the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The CO2 and non-CO2 climate effects of individual flights: simplified estimation of CO2 equivalent emission factors
Abstract. As aviation's contribution to anthropogenic climate change is increasing, industry aims at reducing the aviation climate effect. However, the large contribution of non-CO2 effects to the total climate effect of aviation and their large variability for each individual flight inhibit finding appropriate guidance. Here, we present a method for the simplified calculation of CO2 equivalent emissions, expressed using the physical climate metrics ATR100 or AGWP100, from CO2 and non-CO2 effects for a given flight, exclusively based on the aircraft seat category as well as the origin and destination airports. The simplified calculation method estimates non-CO2 climate effects of air traffic as precisely as possible, without detailed information on the actual flight route, actual fuel burn, and current weather situation. For this purpose, we evaluate a global data set containing detailed flight trajectories, flight emissions, and climate responses, and derive a set of regression formulas for climate effects, which we call climate effect functions, as well as regression formulas for fuel consumption and NOx emissions. Compared to previous studies, this method is available for a larger number of aircraft types, including most commercial airliners with seat capacities starting from 101 passengers, and delivers more specific results through a clustering approach. The climate effects calculated using the climate effect functions derived in this study exhibit a mean absolute relative error of 15.0 % and a root mean square error of 1.24 nK with respect to results from the climate response model AirClim. The climate effect functions are designed for climate footprint assessments, but would not create an incentive in an emission trading system, for which detailed information on the current weather as well as the actual flight route and profile would be required.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(8995 KB)
-
Supplement
(8542 KB)
-
This preprint has been withdrawn.
- Preprint
(8995 KB) - Metadata XML
-
Supplement
(8542 KB) - BibTeX
- EndNote
Interactive discussion
Status: closed
-
RC1: 'Review of Thor et al.', Anonymous Referee #1, 25 Aug 2023
In this study, the authors describe a method to compute the 100-year average temperature response for emissions of carbon dioxide, nitrogen oxides, water vapour, and contrail formation of an individual flight. This is done by first combining the WeCare aviation emission inventory with AirClim emission-climate response relationships, averaged over all weather situations. Then, flights are categorised into three clusters (short flight [<~500 km], tropical, mid-latitude) based on the contribution of different species to total ATR100.
The paper is short and to the point, and well written. The usefulness of the proposed equations is well motivated. Yet I worry that the authors are straying into “too simple” territory, and that the proposed method will lead to large errors in the estimation of the climate impact of individual flights, especially for the longest flights that impact climate the most.
My three main comments are:
- Why is the clustering/regression step needed at all? WeCare+AirClim would be more accurate, although AirClim errors are large already. And that should be inexpensive to run on an individual flight basis. Even if a flight is not in the WeCare inventory, it should be possible to approximate the corresponding AirClim input parameters in a way that is more accurate that the clustered fits described by Eqs 3-5.
- Page 9: “based on the behavior of the respective values” – that part of the methodology needs much clarification. How were the sets of equations 3 to 5 obtained exactly? Which shapes were tested? There are some explanations on lines 205-215, but they are very subjective and difficult to reproduce.
- Figures 3-5, line 208, lines 239-247: Related to the previous comment: The grey dots on Figures 3-5, which correspond to the results of the fitting sets of equations 3-5, clearly only fit a subset of the space, as acknowledged by the authors in Section 3. The climate impact is underestimated for short flights with flight distances over 300 km and for many flights with distances over 4000 km. That is a problem. Why not refine the clusters to decrease errors? Shouldn’t the short-flight threshold be ~300 km instead of ~500 km? Shouldn't new clusters be introduced for the longest flights? Additional clusters would make the tool more complicated but that would be very justified in return of an improved accuracy.
Other comments:
Line 4: The next mention of AGWP100 is in Appendix A. There should be a discussion of the implications of the choice of climate metrics somewhere in the main body of the paper.
Line 28: “e.g. by replacing in-person by virtual meetings” – there are many ways to reduce the number of flights, so the selection of that example feels a bit random. In fact, I do not think an example is needed here.
Line 33: “renewable energy” – the energy only needs to be decarbonized to be carbon neutral. It does not have to be renewable.
