the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 10 Nov 2023
Updated algorithmic climate change functions (aCCF) V1.0A: Evaluation with the climate-response model AirClim V2.0
Abstract. Aviation aims to reduce its climate effect by exploiting the potential of identifying alternative climate optimized aircraft trajectories. Such climate-optimized trajectories require a dedicated meteorological service in order to inform on those regions of the atmosphere where aviation emissions have a large effect on climate, for example, by contrail formation or nitrogen-oxide (NOx)-induced ozone formation. With the algorithmic Climate Change Functions (aCCFs) prototypes of a mathematical formulation for the temporal and spatial climate effects of aviation emissions in the atmosphere is provided, which relies solely on numerical weather prediction at the time and location of emissions. Based on the recently published consistent set of aCCF-V1.0, we here introduce newly derived calibration factors for the individual non-CO2 effects of aviation (NOx, water vapour, contrail cirrus) and establish version V1.0A of aCCFs (aCCF-V1.0A). ACCF-V1.0A represents an updated formulation of aCCF while exploring the current level of scientific understanding of individual climate effects of aviation emissions by evaluating quantitative estimates of climate effects with the state-of-the-art climate-response model AirClim. Individual scaling factors (i.e. AirClim calibration factors) are provided for the respective non-CO2 effects comprising contrail cirrus, water vapour and NOx-induced climate effects on ozone and methane, resulting uniformly in lower estimates in aCCF-V1.0A for all species compared to the earlier version aCCF-V1.0.
- Preprint
(1784 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
CEC1: 'Comment on gmd-2023-92', Juan Antonio Añel, 20 Dec 2023
Dear authors,
Unfortunately, after checking your manuscript we have concluded that it does not comply with our code policy. We know that MESSy qualifies for an exception according to our policy; however, we request that the code and data, despite not being public, is deposited in a permanent repository with a DOI. In this way, you must deposit the code of AirClim and the emissions data in a long-term private repository. For example, Zenodo offers such possibility (private repositories), and MESSy itself is stored in such way. Therefore, please, store AirClim in an acceptable repository, private if you need it, and publish the DOI and link for it as a reply to this comment.
Also, please, do not forget include the same information in the "Code and Data Availability" section of any potentially reviewed version of your manuscript.
Best regards,
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/gmd-2023-92-CEC1 -
AC1: 'Reply on CEC1', Sigrun Matthes, 12 Jan 2024
Dear Juan Añel,
thank you for your comment.
Two zenodo repositories have been established since, one for AirClim and one for emission inventory.
[1] AirClim Version 2.0: , The non-linear climate chemistry-response model AirClim 2.0 used for estimating CO2 and non-CO2 climate effects, zip-File of version used https://zenodo.org/record/8410785 (private)
[2] Emission inventory: EMAC/MESSy v2.54.0 model (EMAC) aviation emission inventories as calculated by AirTraf 2.0, nc-file with 3d-data (lat, lon, alt, fuel, NOx, dist). https://zenodo.org/record/8410729 (open access)We will include this information in the “Code and Data Availability” section of our reviewed version.
Best regards,
Sigrun MatthesCitation: https://doi.org/10.5194/gmd-2023-92-AC1
-
AC1: 'Reply on CEC1', Sigrun Matthes, 12 Jan 2024
-
RC1: 'Review of Matthes et al.', Anonymous Referee #1, 11 Jan 2024
In this study, the authors calibrate a previous version of the algorithmic climate change functions (aCCFs), obtained by fitting climate model simulations to a small selection of meteorological parameters, to the impulse response model AirClim. Both models aim at estimating the average temperature response to CO2 and non-CO2 emissions of flights, but follow different approaches and assumptions. The calibration produces an alternative set of aCCFs, with sizeable differences in the estimated average temperature response of flights compared to the original aCCF set.
The study is well written, and the figures and tables illustrate the discussion well. My comments below aim at making the description clearer and more accurate in places and should amount to minor revisions because new analyses are not required.
Main comment:
- Section 3 and 5 need to be more upfront on two important points. First, it is important to point out in section 3 that the “calibration factors” (Table 2) that translate aCCFv1 into functions comparable to AirClim imply very large changes. Those are not small corrections. So the two models, aCCFv1 and AirClim represent two very different views of the climate impact of flights, for the reasons listed (only qualitatively unfortunately) in the conclusion. Second, Section 5 does not answer the basic question of why one would want to calibrate aCCFv1 to AirClim. As a measure of uncertainty? Not really because the two model are unlikely to cover the whole uncertainty range. As a way to choose between different philosophies (tagging/ perturbation and climatological/synoptic)? But what would be the rationale for such a choice?
