From Climate Scepticism

By Jit

We’re all going to have to give up breakfast, or something.

A week ago, the compliant media was happily swallowing, and regurgitating, yet another “worse than we thought” study, this time on the topic of how future climate change was going to starve humanity.

“Worse than we thought” is, when you pop your ten-gallon sceptic’s hat on, a synonym for “exactly what we thought.” By which I mean, the news has to be terrible, and worse than the previous news, for it to get traction in today’s world of transient thought bubbles.

The study, “Impacts of climate change on global agriculture accounting for adaptation,” was published in the once-great magazine Nature, and is authored by Andrew Hultgren and a motley crew of divers others, too many to count at a blink (i.e. more than nine). One of the authors, Soloman Hsiang, was quoted in one story as saying the losses due to climate change were going to be equivalent to us all giving up breakfast, or words to that effect – though I can’t find that page any more, it is now buried in the others that I visited to read up on how the study was being reported.

According to this press release by Rutgers, the more than nine researchers spent eight years on the research. Time well spent, I say, especially as the answer was so surprising.

What of the study itself? As we know, most of us don’t read the study in “study says…”: we simply take the media’s regurgitation of its talking points at face value. Well, I wanted to (skim) read it, because my sceptic’s radar was picking up strong echoes of bullshit. Was 50% loss of yield remotely plausible? Instinct says no. What of the CO2 fertilisation effect? This should counteract some of the “climate” yield losses. What of the climate model used? RCP8.5, or something remotely plausible?

Hultgren et al, et al, can be found here, and some respect is due to the authors for making the study free to read. (I have long been of the opinion that all science should be free to read. Of course, it means that the gatekeepers, Nature et al, would no longer have a business case. How would we differentiate gold-standard science from frass? Perhaps a system of voluntary readers could be set up, and studies could get reviews and star ratings, like in Amazon. Impossible of course, owing to the politicisation of science today.)

Skim skim skim…

Search for “fertilization” with a z…

Several references to fertiliser of the non-gaseous kind, then a mention in Figure 2’s legend. The figure is the first showing results. Its title is: Projected end-of-century change in crop yields resulting from climate change, accounting for adaptation to climate and increasing incomes. [I’m only showing the wheat panel here. Looks bad for the U.S., huh? But at least we can grow wheat in Mongolia, and on the north coast of Norway….]

The relevant parts of the legend read:

Colours indicate central estimate in a high-emissions scenario (RCP 8.5), net of adaptation costs and benefits, for maize (a), soybean (b), rice (c), wheat (d), cassava (e) and sorghum (f) for 2089–2098.

And

See Extended Data Fig. 7 for a moderate-emissions scenario (RCP 4.5) and Supplementary Information, section J for results adjusted by CO2 fertilization.

So, if we are to get what might be termed, “plausible” results, we cannot even read the paper itself: we have to delve into Supplementary Information, and Extended Data. The authors have presented something here that they must know is wrong, and that any reviewers must have known was wrong, and that Nature’s editors must have known was wrong. Who cares? It got a headline. Or eleventy-six.

The Supplementary Information runs to nearly 100 pages. Am I alone in thinking that Supplementary Information is out of control these days? Anyway, searching for “fertilization” again finds table S11 on page 66. The column of numbers to take notice of is the one furthest right, showing the model’s prediction of yields under the realistic RCP4.5, with CO2 fertilisation, and with farmers actually doing their job, not growing crops that won’t make them money. The first number next to each crop is the prediction of average yield changes. The numbers in square brackets are the 5% and 95% ends of the distribution. [Note: this means that 90% of the distribution is within the brackets.]

Here we see that the predicted yield distributions for five of six crops overlap 0. In other words, they are not significantly different from zero, and that no change in crop yield is a viable interpretation of the results. The exception is wheat, which is negative (but we can’t say for sure it is significantly negative, thanks to the extra margin on the confidence interval).

So: production could drop by 50%.

Or: production could remain the same.

Next: is the fertilisation effect realistic?

We know that CO2 is plant food, although the alarmists hate the idea with zeal. It’s easier to photosynthesise when there is more CO2, and that is not a controversial statement. It’s also non-controversial that C3 plants like wheat and rice will benefit more from CO2 enrichment than C4 plants like corn. [C4 plants can be thought of as plants that have evolved to eke out a living in a world where the food supply – CO2 – has dwindled over aeons until plants are half-starved.] The authors’ treatment of CO2 fertilisation is described on p.69 of the S.I.: does it pass the fairness test? Or did they pick a fertilisation model that downplays the likely effect?

