I make models for electric vehicles and the transition to renewable energy at the Eindhoven University of Technology and that means I follow electric vehicle developments closely. Recently, I took the ADAC (largest German automobile club with 18 million members) to task for using their new “expert tool” to paint an unrealistically negative picture of the electric vehicle (EV). (See here on Twitter and here for another blogpost on Innovation origins.)
ADAC defended itself by saying these were only the preliminary conclusions and the study that was still in the making would explain everything. Last week, the study was released and the bad news is: they didn’t react to the criticism in any substantial way and they didn’t change a thing. The good news is, we no longer have to second-guess the mistakes they have made because the firm that made the expert tool (Joanneum Research Life or JRL for short) spelled almost everything out in a background document.
In this blogpost I’ll point out what exactly they did wrong by using my tried and tested format of:
The Top-6 errors people make, when they trash talk electric vehicles
- The new ADAC / JDL study helpfully lends itself to demonstrating this because they make 5 of the 6 errors.
- Correcting these errors and showing the result in a graphical format looks like this:
As you can see, taking the original ADAC figures into account, choosing diesel over electric makes hardly any difference as far as CO2 emissions are concerned. If we correct some errors (I think that’s more accurate than saying ‘change some assumptions’), the electric vehicle emits less than 100 grams of CO2 per km on the German mix during its lifetime, while the diesel emits over 250 grams per km. Everything is explained in detail in the (English) report below, but let’s start with the management summary:
- Error 1 is to exaggerate greenhouse gasses emitted during battery production. JDL estimates emissions based on one outdated source while acknowledging production in very large factories uses ten times less energy than they assume. This metric is expressed in kg carbon dioxide equivalent emitted per kWh of battery. JBL assumes 163 kg per kWh. I show why more current sources has led me to propose 65 kg per kWh.
- Error 2 is to underestimate battery lifetime. JDL estimates that the battery needs to be replaced after 150 thousand kilometers. But stats from actual Tesla drivers show that the battery will last 500 to 800 thousand kilometers and new research shows that the lifetime of lithium batteries is still rapidly improving. Research also shows cars in Germany are in use for about 225 thousand kilometers. Let’s ignore for now that many cars get another lease on life in countries like Poland. Or that recycling and second life expectations for batteries as stated in their study are far beyond conservative.
- Error 3 is to pretend the electricity will not get cleaner during the lifetime of the electric vehicle. ADAC assumes that the electric vehicle will drive on the same mix forever. This is simply a false assumption as is explained in the report below. They also use a very high figure. This means that the ADAC assumption of 608 grams of CO2 per kWh of electricity should be replaced by 295 grams of CO2 per kWh.
- Error 4 is to use unrealistic tests for energy use. JRL assumes a diesel vehicle will use 4.7 liters per 100 km. Although this might be true in the brochure for some cars, using real world measurement takes you closer to 7.3 liters per 100 km.
- Error 6 is lack of system thinking. With diesel you might reduce emissions around 40% in a renewable future, but with electric you can achieve reductions of around 95%.
So, the ADAC study is the anti-electric vehicle lobby at its finest, using every trick in the book to make the combustion vehicle look good and the electric vehicle look bad.
Are you still reading? Interested in proof and details? Now I’ll dive into the sources and reasoning behind each correction in detail.
1. Exaggerate greenhouse gasses emitted during battery production
An electric vehicle has a battery, something combustion cars don’t have. To produce it you have to use energy, and this emits CO2. Every serious EV researcher agrees on that. The interesting bit is determining how much. The measurement used is mostly greenhouse gasses emitted per kWh of battery.
The problem in determining this is that manufacturers don’t want to tell this because they don’t want to tip off the competition while the scientific literature is usually far behind the facts. The ADAC (or rather Joanneum Research Life or JRL for short that provided them with the tool) said they used 11 literature sources. That seems to be mostly for show as most sources are not referenced in the text. (That’s a big no no in science by the way: it’s like padding your resume with jobs you did not do.)
But I tried to trace back the three sources that are referenced in the main text.
The first is Ellingsen et al 2014 who base themselves on Majeau-Bettez et al (2011) who in turn base themselves on Rhydh and Sanden (2005). So this line takes us back to a theoretical exercise from 2005 that was not even aimed at car batteries. I think it shows the approach of JRL takes them a couple of years behind the times.