Other comments:
- Line 12: “climate effects” Be specific: the aCCFs give the Average Temperature Response.
- Lines 19-20: Clarify: “lower estimates” of the ATR?
- Line 49: Not sure that “short-cut” is the right word, because there is also a loss of accuracy. “Approximation” would be a better word.
- Lines 55 and 62: Need to define what is meant by “perturbation approach” and “tagging contribution approach”. It might be useful to add a paragraph before this one to clearly define those two approaches to attribution, and what they mean in terms of radiative forcing. As you found in Grewe et al. (2019) https://doi.org/10.1088/1748-9326/ab5dd7, the impact of the choice of methods is very large: a factor 6.7 according to Table 1 in that paper.
- Line 61: I do not understand “for eight specific days”. Do you mean eight specific weather situations?
- Line 68: “to local meteorological conditions” is too broad. It is in fact “to a selected set of variables that represent the local meteorological conditions”.
- Line 77: Need to clarify what is meant by “effectiveness” here.
- Line 83: The meaning of “the room for a calibration process is open” is only clarified in the next paragraph (line 88) by “one realization within the range of plausible values”, so I would suggest merging the two paragraphs to clarify what you are doing. And it should be noted that such calibration does not replace the need for a way to properly account for uncertainty in both AirClim and aCCFs.
- Line 109: “it is equivalent to selecting another value from this interval”. Is that true? If there’s a probability distribution, then some values are more probable than others, although I agree that the most probable value is not necessarily the mean.
- Lines 126-127, lines 150-151, line 154: Again, avoid using the imprecise “climate effect”. Here, I assume that it is the ATR that is calculated.
- Line 140: “as boundary conditions” – ambiguous. Do you mean that there is no nudging inside the model domain? By the way, is the model configured to simulate Europe only, or is it global?
- Line 149: The aCCFs are made out of the CCFs for 8 weather patterns. If I understand well that synoptic information is lost in the aCCFs, so it is not possible to relate aCCFs to specific patterns in the North Atlantic. Is that correct? It would be good to clarify that here. And by the way, are the aCCFs fitted over the 8 weather patterns together? Or are they fitted over some mean of the 8 weather patterns?
- Line 173: Why A330 flights only?
- Lines 182-186: Again, no need to keep the suspense on the actual calculation by using the imprecise “climate effect”. Say that you are calculating P-ATR20 from the start of the paragraph.
- Caption of Table 1: The “without forcing efficacy” comes as a surprise. There should be an explanation of what that means in the main text.
- Line 227: Converting to F-ATR20 when Section 3 has been all about P-ATR20 seems an unnecessarily confusing step. Why do it? And here again, the reader would benefit from a reminder of the notion of efficacy, and how it might change the results compared to Section 3.
- Line 245: I would expect variability in the contrail aCCF, but I would also expect some correlation between time steps, since ISSRs are not advected randomly. But then the aCCFs do not have the concept of ISSR, so what do the patterns shown in Figure 2c tell us? It is also surprising to have pockets of cooling in the middle of warming zones – that does not look like the day/night contrast alluded to in the text.
- Line 307: What is the implication of that narrower distribution for climate-optimised routing?
- Lines 366-367: What do those confidence intervals look like? How do you go from your calibration factors to confidence intervals?
Technical comments:
- Line 26: Suggest rewriting to “long-term background ozone is reduced by the NOx-induced methane decrease”.
- Line 43: Presenting the three types of studies as bullet point would make that paragraph easier to read.
- Line 88: I have never seen “event horizon” used in this context. Is that a correct use of the term?
- Line 115: Typo aligned
- Line 117: “is corresponding” -> corresponds
- Figure 1: Typo relvant (and it would be good to disable the spelling checker to avoid those words underlined in red)
- Line 164: green function -> Green’s function
Citation: https://doi.org/10.5194/gmd-2023-92-RC1 - AC3: 'Reply on RC1', Sigrun Matthes, 19 Apr 2024
-
RC2: 'Comment on gmd-2023-92', Anonymous Referee #2, 16 Jan 2024
In this manuscript the authors compute the non-CO2 effects of aviation using their algorithmic climate change functions (aCCF, V1.0), which themselves are defined from a suite of models, and the AirClim v2.0 model for a set of city-pairs in Europe and a full seasonal cycle. From these simulations they compute renormalization factors (one factor for each non-CO2 effect) to derive calibrated aCCFs (V1.0A). The authors also present PDFs of the aCCFs over a European domain and a comparison between the uncalibrated and calibrated aCCF.
