Refine GHG emission factors by regions

[brunolajoie]

The aim of this issue is to discuss the possibility/find peer reviewed studies of a standardized approach to refinie LifeCycle GHG-intensity factors of the various power plant category per geographical zone (country, zone, region...). We don't want to re-invent the wheel, and would rather use an external, trustable analysis that uses the same assumption between all country covered.

[Dec 2019 update]
We (electricitymap team and other) have published a peer reviewed paper called "Real-time carbon accounting method for the European electricity markets" free donwload here that contains in Table 1 of appendix an interesting list of GHG emission factors, LCA based, computed for each EU country.

Other interesting leads include

• the following See & download data: Read sci paper:
• http://www.reliable-disclosure.org/upload/259-D5.2_Best_Practice_Environmental_Data.pdf
• https://link.springer.com/article/10.1007/s11367-013-0665-2
• http://iopscience.iop.org/article/10.1088/1748-9326/aa6047
• http://www.sciencedirect.com/science/article/pii/S1364032117304215

[thomasgibon]

Hi @brunolajoie, great that you mention further refining of the GHG factors and the potential addition of other indicators.

I'm a co-author of the paper you refer to, and I would be glad to provide the underlying data for the various technologies we considered. We published two more papers containing such data, to extend the range of technologies and indicators.

To complete your suggestions, we can indeed:

1. Refine the factors per region, based on how efficient technologies are (based on local characteristics that alter the load factors, such as insolation and wind),

2. Refine the emissions factor based on the technologies used in various regions (we can use ecoinvent data, which has become very spatially-explicit in its latest version, the data team did the work of associating technologies with regions),

3. For me, the most interesting would be to add further indicators (land use and material requirements come as obvious ones because they are potential trade-offs of the energy transition). There is an issue of introducing high uncertainty for certain indicators (namely, for toxicity) but indicators of direct emissions like particulate matter and smog emissions could be added.

[brunolajoie]

Thanks a lot! let me have at it more closely! I'm posting a study here so we don't forget it, which gives a breakdown at the country level for Europe. http://www.reliable-disclosure.org/upload/259-D5.2_Best_Practice_Environmental_Data.pdf

[thomasgibon]

Perfect, I think the ecoinvent data reported in this paper is excellent for the electricity map. The paper detailing the update of electricity inventories in the latest ecoinvent database is a good read: https://link.springer.com/article/10.1007/s11367-013-0665-2

I've had some thoughts on what can be achieved with the app. There are two main reasons why factors change across regions/grids:

• Intrinsic factors. For the baseload technologies (fossil, nuclear, biomass, hydro), the variation across regions is a matter of technology differences, load factors and efficiencies -> factors can be directly applied from the paper.
• Extrinsic factors. For the intermittent technologies (principally solar and wind), the LCA results per kWh are derived from the total impacts of a power plant construction, maintenance and decommissioning, divided by the plant's lifetime production, in kWh. To cite the ecoinvent paper:

Annual wind load hours of wind turbines and annual yields of photovoltaic plants are key factors for the environmental life cycle burdens of wind and solar power

...which is precisely the kind of data that the app fetches with a high resolution. After collecting data over a year, the app could eventually provide yield (load factor) data that is more accurate than the one used by ecoinvent to calculate solar & wind GHG emissions per kWh. It's probably not for now, but I would be interested in knowing whether the load factors assumed in ecoinvent and those from the app align.

And if you want more than GHG data, the full range of environmental impacts from the ReCiPe2016 can be calculated easily for each technology/region, see Table 1.1 here: http://www.rivm.nl/bibliotheek/rapporten/2016-0104.pdf

[brunolajoie]

Hello @thomasgibon, sorry for the late answer:

• Yes, in the future we will be able to add other variables to the map than GHG, but let's start with this one.

• Very good idea, we should cooperate with ecoinvent to improve their LCA analysis based on our identified load factors globally.

• Back to this issue: we're looking for "Lifecycle Analysis CO2eq/kWh values per country/states and per fuel type. We don't need very detailed factors per sub)categories of power plants as we don't have the production data with such breakdown. We would need a standardized approach across many countries in order to make the EMap compare regions in a fair manner.

