the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Estimating the uncertainty of the greenhouse gas emission accounts in Global Multi-Regional Input-Output analysis
Abstract. Global multi-regional input-output (GMRIO) analysis is the standard tool to calculate consumption-based carbon accounts at the macro level. Recent inter-database comparisons have exposed discrepancies in GMRIO-based results, pinpointing greenhouse gas (GHG) emission accounts as the primary source of variation. A few studies have delved into the robustness of GHG emission accounts, using Monte-Carlo simulations to understand how uncertainty from raw data propagates to the final GHG emission accounts. However, these studies often make simplistic assumptions about raw data uncertainty and ignore correlations between disaggregated variables.
Here, we compile GHG emission accounts for the year 2015 according to the resolution of EXIOBASE v3, covering CO2, CH4 and N2O emissions. We propagate uncertainty from the raw data, namely the United Nations Framework Convention on Climate Change (UNFCCC) and EDGAR inventories, to the GHG emission accounts, and then further to the GHG footprints. We address both limitations from previous studies. First, instead of making simplistic assumptions, we utilise authoritative raw data uncertainty estimates from the National Inventory Reports (NIR) submitted to the UNFCCC and a recent study on uncertainty of the EDGAR emission inventory. Second, we account for inherent correlations due to data disaggregation by sampling from a Dirichlet distribution.
Our results show a median coefficient of variation (CV) for GHG emission accounts at the country level of 4 % for CO2, 12 % for CH4, and 33 % for N2O. For CO2, smaller economies with significant international aviation or shipping sectors show CVs as high as 94 %, as seen in Malta. At the sector level, uncertainties are higher, with median CVs of 94 % for CO2, 100 % for CH4, and 113 % for N2O. Overall, uncertainty decreases when propagated from GHG emission accounts to GHG footprints, likely due to the cancelling out effects caused by the distribution of emissions and their uncertainties across global supply chains. Our GHG emission accounts generally align with official EXIOBASE emission accounts and OECD data at the country level, though discrepancies at the sectoral level give cause for further examination.
We provide our GHG emission accounts with associated uncertainties and correlations at https://doi.org/10.5281/zenodo.10041196 (Schulte et al. 2023) to complement the official EXIOBASE emission accounts for users interested in estimating the uncertainties of their results.
- Preprint
(1842 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on essd-2023-473', Anonymous Referee #1, 13 Feb 2024
The authors did a great job, they processed a huge amount of badly structured information from the national inventory reports in pdf format and other datasets. The data they compiled and estimated on uncertainties of the National GHG emission estimates submitted by countries to the UNFCCC and for the Global Multi-Regional Input-Output analysis are new, at least I did not find similar data. The data are useful for researchers as well as for policy makers because knowing the uncertainties is as important as knowing the values themselves.
The manuscript contains a very detailed description of the method, results, and a good discussion of limitations. However, the manuscript contains so many details (e.g. explanation of what is Annex-I parties) that sometimes it is challenging to track the most important information. I would suggest moving all the auxiliary information to the footnotes. Otherwise, everything is described very well.
I have two specific comments:
Could you please clarify what is WM in paragraph 485?
Could you please correct the legend in Figure E2 (p.38)?Data quality.
Unfortunately, I failed with running the R script developed by the authors. I could not find the package mRio. Probably, it is my fault as I'm not an experienced user of R.I looked at some of the data separately. However, it would be good if the authors developed a script for exploring the data that does not require special packages.
I could not estimate the quality of the data comprehensively. I put major revisions because of my problem with the R script and, as a result, missing comprehensive data quality assessment.
Citation: https://doi.org/10.5194/essd-2023-473-RC1 -
AC1: 'Reply on RC1', Simon Schulte, 19 Feb 2024
We thank the referee for their valuable comments. While we will respond in detail on each of them once the discussion phase is closed, we want to make a quick note on our R script that the referee failed to run:Â
First, I apologize for the issues the referee had running the code, which is entirely my fault. I removed the dependency on the 'mRio' package (which is only available on my Github) and included all functions in the repository itself. Moreover, I added two missing data files to the data folder (they accidentally landed in .gitignore), and added all third-party packages to the renv lockfile.Â
The updated version of the R-scripts can be found here: https://github.com/simschul/uncertainty_GHG_accounts/releases/tag/v0.1.1
Citation: https://doi.org/10.5194/essd-2023-473-AC1 - AC2: 'Reply on RC1', Simon Schulte, 12 Mar 2024
-
AC1: 'Reply on RC1', Simon Schulte, 19 Feb 2024
-
RC2: 'Comment on essd-2023-473', Anonymous Referee #2, 19 Feb 2024
This is a good paper. Uncertainties in GHG accounts (production- and consumption-based) is hot topic and very high on the agenda of modellers and decision makers alike. The authors do a great job in shedding light on various issues. I was not aware that uncerainty estimates of the UNFCCC data set did exist in the NRI. This paper is clearly contributing to leveraging this infromation and providing it to a wider audience. The findings are relevant in highlighting the sectors and regions that are of concerne. Using Dirichlet distributions to capture correlations in data disaggregation is a smart move. Using entropy approaches in setting the paramters of the Dirichlet distribution is even smarter. I like that approach. The GHG extensions and uncertainties and meta data provided by the authors is very useful. I suggest this paper for publication after a minor revision.
