Chapter 2 Why climate models are failing by Patrick Michaels
Patrick J Michaels is a key figure in the small community of scientists who doubt that global warming is happening. Attached to various industry groups and right-wing think-tanks he has been a regular source of criticism of current climate science.
He starts the chapter by making himself a hostage to later events by characterising political woes of some politicians in Australia and the US as a consequence of their stance on climate change. He mentions Malcolm Turnbull being deposed as leader of the Australian Liberal Party and just two years later the man is Prime Minister. He also points out:
“Environmental policies are also critical issues in state elections, where only one state premier remains from the Labor Party.”
But by early 2015 Labor had won elections in both Victoria and Queensland. He couldn’t have foreseen that but it just goes to show the dangers of interpreting complex shifts in political fortunes with one issue.
He then goes onto to mention the US 2010 House of Reps election and then onto Canada:
“In Canada, a newly-elected and popular government has basically shut down any significant global warming policies.”
Ouch. Reading this celebratory message of the right from the depths of 2014 almost makes me quite like 2016. Stuff changes and Pat Michaels would have been wise to not set to0 much store on the current state of affairs.
With that misstep out of the way, Michaels starts on the essay proper and like Plimer wants to tell us about how science works.
“In its most basic form, science consists of statements of hypotheses that are retained by critical tests against observations. Without such testing, or without a testable hypothesis, Karl Popper stated that what may be called ‘science’ is, in fact, ‘pseudo-science.’”
It looks like Michaels is going to take a Popperian tack and given the title this could be interesting. Sure enough, he initially veers towards discussing the notion that all sorts of weather conditions have been blamed on global warming. It is the start of an argument where modern climatology could be portrayed as a pseudoscience as (according to its critics) it predicts anything and everything and hence nothing. But Michaels, perhaps anticipating that such an argument would misunderstand both Popper and climate science doesn’t actually go very far down this path.
Instead, the bit about Popper is part of a waffle sandwich. There is a second waffle later, equally weakly connected where Michaels discusses Thomas Khun’s concept of paradigm shift. The actual meat/jam/peanut-butter in the sandwich is a study of some climate models.
The study and resulting graph are a bit odd and it is more than possible I’ve misunderstood but I’ll explain as best I can.
Michaels says that:
“We examined data since 1950 for the 108 model runs used in the Working Group I (Science) 2013 IPCC Fifth Scientific Assessment available from KNMI Climate Explorer (climexp.KNMI.nl). We calculated the model trends for periods beginning at ten years (i.e. 2004-2013), eleven years (2003-2013), etc., all the way back to 1951-2013.”
So he has a whole bunch of output from climate models. That makes sense.
From this bunch of climate model data, he calculates trends of progressively longer length. But why? Never mind, let’s just go with that.
“For each trend length, we ranked the 108 trend values from the individual model runs. From this ranked data set, we determined percentiles.”
OK not sure of his method but essentially he is trying to define a broad range of results. By their nature the models used can give a variety of results, so this is an attempt to capture not just an average result but a broad region that we will later see in a graph. Seems like an odd way of doing it, though.
Then some other stuff etc etc and then:
“We then compared these trends to the observed trends in the HadCRUT4 temperature history.”
HadCRUT being one of several global temperature sets used. It is an interesting choice in some ways because of its association with the University of East Anglia and the so-called ‘Climategate’ scandal. Ironically, it is often the preferred data set of many on Michaels’s side of the argument because it sometimes shows slightly cooler trends.
The punchline is this graph.
It is a very odd bit of graphing. The dots to the left of the graph towards the temperature axis represent a summary of more data than the dots to the right. So going from left to right as one would normally do, you aren’t going forward in time but rather seeing trends calculated over shorter and shorter periods. So for both the actual temperature data and the model data, you get a sort of trumpet shape. Trends for shorter periods are noisier and less relevant to the issue of global warming.
The graph shows increasing departure from the model data versus the temperature data for trends calculated with less data. Which, well, maybe I’m missing something but is that not bleedin’ obvious? Weirdly the very last dot which represents 2004-2013 actual wobbles back into the lower range of the models but that result is just as pointless.
The whole thing shows that the models tend to agree with actual observation more who looked at over long periods and less when looked at over short periods. This is almost apriori true and Michaels’s technique is a very round about way of showing this.
But this is not some piece of statistical illiteracy. This graph has been very carefully designed. The vertical axis says “Temperature” when actually it is a temperature rate (a change in temperature over time i.e. it needs a time unit as well). The horizontal axis is labelled “Trend length” but it is meant to suggest a movement from the past to the future, whereas it actually runs from long to short. This is the reverse of the normal convention* on a graph of smaller values at the left with larger values to the right but it fits well if the visual impression wanted is one of moving from the past to the future. Combining the two and Michaels gets to make a graph where a value (incorrectly) labelled “temperature” drops precipitously towards the right of the graph, while falling outside of something that look sort of like confidence intervals or measures of error but aren’t.
Let’s fix it up a bit:
Arguably it does support the claim that the climate models could be better but there are easier ways of showing this. Intrinsically it shows that the climate models used show better agreement with longer term measures of temperature but this is an unremarkable finding.
In the results section Michaels says:
“If policies were based upon climate science rather than climate studies, this simple, straightforward analysis would spell the end of any onerous climate policy. However, while our similar studies can be scientifically cited, to date, there has been an understandable reluctance to publish this in the tier-1 scientific literature, such as Nature or Science, as that would indicate a massive, unexplainable, and persistent failure of the studies driving global climate policy.”
Well, the reluctance for any tier publication to want to publish that graph certainly is understandable because it is at best pointless and at worse actively deceptive.
Now it is time for Michaels to quote Thomas Khun:
“Kuhn attempted to explain the reluctance of a scientific community to abandon a failing paradigm as a function of the profession, in which the vast majority of practitioners advance professionally by trying to explain minutiae or anomalies within the paradigm.”
The irony is palpable, as Michaels tries to maintain the belief that global warming just somehow isn’t happening, he clinging to an old paradigm and ignoring the shift in understanding that occurred in the second half of the twentieth century.
Overall, this is an unsatisfying chapter. Climate models really do have issues and really raise important questions about what constitutes evidence and how we should conduct science in an age of increasing complexity of understanding.
*[this is only a convention and there can be good reasons for not following it but it is an important to be clear when a graph isn’t following that convention because of people’s expectations.]