To your opinion, which database is preferable to investigate? Econinvent? If so, would you know where we could access such data? It seems that your suggested paper only gives direct emissions, not LCA. Any contacts at ecoinvent we could talk to?

[brunolajoie]

Update on ecoinvent:
1/ It's not free
2/ "Licensee is not entitled to reproduce, disseminate or publicly display any significant portions of the ecoinvent Database or the ecoinvent Datasets."

We want to remain transparent, and show in the electricitymap the c-intensity factor used to compute country's c-intensity.

I'll contact them to figure out in detail if we can collaborate on this, hopefully we can work something out!

[alixunderplatz]

@brunolajoie I'd like to leave this chart here based on LCA for Germany. I had posted another part of this before (in the lignite vs. hard coal emission issue #162).

The study can be found here:
https://www.vdi.de/fileadmin/vdi_de/redakteur_dateien/geu_dateien/FB4-Internetseiten/CO2-Emissionen%20der%20Stromerzeugung_01.pdf

It was released 2007. There are many examples per production type and several sources given.

• For fossil and hydro generation, numbers will still be valid.
• As the majority of PV was installed between 2009-2013 after the relase of the study, I'd stick with IPCC for that category (as well as for the rest of renewables)

[brunolajoie]

Thanks. Same remark than for moldova: lets keep that in mind if we don't find standardized approach valid for all ocuntries.

[thomasgibon]

@brunolajoie

On the database. Ecoinvent is probably the best reference in terms of life cycle inventory databases. As you remarked, it is unfortunately not free BUT I am sure we (e.g., I) can compile life-cycle factors for the electricity-producing processes that can be found in ecoinvent without infringing the copyright issue, as long as (1) we don't disclose the disaggregated inventory data (i.e. only one number per kWh, without detail on the impact contribution) and (2) we remain reasonable in how much data we extract (I guess ~100 technologies would be fine, the database contains >15000 processes). The best would be to get in touch with the ecoinvent people, I can suggest Christian Bauer and Laurent Vandepaer, who are working on refining the electricity inventories for the next ecoinvent version.

On direct emissions. You are right, the regional detail is for direct emissions only, I think the assumption is that all plants are built and decommissioned in the same way everywhere, so the variable share is only the direct emissions. The contribution from infrastructure is usually very low as long as we only look at CO2 (see below the detail for 1 kWh from coal).

@alixunderplatz We have made the same comparison for my PhD thesis work, see Papers III and IV here: https://brage.bibsys.no/xmlui/handle/11250/2469193

Especially:

...for CO2:

and for other environmental impacts:

[brunolajoie]

Great news @thomasgibon. I've contacted ecoinvent last week through their contact-us generic email, (no answer yet) and would be happy to try again if you could give me in private the contacts of Christian Bauer and Laurent Vandepaer that you mentioned. Else i'll try to find them online thanks!

[thomasgibon]

@brunolajoie I actually talked to Laurent yesterday, who said he'd contact you directly. He's a PhD student working with marginal electricity mix data in ecoinvent, and Christian Bauer is one of his supervisors.

[brunolajoie]

Thanks! I'm in discussion with Econinvent. Will let you know

[alixunderplatz]

The following document comes from the scientific service of the German parliament and is a collection of carbon emission factors for different generation types (it's from 2007!).

https://www.bundestag.de/blob/406432/70f77c4c170d9048d88dcc3071b7721c/wd-8-056-07-pdf-data.pdf

I want to point at the chart on page 25. It is a list of related emissions from nuclear power plants. It says that emissions highly depend on where (country) and how (process) the uranium enrichment took place before delivery to the NPP. France mainly uses domestic "nuclear" electricity for this process, so CO2eq is at 8 g/kWh for NPPs using French fuel elements. For South African uranium fuel it is at 125 g/kWh, because of high coal share in their electricity mix and using a more energy consuming process for uranium enrichment.

[brunolajoie]

@thomasgibon, I'd be happy to have a chat with you about your proposed approach, ping me on slack anytime!

[comready]

For info re ecoinvent the paper https://www.sciencedirect.com/science/article/pii/S030142151630091X I referred to in #1309 for Turkey says: "The background life cycle inventory data have been sourced largely from Ecoinvent v2.2 (Dones et al., 2007) but have been adapted as far as possible to Turkey's conditions"

[VIKTORVAV99]

Any update on this?