The second is a what JRL calls a “meta study” from the ICCT (2018). I wonder if they did more than skim the summary for a number to use. Because firstly it is not a meta-study but a policy brief. Secondly, they claim this study points to emissions of 175 kg CO2-eq per kWh while in reality this study says the ICCT doesn’t know because estimates are all over the place and emissions are dropping rapidly as manufacturing scales up and electricity becomes greener. Thirdly, the only time the ICCT uses 175 kg CO2-eq per kWh, it calls this “the central estimate of Romare et al.” which is actually the third and last source Johanneum mention in the main text. So they present a brief as a study, say it concludes 175 kg when it doesn’t and claim this is an independent number when it isn’t. You can’t get it much more wrong than that.
The third and last source is Romare (2017). This is indeed a meta-study. Yet to choose this as a basis for the most important number in the whole tool is peculiar because this study has gotten a lot of push-back and the authors themselves warn against using it for large scale production. Elon Musk, the CEO of Tesla and owner of the world’s largest lithium battery factory said “Calling this clueless would be generous. Much less energy required for lithium-ion batteries.”
Another EV CEO explains a lot is wrong with the study, one thing being outdated numbers. An article in Handelsblatt explains how unreliable these numbers are. I found the briefing of FFE called “Carbon footprint of electric vehicles – a plea for more objectivity” to be especially good. They explain (with support from the authors of the maligned study itself!) that this was just an overview of past studies and using this number as-is ignores the rapid pace of development that is bringing emissions down quickly as production scales up and electricity becomes greener.
So there you have it: everybody agrees the 175 kg CO2-eq is untrustworthy and outdated and that battery production emissions are dropping rapidly. You know what: even JRL seems to agree! But then they decide to ignore it. How do we know? In my criticism I introduced some new sources that JRL have now discussed in a separate chapter called “significant influences”. What do we find in that chapter (on page 192)?
I was flabbergasted when I read that. It’s like saying: “We now know our estimate of energy use for battery production is probably ten times too high. But we are going to use it anyway. Get over it.”
Now just to be clear: this approach by JRL is not unusual. Many large organizations (JRL mentioned the IEA) us it. But that does not make it right. It only shows many large institutions have trouble dealing with change.
So how do I determine emissions? I do not base myself on scientific publications with production numbers that go back more than 4 years because that’s simply irrelevant. So I certainly don’t use meta-studies. I either use recent studies containing original research or industry sources. In the article that I linked to previously this leads me to conclude 65 kg CO2-eq per kWh is currently a good average and it will drop further in the future. This is bolstered by new (June 27, 2019) market research from Bloomberg New Energy Finance (publicly referenced in this podcast) that puts the emissions between 20 and 80 with the average around 40 kg CO2-eq per kWh. However, this is excluding mining so I think 65 is still a good estimate.
By the way: all this is excluding recycling and second life. JRL does include some assumptions but they are very pessimistic: recycling reduces emissions only 2% and second life reduces them a lot but is only assumed in 5% of cases. In effect the result is negligible. There is so much wrong with these assumptions that I will let them be for the moment, but I will just say that in 15-20 years’ time (when current car batteries are scrapped) it is ludicrous to assume the emissions needed for that will be almost the same as for producing new batteries.
2. Underestimate battery lifetime
Many studies limit the battery lifetime to 150 thousand kilometers. When the car drives more, the batteries have to be replaced. ADAC and JRL do this. However, my good friend prof. Maarten Steinbuch publishes a well known blog that shows the data that Merijn Coumans gathers from hundreds of Tesla drivers to see how their batteries are degrading. The numbers suggest that after 800 thousand kilometers you will still have 80% of range left. There are many other public sources that point towards batteries outlasting any realistic use of the electric vehicle. Since the electric motor is also good for such a long distance we might well see electric vehicles being used for more kilometers than currently with diesel vehicles.
I recently supervised a master study that did a deep dive into battery degradation for me and it turns out most research on batteries is from the last five years: research is booming and it’s paying off, especially with regard to lower degradation. It is now better understood how active cooling, smart charging and doping of the electrolyte can make the lithium battery last much longer so it will be easy to make batteries that last over one million kilometers within ten years. There’s literally hundreds of relevant papers but a good introduction is this lecture by leading professor Jeff Kahn.
Many people voice opinions about how long cars will last. Few have facts to back their opinions up. Gniewomir Fils was kind enough to look into this and concluded that 150 thousand kilometers was clearly not enough and 15 to 20 years was more realistic. In more affluent countries, cars are driven shorter but even so in Germany people drive for about 16 years and with 14 thousand km per year that means 225 thousand kilometers. But life for a car bought in Germany often doesn’t end in Germany. Many cars are exported to countries like Poland where cars are much older on average and where electric vehicles can continue to reduce CO2 emissions.
