Guest essay by Alberto Zaragoza Comendador
CO2 is food for plants: a simplistic statement, but also a correct one. The factoid is so well-known, so basic, bringing it up in a discussion may only elicit a shrug or an eyeroll. Nevertheless, I have found that some very prominent papers on the topic of climate change impacts ignore it – just one reason the results they get are dead wrong.
Warning: this is a serious i.e. boring article. If you want to have a laugh check this out.
Yes, CO2 also causes other stuff but the net impact on plants is positive
Let’s get this out of the way. There is some hand-waving about how the CO2 fertilization effect will ‘peter out’ (at some unspecified point in, uhm, the future – it won’t peter out in the past); pessimists also argue that CO2 fertilization is ‘countered’ by other factors, such as the warming induced by CO2 itself (heat stress), or possible weather changes caused by this warming. But such statements are meaningless without quantification: every force in this world is ‘countered’ or ‘offset’ by other forces. If you jump off a tenth floor the force of gravity will be ‘countered’ by the air drag, which will make you fall slower than you would have if affected by gravity alone; and yet the force of this air drag is utterly irrelevant when assessing the impact of jumping off a cliff. In other words, some effects and forces matter more than others.
The bottom line is more CO2 helps plants grow more, the effect will not ‘peter out’ until we reach CO2 concentrations several times higher than now (at least for C3 plants, which is to say 95% of the world’s plants), and there really is no reason to think commercially-grown plants (agriculture) behave differently than wild ones. The last point is critical. When discussing the increase in agricultural yields, there is of course a confounding factor in that agricultural technology and technique change over time (usually for the better). Thus CO2 fertilization cannot be thanked for the entirety of this increase. But forests don’t rely on technological contraptions such as genetically-modified seeds: the fact that they, too, are growing almost everywhere is a strong indication that the effect of CO2, net of any change in technology and technique, is helping plant growth. And yes, a bigger plant will usually produce more – hence it should also be helping to increase yields.
In other words, the increase in global tree cover, green area, leaf area or whatever you call it is evidence that CO2 fertilization is more than offsetting the alleged effects of increased drought and so on. If it works this way with wild olive trees, why wouldn’t it do the same with commercially-grown ones?
The papers in question
I’ve found this problem in:
· Global non-linear effects of temperature on economic production, by Burke, Hsiang and Miguel. I will refer to it as BHM15. Cited 66 times according to Google Scholar.
· Nonlinear temperature effects indicate severe damages to US crop yields under climate change, by Schlenker and Roberts (SR09). Cited 883 times.
· Rising temperatures reduce global wheat production, by Asseng et al (AEA14). Cited 101 times.
BHM15 describe the methodology as: ‘We estimate how economic production changes relative to the previous year—that is, annual economic growth—to purge the data of secular factors in each economy that evolve gradually’. There follows a paragraph providing more detail, but the key is that the comparisons in temperature and production are made between 1970 and 1971, between 1995 and 1996, etc.
If you only compare the year-on-year change in economic activity you’ll eliminate the long-term trend. In principle this sounds good, as we don’t want the effects of an increase in temperatures to be conflated with, for instance, a secular slow-down in economic growth (or the aforementioned improved technology, in the case of agriculture). But if you remove the long-term trend you also remove the CO2 effect. There is no mention of adjusting the data to account for this – in fact the words CO2, carbon, and fertilization appear in the text 0 times.
Let’s use a simplified example. Yearly temperatures tend to zig-zag: rather than every year setting a new record high by 0.01ºC or so, the long-term positive trend is marked by many ups and downs – and when a new record is set, it is usually by more than the average yearly increase. A way to make more sense of what the current temperature level ‘is’ is to use temperatures averaged over more than a year (the longer the better as we reduce year-to-year noise). Let’s assume a decadal average removes all the year-to-year noise. The question is, when will decadal temperatures increase 1ºC from the 2011-2020 average?
(HadCRUT4, monthly values; courtesy of Wood for trees. The yearly changes are smaller than those seen here).
Imagine the data shows that, in years that were 1ºC hotter than the previous, yields fell on average 10%. Yes, I know that in the real world single-year changes are never that big, but it serves to illustrate the example; the authors somehow extrapolate from yearly changes of 0.1 or 0.2ºC to century-scale changes about 10 times bigger and I’m not qualified to comment on how valid their methods are.
The key is that technology in both years should be almost the same so we’re pretty sure that the change in production was due to the weather. Therefore when decadal temperatures have risen 1ºC, you would also expect production to decline 10%.
What the authors forget is that by the time decadal temperatures have risen 1ºC, CO2 levels will be much higher. Thus, the effects of a 1ºC increase from 2015 to 2016 will be radically different from the effects of the same temperature increase from 2011-2020 to, 2091-2100, for instance.
How different? Imagine we start with 400 parts per million (ppm), which is about right for 2015. If transient climate response is 1.5ºC, which is lower than in most climate models but higher than in recent studies of the thermometer record (see slide 51), then a temperature change of 1ºC is equivalent to 2/3 of a doubling in CO2 concentrations. There are other greenhouse gases, and going by the recent ratio of CO2-to-other-GHGs radiative forcing (historically 65%, but about 80% in recent years – see table 2), we can assume that instead of 2/3 of a doubling, this would require 0.67 x 0.8 = 0.536 doublings. There is also a warming effect from the aerosol forcing, which is becoming less negative over time, plus possibly additional warming from a decline in the radiative imbalance (heat released from the ocean).