[corradio]

We're starting to take a look at this but will probably only have a release ready in Q1.
Does anyone know of good open databases for plant-level emission rates that we could use?
For now we've found eGRID in the US (https://www.epa.gov/egrid/download-data)

[VIKTORVAV99]

@corradio EDP International has some very detailed data on a few power plants but you would need supplemental data from elsewhere for the rest.
(Use "Electricity, steam & fuels" in the product category to filter them out)

[thomasgibon]

@corradio I couldn't find a nicely harmonised database but this UNECE portal is a good access point to emission registries in most countries. It's open, but I couldn't find anything standardized, energy-charts.info seems to be parsing some of the German database on the annual level: https://energy-charts.info/charts/emissions/chart.htm?l=en&c=DE&year=2018&source=lignite

Examples of power plants from national databases found on the UNECE portal...

• Germany: emissions of the Neurath lignite power plant in 2018,
• France: Cordemais coal power plant.

[corradio]

Very interesting @thomasgibon and @VIKTORVAV99. Thank you for starting to gather relevant datasets. These will for sure become handy. I'll report back here when we start working on this. Maybe we need to build a standardised and harmonised database that centralises all this information.

[VIKTORVAV99]

I found the following document when searching for data regarding #3611 but I think it would be helpful here too.

There is a lot of data in this publication for worldwide and EU28 power production LCA emissions with a total of 147 references. However I did not see country specific emissions but I think it will be of interest anyway.

Landing page: UNECE - Life Cycle Assessment of Electricity Generation Options
Direct PDF download: LCA_final-FD_0.pdf

[archie-bal]

The numbers for GHG-emissions for germany are estimated too low by electricityMap.
Study for GHG-emissions from the Umweltbundesamt: 22,37 CO2-Äq. (g/kWh)
The World Information Service of Energy (WISE) is calculating in this study with 117 g/kWh CO2-Equivalent.
You might argue, WISE is not agnostic to atomic energy, but there are other independant studies which show similar results: Stanford University Study is calculating 68-180 g/kWh in this study.

Please use realistic and regional green-house gas emissions for nuclear energy, as the current used values which are derived from 2014 from IPCC (12 g/kWH) are not realistic.

[thomasgibon]

Hi @archie-bal, thanks for providing these very interesting sources.

TL;DR: Check these studies' methodology and data if you really want to use these values.

The first number (22.37) seems to actually be per kWh of primary energy, which means the actual GHG factor for a nuclear power plant with 33% efficiency would be about 68 g. I looked at the reference mentioned (GEMIS 2016), which is actually a model developed by IINAS. The original model is from 2002, and the data collection probably from even before that. Some technologies are therefore outdated, for example, the uranium enrichment process today is almost exclusively centrifugation (see Table), while GEMIS assumes 100% gas diffusion. It's important because gas diffusion consumes 50 times more energy per unit of uranium, than centrifugation. Then, all electricity inputs are lower-carbon today, and the assumed plant lifetime is 30 years, which are two more parameters explaining the higher number.

The second number is actually quite old too. It is mentioned in this NEA report. For the plant construction, GHG values have been estimated not from physical accounting (basically the bill of materials) but from financial data – and then multiplied by the average emission factor of the economy (in g CO2 eq./€). Needless to say this introduces a lot of uncertainty, as materials used
in a nuclear power plant are probably more expensive (per kg) than average, for safety or quality reasons. Downstream processes are then calculated as a share of construction, therefore propagating the same overestimation. Upstream processes amount to a few 10s of grams, and I believe they also assumed gas diffusion (instead of centrifugation).

Now just one word on the Stanford study, since Prof. Jacobson has taken a few liberties with the scope of the life cycle assessment he proposed. First, they added to any technology that was not wind power a penalty for delaying the construction of... wind power. They called this "opportunity cost emissions". Basically if it takes 10 years to build an NPP, that's 8 years longer than building a wind farm, which could have decarbonized the economy in the meantime. And so every kWh produced from nuclear (or anything else) gets this penalty. Second, and that's even more creative (and very minor anyway, but for the anecdote), he also includes in the scope the consequences of nuclear warfare. The rationale is that the use of nuclear power makes a country able to develop nuclear weapons, and the burning of cities that may result of that should be allocated to each nuclear kWh produced. Also the study is from 2009 and the Guardian made an article on it.

