Avoid restricting the SCI by prescribing a specific metric. #353
Conversation
…nularity. There is currently no scientific consensus on which decision-making signal (marginal, average, potential profiles, etc.) performs best. On top of that, there is no ground truth for marginal - it is impossible to verify whether the marginal value is true or not. As long as the debate is ongoing, it is risky to prescribe marginal emissions as the only valid signal in the SCI and thus we should be open to other metrics. Any metric(s) included in the SCI need to be verifiable and lead to real emissions reductions and with it decarbonise the electricity system. Additionally, this specification focuses on temporal granularity, which is why "hourly" was added. Sources: https://www.cell.com/iscience/fulltext/S2589-0042(22)00185-7?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS2589004222001857%3Fshowall%3Dtrue https://www.pnas.org/doi/10.1073/pnas.2211624119 https://media.rff.org/documents/Report_22-08.pdf https://cebi.org/wp-content/uploads/2022/11/Guide-to-Sourcing-Marginal-Emissions-Factor-Data.pdf
The marginal emission rate is defined as the carbon intensity of the marginal generator. While the marginal emissions rate does indeed represent a value with units gCO2/MWh, it does not necessarily represent the change in emissions with a change in load. There are two main scenarios: 1. if the load is very small, no additional generation will be deployed and the additional load is reflected in a slight change of grid frequency. No additional emissions are emitted as a result of the added load. 2. If the load is large, the marginal emissions rate is not guaranteed to represent the change in emissions, since an additional marginal generator (with a different MER) could be called upon to service the large additional load. The marginal emission factor therefore does not generally apply to consumers with different intervention sizes. Marginal is removed as there is currently no scientific consensus on which signal is the best for short-term decision-making. As long as the debate is ongoing, it is risky to prescribe marginal emissions as the only valid signal in the SCI. Refer to the previous commit regarding hourly granularity.
…ifferent metrics. As highlighted in previous commits, prescribing marginal would pose a risk since past and currently ongoing research indicates doubts whether MERs are able to properly quantify impacts of interventions. Moreover, there is currently no ground truth on MER as it is practically impossible to verify the marginal generator. We advocate to not prescribe a specific carbon intensity metric until there is scientific consensus on it being superior than others. Any metric(s) included in the SCI need to be verifiable and lead to real emissions reductions and with it decarbonise the electricity system. Sources: https://www.cell.com/iscience/fulltext/S2589-0042(22)00185-7?_returnURL=https%3A%2F%2Flinkinghub.elsevier.com%2Fretrieve%2Fpii%2FS2589004222001857%3Fshowall%3Dtrue https://www.pnas.org/doi/10.1073/pnas.2211624119 https://media.rff.org/documents/Report_22-08.pdf https://cebi.org/wp-content/uploads/2022/11/Guide-to-Sourcing-Marginal-Emissions-Factor-Data.pdf See here for further discussions: Green-Software-Foundation#351 (comment)
…sequential approach does not conform with climate targets set out by SBTi. It is worth mentioning there exist two different accounting frameworks (attributional and consequential) and two different carbon metrics (marginal and average), however they are not necessarily interconnected. Following a consequential framework does not imply having to use a marginal metric. Marginal emission factors can also be used as metrics in attributional accounting frameworks.
…ifferent metrics. Please see the rationale in previous commits.
Please see the rationale in previous commits.
…ble. As the SCI is aiming to be a credible metric, it needs to be possible to explain and reliably verify the impact of an action.
