Michael Daw has written a blog post that criticises my arguments concerning the energetic implausibility of manufactured food (or ‘precision fermentation’ to use the biotech industry’s preferred term). I don’t think his arguments stack up, as I’ll explain below, but this seems like a good opportunity to run through the relevant issues, which is the aim of this post. It will be followed by a few more posts on various issues relating to my book Saying NO to a Farm-Free Future and some of the criticisms of it, before I turn to other issues.
One thing I must do at the outset is thank Michael for the polite and good-natured tone of his post, which alas I’ve found all too rare in this debate. If only other major contributors identified points of intellectual difference with the same measured language that Michael uses.
Michael addresses (1) my analysis of the energy costs of protein production by hydrogen-oxidising bacteria in Saying NO…, and also discusses (2) the production of a dairy substitute whey protein from a genetically modified fungus by a company called Perfect Day, putting these techniques (3) into the wider context of the case for veganism as a way to safeguard nature.
I’ll follow that three-part structure in my response. But some of the details of my case in relation to point (1) may be too nerdy for the casual reader, so I’ve put these in an appendix at the end. I’ll try to cut to the chase in the main body of the post.
1. Hydrogen-oxidising bacteria
How much energy does it take to manufacture edible protein with hydrogen-oxidising bacteria? George Monbiot says 16.7 kWh of electricity per kilo of bacterial protein, but this is demonstrably wrong (see Appendix). A paper by Natasha Järviö et al says 18 kWh/kg bacterial biomass, which translates to 27 kWh/kg bacterial protein (it’s important to note the difference between biomass and protein – see Appendix). The figure I suggest in Saying NO… is at least 65 kWh/kg bacterial protein, which I calculated from a paper by Dorian Leger et al (Dorian Leger confirmed in a personal communication that my derivation was a reasonable low-end figure).
It’s worth adding that an application for EU regulatory authorisation for a bacterial protein product from Solar Foods – a pioneering industry player in this area – appears to suggest that the digestibility of their bacterial protein powder is much less than agricultural-origin alternatives like soy powder. On the face of it, that suggests my 65 kWh/kg protein figure might effectively be a considerable underestimate in relative terms.
Michael thinks I shouldn’t dismiss the Järviö figure, which he says is an actual measurement of small-scale production and not an assumption. But the provenance of this figure is unclear. Järviö et al say that “this study assumed an electricity requirement of 18 kWh per 1 kg product produced” (emphasis added) and states that data were gathered from current pilot-scale production performed by Solar Foods, as well as “expert interviews, and the literature”, but it doesn’t clearly specify the exact derivation of the 18 kWh figure and what’s included and excluded in it (see Appendix).
By contrast, the Leger et al study that I used is much clearer about the sources of its data, the parameters of its analysis and the mathematical basis of its energy derivation. It still omits some significant energy costs, but it was the clearest, most rigorous and most comprehensive study I could find, and that’s why I chose it.
Here, I think I have to bounce Michael’s query back to him. If my figure is wrong, I think that implies the Leger study is wrong, and my question to Michael would be where exactly has it erred? Until somebody can explain satisfactorily to me why the Järviö figure is correct and the Leger data are incorrect, I see no reason to recant my 65 kWh/kg derivation. The Järviö analysis addressed a wider set of questions than the Leger analysis and I think it probably just didn’t probe the energy input aspects as thoroughly as Leger, because it was doing a lot of other things. Fair enough, but energy input is important. In the Appendix, I mention other studies that suggest energy costs in a similar ballpark to mine. On balance, the Järviö figure looks like it’s too low.
Michael says he finds it odd that I draw my conclusions from two papers that seem broadly sympathetic to precision fermentation. I’d argue on the contrary that this is a strength of my analysis and methodologically appropriate. Even with the best will in the world it’s easy to introduce biases into an analysis that slant it toward one’s preferred approach, so using research that favours a different approach is a good defence against cherry-picking. I’ve drawn on studies that indeed are broadly sympathetic to precision fermentation, and yet they still show the high energy costs involved.
Stepping back from the details of this or that figure, it’s worth just considering what’s involved in bacterial protein manufacture. At a time when we desperately need to decarbonise the global energy system, and are largely failing to do so, manufactured food proponents are suggesting that we stop using a free, zero-carbon source of energy (the sun) to provide our dietary protein, and use costly generated electricity instead. Even Monbiot’s incorrectly low figure implies that we’d need to use the world’s entire current solar energy consumption more than twice over to produce sufficient protein globally, when it’s sorely needed for other things. I believe this technology is a clear non-starter as a mass food source.