Line 47: The link between CO2e and species-wise ATR100 values, as calculated by Eqs 3-5, is unclear. It would be useful to give the equation linking the two somewhere in the paper.
Lines 72-73: “[…] we here use increasing emissions over the next 100 years. In both studies, the effects of historical emissions are neglected.” It would be useful to clarify here which step of the calculation is affected by that choice, and what are the implications for the climate impact of an individual flight (pre-empting some of the discussions starting line 259).
Lines 132-135: Scenario Fa1 is very old. Is it still relevant?
Line 137: “We quantify the climate impact as the ATR100.” It would be good to briefly justify this choice of CO2 equivalence metrics, and its implication. ATR100 puts more weight on non-CO2 effects than GTP100, for example. It is also a good place to refer to the possibility of using GWP100, as indicated in Appendix A.
Lines 225-228 should be moved to the beginning of section 2.5 to explain to the reader why the topic changes suddenly.
Figure 8: It would be useful to add the 10, 20, and 50% error lines to calibrate the figure and make it easier to interpret.
Technical comments:
Line 40: marked -> market
In acknowledgments: through -> thorough
Citation: https://doi.org/10.5194/gmd-2023-126-RC1 -
RC2: 'Comment on gmd-2023-126', William Collins, 31 Aug 2023
This paper describes a simple parameterisation of a more complex model of the climate effects of flight routes. In essence, this is a relatively simple paper fitting polynomials to model outputs. In spite of the underlying simplicity, it was not always easy to understand what the authors had actually done.
This paper needs to be explained more clearly. A schematic showing the connections, inputs and outputs of the various models would help this. I think the core of the paper is that polynomial fits to a model can reproduce that model to within 15%. The paper would have been much more useful if it showed a use case for these polynomial fits beyond the simple presentation in table 5.
The wide spread in distributions of the climate effects shown in figure 3-5 seems to suggest that the clustering is not selecting subsets of the data that have similar behaviours. The functional forms chosen do not seem appropriate to fit the distributions, for instance using continuous functions when the data has discontinuities. This is most evident in figure 3 where the distribution is obviously bi-modal, but the fit takes an average of the two modes, although many of the fitted points are below all the model data. The mean absolute error of 15% in the climate effects presumably includes the CO2 contribution which is perfectly fit by definition. Hence the errors in the non-CO2 contributions must be considerably worse than 15%. I recommend revisiting this clustering and fitting as it looks by eye as if it could be done a lot better.
The definition of ATR100 is poorly explained. I think the units are mK per flight, but this is not stated. Why is the average temperature change over 100 years a good measure? Is this just the integrated global temperature potential of Peters et al. 2011 divided by 100? How sensitive is the ATR to the time horizon? Lee et al. 2021 find that the GWP for aircraft NOx varies by a factor of 5 between GWP20 and GWP100. How does ATR100 compare to the more usual GWP100? The appendix states that AGWP100 calculations are also available. How are these calculated? Are the clusters and coefficients the same as for ATR100?
I don’t understand how the scenarios are used. I think the flights in figure 1-5 are for 2012. Do the calculations assume the emissions are instantaneous pulses in 2012? In which case do the scenarios simply provide the background conditions of CO2 and CH4 for the 100 year average temperature? i.e. for the decay of the pulses from 2012 to 2112 (does RCP4.5 extend to 2112?). I don’t see how the scenario Fa1 fits in if the flights are for 2012. Why use such an old scenario anyway, our projections of future aircraft types, fuels and activity will have changed drastically since 1992. How do the scenarios from WeCare fit into this. It is stated that they are from 2015 to 2050 in 5-year steps, so these don’t seem appropriate for calculating 100-year average temperatures.
Abstract
Line 4: There is no discussion of GWP100 in the text.
Line 6: What does “as precisely as possible” mean? It seems that the precisions of the polynomial regressions depend on the functional forms and choices made in the clustering. There doesn’t seem to be any evidence that alternatives were examined to see if they were more or less precise.
Line 7: I don’t think “evaluate” is the right word as I don’t see any evidence that the data set was evaluated.
Line 10: I think just one aircraft type was chosen for each seat capacity.
Line 11: What does “more specific results” mean? What are they more specific than?
Line 12: Note this 15% error includes a large contribution from CO2 which is exact. The error from the non-CO2 components will be larger.
