An October 25 news article entitled “ Not enough fruits, vegetables grown to feed the planet, study reveals,” caught my attention, partly because of its title and its University of Guelph origins (my alma mater and former employer), but also because of some statements/conclusions seemed false. I read the referenced research paper in detail (Krishna Bahadur KC et al, https://doi.org/10.1371/journal.pone.0205683, October 2018) and am providing this overview and critique for those who may be interested.

The paper claims to show that the world is unlikely to be able to meet its need for a quality diet in 2050 without major increases in land usage for agriculture and greenhouse gas emissions. However, I conclude that dubious assumptions and apparent flaws in the calculations cast major doubt on the validity of this conclusion. I explain why below:

In brief, here are some of the major findings and conclusions:

I’ll discuss them in more detail later.

1. If the world continues to consume its present average diet, we’ll require almost no increase in arable land to feed 9.8 billion in 2050, but a 40% increase (about 1.4 billion ha more) in pasture land. That’s based on an assumed 1% annual increase in the productivity of arable land but 0% for pastureland and for the conversation efficiency of livestock (discussed below).
2. If we switch to the superior diet which the authors have designed based partly on Harvard Health Eating Plate (HHEP), this would mean a small (4%) decrease in arable land required by 2050 but a 59% increase in pastureland. This, in turn, would mean a 108% increase in greenhouse gas (GHG) emissions from agriculture. That does not include any emissions associated with the conversion of non-agricultural land (presumably forest, natural grass and conservation/park lands) to pastureland. The authors propose converting the 4% ‘saving’ in arable land usage into biologically diverse habitat rather than using it to diminish the added requirement for pastureland – stating this is for a net environmental benefit. However, the rationale is not explained. The apparent assumption is that converted arable land makes better natural habitat than the equivalent hectares (or more) of additional forest and/or natural grassland converted into grazing land.
3. If we switch to a vegetarian diet which includes milk but no meat consumption, which the authors suggest as an option, that would mean an 8% increase in arable land usage compared to present (an increase equivalent to about twice the current arable ha in Canada) but a 10% decrease in pasture land by 2050, and also a 12% reduction in GHG emissions. This does not include any GHG emissions associated with the conversion of non-arable land to arable, and assumes that the productivity of the converted land will be the same as existing arable hectares. The authors’ calculations assume the elimination (burying, burning, or other disposal) of the bodies of cull ruminant animals raised for milk, but not used for human meat consumption, on 3 billion ha of pastureland. More on that below.
4. The authors considered a fourth alternative that involves a major reduction in meat consumption (20% of the ‘protein’ part of diet from meat, versus 65% as calculated for HHEP). They estimate this will require a 5% and 6% respective increase in arable and pastureland needed in 2050, compared to 2011 – and about a 56% increase in GHG emissions, if I interpret their data correctly. (The relevant supplementary table contains some apparent labelling errors.) The unlikelihood noted in the previous paragraph may also compromise these calculations – in fact to a large extent – if meat from dairy ruminants is not included.
5. The authors also discuss three other options for meeting nutritional goals with available land resources: Improving the GHG efficiency of animal, crops and seafood production; increase research to increase yields of fruits and vegetables by 8% per year and livestock meat by 3% per year; and reducing food wastage. The authors provide limited insight as to the likelihood of achievement with any of these.
6. In the above, authors assume that the world is one homogenous pool of production and consumption, and that global needs can be calculated without considering differences in regional balances of supply and demand/need – nor difficulties and costs involved (including GHG emissions) in shipping surplus food ingredients from one region/country to another. More on that below, as well.
7. The news release contains a claim that vegetable and fruit crops are higher yielding than grain, oilseed and sugar crops – with the implication that arable hectares required for food production can be reduced by shifting land from the latter to the former. (The paper provides no supporting data or references for this claim, which in fact is at odds with contents of the paper itself, and is almost certainly false.)
8. The news release also implies that government programs for grain and oilseed crops are a major reason why farmers grow too much of them and not enough fruits and vegetables. There is no consideration of market place realities and that farmers generally grow food crops for which consumers are willing to pay (at least enough to cover costs of production). This issue is not raised in the paper itself.

A quick review of methodology:

As a first step, the authors calculated the daily intake of various food groups needed for a healthy diet, based partially on the Harvard (university) Healthy Eating Plate (HHEP). Authors then calculated the amount of each food group produced globally using an FAO data base, and compared the land need for world population of 7 billion in 2011 and an estimated 9.8 billion in 2050. The authors assumed an average annual 1% increase in per-ha productivity for all arable crops but no increase for pasture crops or in the efficiency of converting crops into milk and meat. (The assumption about lack of change in pastureland and livestock productivity is not stated in the paper but was provided in conversation with the senior author. This assumption is the reason for calculations showing a far greater future need for more pastureland compared to arable land.)

The paper also includes an allowance for food provided by ocean fisheries and assumes 20% household food wastage.

The rationale for choosing the HHEP guidelines and for the way in which they used it is murky. They state that they considered using Health Canada recommendations, but didn’t because of a published suggestion from one writer, Dr. Marion Nestle (a professor at New York University, known for her very negative views about food companies), that these could be biased. There is no indication that sources other than HHEP and Health Canada were considered.