I have a number of major issues with this manuscript. In my opinion each one of these issues could prevent publication in its own right so I urge the authors to consider them carefully.
1/ Calibration implies reference.
Calibration is the process by which a measurement is adjusted to a reference. It is not clear to me why AirClim V2.0 (as described in Dahlmann et al. 2016) provides such a reference. It is claimed that AirClim is “comprehensive”, “state-of-the-art”, “well-established” but this is not demonstrated. I would welcome elements that prove AirClim is a reference. To which extent is it validated against observations? How is it documented? How does it compare to other models? Dahlmann et al (2016) is now 8 years old, is it still state-of-the-art?
2/ Calibration parameters are beyond reasonable.
The calibration parameters (f_AirClim in Annex A) are much lower than 1 for some of the non-CO2 effects, i.e. 0.333 for contrail-cirrus, 0.179 for O3 and even 0.058 for CH4. Rather than just rescaling their parametrisation, the authors should go back to their models, understand the root causes of the differences, reformulate the model and try to narrow down the discrepancies. I personally have little trust in the structure of a model that requires a scaling factor of 0.058 against a reference.
3/ Tagging vs perturbation approach.
As stated on AirClim is based on a perturbation approach (as stated on lines 54-55), i.e. it computes the marginal change in RF due to the perturbation of NOx from aviation emissions. My understanding is that aCCFs V1.0 are based on the tagging approach which places all the NOx emissions (from different sectors) on a same footing. The perturbation (or marginal) approach has the advantage of representing the change expected from an action (e.g. rerouting) but has the disadvantage of not being additive and the sum of the relative perturbations from different sectors or regions usually do not add up to 100%. The tagging (and other similar methods) have the advantage of being invariant to disaggregation and recombination and are a better measure of the contribution of a sector to the total forcing but are not adapted to quantifying the impact of a change when everything else is fixed. My understanding is that the calibration process shifts the aCCF from the tagging to the perturbation approach (lne 386). For the methodology to be valid, the authors need to show that there is a proportionality coefficient between the two approaches that holds for individual flights and not just on average. Otherwise there should be a clear warning that aCCFs V1.0A should not be used for individual flights.
4/ There is a serious lack of treatment of uncertainties for aCCF V1.0A.
In their present form aCCFs do not come with uncertainties. To be useful, aCCFs should be associated with an uncertainty range (e.g., 1 or 2 sigma or 90% uncertainty range). This manuscript offers the opportunity to address this lasting deficiency of aCCFs by tracking down and combining the different sources of uncertainties. First of all, the aCCFs are expressed as a function of a very limited set of predictive variables and there is a significant dispersion around the average relationship. This implies an uncertainty range that needs to be documented and propagated into aCCFs v1.0A. AirClim V2.0 also has uncertainties that probably add quadratically to those of the aCCF formulation. A third source of uncertainty comes from the representativity of a climatological calibration for individual flights. This can be diagnosed from the dispersion of the calibration coefficients estimated on a flight-by-flight basis and should also be added quadratically to other sources of uncertainties. Finally, there are other sources of uncertainties (e.g., the masking of source regions for the contrail aCCF) that need to be discussed if they cannot be estimated properly.
5/ aCCFs V1.0A lack traceability and transparency
aCCFs lack the traceability and transparency needed by users. aCCFs V1.0A make the situation even worse as they result from a long suite of models (some of which are not freely available) and simulations (some of which not accessible). I am worried that aCCFs V1.0A will be used without any consideration of their inherent uncertainties and/or outside their validity range. The authors need to provide a more traceable workflow with the input and output of the models that are used to produce the aCCF V1.0A. Only in this way can the aCCF be compared to other approaches and become a trustworthy source of information for users.
6/ The choice of metric is unusual and misleading.
The default version of aCCFs comes are for the ATR-20 climate metric (temperature change averaged over a 20 year period) and aCCFs V1.0A are also illustrated for the ATR-20. This is an unusual choice within the large body of literature on climate metrics. It also contrasts with policy choices made in UNFCCC (see also recent SBSTA decision to retain GWP100 in the wake of the Paris Agreement). The ATR-20 metric puts a lot of weight on the climate effects of short-lived species. Sure there is a policy dimension in climate metrics but there are also scientific and economical considerations that favour putting more weight on long-lived greenhouse gases, at least until CO2 emissions are curbed significantly. What is so special to the aviation sector to favour ATR20 over metrics with longer time horizon (e.g. GWP100 or GTP50) that are more consistent with the current mitigation levels? At the very least aCCFs V1.0A should be presented with several climate metrics. Highlighting ATR20 can be very misleading for users who are not very knowledgeable on climate metrics.