3. Assume electricity will not get cleaner during the lifetime of the electric vehicle
Imagine somebody gave you a loan for a house over a 15-year period. Imagine further that the costs started at 402 euro per month in the first year and 9 euro per month less every following year. See the table below. What would you say is the cost per month over this 15-year period?
Almost anybody would say: you take the average over 15 years. I calculated that at 321 euro. That is your average cost when you live in your house for 15 years. Right? Well not according to JRL and the ADAC.
These numbers are not random. They are the expected CO2 emissions per kWh of electricity in the EU. And what most studies of electric vehicles do wrong is that they calculate with the number in the first year. They pretend the coal fired power plants that are closed and the green energy that is added does not matter for the emissions of the electric vehicle. They pretend a vehicle bought in Germany in 2020 will drive on the energy mix of 520 grams of CO2 per kWh forever. But the average is 295. That’s 43% less!
I must say I could not find this piece of information in the background document but the graph in the press release gives us a clue. Eyeballing the graphic from the study/press release we can see that driving the electric vehicle on the German mix for 225 thousand km increases emissions from 12 to 37 tonnes. So, 25 tonnes for 225 thousand km gives 100 grams per km. The tool assumes the electric vehicle uses 0.19 kWh per km. That means they assume that producing 1 kWh of electricity produces 100/0.19=526 grams of CO2. So that is indeed the current German mix. Ergo they assume this dramatic situation will last for another 15 years.
Now maybe you will say: “They are just talking about how green electric vehicles are now and they acknowledge that electric vehicles will be greener in the future.” But that’s not how this works. All the emissions per kilometer numbers that ADAC and JRL are using are based on an evaluation of the car over its entire lifetime. So, they really take the high emissions in the first year and use that number every following year.
4. Use unrealistic tests for energy use
JRL and ADAC assume a diesel vehicle uses 0.52 kWh/km or 4.7 liter per 100 km based on “research by JRL” (page 62). However, if we look at users documenting their real energy use in Germany on spritmonitor.de and select diesel cars sold in the last 3 years the average is 7.3 liter per km. That’s 55% more.
What often happens in Europe where politicians and automakers have long enjoyed a cozy relationship is that real numbers are replaced by very unrealistic tests so politicians and carmakers can say they reduced emissions while in effect they only cheated on the test. Under the NEDC, real emissions where 40% higher than in the test. The new WLTP is also problematic.
For electric vehicles, 0.19 kWh is assumed (page 64). This is about what the EPA (the best source for vehicle emissions including charging) estimates for a car like the Tesla Model 3, Volkswagen eGolf and Nissan Leaf so let’s go with that.
5. Exclude fuel production emissions
If we take diesel, JRL assumes that it will emit around 33 tonnes of CO2 over 225 thousand kilometers. That is 147 gram per kilometer. Taking into account their diesel consumption of 4.7 liters, that implies emissions of around 3120 grams per liter of diesel.
Here they have not made a material mistake. The direct emissions from diesel are around 2600 grams per liter (it differs a bit according to exact composition). Indirect emissions for things like refineries and transport adds around 620 grams in Europe (see e.g. here and here). That brings the emissions per liter to around 3220 grams which is pretty close to 3120 grams – so no complaints here.
You could argue about the exact height of the numbers. However, refining and transport alone added at least 18% in 2010 and this percentage has probably increased since oil takes more and more money in order to extract it.
Also, they could have included biofuels to get at the slightly reduced number. That’s a discussion for another time but modern scientific insights are that using agro-based biofuels causes more emissions than fossil fuels in most situations (waste and some double cropping and fallow grasslands excluded) and put enormous pressure on worldwide food systems and the natural environment. That is why I advise against it for regular cars and why I certainly don’t use it to reduce the carbon footprint of fossil fuels.
6. Lack of system thinking
By this I mean that most critics that claim electric vehicles are not much better than combustion cars seem to miss the “the big picture”. Let me take the table from my article to illustrate:
As you can see, in a Renewable Future scenario, the diesel car could achieve some improvements to total emissions but it’s very limited. The electric vehicle, on the other hand, can achieve a further tenfold reduction over the current twofold reduction compared to diesel vehicles. This last scenario is speculative, but these numbers are estimates that would result when recycling is done using current best practices and when production and operation only use low carbon sources.