As an aside, I should mention that promoters of bacterial food often stress its superiority to farmed plants on the grounds that its underlying chemical pathway for carbon fixation is more efficient than the photosynthetic pathway of plants. There are various problems with this line of argument. Not the least of them is the fatal muddling of efficiency and cost, given that sunlight costs nothing – which is not the case with generated electricity. I plan to look at this in more detail in my next post.
Anyway, getting back to my main theme, Michael wisely says that he doesn’t think manufacturing processes should be used to produce all the world’s protein: “I for one like my lentils, beans, nuts, and soya”. I’m with him there. My critique isn’t really directed at the odd bit of microbial protein manufacture here and there (though see section 3 below). It’s directed at people who profess manufactured food as a disruptive Counter-Agricultural Revolution that’s going to spell the end of most agriculture and challenge the place of plants in the human diet. Quite simply, it won’t. The cheerleading for it strikes me as just another bit of ecomodernist hopium which allows us to imagine we can lower our ecological impact sufficiently to avoid earth systems breakdown without fundamentally changing the present high-energy, high-capital, growth-oriented global economy and our place within it. I don’t think that’s an option, and we need to get real.
Non-dairy whey protein
The process used by Perfect Day to produce non-dairy whey protein is different to the bacterial one just described. In this case, sugars derived from crops like maize or beet are used (along with other inputs like ammonia) to nourish a fungus that’s been genetically modified to excrete the whey protein. This protein constitutes only 22% of the biomass produced in the process, the remaining 78% being unsuitable as a human food source, albeit with other potential uses.
I don’t have a problem with the use of these agricultural feedstocks as such (I mean, I do have a problem with the industrial production of commodity crops like maize, but I’ll save that for another time). But inasmuch as the case for manufactured food rests on a land-sparing argument about eliminating the agricultural footprint, the non-dairy whey technique looks weaker than the hydrogen-oxidizing bacteria approach, because it’s agriculturally based.
Drawing on a Life Cycle Assessment (LCA) of their process commissioned by Perfect Day, Michael states that the process consumes 13 kWh of electricity to produce a kilo of protein, and notes how favourably this compares with my figure of 65 kWh/kg for hydrogen-oxidizing bacteria. But this isn’t comparing like with like, for various reasons. For one thing, there are other energetic inputs besides electricity, for example in producing the ammonia. The LCA Michael links reports a primary energy use of 56.3 MJ/kg protein, which converts to 15.6 kWh/kg. Granted, this isn’t much higher than the electricity figure, but it only refers to the non-renewable energy used in the process (from which it excludes nuclear energy, which is surprising since nuclear isn’t conventionally counted as ‘renewable’).
Further, this figure is based on allocating the energy between the whey and the biomass byproduct in their 22/78 mass proportions. If you allocate it wholly to the whey – as you probably should, since getting this product is the whole point of the process – the non-renewable energy input as defined by Perfect Day’s LCA comes to around 72 kWh/kg. That’s a lot of precious primary energy to throw at substituting for milk. And, as with almost all the studies in this field, this excludes the considerable energy costs of building, maintaining, decommissioning and rebuilding the industrial and electricity generating plant needed in the process.
I haven’t looked at this genetically engineered fungus to whey protein technique in as much detail as I have the hydrogen-oxidising bacteria to protein powder one, but if my analysis is correct I think there are grounds to be equally or more sceptical about its energetic implications. As I said above, the LCA that Michael links was commissioned by Perfect Day, and arguably this shows in the way it defined its terms – exactly the problem I mentioned of relying on studies supportive of one’s favoured approach.
There has in fact been a more recent LCA which is explicitly critical of LCAs done by Perfect Day on their technique, and finds claims about the environmental benefits of the technique made by Tony Seba, who Michael enthuses about in his post, to be “extremely unlikely”. This study found that the environmental footprint of the Perfect Day technique may be no better than and possibly even worse than animal-based dairy systems:
For many dairy products such as fluid milk, yoghurts and cheeses, it is reasonable to use raw milk as a raw material instead of extracted milk proteins. For these products and in countries with developed dairy chains, dairy proteins within the raw milk are likely to create a smaller footprint than the use of rBLG would create.
(rBLG refers to the recombinant whey protein excreted by the genetically modified fungus).
