European carbon intensity for nuclear

[OLIVIER-LUMINAIS]

Bonjour,

In France according to the IPCC, the quantity of CO² per Kwh is 8 grams of CO² equivalent / Kwh!

more now according to ADEM in France it is 6 grams of CO² / Kwh equivalent produced for French Civil nuclear electricity.

would it be possible to change the CO² produced by civilian nuclear power for France?

Extra information: https://www.youtube.com/watch?v=Bcg9yDp4Azs

to check this information go to 11 minutes and 27 seconds either there: https://youtu.be/Bcg9yDp4Azs?t=687

thank you.

[jarek]

Hello, thank you for the note!

Am I getting this right - looks like the source for the 6 gCO2eq/kWh figure for French nuclear is https://www.bilans-ges.ademe.fr/documentation/UPLOAD_DOC_FR/index.htm?conventionnel.htm which came in 2015 from https://eplca.jrc.ec.europa.eu/ELCD3/index.xhtml which has since been discontinued? Do you know if there was a paper written for/by ELCD that would have this figure?

[jarek]

On a practical note, looks like reducing the French nuclear carbon intensity from the current IPCC 12 gCO2eq/kWh to 6 doesn't really make much of a change in overall carbon intensity right now. This is because most of the CO2eq emissions at the moment are anyway coming from the ~9% generated from natural gas.

Current figure is 73 gCO2eq/kWh, reducing nuclear to 6 would make the figure about 69 gCO2eq/kWh

[checkxp]

The UN ECE has published a new report, where it says, on average in Europe, nuclear provides 5.6g/kWh of CO2eq of electricity generation.

https://energyindustryreview.com/power/unece-nuclear-is-the-lowest-carbon-electricity-source/

I'm trying to find the source, but I can only find un article about it:
https://news.un.org/en/story/2021/08/1097572

[archie-bal]

@checkxp Other sources calculate with much higher GHG-emission factors.
Stanford University:
Jacobson (2009) presented life cycle analyses for different electricity generation technologies and included also so called “opportunity cost emissions”, i.e. CO2 costs due to delays from planning to operation where a faster deployable technology could have avoided emissions. He referenced opportunity cost emissions for nuclear power of 59–106 g CO2 /kWhe in addition to the CO2 costs from the nuclear lifecycle of 9–70. In total he stated for nuclear energy costs of 68–180 g CO2 /kWhe.

German Umweltbundesamt:
22,37 CO2-Äq. (g/kWh)

[pierresegonne]

Hey,

Your first link seems broken and the second is very unspecific, could you pinpoint where they mention this please?

Overall it seems like the methodology from the first article is non-standard. The emission factors we use must come from reputable sources (and a single article is not really), that aggregate multiple studies (such as the IPCC or UNECE) or result from a comprehensive analysis of all lifecycle impacts.

Would be happy to discuss further if you can provide more specific elements

[archie-bal]

Hello @pierresegonne,

sorry, first link was broken.
It was taken out of a book from Author Marc Jacobsen from Stanford University: "100% Clean, renewable energy"

Another source (scientific study):
A "balanced" study by Beerten et al. (2009), which results in green-house gas emission of 32 gCO2/kWh for a european context:
https://www.sciencedirect.com/science/article/abs/pii/S0301421509005102

[jarek]

See also #738 (comment) for comments on the Jacobson and Umweltbundesamt studies linked above

[archie-bal]

Hello @pierresegonne @jarek
see also
#738 (comment)
for details about a study from 2021.
This finds value ranges of carbon intensity for EU nuclear reactors:
16.55–17.69, 18.82–35.15, and 24.61–32.74 gCO2e/kWh
It seems, electricitymap is using a value for carbon intensity which is much too low.

[maxhille]

I also think the current value of 5g/kWh is too low.

The source listed (at the time I am writing this comment) is "UNECE 2022" which I guess refers to this document. The lead author @thomasgibon also engaged in related discussions here, eg in #738. I am wondering if that document actually counts as a scientific publication. I only looked into the sources for the nuclear lifecycle and a lot of the numbers cited "consultation with the World Nuclear Association" as a source instead of pointing to published articles.

The paper referenced to by @archie-bal looks like a better source to me.

[Icchy92]

Any news about this argument? It doesn't look very professional to use the lowest possible value only for nuclear when most of the studies indicate it as on par with wind power or even slightly above in terms of carbon intensity. I know the difference is not huge, but I see too many nuclear fanatics that use this map to praise France against Germany, when in days like today with more than 62% of total consumption generated by wind I think the gap should not be ad big as it is showed.