Line 13: Units are 1.24 nK per flight (I think).
Lines 49-50: Need to define CO2,e factors i.e. that they are a scaling of the CO2 contribution.
Line 51: The large variation with time applies to the GWP as well (table 5 of Lee et al.) and presumably to ATR too. This is because the non-CO2 effects are all short-lived in contrast to CO2.
Line 73: I don’t understand how “increasing emissions over the next 100 years” are used? It seems that the clustering and polynomial fits are based on pulse emissions 2012 values. These then produce an ATR100 in mK per flight. I can see that if you wanted to calculate the total climate effect of future aircraft you would then apply the ATR100 to the flight scenario, but I don’t think that is done here?
Line 74: I don’t understand what is meant by “the effects of historical emissions are neglected”? The ATR100 metrics are applied to individual flights. There isn’t a concept of “historical” or “future” in an emission metric. The historical and future do determine the background atmosphere into which the emissions are emitted.
Line 75: These aren’t precise estimates, the uncertainty is 15%.
Sections 2.1 and 2.2: These were difficult to follow, maybe a schematic showing how the different models are used would help.
Line 92: Are these Grewe et al. (2017) emission inventories for future scenarios as well?
Line 105: Are the WeCare inventories different to the Grewe et al. ones? This text says they start in 2015, yet later on (line 132) the emissions start in 2012. Are all of them integrated for 100 years? I.e. the 2015 emissions out to 2115 and the 2050 emissions out to 2150?
Lines 115-120: What does it mean that “only aggregated flight and emission inventories were available”? Does that meant that the WeCare 2015-2050 weren’t used? How is “re-calculating the emissions on a per-flight basis” different from running the model all over again?
Lines 124-130: Have these 85 steady-state simulations been run for this study or are the authors describing the input data that are included in the AirClim model?
Line 133: Why is such an old scenario used, it must be entirely out of date?
Lines 139-141: Are the 57631 flights for the whole scenario i.e. covering the 2015 (or 2012) to 2050 period? Or even for the 100 years from 2012 to 2112? Or are they a 2012 snapshot? If they are a snapshot, how are the WeCare or Fa1 scenarios used?
Line 147: PMO is mentioned here, but it isn’t one of the species considered.
Line 153: PMO isn’t included, so there are only 5 components.
Figure 1 and 2: These need to have colour legends on the plots, not in the captions.
Equation (2): Need to explain what the units of cCO2, cNOx, cH2O and cCiC are.
Equation (3). It seems extremely unwise to fit a quartic expression since they can give wild variations for small changes in parameters. This may be why many of the grey points lie beneath all the model results in figure 3(b).
Equation (4) This doesn’t seem to include any latitudinal dependence for c_NOx. Why is that? The southern hemisphere is more sensitive to NOx than the northern. Within the northern hemisphere there is an increase in ATR100 NOx with latitude.
Line 217: Need to give units for EI(NOx). I think it is g(NOx) per kg(fuel).
Table 1. All these coefficients need units otherwise it is difficult to understand what the units of cNOx, cH2O and cCiC are.
Equation (8) a0, a1 and a2 are already used in Eq (7).
Line 225-230: This need to be explained better. Surely the formulas 3-5 need to be combined with BF and EI(NOx) to get climate effects as functions of distance and latitude.
Line 239-240: Do these MARE values include the CO2-effects? What are the MARE values for non-CO2 only? Presumably these would be significantly higher?
Lines 263-267: I think these sentences are confusing different issues and need to be explained better. The RFI and CO2,e factors are different quantities. RFI does indeed depend on the historical emission and CO2,e factors do not. But is isn’t that you have chosen not to consider it. It is a different calculation. The value of RFI depends entirely on the past trajectory of historical emissions and the fact that grown has been roughly exponential (See Lee et al. section 6). A flatter trajectory would give a lower RFI and a steeper trajectory a higher RFI. So the sentence “As the climate impact of …” isn’t quite correct. RFI and ATR100 quantify different effects. You would get much higher factors if you chose ATR50 for instance.