At end-of-century, the CO2 fertilization effect under RCP8.5 increases yields by 9.8% and 5.2% for C3 and C4 crops. Under RCP 4.5 end-of-century yields increase by 5.8% and 3.5% for C3 and C4 crops.

I’m sorry – what? RCP8.5 increases yields by a mere 10% for C3?

Hultgren et al draw their CO2 fertilisation model from a meta-analysis by Moore et al. 2017. [The link takes you to their response function.]

Moore et al present revised damage functions from climate change – and you guessed it! They told us what we already knew. Things are “worse than we thought.” Is Moore et al’s CO2 fertilisation model fair? Only a cynic would suspect otherwise, right?

Well, it might be fair, I’m not sure about that, but it’s clearly wrong. Here’s how the modelled relationship works:

The baseline CO2 is 360 ppm, which is not “pre-industrial,” usually given as 280 ppm. The result is that at 360 ppm the change in yield is 0, by definition. The parameters used mean that half of the CO2 fertilisation for C3 plants occurs in the first 100 ppm (from 360 to 460 ppm) and rapidly tails off thereafter. Starting at 360 ppm is mad, because if you trace the curve back to 280 ppm, you end up with a yield of about -70% of the “modern” baseline of 360 ppm. In other words, if this figure represented reality, everything was dead in the pre-industrial era. Here’s my emulation of their figure for C3 plants, but going back to 280 ppm.

Now, you might say that the figure is and should only be valid in the range it is modelling. Maybe; but it clearly shows that the beginning of the curve is too steep. Here’s a different view of CO2 fertilisation, what you might call an ecologist’s view, as annotated by Dave Burton of sealevel.info:

Here, C3 have plenty of gains in their pockets as the ambient concentration of CO2 goes up.

Note also that using 360 ppm as a baseline for yields means you have banked quite a bit of CO2 fertilisation and effectively zeroed it out: the benefits of increasing CO2 from 280-360 ppm are null.

Here’s what Hultgren et al did:

To apply these estimates of CO2 fertilization to our projections, we estimate this CO2 fertilization effect for each future year using RCP-specific CO2 concentrations from [201], and we subtract off the CO2 fertilization effect for the year 2015 (the end of our baseline period).

So most of the CO2 fertilisation has been banked and zeroed out, and what is left is not enough to balance out the “climate” impacts.

At end-of-century, the CO2 fertilization effect under RCP8.5 increases yields by 9.8%

This figure shows the CO2 concentration in 2100 under the RCP8.5 scenario (Figure reproduced from Denierland, data from climatechange2013.org (AR5). [Ignore “jit’s simple model”; one of these days, I’ll check to see how it is doing prediction-wise.]

The 2100 concentration under RCP8.5 is 936 ppm, up from the baseline year (2015 for Hultgren et al) of 400 ppm. And the effect of CO2 fertilisation, they say, for that change (more than a doubling) is only +10%? I think this is low.

This NBER paper (“not peer reviewed”) by Taylor & Schlenker (2023) found that we are still firmly in the linear response phase (cf. the figure annotated by Dave Burton above):

We consistently find a large CO₂fertilization effect: a 1 ppm increase in CO₂ equates to a 0.4%, 0.6%, 1% yield increase for corn, soybeans, and wheat, respectively.

Now, Taylor & Schlenker’s numbers may be optimistic. I wouldn’t be surprised (their response curve is more of a cloud – it’s one of those where if you have enough data, a cloud can have a significant slope). But this range in interpretations is somewhat concerning. After 40+ years of climate research, we still don’t know what effect CO2 will have on crop yields.

As to the story about agricultural apocalypse, it’s another case of paper published, hysterical headlines published, good job well done, move on to the next “worse than we thought” headline-hunting journal article.

Meanwhile, what are Hultgren et al, et al’s results for the more realistic RCP4.5, with their (in my eyes) questionable CO2 fertilisation bonuses?

[Same key as above.] Thankfully, under the realistic scenario we can grow wheat on the Himalayas, Svalbard, and Kaffeklubben Island off the north coast of Greenland. [Snipped from Supplementary Figure S.14, p.71 of the S.I.]

/message ends