[pierresegonne]

It doesn't look very professional to use the lowest possible value

Indeed it wouldn't :) That's why we don't, we trust the scientific literature to run lifecycle analysis and to provide given different inputs (that is ore origin, refinement, reactor type etc) an average value, that's where we get that number from. See the original paper for yourself if you're curious to learn more. The author also published a follow-up paper to show what inputs have the most influence on the final emission factor. The result, ore concentration and enrichment method are most influential and explain why countries like France or Sweden have such low emission factors.

[OLIVIER-LUMINAIS]

France enriches its uranium with centrifuges which are 100% powered by a nearby nuclear power plant.

[Icchy92]

As I said, most of the studies I have read in the past few years put nuclear energy at 12 gCO2/kWh on average, then there are some like the UNECE's one used for this map that have a lower value, but there are others which raise it even at 50 gCO2/kWh or more. So why choosing specifically that? I think a more scientific approach should be to select something in the middle, the most recurring range, or at least add the possibility to switch between 2-3 data sets for comparison.

[OLIVIER-LUMINAIS]

France enriches its uranium with centrifuges which are 100% powered by a nearby nuclear power plant.

In fact, they write 12 grams of CO² equivalent/kWh on average.

Average was the word that needed to be understood.
Average means it can be lower or higher.

[archie-bal]

Exactly. And 5 grams of CO² equivalent/kWh for ALL european reactors would be an invalid assumption. But this is what electricitymaps does: Making assumptions, while we have better numbers available.

[archie-bal]

It doesn't look very professional to use the lowest possible value

Indeed it wouldn't :) That's why we don't,

@pierresegonne Yes, you do. Just admit, that you cherry-picked the lowest possible value from ONE cherry-picked paper and then take this value (which might be valid for France) and apply it to most of the european nuclear reactors (they all have stated 5 g CO²/kWh for nuclear, whereas for gas/coal it varies for each country).

In contrast to this, you ignored the fact, that many countries have published their specific CO² equivalent/kWh for their nuclear reactors.
As mentioned before: The german Umweltbundesamt had calculated for the german reactors an equivalent of 22 g CO²/kWh. You neglected this and instead took the lower value from France. Now it's too late, as nuclear energy production is finished for germany anyways...
For europe, still a value of 32 g CO²/kWh is calculated in a peer-reviewed study.

[pierresegonne]

Right now our whole engine cannot support having two different values :/ So it would be a lot of work to support that.

From a user perspective, it would make the whole information less digestible (how would one visualise the map if the carbon intensity is a range??) Cheers

[leon-hard]

There is a study published by Pomponi & Hart in Applied Energy from 2021 discussing published values and also proposing a multi-method, scenario-based approach:

https://www.sciencedirect.com/science/article/pii/S0306261921002555

"The quoted carbon intensity of 5 gCO2e/kWhe is significantly lower
than values calculated in most previous studies of nuclear power, and
the proposed efficiency and long life of the plant are significantly higher
than historic data suggests. Therefore, here we set out to provide a world
first, multi-method transparent and replicable study with the aim of
quantifying the likely carbon impacts and benefits of HPC as viewed
through different prisms. Our aim materialises into two different objectives.
Firstly, we provide an overview of how life cycle GHG emissions
of nuclear power have been assessed in the literature to date; the
numbers reported, the factors driving them, and how they will likely
change over time. Secondly, we estimate the life cycle (i.e. cradle to
grave) GHG emissions of HPC, using the three main LCA methods and
publicly available information that is relevant to the plant or the wider
context. In addition to the contribution that our work makes to the
literature on nuclear energy, it is useful to note that the carbon-intensity
of nuclear energy, is also used as input data in other studies, for instance
in investigating future scenarios for building renovations [12], thus
enhancing the importance of reliable and accurate numbers."

[hno2lr-xcdx]

@leon-hard Indeed.
@pierresegonne why not use those updated numbers from Pompini & Hart?
As soon as implemented, I will trust the calculation of electricitymaps again.

[thomasgibon]

@hno2lr-xcdx Pompini & Hart use a hybrid-LCA approach, i.e. they use monetary coefficients for a part of the life cycle inventory. I don't mind if electricitymaps is using their results, my only issue is that it's just not consistent with literally every other emission factor used by electricitymaps. There are precedents of using monetary factors to estimate nuclear power's impacts, unfortunately such numbers have made their way to (partisan) popular outlets so I would consider hybrid LCA results with caution – unless the same method is used for all electricity sources.

A new article came out last month, proposing a meta-review of LCA of nuclear power, it's in my comment below. Pomponi & Hart's single data point would tend to be in the high end of the range of values the authors collected.