Citation: https://doi.org/10.5194/gmd-2023-126-RC2 -
EC1: 'Comment on gmd-2023-126', Sophie Valcke, 25 Sep 2023
Dear author,
You have contacted me, asking for a delay for the revised version of the manuscript, arguing in particular that the first author has left the scientific arena. You also state that to appropriately address the comments you would like to have the time to look into a) different clustering mechanisms and b) perhaps also different AI based response function analysis in order to enhance the accuracy of the formulas. Furthermore, the two reviewers asked for major revisions.
Given all this, I would encourage you to withdraw the current manuscript, and to submit a new paper later, whenever you will be ready.
Thanks for your feedback on this proposition, which seems the most reasonable to me at this point.
With very best regards
Citation: https://doi.org/10.5194/gmd-2023-126-EC1
Interactive discussion
Status: closed
-
RC1: 'Review of Thor et al.', Anonymous Referee #1, 25 Aug 2023
In this study, the authors describe a method to compute the 100-year average temperature response for emissions of carbon dioxide, nitrogen oxides, water vapour, and contrail formation of an individual flight. This is done by first combining the WeCare aviation emission inventory with AirClim emission-climate response relationships, averaged over all weather situations. Then, flights are categorised into three clusters (short flight [<~500 km], tropical, mid-latitude) based on the contribution of different species to total ATR100.
The paper is short and to the point, and well written. The usefulness of the proposed equations is well motivated. Yet I worry that the authors are straying into “too simple” territory, and that the proposed method will lead to large errors in the estimation of the climate impact of individual flights, especially for the longest flights that impact climate the most.
My three main comments are:
- Why is the clustering/regression step needed at all? WeCare+AirClim would be more accurate, although AirClim errors are large already. And that should be inexpensive to run on an individual flight basis. Even if a flight is not in the WeCare inventory, it should be possible to approximate the corresponding AirClim input parameters in a way that is more accurate that the clustered fits described by Eqs 3-5.
- Page 9: “based on the behavior of the respective values” – that part of the methodology needs much clarification. How were the sets of equations 3 to 5 obtained exactly? Which shapes were tested? There are some explanations on lines 205-215, but they are very subjective and difficult to reproduce.
- Figures 3-5, line 208, lines 239-247: Related to the previous comment: The grey dots on Figures 3-5, which correspond to the results of the fitting sets of equations 3-5, clearly only fit a subset of the space, as acknowledged by the authors in Section 3. The climate impact is underestimated for short flights with flight distances over 300 km and for many flights with distances over 4000 km. That is a problem. Why not refine the clusters to decrease errors? Shouldn’t the short-flight threshold be ~300 km instead of ~500 km? Shouldn't new clusters be introduced for the longest flights? Additional clusters would make the tool more complicated but that would be very justified in return of an improved accuracy.
Other comments:
Line 4: The next mention of AGWP100 is in Appendix A. There should be a discussion of the implications of the choice of climate metrics somewhere in the main body of the paper.
Line 28: “e.g. by replacing in-person by virtual meetings” – there are many ways to reduce the number of flights, so the selection of that example feels a bit random. In fact, I do not think an example is needed here.
Line 33: “renewable energy” – the energy only needs to be decarbonized to be carbon neutral. It does not have to be renewable.
Line 47: The link between CO2e and species-wise ATR100 values, as calculated by Eqs 3-5, is unclear. It would be useful to give the equation linking the two somewhere in the paper.
Lines 72-73: “[…] we here use increasing emissions over the next 100 years. In both studies, the effects of historical emissions are neglected.” It would be useful to clarify here which step of the calculation is affected by that choice, and what are the implications for the climate impact of an individual flight (pre-empting some of the discussions starting line 259).
Lines 132-135: Scenario Fa1 is very old. Is it still relevant?
Line 137: “We quantify the climate impact as the ATR100.” It would be good to briefly justify this choice of CO2 equivalence metrics, and its implication. ATR100 puts more weight on non-CO2 effects than GTP100, for example. It is also a good place to refer to the possibility of using GWP100, as indicated in Appendix A.
Lines 225-228 should be moved to the beginning of section 2.5 to explain to the reader why the topic changes suddenly.
Figure 8: It would be useful to add the 10, 20, and 50% error lines to calibrate the figure and make it easier to interpret.
