Guest Post by Willis Eschenbach
Well, the BBC, which as I understand it is an acronym for “Blindly Broadcasting Cra- ziness”, gives us its now-standard tabloid style headline, that
Climate change is ‘killing penguin chicks’ say researchers
Of course they’ve included the obligatory “awwwww-inspiring” picture, viz:
Naturally, the researchers didn’t say what the Beeb claimed. What they said was in their paper, Climate Change Increases Reproductive Failure in Magellanic Penguins, viz:
We tested whether chick age, amount of rain, or low temperature affected a chick’s probability of dying during a storm using our 28 years of data with multiple logistic regressions.
Mmmm … testing to see whether more young chicks die in extremely cold, rainy weather … seems to me that even city kids would know the answer to that one.
In any case, how does this blinding insight into penguin mortality tie into climate? Glad you asked. It has to do with their model … or rather their models.
Figure 1. A list of the combinations of three predictor variables used in their twenty-one different models. These are used to model the odds of a penguin chick dying in a storm. The three predictor variables are age (a), amount of storm rain (r), and low minimum temperatures (l). Sadly, they did not archive their data … so this is just pretty pictures at present. Click the image to embiggen.
Their logic and observations go like this. They’ve noticed that the period during which the penguins lay their eggs has gotten longer over the last 30 years. Their hypothesis is that this will make them more vulnerable to the storms. Only thing is, how to prove it?
Why, make up a bunch of computer models of chick mortality, of course. Why not? Or as they say:
We simulated the effects of breeding synchrony on chick mortality in storms. We simulated the proportion of chicks likely to die in a storm on a given day by the hatching spread: for 13 days (the mean for 1983–1986) and 27 days (the predicted value for the early 2080s, based on an increase of 0.15 days per year; see results).
I do love the “extend a trend to infinity” logic of saying that by 2080 (or to be exact, the “early 2080s”) the Magellanic penguins will have a 27 day spread in their egg-laying dates … and using that same logic, we can be sure that by the year 2500 they will be breeding randomly throughout the year … but I digress …
So they simulated the chick deaths from storms, and then to connect that to climate change, they say:
Climate models predict that the frequency and intensity of storms will continue to increase.
Hey, that settles it for me. Since the data says there’s been a change in the length of their laying season, and since models say that the storms will kill more chicks if their laying season gets longer, and since they’ve included one sentence to establish that climate models predict more storms in the future, heck, their work is done.
It’s a beautiful chain of imaginary causation, the scientific version of the bumper sticker that says, “God said it – I believe it – That settles it!”, with “Models” in place of the Deity.
I have to say, this all seems to me like a huge waste of good data. These fine folks have done a solid, workmanlike job of collecting a very large mass of data over 28 years … but then they simply waterboarded the data until it confessed. One example of this is their choice of models.
First, while it is legit to try 21 models, at the end of that process the model you find should be pretty amazing, or else you’re just flipping coins until you get seven heads in a row and declaring victory … especially when you just keep adding parameters.
Next, they make a laudable effort to only use real-world variables in their models. For example they say:
We included all 2-way interactions except age × age squared because we did not want to include a cubic fit for age which is unlikely to have biological meaning.
I like that point of view, that the predictor variables should be real-world variables with physical or biological meaning, and age, rain, and low temperatures certainly fit the bill. Now that seems legit until you get to some of the combinations they use. For example, the model that they finally chose has the predictor variables of the following form.
A + A2 + R + A*R + A2*R + A2*L + A2*R*L
where “A” is age, “R” is rain, and “L” is low temperatures.
And that all looks logical … until we factor and simplify it, and we get
R + A (R + 1)+ A2 (L + 1) (R + 1)
So in fact, rather than the 7 variables they say they are using, in fact they are only using 5 variables:
A, R, A2, (R + 1), and (L + 1)
Unfortunately two of these variables that they are using, “rain plus one” and “low temperatures plus one”, have no conceivable physical meaning.
And that, in turn, means that their best model is actually nothing more than curve fitting using unreal, imaginary parameters without biological or physical meaning.
It is for this reason, among others, that I’m very cautious when I make models, and in general I don’t like combination additive-multiplicative models of the type they use. Yes, I’m sure that people can make an argument for using them … I’m just saying that such models make me nervous, particularly when they end up with eight or ten parameters as in their models.
Here’s the strange part for me. Since they have good data on the length of the egg laying season, and good data on storms and chick deaths, why not just use the data to actually calculate the relationship between storm-related chick deaths and the length of the egg laying season? Perhaps I missed it, but I couldn’t find that calculation in all of their work. Instead, they make a complex model of the situation for which they already have data …
I see this as another tragic casualty of the ongoing climate hysteria. But I suppose I’m just being idealistic, and I’m overlooking the fact that in this current insane situation, it’s much easier to get funding if you say “Hey, I’m not just studying a bunch of birds that are too dumb to remember how to fly, I’m doing vital work on the climate crisis! Think of the grandchildren!” …
Finally, despite their whizbang model, I strongly doubt the researchers’ conclusion that the change in the length of the breeding season will lead to more chick deaths. Natural species survive in part because their methods of living and eating and giving birth are flexible, and they are able to change them in response to changing circumstances in such a way as to increase their odds of survival. The idea that the penguins are changing their breeding habits in the direction of communal suicide seems like … well, like an unusual claim that would require supporting evidence that is much more solid than a computer model with imaginary parameters to make me believe it.
Ah, well … onwards, ever onwards …
N.B: If you disagree with me, please quote EXACTLY what it was that I said that you disagree with. A claim that I don’t know what I’m doing, or that I’m just wrong, or that I should go back to school, any of that kind of vague handwaving goes nowhere because I don’t have a clue what has you (perhaps correctly) upset … you could be right and no one will ever know it. So quote what you object to, that way we can all understand what you are referring to.