Other comments:
Title: the title does not describe the content of the manuscript. There is no evaluation of aCCF against AirClim V2.0. Performing such an evaluation would imply to show a range of score of aCCF against AirClim V2.0 for individual flights. As mentioned above, I am also dubious that this is a calibration. In any case the title needs to be changed and reflect the content of the manuscript.
Line 24: this needs to be rephrased as it is not “emissions” per se that are radiatively active or not but the molecules in the atmosphere.
Lines 27-28: it would be useful to say a bit more about contrails and induced cirrus at this point.
Lines 29-32: the two sentences repeat rather than complement each other. The “accordingly” does not read well. The first sentence omit the dependence on environmental factors (position of the Sun, surface albedo, clouds, …).
Lines 63-64: see above for an alternative view.
Line 69: this may be a strength but this is also their main weakness.
Lines 75-76: given the bold assumptions made in aCCF and the absence of characterization of uncertainties, it is dangerous if not fallacious to encourage climate-optimized flight trajectories.
Lines 83-84: I agree that the uncertainty of aCCF estimates is missing but this manuscript does nothing to better quantify such uncertainties.
Lines 107-112: this paragraph is not clear and I could not link it to the calibration approach described later in the manuscript.
Figure 1: why are the bottom arrows pointing to the green boxes rather than starting from there?
Line 140: I understand MESSy2 is nudged to ERA5 but why is this considered as “boundary conditions”? What does it imply for the simulation of ISSR? How does it compare to observations?
Line 153: see above. Considering great circles is an issue because there are co-variations between the jet stream (which airlines will try to avoid or use depending on the direction) and the presence of ISSR. It is well known that there are significant departures from the great circle. If aCCF are not computed for real trajectories, how can it be assumed that it is useful for designing climate-optimized trajectories?
Line 164-165: How does the Green function for CO2 from Hasselmann et al (1997) compare to more recent estimate? To which extent does it depend on future emission scenarios?
Line 169: is it a “calibration”, a “comparison” or an “evaluation” (as per the title)?
Table 1: SI units are m and kg rather than km and g. EGU journals recommend to use SI units unless there is a good justification not to. Is there a good reason to deviate from SI units in Table 1?
Table 1: I assume the 85 flights operate daily during a year, hence the ~50 millions km flown. Unless I missed the information, the fact that the flights are assumed to operate daily is missing.
Line 229: Kg should read kg.
Line 229 and elsewhere: for the sake of clarity, it should be stated that this is kg NOx as NO2.
Figure 2: the caption should state whether the figure shows the uncalibrated (V1.0) or the calibrated (V1.0A) aCCF.
Figure 2: the figure shows a filament of elevated aCCF for water vapour. I guess this corresponds to a filament of larger PV values indicative of a tropopause folding. If so, the residence time of the water vapour in the stratosphere is short and the non-CO2 effect is probably much less than indicated by the aCCF. What is the confidence level / uncertainty on these values?
Lines 267-268: why not show or at least give the fraction of aCCF that is zero?
Figures 3 and 4: the caption should make it clear that the PDFs are for calibrated aCCF.
Figures 3 and 4: I could not find the % of non-zero aCCF on panels c.
Figure 5: V1.1 should read V1.0A on top of panel b.
Line 317: how is AirClim “well established” ?
Lines 331-334: does this mean that the aCCFs depend on the choice of trajectories for a given set of city pairs? If the calibration depends on the assumption of great circles between city pairs then do the calibration coefficients hold for climate-optimized trajectories?
Line 357: can the authors elaborate on the issue here?
Lines 373-375: I disagree here. A parametrized approach is required for the medium- to long-lived non-CO2 and CO2 effects, but not necessarily for the short-lived effects such as contrails and contrail-cirrus.
Lines 418-422: this is indeed a limitation and the associated uncertainty should be quantified.
Line 456: I agree with this statement, therefore the purpose of the manuscript should be to decrease uncertainty not rescale uncertain parameters.
Line 470ff: This paragraph contradicts lines 72-73 that specify aCCF is also available as an open source Python library of CLIMaCCF V1.0. Why not mention this in the “code availability” section? Or is it not the version used for this manuscript?
Lines 511-514: please provide details of the final version (rather than the submitted version).
Lines 694-696: please provide details of the final version (rather than the submitted version).
References: please make sure the format of the references is consistent through the list.
