GWPs need to be corrected, but not in this way.
The GWP corrections introduced here may be needed, but they are not well founded, nor adequately documented. Thus, they fail to provide a traceable or reproducible chain of uncertainty, which is critical for their use in high level policy as intended for IPCC or WMO O3 assessments.
Moreover, the work here continues to commit errors in deriving metrics for the short-lived gases based on use of a standard lifetime without recognizing that the lifetime depends on the spatial-temporal emission patterns (this is now noted briefly but not addressed in the calculations here). The author’s rebuttal to the review highlights the errors in the approach used here that would need to be corrected before it is accepted as a standard for IPCC or WMO O3 assessments. Some of these failings are highlighted below.
(1) The authors have an obsession with using the global atmospheric mean mole fraction, x_atm, to describe the radiative properties. This makes no sense, they should be using the tropospheric mean mole fraction, x_trp, as used in the IPCC for observations and scenarios since 1994. The value of x_atm is the key factor in determining the burden (Tg) of a gas which is critical when connecting emissions, lifetime, and burden. The problem is that the x_trp values more accurately reflects the RF of a gas as stated in their Line 54-55. For the well mixed GHGs it is x_trp that proportional to the RF and also directly predicts the mole fraction of Cl or Br entering the stratosphere. Also, the value x_trp is readily measured by the remote marine surface sites or transect campaigns like NASA's ATom. The values of x_atm cannot be simply measured without measuring the falloff of x_ in the stratosphere. Likewise, the average surface value x_sfc is not really measurable since we cannot average over the large spatial-temporal heterogeneity of boundary layer mole fractions over the large emission regions. Similarly, x_sfc is very hard to model consistently across our different models.
What is vexing to this reviewer is that the authors claim that all the modern RF calculations used a surface value that was different from the mean tropospheric value, but they cannot document this, nor even recover what was used. Using their prime example of Etminan 2016, I tried to recover the GHG profiles used but could find no reference to profiles or surface values anywhere in the main text or supplemental data. In fact, the supplemental tables imply that a constant GHG profile is used since only a single value of x_ is given. Given their participation in these papers, they must have inside information that was never documented. This speaks to the need to redo these RF calculations from scratch with proper documentation, rather than trying to correct something (x_sfc/x_atm) that you are unsure of exactly what was used in the original calculation.
(2) The authors assert that this work is important for future assessments, but without proper uncertainty assessment it really cannot be a mainstay of any assessments. Furthermore, without documentation of the testable intermediate steps in the calculation as requested in the first review, no one can use or test the results using different modeling tools. Calculating “quick & dirty” correction factors to climate metrics that were calculated incorrectly (or were undocumented) in the first place is not a way to provide traceable metrics. It is not clear what the RF radiation models used for profiles; same problem applies to the CTM used for ‘correcting’ them?
(3) The author’s rebuttal states that the correct balance between greenhouse gas content (burden) and mean tropospheric abundance (dry-air mole fraction, ppb) has not been used by the IPCC previously in “climate metrics”. This assertion is highly misleading and wrong. In the 2001 IPCC TAR, the many chapters dealing with greenhouse gas emissions, burdens, observations, and future scenarios all used a near-correct conversion factor relating these (see details in my review). I would call the calculation of the mean tropospheric abundance of greenhouse gases a climate metric. Their use in the TAR chapter 4 was clear, but not well documented as to how they were calculated. Thus, I wrote up a 2007 Phil. Trans. Roy. Soc. Monograph on lifetimes (“Lifetimes and time scales in atmospheric chemistry”, Prather) and then we prepared a detailed analysis of the factors and uncertainties in the 2012 GRL article (“Reactive greenhouse gas scenarios: Systematic exploration of uncertainties and the role of atmospheric chemistry”, Prather, Holmes, Hsu). I know these studies and their uncertainties were carried through to the 2017 IPCC AR5, since I edited the Annex II summary tables. That 2012 paper needs updating. All the greenhouse gas abundances in the IPCC are calculated as the tropospheric mean, not “atmospheric” mean; the latter requires one estimate how much the gas fills the stratosphere and this gives the standard conversion factor (CF, Tg/ppb) used since 2001. It appears that only the chapter calculating the damage metrics (GWPs, ODPs) was incorrect. It is interesting to note that chemistry-climate models calculating RF from their own GHG distributions were correct. Please correct the manuscript with an accurate history and avoid misleading the readers, i.e., that we did not know how to calculate the abundance anddistribution of the greenhouse gases.