1. So how did the EXIOBASE team construct the GHG extensions? I am sure the answer to that question is already contained in the manuscript but I would love to see this stated more clearly. Is the EXIOBASE extension based on UNFCCC alone? This is not clear to me after reading the text and when MC results are compared to the original EXIOBASE extension are was wondering what I am lookig at here.
2. You have all this information on uncertainties. Would it be possible for you to suggest a sector classification for GHG footprins from EXIOBASE that some incorporates that insights on uncertainties. I mean you implicitly say that assessing GHG footprints on the level of 163 sectors, is for many sector not a good idea due to the uncertainties. And yes, I am aware that such a decision depends greatly on the research question. However, if one aims at an analysis of the GHG footprints of sectors for a given country, what would be a robust sector aggregation that somehow "outsources" or "collects" the more problematic sectors in a lets call it "highly uncertain" sector group (something like products nec). This would be of high value for analysts. Maybe that is beyond the scope of your paper, but I think you are highly qualified to give such a suggestion or at least some reflections.
3. You provide a good overview on the different approaches to estimating uncertainties of the production-based emissions accounts (heuristics vs. modelling with power series regressions): Am I right in assuming that all power series regressions actually rely on the "law of large numbers"? Moreover, looking the MC results for sectors and countries, it seems to me you arrive at similar findings that could be understood as supporting the law of large numbers assumption: The larger a sector or region the smaller the uncertainty in relative terms. I don't want you to reframe the introduction but I was wondering whether to a ceratin degree the law of large numbers and more simplistic approach are actually justified. Would be interesting to see some reflections how your findings connect to that.
4. I am confused by the terms "classification" and "categories" in the context of the UNFCCC data. See Figure 3 for example. Does category stand for process/industry and classification for the type GHG emissions i.e. flow? I am confused. Please clarify that somehow a bit better.
5. You make a great job in detailing the importance of how to allocate international road transport, especially for european countries. I would love to read a bit more about that in your reflections. Is there something you would recommend to modellers that are dealing with the same problem (best practice)? This would be a nice to have, not a must-have. I think this is really another important intervention point for dealing with uncertainties in the future, which deserves a bit more attention.
Â
The DOI link you give to your GHG extensions (https://doi.org/10.5281/zenodo.10037713 ) is forwarding me to the UNFCCC uncertainty data set (https://zenodo.org/records/10037714 ). Is this meant to be like that? Please check the links.
Â
All in all, a well written and interesting piece of work. I congratulate the authors.
Citation: https://doi.org/10.5194/essd-2023-473-RC2 - AC3: 'Reply on RC2', Simon Schulte, 12 Mar 2024
Status: closed
-
RC1: 'Comment on essd-2023-473', Anonymous Referee #1, 13 Feb 2024
The authors did a great job, they processed a huge amount of badly structured information from the national inventory reports in pdf format and other datasets. The data they compiled and estimated on uncertainties of the National GHG emission estimates submitted by countries to the UNFCCC and for the Global Multi-Regional Input-Output analysis are new, at least I did not find similar data. The data are useful for researchers as well as for policy makers because knowing the uncertainties is as important as knowing the values themselves.
The manuscript contains a very detailed description of the method, results, and a good discussion of limitations. However, the manuscript contains so many details (e.g. explanation of what is Annex-I parties) that sometimes it is challenging to track the most important information. I would suggest moving all the auxiliary information to the footnotes. Otherwise, everything is described very well.
I have two specific comments:
Could you please clarify what is WM in paragraph 485?
Could you please correct the legend in Figure E2 (p.38)?Data quality.
Unfortunately, I failed with running the R script developed by the authors. I could not find the package mRio. Probably, it is my fault as I'm not an experienced user of R.I looked at some of the data separately. However, it would be good if the authors developed a script for exploring the data that does not require special packages.
I could not estimate the quality of the data comprehensively. I put major revisions because of my problem with the R script and, as a result, missing comprehensive data quality assessment.
