Uncertainty in temperature response of current consumption-based emissions estimates

Karstensen, J.; Peters, G. P.; Andrew, R. M.

doi:https://doi.org/10.5194/esd-6-287-2015

Articles | Volume 6, issue 1

https://doi.org/10.5194/esd-6-287-2015

© Author(s) 2015. This work is distributed under
the Creative Commons Attribution 3.0 License.

https://doi.org/10.5194/esd-6-287-2015

© Author(s) 2015. This work is distributed under
the Creative Commons Attribution 3.0 License.

Articles | Volume 6, issue 1

Research article

|

27 May 2015

Research article |

| 27 May 2015

Uncertainty in temperature response of current consumption-based emissions estimates

J. Karstensen, G. P. Peters, and R. M. Andrew

Download

Final revised paper (published on 27 May 2015)
Preprint (discussion started on 09 Sep 2014)
Supplement to the preprint

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC C510: 'Review for 'Uncertainty in temperature response of current consumption-based emissions estimates'', Joshua Elliott, 15 Oct 2014
- AC C710: 'Response to reviewer', Jonas Karstensen, 31 Jan 2015
RC C562: 'Review of "Uncertainty in temperature response of current consumption-based emissions estimates', Anonymous Referee #2, 14 Nov 2014
- AC C711: 'Response to reviewer', Jonas Karstensen, 31 Jan 2015

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Jonas Karstensen on behalf of the Authors (26 Feb 2015) Author's response Manuscript

ED: Referee Nomination & Report Request started (09 Mar 2015) by Daniel Kirk-Davidoff

RR by Joshua Elliott (23 Mar 2015)

Suggestions for revision or reasons for rejection

Review 2

I am mostly satisfied by the response to the correlations issue. I’m still disappointed that even a synthetic example couldn’t be included to show the importance of correlations, but the caveats included in the paper are sufficient.

In general I would suggest that the authors take advantage of high-performance computing in some capacity. I appreciate that on a single desktop machine the scale of these computations is large, but this problem is “embarrassingly parallel” and using even a small cluster of nodes I suspect the authors would be able to overcome all the computational limitations frequently mentioned, including to do the correlation analysis, rebalancing matrices in all the MCs and etc. I don’t think the authors ever mention what software package they are using for this. I would assume they are using GAMS. Since GAMS is not open source it can be challenging to get it installed on large clusters, but I’ve always found the folks at GAMS to be happy to help. We even installed and ran GAMS simultaneously on one million cores of a supercomputer at Argonne once. Might even still be installed if the authors would like to try and get access to use it. Alternatively of course, the authors could use an open package like AMPL to get around this constraint. Either way, computational limitation for an easily parallizable problem like this should never be a limitation in this day and age.

Regarding the relationship between sector sizes and uncertainty, I think the issue here is again one of ignoring correlations. When talking about real error from all the possible sources:
errors related to “unreliable measurements, estimates and assumptions, bias in source data, temporal, geographical, and technological miscorrelation, and lack of knowledge about the system” – Lenzen 2000
the argument for why large sectors would have smaller errors is again relying on the assumption that these types of errors within the sector are cancelling. However, this assumption/caveat isn’t stated. Instead, the authors are using the relationship from the GTAP “uncertainty” which shows smaller errors for larger sectors. But this is ONLY the error introduced from matrix rebalancing. Its not at all surprising that this one source of error should show this behavior but to suggest that all economic error should show this behavior because of this fact seems going much too far. I think this is plainly an incorrect assumption and since it wouldn’t increase the computational requirement or difficulty to relax it, it seems like its worth evaluating the consequences. At the least this caveat needs to be added, in a way similar to Lenzen-2000 (in the introduction).

Once again, I would strongly suggest that the section “general uncertainty relationships” needs an example. The process as it stands looks incredible ad hoc. The authors choose a reasonable approach but then add various adjustments and procedures that make it very difficult to intuit exactly what the uncertainty is. The inclusion of a couple of short examples that said “the concrete industry in china is X billion dollars”, its non-annexB so we see rmin = x rmax = y and then we get an uncertainty range like z but then we rescale it to reflect the percentage of Chinese GDP that comes from concrete so the final error estimate is Q, which is J% of the total sector size.

Hide

ED: Publish subject to minor revisions (review by Editor) (23 Mar 2015) by Daniel Kirk-Davidoff

AR by Jonas Karstensen on behalf of the Authors (08 Apr 2015) Author's response Manuscript

ED: Publish as is (13 Apr 2015) by Daniel Kirk-Davidoff

AR by Jonas Karstensen on behalf of the Authors (14 Apr 2015)

Short summary

We quantify uncertainties in estimates of global temperature change from regional and sectoral territorial- and consumption-based emissions. We find that the uncertainties are sensitive to the emission allocations, mix of pollutants, the metric used and its time horizon, and the level of aggregation of the results. Uncertainties in the final results are dominated by metric parameters and emission uncertainties, while the economic data appear to have small uncertainties at the national level.