(4) The obsession here with the “surface” abundance is confusing and misleading. Regarding ‘surface’ values, the author’s rebuttal shows a lack of experience in analyzing NOAA or AGAGE surface measurements. Yes, these are surface sites (for the most part) but they have very strict protocols for taking flasks or excluding polluted data. They are meant to represent the remote marine troposphere and thus their reported surface measurements are systematically biased to represent the free troposphere. One cannot compare these observations with a CTM averaged surface value because the surface data representing emissions has been removed or avoided. Cape Grim Observatory is an obvious example. I know this because I had to ask ALE/GAGE to provide the full, unbiased data set for my 1985 Nature paper based on the polluted surface layer (“Continental sources of halocarbons and nitrous oxide”). The “remote marine clean air” observations, when latitudinally averaged, are typical of the tropospheric mean. Since we cannot really observe the surface abundance with any accuracy, we must also admit that we cannot model it very well either since the results will depend on the height of the nighttime boundary layer and hence on the diurnal cycle of emissions. Thus, the ratio of “surface” to “atmospheric” mole fraction developed here using one set of emissions and resolution with OsloCTM2 (and used as a correction factor) is irreproducible in other models. Further, the curve fitting on a log scale does not help address uncertainties. On the other hand, the mean tropospheric abundance for a GHG like CH4 is easy to estimate or observe – even the Atom mission transects provide a good estimate.
(5) A much broader issue that could be addressed here - and really should be addressed by the community as a whole - is that we must stop reporting these damage metrics (e.g., GWP, ODP) for short-lived gases without also documenting how they were derived. Defining short-lived as <1/2 yr (i.e., not well mixed within a hemisphere) we must recognize that their lifetimes – a critical factor in these metrics – will vary depending on where and when it is emitted. A recent quick-study of CO, showed lifetimes ranging from 55 to 85 days. While the lifetime for a well mixed GHG is about the same for any emissions within the troposphere, for stratospheric emissions (e.g., some proposed mechanisms for photochemical production of N2O) it is much, much shorter (1998 Science paper, “Time scales in atmospheric chemistry: coupled perturbations to N2O, NOy, and O3”). More applicable to this problem, we can look in our 2024 Elementa paper (“Lifetimes and timescales of tropospheric ozone”, Prather & Zhu) where we calculate the lifetime of a tropospheric O3 emission to vary from 6 days (E. Asian pollution source) to 26 days (aviation or strat-trop exchange). Thus, if we calculate the RE of an ozone source we have a large factor-of-four range, but fortunately we usually calculate the ozone RE from a modeled change in burden, and the lifetime is factored into the burden. This same range, or potential error in GWP, could apply to short-lived HFCs. The authors feel that giving a “best estimate” for HFC-152 is appropriate since people are using it, but I would contend that they do not necessarily know what “best” is, and they certainly have not documented that choice. I find the revised wording to be inadequate since their correction of 10% is probably dwarfed by the uncertainty they should report on the GWP.
I find the authors show very little understanding of what the “lifetime” of a gas depends on, they seem to think it can be scaled from knowledge of chemical kinetics alone. With these short-lived gases, a lot of the build up of the burden and the chemical rates of destruction depend on tropospheric processes such as boundary layer venting, scavenging, clouds and convection, which are not consistent across models and introduce unknown uncertainties. The range of emission location/timing determines lifetime and is not a scientific uncertainty but a societal choice. All we can do is to calculate the scientific uncertainty for a specific emission and then include the possible range from emission choices.
The publication of this article will not help the IPCC or WMO O3 assessments, it will only confuse. Approximate correction factors, parameterized by a curve fit on a log scale, for an incorrect on unknown original value with unclear documentation does not lend confidence, but erodes it.
Michael Prather
UC Irvine
28 May 2026 |
I have high esteem for the authors of this Letter, they are good colleagues and friends, and they have decades of experience in quantifying the impact of different greenhouse gases (GHG). In my personal view, unfortunately, this Letter is mistaken and adds more confusion than elucidation to our understanding of atmospheric trace gases, their distributions, their lifetimes, and how we assess their climate impacts. The letter seems to explain a small mistake or confusion that was made by the authors in calculating the global warming potentials (GWPs) for the assessments. Thus, maybe it should appear as an IPCC erratum on /ipcc.ch rather than as a scientific advance in ACP, although maybe corrections do belong in ACP.