[hno2lr-xcdx]

@thomasgibon Sorry, but you did not read the Pompini & Hart study correctly, or you give at least this impression in your statement. In the study they do indeed use a hybrid-LCA approach, but only as one out of three methods.
Quote:

Our research is based on life cycle assessment (LCA) and we apply all of its three main methods in our study, notably process-based, input-output, and hybrid analyses. To further strengthen the reliability of our findings, for each method we develop a set of scenarios to mitigate the uncertainty in the input data or in the publicly availability information, and to factor in both the complexity and long-term nature of building, operating, and decommissioning nuclear power plants.

[thomasgibon]

Hi all, I'd like to draw your attention to this new article, which came out last month:

Le Boulch, D., Buronfosse, M., Le Guern, Y., Duvernois, P. A., & Payen, N. (2024). Meta-analysis of the greenhouse gases emissions of nuclear electricity generation: learnings for process-based LCA. The International Journal of Life Cycle Assessment, 1-16.

Here are the results

The life cycle GHG emissions of nuclear electricity found with the meta-analysis were 3.09 g CO2 eq./kWh (min), 6.36 g CO2 eq./kWh (median), 12.4 g CO2 eq./kWh (average excluding extrema), and 43.2 g CO2 eq./kWh (max), although extremum values were also identified at 53.4, 60.0, and one outlier, based on theoretical scenarios. Using principal component analysis (PCA), the two most influential variables of the environmental performance of nuclear electricity were identified: GHG emissions intensity of the electricity consumed during the enrichment of uranium and energy demand for the extraction of uranium ore.

[leon-hard]

Thank you for your detailed response and for sharing your comprehensive analysis. I appreciate the insights and concur with several points you've raised, especially the need to include lifecycle assessments that account for the entire lifespan and potential upgrades of nuclear facilities—a factor not yet considered in the studies referenced by EM.

It's great to see the discussion progressing from thread #6743. I hope we can maintain a constructive dialogue here, focusing on sharing knowledge rather than suggesting others need to "study and come back."

Regarding the comment from @thomasgibon on the hybrid LCA approach:

Pompini & Hart use a hybrid-LCA approach, i.e. they use monetary coefficients for a part of the life cycle inventory. I don't mind if electricitymaps is using their results, my only issue is that it's just not consistent with literally every other emission factor used by electricitymaps.

This statement needs clarification. EM does use LCA results based on the exact same methodology and also monetary coefficients. Let's have a look at the emission factor of wind power. EM is using the UNECE 2022 factors, average of wind on- and offshore. The study is available here:

https://unece.org/sed/documents/2021/10/reports/life-cycle-assessment-electricity-generation-options

Why is the value for offshore higher than for onshore, even though the capacity factor is way higher? It's because the maintenance and operation is taken into consideration, which is much more challenging in offshore conditions. And how is this modeled? With a hybrid LCA approach using monetary coefficients. UNECE is using this publication for the offshore inventory:

https://pubs.acs.org/doi/10.1021/es304509r

You can find the same also for other LCA studies used within EM.

Regarding your second comment on "handpicking". I made the statement, that the limits of the sensitivity analysis (aka "pushing to the limits") should at least include the operational parameters observed in the field, if not even more tight/severe assumptions. Here is a list of CFs of the nuclear fleet in France from 2015 onwards:

2015: 75.6%
2016: 71.0%
2017: 69.0%
2018: 71.6%
2019: 69.5%
2020: 61.0%
2021: 65.5%
2022: 50.7%
2023: 58.9%

I also analysed the different sceanios for consumption and installed nuclear power of the RTE Futurs Énergétique study:

https://rte-futursenergetiques2050.com/

The numbers can be translated into capacity factors as installed power, consumption and percentage of production are defined:

These figures demonstrate that for the reference consumption scenario, all CFs are below 70%, except for the N03 scenario, which foresees the construction of 14 new EPRs and additional SMRs until 2050. And all are way below the 90% used in most of the LCAs.

The assertion that low capacity factors are economically nonviable and thus unlikely might miss subtler dynamics in energy markets, especially as the global energy mix is rapidly changing with increased renewable integration. Economic viability can fluctuate significantly with policy changes, technological advancements, and market dynamics.

While pointing out that newer reactors are designed for longer lifespans with fewer needs for upgrades, this overlooks the historical necessity for upgrades due to unforeseen issues, regulatory changes, or advances in technology.

The dismissal of concerns about uranium ore grade based on historical underestimations of resource limits may not be prudent. Resource exhaustion and the energy cost of extraction from lower grade ores are critical factors in sustainability assessments and should not be underestimated.