Also puzzling is the fact that, despite specific instructions in the HHEP reference that these guidelines “are not meant to prescribe a certain number of calories or servings per day,” authors of the current paper do just that. Then they go further in making some estimates of what the HHEP guidelines mean; for example, the HHEP statements of “protein power – ¼ of your plate” and “healthy plant oils in moderation” are interpreted by the authors (no reference cited) to mean “1 serving of fat/oil, 1 serving of milk/dairy, and 5 servings of protein.” The authors further divide ‘protein’ into plant and animal protein sources (65% animal according to an appendix table).

The HHEP guidelines are clear that potatoes are not to be included in the vegetable portion of the ‘heathy plate’ because of the nature of their carbohydrate composition (too easily digested). However, the authors of the paper included potatoes in their calculations anyway. The authors show that potatoes represent over 18% of the world’s current hectares devoted to vegetable production for direct human consumption. (This increases to 31% if cassava, with a carbohydrate composition similar to potatoes, is included.)

The authors state that HHEP guidelines say red meats should be restricted to two portions per week, though I could not find that advice stated anywhere in the literature citations, and it doesn’t appear to figure directly into their calculations.

These assumptions and omissions seem very significant to me in that modest changes in diet might be expected to mean very large changes in the predictions. For example, the authors note that use of Health Canada’s dietary guidelines would have meant “27% fewer servings of fruits and vegetables, 34% fewer servings of meat/protein, but 60% more servings of dairy products and 25% more grains,” compared to their idealized diet based in part on HHEP.

A few other comments:

The assumption of a zero increase in pastureland productivity between 2011 and 2050 seems extreme and unrealistic. Indeed, the authors cite an example of how pasture management can be improved, and another describing research to improve the efficiency of livestock feed conversion. A quick Google search found this review detailing how feed conversion efficiency has improved for dairy cattle and outlining opportunities for continued improvement.

A weak understanding of livestock, especially ruminants, seems apparent in the paper. I’ve already mentioned the assumption of milk but no meat from dairy animals. I know that there are some countries/regions where cows are milked but not used for meat but these represent a minute share of world production. India, where cows are prominent but bovine meat consumption is not, is actually the world’s largest or second largest exporter of beef meat. The authors present calculations that only 30% of present world pastureland is used to produce meat and the rest all for milk/dairying milk. I also doubt the accuracy of that that figure. The concept of dual-purpose ruminant animals (milk plus meat) is not mentioned.

If meat from dairy animals is included in calculations of global food consumption, as would seem the most logical assumption, there may also be critical flaws in their calculations for a low-meat-but-full-dairy diet. (A check of their data suggests that the authors have likely ignored meat from dairy animals, but it’s ambiguous.)

I question the authors’ calculation of 1% average annual increase in productivity of arable land, for it turns out that this assumes no trend for increased annual improvement between now and 2050. The increment each year until 2050 is assumed to be 1% of crop productivity in 2011. Hence, the productivity in 2050 is calculated to be that of 2011 multiplied by 1.39 (39 years). But what if that average 1% is accumulative, as seems more logical, i.e., 1% of each previous year? The annual productivity in 2050 would then be that in 2011 multiplied by 1.01³⁹. This calculation would mean 8% greater arable production in 2050 or equivalent to the production from about 80 million ha. Compare this with a quote from one of the authors in the news release, “Without any change, feeding 9.8 billion people will require 12 million more hectares of arable land.” I suggest the 12 million is like a rounding error, compared to the uncertainty in assumptions.

The authors ignore options for meeting nutritional needs in 2050 other than shifting portion sizes among various food groups. It may make sense to eat more carrots to get vitamin A in Canada, but in Southeast Asia? Why not genetically bio-fortified rice or the similar advances now in development with cassava and bananas in Africa? There is lots of research underway globally on improved nutritional composition of agriculturally based foods.

A similar comment applies to assumed primacy of the HHEP diet. My impression is that the developers at Harvard created this for Americans or others who eat similar foods. But why assume this idealized diet applies to someone in Uganda or Bangladesh?

My biggest concern with the study

It’s the assumption that world agriculture and food supply/demand is one big pool and that regional distinctions are of negligible significance.  A proper analysis, in my view, should allow for inter-regional shipments, of course, but must also recognize need for a substantial degree of regional self-sufficiency. (The experiences of 2008 and its food supply/price panic showed the importance of that.) It should also recognize regional differences in what is considered a proper, balanced diet. If that were done, the conclusions in this paper might change very substantially.

So what do we learn from this paper with everything considered?

The main conclusion is that if the world’s citizens shift to a diet which contains fewer grains, oilseeds and sugar, and more fruits, vegetables and protein, then farmers will need to grow more of the latter and less of the former. Will this require more or less land? We don’t really know from this paper. It all depends on assumptions.

The paper would appear to show that it will be almost impossible for the world to meet future nutritional needs without animal agriculture. That’s a huge issue considering the abundance of attacks these days on animal agriculture. Unfortunately, given the weaknesses in the paper’s calculations, especially concerning the use for meat of dairy animals, no conclusion is possible using the analyses in this paper.

As for effects on GHG emissions, it depends largely on assumptions about relative dependence on livestock production – as we already know from many previous publications and analyses.

I do thank the authors for stimulating lots of thought – for both me, for sure, and I expect many others. A special thank you to senior author, Dr. Krishna KC who was very gracious and open in discussing the paper with me, clarifying some ambiguities.

In closing: Before posting this critique, I provided a draft copy to key several key authors requesting their advice on where my comments were in error or unjust. The responses identified neither errors nor unfairness, but stated simply that their paper was intended to stimulate discussion.