So it’s unclear whether the additional 1ºC would require increases in CO2 of 30, 40 or 50%. But any of these increases is huge. In the lowest case, we’re talking about an additional 120ppm – about the same increase as took place from the ‘preindustrial’ era, circa 1750, to 2015. The study I linked to at the beginning found a strong greening trend between 1982 and 2009, when CO2 levels rose only by 30ppm. It’s safe to say the greening from a CO2 increase 4, 5 or 6 times bigger would be much greater.
In short: the authors looked at the effect of warming on plants, but failed to factor in the fertilization effect of the very CO2 that will cause the warming. This is similar to the mistake in a paper by McLean and two other authors: they detrended temperatures and saw that the main factor explaining the remaining variance was El Niño, thus concluded that a change in El Niño patterns could have caused the warming seen in recent decades. But of course, if you eliminate the trend you eliminate the CO2 effect, whether we’re talking about agricultural yields or temperatures.
Now it’s time to see SR09. This paper is focused on the growth rates of specific crops grown in the US under different temperatures. It uses daily temperature data rather than year-on-year changes. Having arrived at it after reading BHM15, which cited it, one of the first things I did was search for the word ‘fertilization’ and boom:
‘An important caveat concerns our inability to account for CO2 concentration. Plants use CO2 as an input in the photosynthesis process, so increasing CO2 levels might spur plant growth and yields. Yield declines stemming from warmer temperatures therefore may be offset by CO2-fertilization… We cannot account for CO2 effects in regression analysis of observed yields because CO2 concentrations quickly dissipate throughout the atmosphere, leaving only a gently increasing time trend, which is impossible to statistically disentangle from technological change.’
Emphasis added. On the one hand, it’s admirable the authors themselves note this caveat. On the other, one wonders how many of the people citing the paper are aware of this.
Finally, AEA14 is a study that tries to simulate the past (responses of agricultural yields to temperature increases) to forecast the future. Its headline claim was:
‘Global wheat production is estimated to fall by 6% for each ◦C of further temperature increase’
And how do they know that?
‘We systematically tested multiple models against field and artificial heating experiments, focusing only on temperature responses.’
Emphasis added. There is no deception on the side of the authors, but the decision to run models that only vary temperature is bizarre: it invalidates their results completely, for the same reason as the other two papers. Even more bizarre is the fact that their references show four papers that do take CO2 into account (two of which included Asseng, the lead author of this paper).
Studies refuted by 2015-16 temperatures, not just by CO2 fertilization
I’m just going to comment on what Andrew Bolt pointed out a few days ago with a bit more precision: temperatures over the last two years have surged into what the papers above classify as decline territory, and yet all major crops are at or near all-time highs. In fact, the rapid increase in temperatures is as close to a perfect experiment as the real world could get, because there has been almost no time for adaptation and technological change. One can even dismiss CO2 as a growth factor for the period: concentrations of this gas have risen in the past two years by only 5ppm.
Again checking HadCRUT4 we see that temperatures in the first half of 2016 were about 0.4ºC higher than the average for 2010-2014.
Even this is an underestimate, though. While BHM15 referenced global temperatures, SR09 and AEA14 discuss the impact of field temperatures, i.e. right where the wheat, corn, etc. grow; they also specify that these are the temperatures during the growing season. So even though the last quarter of 2016 will be colder than the first nine months, reducing the gap with average 2015 values, we can ignore that; spring and summer temperatures are what matter most for plant growth.
If the question is ‘what were the temperatures where wheat (for example) was grown’, well, I don’t have an answer. But the fact that a perfect answer is not available shouldn’t keep us from trying to get one that is closer to the truth. At a minimum, we can check temperatures in the Northern Hemisphere, which is where 90% of us live and where most food is grown.
Well, there you have it: the first half of 2016 was about 0.6ºC warmer than the 2010-2014 average. And sorry to repeat myself, but this is probably an underestimate as well – the increase over land was surely greater than over water. But I believe you get the point.
AEA14 expected wheat yields to decline 6% for each additional degree centigrade. So how much did wheat production decline from 2014 to 2016? 2%? 4%?
It’s actually setting a new record…
‘The outlook for wheat also improved (by 1.2 percent), putting this year’s world production forecast above the 2015 record, at 741 million tonnes’
I could go on and on with more crops showing blockbuster results, but I’d be beating a dead horse. We’ve seen the effects of a temperature surge, and they directly contradict the predictions of many papers and reports.
Including the IPCC’s latest report, which is looking more ridiculous by the day.
More research is needed (on why these forecasts fail so hard)
I’m an amateur. I lack the time to pore over the literature and the skill to analyze these papers in detail. But this seems to be an issue that could affect a lot more papers. There seems to be a consistent tendency for agricultural forecasts in climate science to fail miserably, and although it’s not like the world depends on correcting these studies, if we understand why they are so wrong perhaps they’ll be a bit more useful – and we’ll learn something along the way.