Technical comments:
Line 40: marked -> market
In acknowledgments: through -> thorough
Citation: https://doi.org/10.5194/gmd-2023-126-RC1 -
RC2: 'Comment on gmd-2023-126', William Collins, 31 Aug 2023
This paper describes a simple parameterisation of a more complex model of the climate effects of flight routes. In essence, this is a relatively simple paper fitting polynomials to model outputs. In spite of the underlying simplicity, it was not always easy to understand what the authors had actually done.
This paper needs to be explained more clearly. A schematic showing the connections, inputs and outputs of the various models would help this. I think the core of the paper is that polynomial fits to a model can reproduce that model to within 15%. The paper would have been much more useful if it showed a use case for these polynomial fits beyond the simple presentation in table 5.
The wide spread in distributions of the climate effects shown in figure 3-5 seems to suggest that the clustering is not selecting subsets of the data that have similar behaviours. The functional forms chosen do not seem appropriate to fit the distributions, for instance using continuous functions when the data has discontinuities. This is most evident in figure 3 where the distribution is obviously bi-modal, but the fit takes an average of the two modes, although many of the fitted points are below all the model data. The mean absolute error of 15% in the climate effects presumably includes the CO2 contribution which is perfectly fit by definition. Hence the errors in the non-CO2 contributions must be considerably worse than 15%. I recommend revisiting this clustering and fitting as it looks by eye as if it could be done a lot better.
The definition of ATR100 is poorly explained. I think the units are mK per flight, but this is not stated. Why is the average temperature change over 100 years a good measure? Is this just the integrated global temperature potential of Peters et al. 2011 divided by 100? How sensitive is the ATR to the time horizon? Lee et al. 2021 find that the GWP for aircraft NOx varies by a factor of 5 between GWP20 and GWP100. How does ATR100 compare to the more usual GWP100? The appendix states that AGWP100 calculations are also available. How are these calculated? Are the clusters and coefficients the same as for ATR100?
I don’t understand how the scenarios are used. I think the flights in figure 1-5 are for 2012. Do the calculations assume the emissions are instantaneous pulses in 2012? In which case do the scenarios simply provide the background conditions of CO2 and CH4 for the 100 year average temperature? i.e. for the decay of the pulses from 2012 to 2112 (does RCP4.5 extend to 2112?). I don’t see how the scenario Fa1 fits in if the flights are for 2012. Why use such an old scenario anyway, our projections of future aircraft types, fuels and activity will have changed drastically since 1992. How do the scenarios from WeCare fit into this. It is stated that they are from 2015 to 2050 in 5-year steps, so these don’t seem appropriate for calculating 100-year average temperatures.
Abstract
Line 4: There is no discussion of GWP100 in the text.
Line 6: What does “as precisely as possible” mean? It seems that the precisions of the polynomial regressions depend on the functional forms and choices made in the clustering. There doesn’t seem to be any evidence that alternatives were examined to see if they were more or less precise.
Line 7: I don’t think “evaluate” is the right word as I don’t see any evidence that the data set was evaluated.
Line 10: I think just one aircraft type was chosen for each seat capacity.
Line 11: What does “more specific results” mean? What are they more specific than?
Line 12: Note this 15% error includes a large contribution from CO2 which is exact. The error from the non-CO2 components will be larger.
Line 13: Units are 1.24 nK per flight (I think).
Lines 49-50: Need to define CO2,e factors i.e. that they are a scaling of the CO2 contribution.
Line 51: The large variation with time applies to the GWP as well (table 5 of Lee et al.) and presumably to ATR too. This is because the non-CO2 effects are all short-lived in contrast to CO2.
Line 73: I don’t understand how “increasing emissions over the next 100 years” are used? It seems that the clustering and polynomial fits are based on pulse emissions 2012 values. These then produce an ATR100 in mK per flight. I can see that if you wanted to calculate the total climate effect of future aircraft you would then apply the ATR100 to the flight scenario, but I don’t think that is done here?
Line 74: I don’t understand what is meant by “the effects of historical emissions are neglected”? The ATR100 metrics are applied to individual flights. There isn’t a concept of “historical” or “future” in an emission metric. The historical and future do determine the background atmosphere into which the emissions are emitted.
Line 75: These aren’t precise estimates, the uncertainty is 15%.
Sections 2.1 and 2.2: These were difficult to follow, maybe a schematic showing how the different models are used would help.