Lines 703ff: I find many assumptions to be unnecessarily simplistic, especially those related to thresholds (eg time of day for contrails) or maximum values (eg radiation). Space or time integrals should be more appropriate.
Line 736: it is not specified how persistent contrails are defined.
Line 737: Change Yin et al (2022) to (2023).
Lines 753-754: this is an extreme example of the lack of transparency I mentioned above.
Fig B1: is that for uncalibrated or calibrated aCCFs?
Citation: https://doi.org/10.5194/gmd-2023-92-RC2 - AC2: 'Reply on RC2', Sigrun Matthes, 19 Apr 2024
Status: closed
-
CEC1: 'Comment on gmd-2023-92', Juan Antonio Añel, 20 Dec 2023
Dear authors,
Unfortunately, after checking your manuscript we have concluded that it does not comply with our code policy. We know that MESSy qualifies for an exception according to our policy; however, we request that the code and data, despite not being public, is deposited in a permanent repository with a DOI. In this way, you must deposit the code of AirClim and the emissions data in a long-term private repository. For example, Zenodo offers such possibility (private repositories), and MESSy itself is stored in such way. Therefore, please, store AirClim in an acceptable repository, private if you need it, and publish the DOI and link for it as a reply to this comment.
Also, please, do not forget include the same information in the "Code and Data Availability" section of any potentially reviewed version of your manuscript.
Best regards,
Juan A. Añel
Geosci. Model Dev. Executive Editor
Citation: https://doi.org/10.5194/gmd-2023-92-CEC1 -
AC1: 'Reply on CEC1', Sigrun Matthes, 12 Jan 2024
Dear Juan Añel,
thank you for your comment.
Two zenodo repositories have been established since, one for AirClim and one for emission inventory.
[1] AirClim Version 2.0: , The non-linear climate chemistry-response model AirClim 2.0 used for estimating CO2 and non-CO2 climate effects, zip-File of version used https://zenodo.org/record/8410785 (private)
[2] Emission inventory: EMAC/MESSy v2.54.0 model (EMAC) aviation emission inventories as calculated by AirTraf 2.0, nc-file with 3d-data (lat, lon, alt, fuel, NOx, dist). https://zenodo.org/record/8410729 (open access)We will include this information in the “Code and Data Availability” section of our reviewed version.
Best regards,
Sigrun MatthesCitation: https://doi.org/10.5194/gmd-2023-92-AC1
-
AC1: 'Reply on CEC1', Sigrun Matthes, 12 Jan 2024
-
RC1: 'Review of Matthes et al.', Anonymous Referee #1, 11 Jan 2024
In this study, the authors calibrate a previous version of the algorithmic climate change functions (aCCFs), obtained by fitting climate model simulations to a small selection of meteorological parameters, to the impulse response model AirClim. Both models aim at estimating the average temperature response to CO2 and non-CO2 emissions of flights, but follow different approaches and assumptions. The calibration produces an alternative set of aCCFs, with sizeable differences in the estimated average temperature response of flights compared to the original aCCF set.
The study is well written, and the figures and tables illustrate the discussion well. My comments below aim at making the description clearer and more accurate in places and should amount to minor revisions because new analyses are not required.
Main comment:
- Section 3 and 5 need to be more upfront on two important points. First, it is important to point out in section 3 that the “calibration factors” (Table 2) that translate aCCFv1 into functions comparable to AirClim imply very large changes. Those are not small corrections. So the two models, aCCFv1 and AirClim represent two very different views of the climate impact of flights, for the reasons listed (only qualitatively unfortunately) in the conclusion. Second, Section 5 does not answer the basic question of why one would want to calibrate aCCFv1 to AirClim. As a measure of uncertainty? Not really because the two model are unlikely to cover the whole uncertainty range. As a way to choose between different philosophies (tagging/ perturbation and climatological/synoptic)? But what would be the rationale for such a choice?
Other comments:
- Line 12: “climate effects” Be specific: the aCCFs give the Average Temperature Response.
- Lines 19-20: Clarify: “lower estimates” of the ATR?
- Line 49: Not sure that “short-cut” is the right word, because there is also a loss of accuracy. “Approximation” would be a better word.
- Lines 55 and 62: Need to define what is meant by “perturbation approach” and “tagging contribution approach”. It might be useful to add a paragraph before this one to clearly define those two approaches to attribution, and what they mean in terms of radiative forcing. As you found in Grewe et al. (2019) https://doi.org/10.1088/1748-9326/ab5dd7, the impact of the choice of methods is very large: a factor 6.7 according to Table 1 in that paper.