I suggest we clearly distinguish between new and existing reactors in our discussions and analyses. Out of a scientific point of view, Pomponi & Hart are raising many open questions also for the LCA studies used for existing reactors. But these can not be solved easily and should be discussed separately. But it is definitely not a good practice to use these values for new reactors, where we have studies pointing to higher emissions like the one of Pomponi & Hart for the EPR.

My proposal would therefore be to factor this in. But how could this be done in a practical way?

At least for all ENTSO-E countries, the production values for all nuclear power plants is published and accessible in a uniform fromat:

https://transparency.entsoe.eu/generation/r2/actualGenerationPerGenerationUnit/show

For instance, here is the share of Olkiluoto 3 B3 on overall nuclear generation in Finnland from 2021 onwards:

2021: 0 TWh / 21.7 TWh (0%)
2022: 1.8 TWh / 23.2 TWh (7.79%)
2023: 9.9 TWh / 31.4 TWh (31.70%)

You can easily calculate the emissions factors based on the existing values with this data.

carbon_intensity (year) = carbon_intesity_old * (100% - share_new) + carbon_intensity_new * share_new

The intensity for the last year available can be used until updates are available.

Now you could questions if it makes sense to have a different carbon intensity for different years. And I would reason yes, definitely, you have to, especially for fast emerging technologies. If you have for instance a look at the modelling of photovoltaics in INCER ACV (the one used by EM), you can see that they are using an average module efficiency of 17%. At the same time, you can buy easily modules with 23% efficiency or above, which is 35% above the modeled efficiency. However, you need to consider that the LCA should be of course for the "average" PV module installed. Furthermore, the main carbon footprint of photovoltaics is due to its electricity consumption during manfuacturing, which will also significantly decrease with less coal production in the near future.

[postnormaltimes]

I didn't know UNECE used a hybrid LCA approach using monetary coefficients for offshore wind, thank you. However, this still brings up the question if it's preferable to use this methodology for all technologies or none, as I believe you'd also say that using it only for certain sources (even if they actually have higher lifetime inventory requirements than others, like nuclear and offshore wind) would make results even harder to compare.

Regarding France, yes, RTE also models pathways to 2050 with a nuclear CF which in some scenarios resembles those in the last 5 years. However, as I tried to explain and can be clearly seen looking at national data, availability and capacity factors are closely related to the national context, and the French one is even more of a unicum: high nuclear per capita and high nuclear capacity with respect to average load created the conditions for the only fleet in the world which modulates its generation to such an extent. More than an operational parameter of nuclear technology performance, it's an outlier heavily related to fleet management in the French political and energy context.
As I've said in the other comment, two distinctions offer a better framework for a sensitivity analysis: an aggregated lifetime capacity factor (because we shouldn't make predictions only based on 5-year data for a 40+ years asset) still skewed by national peculiarities (looking at global data, the indicator in question went up, not down, signifying how France might be an outlier), thus it might be better to add a differentiation based on reactor design (PWR, BWR, PHWR).

It's very much true that economic viability can fluctuate significantly, although I don't see how this can help in trying to determine boundaries for nuclear's CF. There are many uncertainties regarding future energy investments, as you've said heavily related to policy preferences and technological progress, but market dynamics are actually pretty straight-forward given a policy framework and technologies' assumptions. That's the best we can do so far with the best energy system models out there, but some conclusions are very common: for example, high shares of VRE overall and somewhat high CFs for nuclear if built (since its most valued trait is being a clean firm resource, not just its flexibility). We can discuss this further, but then we'd need to have a parallel discussion about renewables curtailment in different decarbonisation scenarios (more assumptions, more modeling, more uncertainty).

While pointing out that newer reactors are designed for longer lifespans with fewer needs for upgrades, this overlooks the historical necessity for upgrades due to unforeseen issues, regulatory changes, or advances in technology.
I suggest we clearly distinguish between new and existing reactors in our discussions and analyses. Out of a scientific point of view, Pomponi & Hardt are raising many open questions also for the LCA studies used for existing reactors. But these can not be solved easily and should be discussed separately. But it is definitely not a good practice to use these values for new reactors, where we have studies pointing to higher emissions like the one of Pomponi & Hardt for the EPR.