Citation: https://doi.org/10.5194/essd-2023-473-RC1 -
AC1: 'Reply on RC1', Simon Schulte, 19 Feb 2024
We thank the referee for their valuable comments. While we will respond in detail on each of them once the discussion phase is closed, we want to make a quick note on our R script that the referee failed to run:Â
First, I apologize for the issues the referee had running the code, which is entirely my fault. I removed the dependency on the 'mRio' package (which is only available on my Github) and included all functions in the repository itself. Moreover, I added two missing data files to the data folder (they accidentally landed in .gitignore), and added all third-party packages to the renv lockfile.Â
The updated version of the R-scripts can be found here: https://github.com/simschul/uncertainty_GHG_accounts/releases/tag/v0.1.1
Citation: https://doi.org/10.5194/essd-2023-473-AC1 - AC2: 'Reply on RC1', Simon Schulte, 12 Mar 2024
-
AC1: 'Reply on RC1', Simon Schulte, 19 Feb 2024
-
RC2: 'Comment on essd-2023-473', Anonymous Referee #2, 19 Feb 2024
This is a good paper. Uncertainties in GHG accounts (production- and consumption-based) is hot topic and very high on the agenda of modellers and decision makers alike. The authors do a great job in shedding light on various issues. I was not aware that uncerainty estimates of the UNFCCC data set did exist in the NRI. This paper is clearly contributing to leveraging this infromation and providing it to a wider audience. The findings are relevant in highlighting the sectors and regions that are of concerne. Using Dirichlet distributions to capture correlations in data disaggregation is a smart move. Using entropy approaches in setting the paramters of the Dirichlet distribution is even smarter. I like that approach. The GHG extensions and uncertainties and meta data provided by the authors is very useful. I suggest this paper for publication after a minor revision.
1. So how did the EXIOBASE team construct the GHG extensions? I am sure the answer to that question is already contained in the manuscript but I would love to see this stated more clearly. Is the EXIOBASE extension based on UNFCCC alone? This is not clear to me after reading the text and when MC results are compared to the original EXIOBASE extension are was wondering what I am lookig at here.
2. You have all this information on uncertainties. Would it be possible for you to suggest a sector classification for GHG footprins from EXIOBASE that some incorporates that insights on uncertainties. I mean you implicitly say that assessing GHG footprints on the level of 163 sectors, is for many sector not a good idea due to the uncertainties. And yes, I am aware that such a decision depends greatly on the research question. However, if one aims at an analysis of the GHG footprints of sectors for a given country, what would be a robust sector aggregation that somehow "outsources" or "collects" the more problematic sectors in a lets call it "highly uncertain" sector group (something like products nec). This would be of high value for analysts. Maybe that is beyond the scope of your paper, but I think you are highly qualified to give such a suggestion or at least some reflections.
3. You provide a good overview on the different approaches to estimating uncertainties of the production-based emissions accounts (heuristics vs. modelling with power series regressions): Am I right in assuming that all power series regressions actually rely on the "law of large numbers"? Moreover, looking the MC results for sectors and countries, it seems to me you arrive at similar findings that could be understood as supporting the law of large numbers assumption: The larger a sector or region the smaller the uncertainty in relative terms. I don't want you to reframe the introduction but I was wondering whether to a ceratin degree the law of large numbers and more simplistic approach are actually justified. Would be interesting to see some reflections how your findings connect to that.
4. I am confused by the terms "classification" and "categories" in the context of the UNFCCC data. See Figure 3 for example. Does category stand for process/industry and classification for the type GHG emissions i.e. flow? I am confused. Please clarify that somehow a bit better.
5. You make a great job in detailing the importance of how to allocate international road transport, especially for european countries. I would love to read a bit more about that in your reflections. Is there something you would recommend to modellers that are dealing with the same problem (best practice)? This would be a nice to have, not a must-have. I think this is really another important intervention point for dealing with uncertainties in the future, which deserves a bit more attention.
Â
The DOI link you give to your GHG extensions (https://doi.org/10.5281/zenodo.10037713 ) is forwarding me to the UNFCCC uncertainty data set (https://zenodo.org/records/10037714 ). Is this meant to be like that? Please check the links.
Â
All in all, a well written and interesting piece of work. I congratulate the authors.
Citation: https://doi.org/10.5194/essd-2023-473-RC2 - AC3: 'Reply on RC2', Simon Schulte, 12 Mar 2024
Data sets
Uncertainty of EXIOBASE GHG emission acounts 2015 Simon Schulte, Arthur Jakobs, and Stefan Pauliuk https://zenodo.org/records/10041196
Uncertainties from the UNFCCC National Inventory Reports (submission 2017) Simon Schulte and Joshua Heipel https://zenodo.org/records/10037714
Correspondence table between UNFCCC CRF and EXIOBASE industry sectors Simon Schulte https://zenodo.org/records/10046372
Model code and software
Code for estimating the uncertainty of EXIOBASE GHG emissions accounts Simon Schulte https://github.com/simschul/uncertainty_GHG_accounts
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
448 | 104 | 32 | 584 | 19 | 25 |
- HTML: 448
- PDF: 104
- XML: 32
- Total: 584
- BibTeX: 19
- EndNote: 25
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1