- Line 61: I do not understand “for eight specific days”. Do you mean eight specific weather situations?
- Line 68: “to local meteorological conditions” is too broad. It is in fact “to a selected set of variables that represent the local meteorological conditions”.
- Line 77: Need to clarify what is meant by “effectiveness” here.
- Line 83: The meaning of “the room for a calibration process is open” is only clarified in the next paragraph (line 88) by “one realization within the range of plausible values”, so I would suggest merging the two paragraphs to clarify what you are doing. And it should be noted that such calibration does not replace the need for a way to properly account for uncertainty in both AirClim and aCCFs.
- Line 109: “it is equivalent to selecting another value from this interval”. Is that true? If there’s a probability distribution, then some values are more probable than others, although I agree that the most probable value is not necessarily the mean.
- Lines 126-127, lines 150-151, line 154: Again, avoid using the imprecise “climate effect”. Here, I assume that it is the ATR that is calculated.
- Line 140: “as boundary conditions” – ambiguous. Do you mean that there is no nudging inside the model domain? By the way, is the model configured to simulate Europe only, or is it global?
- Line 149: The aCCFs are made out of the CCFs for 8 weather patterns. If I understand well that synoptic information is lost in the aCCFs, so it is not possible to relate aCCFs to specific patterns in the North Atlantic. Is that correct? It would be good to clarify that here. And by the way, are the aCCFs fitted over the 8 weather patterns together? Or are they fitted over some mean of the 8 weather patterns?
- Line 173: Why A330 flights only?
- Lines 182-186: Again, no need to keep the suspense on the actual calculation by using the imprecise “climate effect”. Say that you are calculating P-ATR20 from the start of the paragraph.
- Caption of Table 1: The “without forcing efficacy” comes as a surprise. There should be an explanation of what that means in the main text.
- Line 227: Converting to F-ATR20 when Section 3 has been all about P-ATR20 seems an unnecessarily confusing step. Why do it? And here again, the reader would benefit from a reminder of the notion of efficacy, and how it might change the results compared to Section 3.
- Line 245: I would expect variability in the contrail aCCF, but I would also expect some correlation between time steps, since ISSRs are not advected randomly. But then the aCCFs do not have the concept of ISSR, so what do the patterns shown in Figure 2c tell us? It is also surprising to have pockets of cooling in the middle of warming zones – that does not look like the day/night contrast alluded to in the text.
- Line 307: What is the implication of that narrower distribution for climate-optimised routing?
- Lines 366-367: What do those confidence intervals look like? How do you go from your calibration factors to confidence intervals?
Technical comments:
- Line 26: Suggest rewriting to “long-term background ozone is reduced by the NOx-induced methane decrease”.
- Line 43: Presenting the three types of studies as bullet point would make that paragraph easier to read.
- Line 88: I have never seen “event horizon” used in this context. Is that a correct use of the term?
- Line 115: Typo aligned
- Line 117: “is corresponding” -> corresponds
- Figure 1: Typo relvant (and it would be good to disable the spelling checker to avoid those words underlined in red)
- Line 164: green function -> Green’s function
Citation: https://doi.org/10.5194/gmd-2023-92-RC1 - AC3: 'Reply on RC1', Sigrun Matthes, 19 Apr 2024
-
RC2: 'Comment on gmd-2023-92', Anonymous Referee #2, 16 Jan 2024
In this manuscript the authors compute the non-CO2 effects of aviation using their algorithmic climate change functions (aCCF, V1.0), which themselves are defined from a suite of models, and the AirClim v2.0 model for a set of city-pairs in Europe and a full seasonal cycle. From these simulations they compute renormalization factors (one factor for each non-CO2 effect) to derive calibrated aCCFs (V1.0A). The authors also present PDFs of the aCCFs over a European domain and a comparison between the uncalibrated and calibrated aCCF.
I have a number of major issues with this manuscript. In my opinion each one of these issues could prevent publication in its own right so I urge the authors to consider them carefully.
1/ Calibration implies reference.
Calibration is the process by which a measurement is adjusted to a reference. It is not clear to me why AirClim V2.0 (as described in Dahlmann et al. 2016) provides such a reference. It is claimed that AirClim is “comprehensive”, “state-of-the-art”, “well-established” but this is not demonstrated. I would welcome elements that prove AirClim is a reference. To which extent is it validated against observations? How is it documented? How does it compare to other models? Dahlmann et al (2016) is now 8 years old, is it still state-of-the-art?
2/ Calibration parameters are beyond reasonable.