I agree we should distinguish between new and legacy reactors, but then again the work and considerations by Pomponi & Hardt wouldn't really apply to this framework. Yes, they raised relevant issues regarding LCAs of older reactors and how some of that data has been used in some more recent studies, and the consideration of upgrades in a longer lifetime perspective. However, as I said in the other comment, they're trying to calculate emissions for a French PWR of III gen, the technical characteristics of which are different from older reactors. A 40 years lifetime is just not plausible, and comments on EPR's assumed capacity factor by looking at the indigenous AGR fleet's one don't really make sense from an engineering and operational perspective. I really don't see how the study can point to any conclusions given these premises, but the methodology could be relevant for future studies.

On ore grade, I understand prudency and also believe we should study the situation further. My point is simply that when it comes to uranium resources, differently from others in the past decade, stagnant demand has lead to few new explorations and thus ever decreasing ore grades. Thus, the assumption in most studies looking at energy input and cost of extraction for future uranium demand has been that no new resources would be discovered, and only some reserves tapped into. In my opinion that is an unlikely and pessimistic assumption, which is not used in studies looking at demand and supply projections for other minerals or materials.

Regarding fast emerging technologies, I wholeheartedly agree that it's much more difficult to represent accurate data as progress is fast and some is hard to find from manufacturers (plus, electrical input less polluting over time). My view is that we could have a rolling average of yearly gCO2ew/kWh for a given technology at the national level, adding yearly additions to a database of older installations (thus including progress in many processes over time), or a weighted-average for all technologies at the global level using a periodic repowering (more efficient and better performance) given a specific timeframe.

PS: the snarky comment wasn't related to your interesting and detailed responses which further and better the app, but specifically to the user which raised an issue by linking a Twitter thread with many faulty premises, many of which could have been easily answered by actually looking at EM's methodology (after having been explained, still promotes a wrong explanation of EM's data). Commentators already sure EM wants to manipulate data to promote a certain ideology shouldn't be treated seriously.

[postnormaltimes]

To clarify on Pomponi & Hardt, here's the results their same methodology would have by changing two relatively unreasonable assumptions (although there are many more comments to be made, these were the most important)

Barely over 20gCO2eq/kWh. I'd like to know if the question raised in this reply from @thomasgibon about operational energy inputs might lower the results of the process-based approach enough, such that its and the hybrid approach values would be lower than or around 15gCO2eq/kWh (much more in line with his and the above mentioned meta-analysis).

[madsnedergaard]

We have been eagerly following along here in the EMaps team! We truly appreciate such a constructive and extensive discussion about the topic, it's really great to see! 🙌

For thermal power plants in Europe, we use a mapping of all power plants (see wiki page with details) to get as precise a number as possible. If we can agree on what data source to use for different nuclear power plants (e.g. newly constructured vs older), we could potentially do the same specifically for nuclear (with different values depending on age and type of the plant)? I am however unsure how to best move forward on what numbers to use here as there's clearly pros/cons of the highlighted papers. Any suggestions? :)

[thomasgibon]

I have also been reading the good discussion with great interest!

I do not have much time at the moment, but in short:

• @leon-hard the UNECE study does not use hybrid LCA. Yes we have been using some data from Arvesen et al. but not the monetary inputs, since our database is purely a process-based life cycle inventory database (no monetary data). No problem to send you the inventories if you like.
• @hno2lr-xcdx OK Pomponi & Hart also published "pure" process-based LCA data, my bad, but the highest estimates are found with a hybrid LCA technique, and unrealistic assumptions ("Option A" in the paper), which are not representative of current and upcoming reactors. I trust the corrected values shown in @postnormaltimes's post above.
• Happy to discuss how we can regionalize/temporalize (with rolling averages as suggested by @postnormaltimes?) LCA emission factors for nuclear power, including an update for the material intensity if it is deemed too optimistic (by @leon-hard), after the discussion I admit this is an area we should probably investigate further. I guess the question is: what source would gather most consensus in this thread? UNECE? the Le Boulch review?
• Corollary on supply chains: regionalization would entail distinguishing uranium supply chains by region, this is probably quite difficult to do, unless someone has good data on that? Knowing the influence of extraction techniques and ore grade over the lifecycle emission factor, I think this would be necessary.

[Icchy92]

How about using intervals instead of single values for showing CO2e emissions? Or it would be too complicated to develop?

[VIKTORVAV99]

It's not that it would be complicated it's technically impossible.

If we used a range that means there would be an infinite amount of values between those two extremes and not only would we not be able to calculate emissions properly, people and companies would pick the extreme that suits their needs best.

[Cubikon]

You could do it with propagation of uncertainty. You could use the average value for your calculations. And use the relative difference to the min and max value to calculate the uncertainty propagation and show it as error percentage +-x % https://en.m.wikipedia.org/wiki/Propagation_of_uncertainty