The authors present a corrected methodology for calculating the conversion factor CF relating the increase in atmospheric GHG burden (in Tg) to the mean tropospheric abundance (X, mole fraction with respect to dry air in ppb, parts per billion). This revision results in a set of %-corrections for the recent set of GWPs that the authors calculated in the first place. The authors should be commended for correcting the record.
CF (Tg per ppb) is used with the Radiative Efficiency (e.g., RE in W m-2 per ppb) to calculate the radiative imbalance caused by an increase of 1 Tg in the atmospheric burden of the GHG (W m-2 per Tg). CF values have been published and used extensively throughout the IPCC to convert emissions scenarios (Tg) into tropospheric mean X (ppb). From this Letter, it appears that the GWP calculations have used different, incorrect CF values. The recent IPCC AR6 assessment, Table 7.15 in ref [0], does not list the CFs used in the GWP calculation, and this Letter also fails to list the corrected CFs so that they can be checked by others. The CF is a very simple quantity and can be simply verified from the observed or modeled distribution of the GHG in troposphere and stratosphere.
The central question here is what “ppb” should we be using in CF and RE. For RE it obviously the abundance in the middle to upper troposphere since GHG in the near-surface layer have no thermal infrared radiative effect being indistinguishable from the surface, and GHG in the stratosphere are not so coupled to the troposphere-climate system (as recognized by the stratospheric adjusted RF, SARF). For CF, thus it should relate total burden to the mean tropospheric abundance, since that is a quantity that we typically measure (e.g., NOAA’s marine boundary layer GHG record) or model, and the one that should be used in the RE calculations.
For well-mixed (tropospheric that is) GHG, we must account for the stratospheric fall-off in abundance. I am familiar with this approach since I derived some of the key values, including the CF used in the IPCC SAR & TAR [1]. The 2001 TAR conversion factors, 2.78 Tg-CH4/ppb and 4.81 Tg-N(N2O)/pbb, are used in sections 4.2.1.1 & 4.2.1.2, respectively. I was unhappy with these simplistic TAR numbers and pursued a more careful and quantitative evaluation of the CF (Tg/ppb) values in [2] based on a well-mixed troposphere (from NOAA remote surface sites) and the observed fall-off of stratospheric profiles. These updated 2012 [2] values were published as 2.75 Tg-CH4/ppb ±1% and 4.79 Tg-N(N2O)/ppb ±1% (68% confidence interval), which as expected are notably less than the values 2.84 and 4.96 derived from Equation (1) in the Letter and well outside the expected uncertainty. What is surprising is that based on the references in the Letter (2005 and 2013), one side of IPCC did not apparently know what the other side was publishing. In addition, the 2013 AR5 Annex II GHG Tables [3] use and explicitly reference the ref [2] methodology. These values are used in the scenario calculations by Meineshausen and the MAGICC code to convert Tg to ppb. It appears from the authors of this Letter that the GWP calculations since the TAR have assumed that all GHG are well mixed to the top of the atmosphere and now need to be corrected. It is good to see the GWPs 'corrected' but the Letter could note the problem with a better historical perspective and make sure the AR6 GWPs are updated on /ipcc.ch.
There remain two fundamental disagreements with the approach and derivations presented here: (1) the incorrect use of surface abundance, X_srf; and (2) the derivation of lifetimes for short-lived (<0.2 yr) gases simply as a function of their chemical rates.
(1) The “ppb” used in both CF and RE should be the tropospheric mean, X_trop. The repeated use of X_srf in this Letter is simply wrong. No one seriously uses the mean surface abundance for anything because it is not observed nor modeled correctly. For one, X_sfc globally averaged is impossible to measure. Near source regions (and these GHG do have regions of intense surface emissions) one could never assemble enough observing sites to accurately measure the global mean, and satellite remote sensing cannot really detect surface abundances. That is why NOAA and ALE/GAGE have put together a latitudinal range of remote marine boundary layer sites which are combined to derive a mean tropospheric abundance, X_trop. It is this X_trop that is quoted in the rest of the IPCC as the mean abundance for the GHG, not a mean surface value that would have to include averages over source regions. For two, we really cannot model the global mean X_sfc very well either. In a model like CTM2, this value depends on the model layers and the nighttime boundary layer, including the diurnal cycle of emissions that will control the large build up overnight. While our ability to model X_sfc is very limited, our ability to model and measure X_trop is solid and insensitive to boundary layers, diurnal cycles, etc.