These two last studies also insist on the fact that uranium ore grade might be degrading, and the energy spent in resource extraction may increase proportionally in the future. They both use the conservative estimate for many parameters (the gas diffusion, the construction, plant lifetime, ...). I am not saying they should be disregarded, I only recommend to go look under the hood before taking a number for granted.

In the meantime, many recent LCA studies have consistently assessed nuclear as having a sub-10 carbon footprint, they include:

• Vattenfall (2019), with 4.1 (only 2.5 g/kWh excluding distribution),
• Dones et al. (2007), 5-12 g/kWh,
• Poinssot et al. (2014), 5.3 g/kWh,
• Zhang & Bauer (2018), 5.6 and 9.4 g (two reactor types),
• Warner & Heath (2012), a very complete meta-study, but the data is already >10 years old, with 9-110 g/kWh.

Sorry for the long post, but I regularly see these studies mentioned over and over again, and I just wanted to warn about their content and methods. Now feel free to do what you want with them :)

[VIKTORVAV99]

@thomasgibon I believe I speak for everyone here when I say you have no reason to be sorry for the long post, it is detailed and to the point. We or at least I very much like that.

@archie-bal obviously having regional emission factors would be best but it might not always be possible to find the data needed for that to be implemented in every region so having a average for a bigger region like the EU (EU 28) as a fallback will probably always be needed. As @q-- also mentioned in #4161 (comment) the the article you reference to seem to include “opportunity cost emissions”, i.e. the emissions that could have been avoided by building other (faster) production sources. I believe this makes the values unusable for electricityMap.

[VIKTORVAV99]

The current granularity for this issue seems to be on a country level but for several countries we support higher granularity than that.

Would it be in the scope of this issue to calculate co2eq intensities for individual zones in the zones.json file or is that outside the scope of this project?

I believe there could be some major benefits to this as for example the solar intensity metric discussed in #4253 would vary greatly with latitude. So slitting it by zone rather than country would allow for higher accuracy. (also applies to hydro power to an extent)

This would mostly benefit large countries with multiple zones like the USA, Canada, Australia, India, Brazil and Russia. As well as countries with multiple zones that have a significant latitude difference between North and South like Sweden, Norway and Chile.

[archie-bal]

@thomasgibon
Please investigate this recent study from 2021 (Francesco Pomponi & Jim Hart):
The greenhouse gas emissions of nuclear energy – Life cycle assessment of a European pressurised reactor

Abstract:
Nuclear energy contributes ~10% of the global electricity generation and different views exist on its carbon-intensity and sustainability. Context is crucial to determine the sustainability of new nuclear power generators, making the existence of a global answer to the unresolved question unlikely. This study aims to establish the life-cycle greenhouse gas emissions associated with nuclear energy in Europe given ongoing construction of nuclear generators.
Due to the high uncertainty and complexity that characterise construction and operation of nuclear generators, we adopt a multi-method, scenario-based approach. The three methods used are: process-based, input-output, and hybrid life cycle assessment. Scenarios account for different total energy outputs over the life cycle of the nuclear generator, different end of life options, and different sectoral allocations of costs in the input-output calculus.
Results for the process-based, input-output, and hybrid methods range between 16.55–17.69, 18.82–35.15, and 24.61–32.74 gCO2e/kWh, respectively. These are either well above or at the upper end of the range of possibilities (5 to 22 gCO2e/kWh) stated in a report for the UK’s Committee on Climate Change, and significantly higher than the median value of 12 gCO2e/kWh presented by the Intergovernmental Panel on Climate Change. They are also higher than the values acknowledged by the nuclear industry.
Given the severe potential lock-in effects of today’s energy choices for future generations, this research questions the role of nuclear energy to meet the UN Sustainable Development Goals and calls for further scrutiny on its sustainability and environmental viability.

It seems like electricitymap is using a value which is too low for the carbon intensity of nuclear energy.

