I’m still trapped in this hell-hole of a right-wing think-tank’s attempt to wish climate change away. It almost makes me miss Vox Day.
Intro, Ch 1, Ch2, Ch3, An Aside, Ch4, Ch5, Ch6, Section 1, Ch7, Another Aside, Ch8, Ch9, Ch10, Ch11, …
Kesten C. Green & J. Scott Armstrong are both affiliated with the Ehrenberg-Bass Institute at the University of South Australia. Now that sounds quite impressive but it is an institute of Market Research. No, no, that’s OK – no snide comments – market research is a numerate discipline and is in an interesting place developing underlying theories and models. Still as a discipline it has neither the depth, success or theoretical clout of climatology and meteorology.
Now you may, if you’ve been paying attention, wonder why a discussion of the efficacy of forecasting techniques of climate science is stuck as the last chapter of the ‘economics and politics’ section rather than in the ‘science’ section. Given that this chapter is going to claim to demonstrate that forecasts of global warming are somehow invalid, then you might doubly wonder that. Indeed, given how confident Green and Armstrong are of their finding, I’d be a little disappointed that I wasn’t Chapter 1 and that Plimer, Michaels, Linden, Soon, Carter et al weren’t making a big hullabaloo about this chapter – particularly Soon who apparently has collaborated with Green and Armstrong. It is almost as if even fellow doubters are unconvinced.
Kesten C Green is, or presents himself as, an expert in forecasting as a general discipline. Looking through his publicly available work that isn’t global warming related, his work has grown out of forecasting techniques for businesses. In particular describing the principles that should be applied when faced with complex data (for examples sales data or perhaps political polling data) and attempting to make guesses about what will happen next.
Green describes his view of forecasting like this:
For nearly a century, researchers have been studying how best to make accurate and useful forecasts. Knowledge on forecasting has accumulated by testing multiple reasonable hypotheses about which method will provide the best forecasts in given conditions. This scientific approach contrasts with the folklore that experts in a domain will be able to make good forecasts about complex uncertain situations using their unaided judgement, or using unvalidated forecasting methods.
And that all makes sense. Empirical, data-driven approaches make sense. Intuitive approaches don’t, even when those intuitions are based on people with experience. Such approaches should involve substantive models when possible but collecting reliable data and basing claims (whether they are forecasts or projections or something else) on the combination of sound models, good data and knowledge of changing conditions is wise. It’s also something climatologist already know.
Green goes on to talk about various principles of forecasting and says:
The principles are readily available in the Principles of Forecasting handbook.
Which is, specifically Armstrong’s Principles of Forecasting handbook.
And this is where the chapter gets progressively more odd. Firstly the authors can’t find in the IPCC any reference to (Green & Armstrong’s) validation of forecasting processes (ignoring the actual processes followed within climatology as a discipline). They then sent emails to authors of sections in the IPCC asking them about validation. I’m guessing they go the equivalent of either blank looks or responses akin to ‘do your own homework’.
Green and Armstrong then ‘audited’ the IPCC’s “forecasting procedures” using “Forecasting Audit Software available on ForPrin.com”. Specifically that is the Green & Armstrong Forecasting Audit Software that is available on Green & Armstrong’s ForPrin.com website.
It’s around this point that the weird tone and approach of the chapter suddenly makes sense. This is an advertorial. It is literally marketing. The customer base for “ForPrin.com” is a business audience, and some proportion of the readership of this right-leaning book of disturbed wonkery are a good fit for Green & Armstrong’s market. And good for them! I’ve no objection to self-promotion! It’s a clever use of what is otherwise a giant waste of effort.
“We analysed the IPCC’s forecasting procedures to assess whether they followed the Golden Rule of Forecasting. The Golden Rule of Forecasting requires that forecasters be conservative.”
That’s Kesten C Green’s Golden Rule of Forecasting of course.
“We found that the IPCC procedures violated all nineteen of the Golden Rule guidelines that are relevant to long-term climate forecasting.”
So they set off to do a better job. If you are thinking ‘this is going to be a trainwreck’ then you’ve been reading this blog to long. Go and do something useful with your life 😉
Thye produce this graph:
What is this? You may ask. It is the train wreck you anticipated earlier.
What they’ve done is take the HADCRUT3 global mean anomaly data set from 1850 to 1975 and then tested three “forecasts” that somebody in 1850 could make about future temperature changes.
- Persistence: basically change nothing. 1850 average temperature stays the same.
- Cooling: a steady 0.01 degree C cooling per year.
- Warming: a steady 0.03 degree C warming per year.
These are on the basis of predictions somebody may have made in 1975. Now they then find the absolute difference between each scenario and then plot it.
This is an odd way of finding trends in data and there are better ways of treating time series data, and they are assuming a linear relationship etc. More weird is why 0.03 degrees C per year warming (a projection based on increasing global warming) starting in 1850 when they know in advance that the warming is less than that for the period in question.
They also stop in 1975 before a major increase in warming – for vague reasons.
This what the HADCRUT3 global mean data looks like http://woodfortrees.org/plot/hadcrut3gl/from:1850/to:1975/plot/hadcrut3gl/from:1975/to:2016
Green & Armstrong’s graph covers the red bit and ignores the steeper green bit and tests a projection of warming for increased levels of CO2 that come after the green bit. Quite what golden rule of forecasting this is meant to be is anybody’s guess but it looks a lo like “just make up shit”.
I thought, what the heck, I’ll try and draw their graph and see what they are trying to do. I think this is it:
Yellow is their 0.03 warming, orange is there 0.01 warming and blue is persistence. As they are insistent that good forecasting involves testing multiple scenarios I added in the grey line: 0.01 warming – much closer to what we know occurred. And blow me down! The obvious warming scenario they left off (in violation of their own principles of forecasting) does quite well. Actually I could tinker around and get a warming scenario that does even better.
Conclusion: somebody who predicted mild warming in 1850 would have made a BETTER forecast than Green & Armstrong’s no change rule.
They also say that good forecasts use recent data. So let’s throw in 1975 onwards:
The grey line is actually 0.013 degrees warming because I was tinkering. Notably, persistence now is diverging more and the 0.013 degrees warming is looking better. This isn’t a surprise because we know what actually happened in the 20th century.
But let’s skip forward. After all, nobody in 1850 was making these predictions. What about 1950? That is a more interesting time because global warming from anthropogenic greenhouse gases was being taken more seriously as a hypothesis. There wasn’t a consensus of opinion on it at that point though.
This is less data of course but it is also more recent data.
This time, the grey warming scenario is 0.015 degrees -and doesn’t it do well! Persistence is only a tad better than 0.03 warming as a forecast and cooling is the worst.
The Persistence scenario by Armstrong & Green’s tests would have been a not good prediction in 1950 – particularly compared to warming.
We found that there are no scientific forecasts that support the hypothesis that manmade global warming will occur.
Which is odd, because using their methods I found plenty.
Instead, the best forecasts of temperatures on Earth for the twenty-first century and beyond are derived from the hypothesis of persistence.
Which isn’t even true if you intentionally cut out the period of greatest warming from the twentieth century.
And that’s chapter 12 and the end of the politics section!