The calibration parameters (f_AirClim in Annex A) are much lower than 1 for some of the non-CO2 effects, i.e. 0.333 for contrail-cirrus, 0.179 for O3 and even 0.058 for CH4. Rather than just rescaling their parametrisation, the authors should go back to their models, understand the root causes of the differences, reformulate the model and try to narrow down the discrepancies. I personally have little trust in the structure of a model that requires a scaling factor of 0.058 against a reference.
3/ Tagging vs perturbation approach.
As stated on AirClim is based on a perturbation approach (as stated on lines 54-55), i.e. it computes the marginal change in RF due to the perturbation of NOx from aviation emissions. My understanding is that aCCFs V1.0 are based on the tagging approach which places all the NOx emissions (from different sectors) on a same footing. The perturbation (or marginal) approach has the advantage of representing the change expected from an action (e.g. rerouting) but has the disadvantage of not being additive and the sum of the relative perturbations from different sectors or regions usually do not add up to 100%. The tagging (and other similar methods) have the advantage of being invariant to disaggregation and recombination and are a better measure of the contribution of a sector to the total forcing but are not adapted to quantifying the impact of a change when everything else is fixed. My understanding is that the calibration process shifts the aCCF from the tagging to the perturbation approach (lne 386). For the methodology to be valid, the authors need to show that there is a proportionality coefficient between the two approaches that holds for individual flights and not just on average. Otherwise there should be a clear warning that aCCFs V1.0A should not be used for individual flights.
4/ There is a serious lack of treatment of uncertainties for aCCF V1.0A.
In their present form aCCFs do not come with uncertainties. To be useful, aCCFs should be associated with an uncertainty range (e.g., 1 or 2 sigma or 90% uncertainty range). This manuscript offers the opportunity to address this lasting deficiency of aCCFs by tracking down and combining the different sources of uncertainties. First of all, the aCCFs are expressed as a function of a very limited set of predictive variables and there is a significant dispersion around the average relationship. This implies an uncertainty range that needs to be documented and propagated into aCCFs v1.0A. AirClim V2.0 also has uncertainties that probably add quadratically to those of the aCCF formulation. A third source of uncertainty comes from the representativity of a climatological calibration for individual flights. This can be diagnosed from the dispersion of the calibration coefficients estimated on a flight-by-flight basis and should also be added quadratically to other sources of uncertainties. Finally, there are other sources of uncertainties (e.g., the masking of source regions for the contrail aCCF) that need to be discussed if they cannot be estimated properly.
5/ aCCFs V1.0A lack traceability and transparency
aCCFs lack the traceability and transparency needed by users. aCCFs V1.0A make the situation even worse as they result from a long suite of models (some of which are not freely available) and simulations (some of which not accessible). I am worried that aCCFs V1.0A will be used without any consideration of their inherent uncertainties and/or outside their validity range. The authors need to provide a more traceable workflow with the input and output of the models that are used to produce the aCCF V1.0A. Only in this way can the aCCF be compared to other approaches and become a trustworthy source of information for users.
6/ The choice of metric is unusual and misleading.
The default version of aCCFs comes are for the ATR-20 climate metric (temperature change averaged over a 20 year period) and aCCFs V1.0A are also illustrated for the ATR-20. This is an unusual choice within the large body of literature on climate metrics. It also contrasts with policy choices made in UNFCCC (see also recent SBSTA decision to retain GWP100 in the wake of the Paris Agreement). The ATR-20 metric puts a lot of weight on the climate effects of short-lived species. Sure there is a policy dimension in climate metrics but there are also scientific and economical considerations that favour putting more weight on long-lived greenhouse gases, at least until CO2 emissions are curbed significantly. What is so special to the aviation sector to favour ATR20 over metrics with longer time horizon (e.g. GWP100 or GTP50) that are more consistent with the current mitigation levels? At the very least aCCFs V1.0A should be presented with several climate metrics. Highlighting ATR20 can be very misleading for users who are not very knowledgeable on climate metrics.
Other comments:
Title: the title does not describe the content of the manuscript. There is no evaluation of aCCF against AirClim V2.0. Performing such an evaluation would imply to show a range of score of aCCF against AirClim V2.0 for individual flights. As mentioned above, I am also dubious that this is a calibration. In any case the title needs to be changed and reflect the content of the manuscript.
Line 24: this needs to be rephrased as it is not “emissions” per se that are radiatively active or not but the molecules in the atmosphere.
Lines 27-28: it would be useful to say a bit more about contrails and induced cirrus at this point.
