of CFC-12 and SF6'
This is not true, as the IG-functions always include a long tail towards high ages. This tail is more pronounced for higher Delta/Gamma ratios, leading to the increase of the mean age with the Delta/Gamma ratio. What is true is that this tail cannot be constrained from CFC-12 and SF6, as is correctly stated in l. 335. Please rephrase.Figures:
In general, the maps showing global distributions are too small. This could be changed easily, e.g.:
Fig.1: placing the color bar below the figures and showing latitude labels only on the left figure
would allow to increase the maps itself significantly
Fig. 2 and 5:
These figures stretch over one whole page, but there is a lot of free space between the single maps,
which could be enlarged
Fig.4:
latitude labels could be omitted at the right figure
The labels at the color bar are 'cut off' at the right side (the same holds for figure 2b)Fig. 3a, 6a and 7a: quantity and unit for the color bar are missing
the correlation values 0.95 and 0.9? at the Taylor Diagram overlap with the color barMinor comments:
Title: 'water mass' in two wordsAdditional literature:
Chouksey, M., Griesel, A., Eden, C. and Steinfeldt, R. (2022), Transit Time Distributions and ventilation pathways using CFCs and Lagrangian backtracking in the South Atlantic of an eddying ocean model. J. Phys. Oceanogr., 52(7), 1531-1548, 2022, doi:10.1175/JPO-D-21-0070.1.Steinfeldt, R., Rhein, M., and Kieke, D. (2024), Anthropogenic carbon storage and its decadal changes in the Atlantic between 1990–2020, Biogeosciences, 21, 3839–3867, doi:10.5194/bg-21-3839-2024.
Citation: https://doi.org/10.5194/egusphere-2024-2552-RC2 -
EC1: 'Comment on egusphere-2024-2552', Katsuro Katsumata, 13 Jan 2025
Finally two review comments are available. I apologize again that it took much longer than it should be.
Both reviewers provide constructive comments and please consider revising the manuscript accordingly.
Citation: https://doi.org/10.5194/egusphere-2024-2552-EC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 186 | 34 | 105 | 325 | 35 | 10 | 5 |
- HTML: 186
- PDF: 34
- XML: 105
- Total: 325
- Supplement: 35
- BibTeX: 10
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1