(2) Lifetime is defined correctly as total burden of the entire reservoir divided by source or sink. That definition is critical if we are to use the steady-state lifetime as an integral of the impacts of the emitted gas, see review and derivations in ref [4]. Lifetime depends on the kinetics of the chemical loss throughout the atmosphere, but it also depends on the location and time of the source. For example, in the derivation of the N2O chemical feedback on lifetime [5], it is shown that emissions of N2O at 40 km result in a lifetime of 2.5 years, rather than 120 yr for tropospheric emissions. These facts about lifetime are woefully and incorrectly propagated in this Letter. The authors assume that their corrections and lifetime adjustments for short-lived GHG (<0.2 yr) are meaningful and valid and can be used for GWPs. This assumption is simply wrong and can have very large errors. The burden that accumulates following emission, and the resulting GWP, will vary by large factors depending on where and in what season the gas is emitted. Tropical emissions will have a short lifetime, probably <0.1 yr, and small GWP (lower X_trop), but high latitude winter-time emissions will accumulate and probably have an effective lifetime >0.5 yr and correspondingly greater GWP. Consequently, one cannot scale ERF to emissions with a metric like GWP. For different example, recent studies of tropospheric O3 [6] show a lifetime ranging from 6 to 27 days depending on latitude and season of its production. This result means that the ERF of an O3 “source” has range of a factor of 4 or more, and cannot be simply “corrected” as done here. The idea of an RE (per ppb) is still mostly valid but the CF is now widely variable. To be fair, we should not publish single GWPs for such short-lived GHG unless we deliver a range based on possible emission scenarios.
The current Letter could be revised in its discussion of surface abundances and lifetimes to avoid propagating old science. Then, collectively, we – as authors of the past IPCC chapters on trace gases and radiation – must accept our failure to communicate with one another.
Michael Prather, UC Irvine
[0] see Table 7.15 of Forster, P., T. Storelvmo, K. Armour, W. Collins, J.-L. Dufresne, D. Frame, D.J. Lunt, T. Mauritsen, M.D. Palmer, M. Watanabe, M. Wild, and H. Zhang, 2021: The Earth’s Energy Budget, Climate Feedbacks, and Climate Sensitivity. In Climate Change 2021: The Physical Science Basis. Contribution of WGI to IPCC AR6 [Masson-Delmotte, V., P. Zhai, A. Pirani, et al. (eds.)]. Cambridge University Press, Cambridge, United Kingdom, pp. 923–1054, doi:10.1017/9781009157896.009.
[1] D. Ehhalt, M. Prather, F. Dentener, R. G. Derwent, E. Dlugokencky, E. Holland, I. S. A. Isaksen, J. Katima, V. Kirchhoff, P. Matson, P. M. Midgley, and M. Wang (2001) Chapter 4. Atmospheric Chemistry and Greenhouse Gases, in Climate Change 2001: The Scientific Basis, J.T. Houghton et al., eds., Cambridge U. Press, pp. 239-287.
[2] Prather, M.J., C.D. Holmes, J. Hsu (2012), Reactive greenhouse gas scenarios: Systematic exploration of uncertainties and the role of atmospheric chemistry, Geophys. Res. Lett., 39, L09803, 5 pp., doi:10.1029/2012GL051440.
[3] Prather, M., G. Flato, P. Friedlingstein, C. Jones, J.-F. Lamarque, H. Liao and P. Rasch (section eds.) (2013) Annex II: Climate System Scenario Tables. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge University Press, New York, NY, USA, pp. 1395-1445.
[4] Prather, M.J. (2007) Lifetimes and time-scales in atmospheric chemistry, Phil. Trans. R. Soc., A 365: 1705–1726, https://doi.org/10.1098/rsta.2007.2040
[5] Prather, M.J. (1998) Time scales in atmospheric chemistry: coupled perturbations to N2O, NOy, and O3, Science, 279, 1339-1341.
[6] Prather, M.J. and Xin Zhu (2024) Lifetimes and timescales of tropospheric ozone, Elementa: Science of the Anthropocene, 12 (1): 00112, doi: 10.1525/elementa.2023.00112