[Audiard-Jerome]

@archie-bal
Please investigate this recent study.
https://www.edf.fr/groupe-edf/produire-une-energie-respectueuse-du-climat/lenergie-nucleaire/notre-vision/analyse-cycle-de-vie-du-kwh-nucleaire-dedf

This study was critically reviewed by a panel of independent experts, whose complementarity met the requirements of ISO 14044 and ISO/TS14071. These experts consider "that the results provided respond adequately and credibly to the objectives mentioned and that they have been established in compliance with the standards mentioned".

Results :
4 g CO₂ per kWh

[VIKTORVAV99]

@thomasgibon Please investigate this recent study from 2021 (Francesco Pomponi & Jim Hart): The greenhouse gas emissions of nuclear energy – Life cycle assessment of a European pressurised reactor

Abstract: Results for the process-based, input-output, and hybrid methods range between 16.55–17.69, 18.82–35.15, and 24.61–32.74 gCO2e/kWh, respectively. These are either well above or at the upper end of the range of possibilities (5 to 22 gCO2e/kWh) stated in a report for the UK’s Committee on Climate Change, and significantly higher than the median value of 12 gCO2e/kWh presented by the Intergovernmental Panel on Climate Change.

It seems like electricitymap is using a value which is too low for the carbon intensity of nuclear energy.

While the article is interesting from what I can gather from it it's limited to new nuclear power plants only with a specific focus on UK nuclear reactors, specifically Hinkley Point C. The article also mention that there is large differences between different geographical areas and therefore even if this data was 100% accurate for the new nuclear power plants in the UK it would not hold true for the rest of Europe. There is also several other studies, articles and certification that seem to refute these claims but I have not had the time to review them all in depth but the UNECE article alone, which has been substantially peer reviewed, put the (total) average co2eq at 5.1 g/kWh for the EU28 members (EU27 + UK).

On another note, in the future when including abstracts from articles please include the whole abstract as just including a part of it can be highly missrepresentative.

[archie-bal]

Ok, thanks for the hint @VIKTORVAV99. Added the whole abstract text.

[ericbl]

the study is about the EPR, currently only on production in Taishan, China (twice), in Finland (OL3) and soon in France (FL3) and UK (Hinkley Point C.). The emissions from this rare reactor cannot obviously be taken as average value for all nuclear reactors.
EPR on wikipedia

EPR1 design: two-layer concrete wall with a total thickness of 2.6 m, designed to withstand impact by aeroplanes and internal overpressure
The most notable simplification (of the EPR2) is a single layer containment building with a liner as opposed to the EPR's double layer with a liner.

It is likely that the double layer of the EPR1 with a lot of concrete contributes to higher emissions.

And back to the topic, as for South Africa spotted above by Alex in Feb 2018, the EPR in China will be likely have much higher emissions than those in Europe...

[thomasgibon]

I looked at the Pomponi & Hart article, I'm intrigued by a few things

• First, as noted, the amount of concrete used in HPC, 3 million tons is astronomical. We had plotted the comparison for various designs in the UNECE study. Figure 62 shows the value as a clear outlier (we used the average for the study).
  
  I actually suspect the Hinkley Point C brochure (their source) to be a mistake, the reasoning is that the whole foundation for one unit is "only" 49000 tonnes. Pouring so much concrete at once is literally a record in the UK, and 3 Mt is yet 60 times that...
• Second, there is a significant energy input in the operational phase (namely 10 TWh of natural gas), which they take from the Lenzen study – I still need to look at it.
  

All in all, capital (the construction of the plant) dominates their lifecycle impacts, which is somewhat surprising as most LCAs show that the uranium chain is usually what contributes the most to impacts.

The EDF LCA publication is very good, but should be taken as an absolute low (with Vattenfall maybe?) since France has its own enrichment facility using nuclear power itself, and reprocesses spent fuel, working to close the uranium cycle. So there's a lot of optimisation at play.

Anyway, a colleague of mine and I are working on publishing the nuclear part of the UNECE report, so stay tuned!

[madsnedergaard]

Update: We're now actively working on building support for regional emission factors!

In order to discuss the functionality separately (and keep this issue for discussion of actual data sources), I have created a new issue here: #4737
We'll take all the great information shared here into consideration, so stay tuned 🚀

[corradio]

Fixed by #4739

[brunolajoie]

🥹