atg-abhishek
added
major-3
|
Hello @ciril-emaps I agree with Pieter Gagnon's letter that a short run marginal does not capture the full effect of load and renewable procurement activities. Pieter's letter calls for consideration of long run effects in decision making (a long run marginal signal), he does not advocate for any particular signal. We think that there could be many signals that might approximate that. I've tried to address this in PR #352 by making it clear that the marginal signal should include both short and long run effects. I think there is consensus that a marginal emissions signal is the correct signal for making decisions to reduce emissions (Olivier's blog post agrees with this and correctly identifies that long run marginal is hard to identify and that average emissions in this context is just a proxy for long run marginal). Am I misunderstanding your point here? If not, a discussion of good signals for long run marginal might make sense in the guidance document. Also, more granular data is not necessarily required for compliance with this standard. An organization can still make very informed siting decisions based on less granular data. For example, if an organization is making a decision on where to site computation load between Poland and France, annual data may be sufficient to make that decision, especially if load flexibility is not an option. Granular/hourly data alone should not be required if time-based load shifting decisions are being made. Global annual marginal emissions data (both operating and build margin) can be found from the UNFCCC, making this more accessible in a broader range of regions. I don't particularly understand the bearing that other standards like the EU hydrogen standard has on this standard? I think this standard is attempting to measure something different and drive different action, especially because it includes market measures like renewable energy procurement, which the SCI explicitly excludes? Does this make sense? Looking forward to discussing this in more depth. |
|
Many thanks for your comments. From what you wrote, I actually do not think that we are in disagreement here. This PR was indeed prepared with the intent of encouraging action to reduce global emissions. This is why we'd like the standard to allow the use of multiple metrics such as marginal, average, renewable potential profiles, as long as they, as you mentioned, serve as a proxy to quantifying long-run impact. To elaborate, given what you wrote, we seem to be in agreement that currently, several proxies exist for long-term impact, including, but not limited to:
To make it absolutely clear, this PR is to enable the use of other proxies, alongside the (short-run) marginal metric, as long as they are indeed leading to long-term impact. Offering further clarification, we showcased the example of the EU regulation to highlight that other parts of the world are following other signals than currently allowed in the SCI. This specific regulation is creating a legal requirement about siting and load-shifting of electrolyzers, much in the same way that the SCI is incentivising going with software. Regarding temporal granularity, we did not realise the SCI was not intending to mandate hourly emission factors - we were confused by the mention of “regional, granular marginal emissions rate” line 99. I believe it would be beneficial to provide more clarity around that, but happy to remove this from our pull request if controversial. Hope this clears up the misunderstanding and I am looking forward to our discussion in the WG. |
There was a problem hiding this comment.
I would suggest removing "hourly" as it is in direct conflict with the Core characteristics
|
Were there any actions to be taken on this PR @ciril-emaps based on the discussion in this week's SWG meeting? Thanks! |
|
Hi @atg-abhishek, in the last session, we were busy with discussing the Huawei case study and this pull request, so we have not been able to discuss this pull request yet. |
|
No worries, thanks for the update @ciril-emaps ! |
|
One suggestion would be to list use cases where various signals apply. This would be helpful for partitioners to make decisions. In certain cases, this decision might even be abstracted through APIs-
|
|
Great suggestion @navveenb :) |
This resolves the discussion in the working group and issue #357 and clarifies the definition of I. Coauthored by @Henry-WattTime and https://github.com/ciril-emaps This supersedes #352 and #353 and #337. Signed-off-by: Henry-WattTime <[email protected]>
This resolves the discussion in the working group and issue #357 and clarifies the definition of I. Coauthored by @Henry-WattTime and https://github.com/ciril-emaps This supersedes #352 and #353 and #337. Signed-off-by: Henry-WattTime <[email protected]>
|
WG approved pull request #358 which addresses this. |
We are hereby submitting a formal proposal to amend the SCI standard.
From the evidence gathered, it seems like there is currently little consensus on which metric to use as a decision-making signal, as can be seen in these examples below (some of them are taken from this discussion in this forum):
Even though more evidence could be collected, the examples above highlight that mandating the use of a specific metric in the SCI seems to be an unnecessary restriction that might hinder the adoption of the SCI:
In the interest of a credible, robust and globally applicable SCI, we suggest making the SCI less restrictive, by not mandating marginal as the only decision-making signal. Instead, we suggest allowing multiple, equally relevant metrics in the SCI. Such a change facilitates the adoption of the SCI globally and provides companies with a more broadly applicable tool to decarbonise software.
We recognize the motivation of having a single signal in the standard, and would therefore like to express our intention in helping the group follow and stay aligned with the upcoming scientific consensus, with the ultimate goal to make the SCI more impact-driven and globally applicable.