Lines 29-32: the two sentences repeat rather than complement each other. The “accordingly” does not read well. The first sentence omit the dependence on environmental factors (position of the Sun, surface albedo, clouds, …).
Lines 63-64: see above for an alternative view.
Line 69: this may be a strength but this is also their main weakness.
Lines 75-76: given the bold assumptions made in aCCF and the absence of characterization of uncertainties, it is dangerous if not fallacious to encourage climate-optimized flight trajectories.
Lines 83-84: I agree that the uncertainty of aCCF estimates is missing but this manuscript does nothing to better quantify such uncertainties.
Lines 107-112: this paragraph is not clear and I could not link it to the calibration approach described later in the manuscript.
Figure 1: why are the bottom arrows pointing to the green boxes rather than starting from there?
Line 140: I understand MESSy2 is nudged to ERA5 but why is this considered as “boundary conditions”? What does it imply for the simulation of ISSR? How does it compare to observations?
Line 153: see above. Considering great circles is an issue because there are co-variations between the jet stream (which airlines will try to avoid or use depending on the direction) and the presence of ISSR. It is well known that there are significant departures from the great circle. If aCCF are not computed for real trajectories, how can it be assumed that it is useful for designing climate-optimized trajectories?
Line 164-165: How does the Green function for CO2 from Hasselmann et al (1997) compare to more recent estimate? To which extent does it depend on future emission scenarios?
Line 169: is it a “calibration”, a “comparison” or an “evaluation” (as per the title)?
Table 1: SI units are m and kg rather than km and g. EGU journals recommend to use SI units unless there is a good justification not to. Is there a good reason to deviate from SI units in Table 1?
Table 1: I assume the 85 flights operate daily during a year, hence the ~50 millions km flown. Unless I missed the information, the fact that the flights are assumed to operate daily is missing.
Line 229: Kg should read kg.
Line 229 and elsewhere: for the sake of clarity, it should be stated that this is kg NOx as NO2.
Figure 2: the caption should state whether the figure shows the uncalibrated (V1.0) or the calibrated (V1.0A) aCCF.
Figure 2: the figure shows a filament of elevated aCCF for water vapour. I guess this corresponds to a filament of larger PV values indicative of a tropopause folding. If so, the residence time of the water vapour in the stratosphere is short and the non-CO2 effect is probably much less than indicated by the aCCF. What is the confidence level / uncertainty on these values?
Lines 267-268: why not show or at least give the fraction of aCCF that is zero?
Figures 3 and 4: the caption should make it clear that the PDFs are for calibrated aCCF.
Figures 3 and 4: I could not find the % of non-zero aCCF on panels c.
Figure 5: V1.1 should read V1.0A on top of panel b.
Line 317: how is AirClim “well established” ?
Lines 331-334: does this mean that the aCCFs depend on the choice of trajectories for a given set of city pairs? If the calibration depends on the assumption of great circles between city pairs then do the calibration coefficients hold for climate-optimized trajectories?
Line 357: can the authors elaborate on the issue here?
Lines 373-375: I disagree here. A parametrized approach is required for the medium- to long-lived non-CO2 and CO2 effects, but not necessarily for the short-lived effects such as contrails and contrail-cirrus.
Lines 418-422: this is indeed a limitation and the associated uncertainty should be quantified.
Line 456: I agree with this statement, therefore the purpose of the manuscript should be to decrease uncertainty not rescale uncertain parameters.
Line 470ff: This paragraph contradicts lines 72-73 that specify aCCF is also available as an open source Python library of CLIMaCCF V1.0. Why not mention this in the “code availability” section? Or is it not the version used for this manuscript?
Lines 511-514: please provide details of the final version (rather than the submitted version).
Lines 694-696: please provide details of the final version (rather than the submitted version).
References: please make sure the format of the references is consistent through the list.
Lines 703ff: I find many assumptions to be unnecessarily simplistic, especially those related to thresholds (eg time of day for contrails) or maximum values (eg radiation). Space or time integrals should be more appropriate.
Line 736: it is not specified how persistent contrails are defined.
Line 737: Change Yin et al (2022) to (2023).
Lines 753-754: this is an extreme example of the lack of transparency I mentioned above.
Fig B1: is that for uncalibrated or calibrated aCCFs?
Citation: https://doi.org/10.5194/gmd-2023-92-RC2 - AC2: 'Reply on RC2', Sigrun Matthes, 19 Apr 2024
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 451 | 172 | 44 | 667 | 40 | 38 |
- HTML: 451
- PDF: 172
- XML: 44
- Total: 667
- BibTeX: 40
- EndNote: 38
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