The lower energy costs of animal-based dairy appear to be corroborated in a master’s thesis co-supervised by the aforementioned Natasha Järviö, which calculated that the total energy cost of protein from fresh milk is 37 kWh/kg as compared to 220 KWh/kg from the genetically-modified fungal approach, and 90 kWh/kg for hydrogen-oxidising bacteria (when the electricity comes from the average Finnish grid). That result of 90 kWh/kg for bacterial protein powder is notably higher than the 65 kWh/kg minimum which I derived.
I should probably mention that the LCA critiquing the Perfect Day finding had some co-authors associated with the dairy industry. I can’t comment on the implications, although that LCA was at least published in a peer-reviewed journal, unlike the Perfect Day one Michael linked. But there does seem to be an emerging agreement in the literature that real dairy products require less total energy than analogous non-animal manufactured products. With the current state of knowledge in this area, I’d suggest it’s probably unwise to assume that dairy-type food produced with cellular biotech approaches necessarily has a lower total environmental impact than food from dairy animals.
Veganism and the defence of nature
Turning now to bigger picture stuff, I don’t – just in case there’s any doubt – have a problem with veganism. People opt for veganism for many reasons, but inasmuch as reducing the human impact on nature is one of them, there are some wider contexts I think it’s good to be aware of that can help clarify personal dietary and wider food system choices.
Michael references the inefficiency of meat and dairy production, and it’s certainly true that you can meet most human nutritional needs from a smaller land footprint and with low primary energy costs if you eat an exclusively plant-based diet (for several reasons discussed in my book it doesn’t follow that the land thereby ‘saved’ will actually benefit nature, which is one of the problems with couching the issue in terms of individual consumption rather than structural politics, but that’s another issue).
But the lesson from the energetics of manufactured food outlined above is that if you move away from eating wholefood plants and vegetables towards manufactured fungal or bacterial products, this efficiency argument weakens in terms of energy costs – possibly to the point of comparing unfavourably with animal-agriculture analogues. This parallels the efficiency arguments against animal agriculture. If instead of following the direct plant-to-human-stomach route you introduce intermediary processes for which plants are at most feedstocks, it’s hard to avoid increasing the costs in terms of energy and/or land footprint – and this is true whether the intermediary bioreactor is a cow’s stomach or a stainless steel fermenting vessel in a factory. Either way, you risk losing the efficiencies of the direct plant-to-human-stomach route. So if you’re opting for manufactured protein on environmental impact grounds, it’s worth being aware that it may not be low impact just because it’s livestock-free.
The flipside of this argument is that when livestock are used to cycle nutrients and deliver ‘ecological services’, as in most traditional mixed agricultures, and don’t compete with humans for their food sources, this doesn’t detract from system efficiency, but adds to it – an argument I outline in more detail in my book. The fact is that not much of the animal-based food in modern supply chains derives from such efficiency-augmenting mixed agricultures, at least in rich countries like the UK, so this argument in itself isn’t currently a strong one against veganism.
Still, those modern supply chains have emerged as a result of and largely depend on cheap fossil energy. In that sense, I disagree with Michael when he says “Changing our diets to eat less meat and dairy is the most impactful action anyone can take to tackle our most pressing environmental crises”. The most impactful action we can take collectively is cutting our use of abundant cheap energy in general, and fossil fuels in particular. If we did that, it would inevitably and drastically cut livestock impacts, because we’d have to meet needs for food, fibre, energy and fertility from local bioregions, and any livestock component could only serve that larger need. That’s basically the argument that I outline in my book – not high-energy farm-free food, but low-energy local mixed-farming food.
The issue is no longer our individual consumption choices within an existing global commodity food system, if it ever was. Like it or not, that system is unravelling, and I think the result is going to be the widespread adoption of low-energy local food systems. Those systems will be many and varied, but in general they’ll produce a lot less meat and dairy than people in the rich countries are accustomed to eating. They will not, however, involve no meat and dairy, except possibly in the most densely populated areas. I seriously doubt many of these systems will involve much in the way of high-energy biotech manufacturing of bacterial or fungal protein. That kind of energy and material profligacy will scarcely be affordable in the world to come.
And that, in a nutshell, is why Michael’s blog post hasn’t convinced me that my arguments in Saying NO… are wrong.
Current reading
I mentioned in my last post that I’d start listing my current reading. I’ve got into the habit of reading multiple books simultaneously, some of which take me months to finish. So I think I’m just going to mention any new books that I’ve started reading since my previous post – which will probably give a misleading impression of my actual reading rate. Another biased methodology!
New reading:
Peter Heather Empires and Barbarians
Naomi Klein Doppelganger
Régine Pernoud Those Terrible Middle Ages
A better gender balance, but still no fiction!