Line 92: Are these Grewe et al. (2017) emission inventories for future scenarios as well?
Line 105: Are the WeCare inventories different to the Grewe et al. ones? This text says they start in 2015, yet later on (line 132) the emissions start in 2012. Are all of them integrated for 100 years? I.e. the 2015 emissions out to 2115 and the 2050 emissions out to 2150?
Lines 115-120: What does it mean that “only aggregated flight and emission inventories were available”? Does that meant that the WeCare 2015-2050 weren’t used? How is “re-calculating the emissions on a per-flight basis” different from running the model all over again?
Lines 124-130: Have these 85 steady-state simulations been run for this study or are the authors describing the input data that are included in the AirClim model?
Line 133: Why is such an old scenario used, it must be entirely out of date?
Lines 139-141: Are the 57631 flights for the whole scenario i.e. covering the 2015 (or 2012) to 2050 period? Or even for the 100 years from 2012 to 2112? Or are they a 2012 snapshot? If they are a snapshot, how are the WeCare or Fa1 scenarios used?
Line 147: PMO is mentioned here, but it isn’t one of the species considered.
Line 153: PMO isn’t included, so there are only 5 components.
Figure 1 and 2: These need to have colour legends on the plots, not in the captions.
Equation (2): Need to explain what the units of cCO2, cNOx, cH2O and cCiC are.
Equation (3). It seems extremely unwise to fit a quartic expression since they can give wild variations for small changes in parameters. This may be why many of the grey points lie beneath all the model results in figure 3(b).
Equation (4) This doesn’t seem to include any latitudinal dependence for c_NOx. Why is that? The southern hemisphere is more sensitive to NOx than the northern. Within the northern hemisphere there is an increase in ATR100 NOx with latitude.
Line 217: Need to give units for EI(NOx). I think it is g(NOx) per kg(fuel).
Table 1. All these coefficients need units otherwise it is difficult to understand what the units of cNOx, cH2O and cCiC are.
Equation (8) a0, a1 and a2 are already used in Eq (7).
Line 225-230: This need to be explained better. Surely the formulas 3-5 need to be combined with BF and EI(NOx) to get climate effects as functions of distance and latitude.
Line 239-240: Do these MARE values include the CO2-effects? What are the MARE values for non-CO2 only? Presumably these would be significantly higher?
Lines 263-267: I think these sentences are confusing different issues and need to be explained better. The RFI and CO2,e factors are different quantities. RFI does indeed depend on the historical emission and CO2,e factors do not. But is isn’t that you have chosen not to consider it. It is a different calculation. The value of RFI depends entirely on the past trajectory of historical emissions and the fact that grown has been roughly exponential (See Lee et al. section 6). A flatter trajectory would give a lower RFI and a steeper trajectory a higher RFI. So the sentence “As the climate impact of …” isn’t quite correct. RFI and ATR100 quantify different effects. You would get much higher factors if you chose ATR50 for instance.
Citation: https://doi.org/10.5194/gmd-2023-126-RC2 -
EC1: 'Comment on gmd-2023-126', Sophie Valcke, 25 Sep 2023
Dear author,
You have contacted me, asking for a delay for the revised version of the manuscript, arguing in particular that the first author has left the scientific arena. You also state that to appropriately address the comments you would like to have the time to look into a) different clustering mechanisms and b) perhaps also different AI based response function analysis in order to enhance the accuracy of the formulas. Furthermore, the two reviewers asked for major revisions.
Given all this, I would encourage you to withdraw the current manuscript, and to submit a new paper later, whenever you will be ready.
Thanks for your feedback on this proposition, which seems the most reasonable to me at this point.
With very best regards
Citation: https://doi.org/10.5194/gmd-2023-126-EC1
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
782 | 348 | 50 | 1,180 | 93 | 35 | 50 |
- HTML: 782
- PDF: 348
- XML: 50
- Total: 1,180
- Supplement: 93
- BibTeX: 35
- EndNote: 50
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
Malte Niklaß
Katrin Dahlmann
Florian Linke
Volker Grewe
Sigrun Matthes
This preprint has been withdrawn.
- Preprint
(8995 KB) - Metadata XML
-
Supplement
(8542 KB) - BibTeX
- EndNote