Appendix
If you read around the literature on manufacturing edible protein from hydrogen-oxidising bacteria, you’ll come across various figures for the energy cost of the process. Here are a few:
- 9.86 and 10.96 kWh of electrical energy input per kilo of bacterial biomass reported by Sillman et al in two papers here and here
- 16.7 kWh/kg bacterial protein reported by George Monbiot in his book Regenesis (p.190)
- 18 kWh/kg bacterial biomass reported by Järviö et al
- 17.8 kWh/kg bacterial protein as the lowest theoretically achievable energy input required by the chemical reactions involved in producing the protein reported by Wise et al
- 65.3 kWh/kg bacterial protein reported by me in my Saying NO… book, calculated from a paper by Leger et al on the basis of a methodology I explain here
I’ll now try to make sense of these figures, and specifically why they differ.
One thing to note: some of these figures specify total bacterial biomass, while others specify only the protein component, which is typically 60-65% of the biomass.
Monbiot (eventually) revealed that his 16.7 figure was derived from Sillman et al’s 9.86 figure (9.86 rounded up to 10 and then divided by the 60% protein component = 16.7). Note that this 9.86 kWh figure refers only to the electrical energy input into the bioreactor and not to other electrical and other energy costs that Sillman et al report. It’s demonstrably not the total energy cost of the process and isn’t even theoretically possible in view of the 17.8 kWh minimum established by Wise et al. The Wise study suggests that this minimum figure could only be achieved by genetically modified bacteria (not the strain used by Solar Foods) and only then momentarily during the production process – not as the full energy cost of the manufactured protein, which will inevitably be higher.
Even Sillman et al’s 10.96 kWh/kg figure appears not to be a real-world one. They describe their analysis as a “quantitative literature review”. In the words of Järviö et al, Sillman’s figure was “based on theoretical assumption using currently available but limited literature values”.
Now, Michael argues that I should have used the figure from Järviö et al (they report 18 kWh/kg bacterial biomass, so at 65% protein content that would be 27.7 kWh/kg protein). The trouble is, the study is vague about which energy costs (beyond electricity) it considers for the various processes involved. Michael says the paper makes it clear that the 18 kWh/kg is an actual measurement and not an assumption. What the paper actually says is this:
Whereas this study assumed an electricity requirement of 18 kWh per 1 kg product produced, Sillman et al. (2020) estimated 10.96 kWh per 1 kg product produced. This difference could mostly be explained by the fact that the estimate of Sillman et al. (2020) was based on literature values whereas this study was based on empirical data.
So the Järviö study ‘assumes’ a requirement of 18 kWh per 1 kg of biomass (or 28 kWh per 1 kg of protein), which is ‘based’ on empirical data, with some of the data apparently from the pilot-scale production of Solar Foods. The paper briefly mentions the assumed electricity requirement while leaving out other energy requirements. The supplemental information document gives some details which add up to 17.83 kWh of electricity for the fermentation step, before the separation and drying steps, without including all the energy required for the ammonia, carbon dioxide, steam, and mineral inputs. The 18 kWh figure seems to leave out or underestimate some energy costs which the Leger et al. study includes and substantiates. The Leger study devoted much more detailed consideration to the total energy costs.
Another study that looked at hydrogen-oxidising bacteria as human food shows an energy requirement of 37.8 kWh/kg for bacterial biomass with 65% protein content, equivalent to 58 kWh/kg of protein, which is pretty close to the 65 kWh/kg I derived from the Leger et al. study.
In fairness, the Järviö study has a wider focus than the energy input into the process and it seems to me that the authors didn’t probe the issue all that thoroughly. By contrast the Leger et al study I used does clearly specify the energy inputs it considers (it omits some things which could be pretty important, such as the energy costs of manufacturing, installing, maintaining, decommissioning and replacing the PV generation and other facilities). I concluded that the Leger paper was more reliable in relation to energy costs than the Järviö one. Unlike the latter, it provided a clear mathematical rationale for its energy derivation. So for my purposes, it was unquestionably the stronger paper to use.
The thesis mentioned earlier took a closer look at the Järviö study and calculated that the resulting total energy cost for hydrogen-oxygenating bacterial protein is at least 55.5 kWh/kg protein, when using electricity from renewable sources only. When using the average Finnish electricity grid, which has some additional energy requirements due to generation losses from non-renewable sources, the data from the Järviö study resulted in a total energy requirement of 108 kWh/kg of protein.
