You create a model which predicts the sun will rise in the west. The sun fails to cooperate and rises in the east. Do you:
- Admit failure and return your remaining unused grant monies,
- Call for the firing of the comedians who point out the discrepancy,
- Claim that your model was right and the observation naught but “natural variations”?
If you said (C), you should consider submitting your ploy to Nature; they’ll likely publish it. Just like they did the peer-reviewed “Forcing, feedback and internal variability in global temperature trends” by Jochem Marotzke and Piers M. Forster.
As summarized by Phys Org in the article “Global warming slowdown: No systematic errors in climate models“:
Sceptics who still doubt anthropogenic climate change have now been stripped of one of their last-ditch arguments: It is true that there has been a warming hiatus and that the surface of the earth has warmed up much less rapidly since the turn of the millennium than all the relevant climate models had predicted. However, the gap between the calculated and measured warming is not due to systematic errors of the models, as the sceptics had suspected, but because there are always random fluctuations in the Earth’s climate. [emphasis added]
I wept when I read that. Real tears. That sorry excuse is no different than you saying the sun’s departure from your prediction was due to “natural variability” and that any skeptics who point our your model is a bust were wrong.
No. No. No. It is as simple and no more difficult than this. A climate model consistently says the temperatures will be way up here, and reality just as consistently fails to cooperate and puts temperatures way down there. That model is therefore a failure. It is busted. It is broken. It is not right. It should not be trusted. It should not be used as a basis for any decision. It is wrong.
The job of a climate model is to predict the climate, no? And if the actual climate does X when the model says Y, the model is wrong. “Natural variability” is what the climate does. It was the model’s function to predict that “natural variability.”
If your model and reality don’t match you cannot claim that your model was right all along and the observation not the real observation. Such as action is preposterous!
The authors perform some dreary regression analysis, the details of which are too depressing to recount, and conclude “The claim that climate models systematically overestimate the response to radiative forcing from increasing greenhouse gas concentrations therefore seems to be unfounded.”
Oh Lord. The claim that climate models systematically overestimate the response to radiative forcing, or are busted in some other way, is entirely well founded. Founded on what? Founded on model failure. Years and years of model failure, too.
How is it that we have come to the point where such a basic scientific principle that models which make lousy predictions are wrong yet are still considered right? Not only that, but the authors are, in the modern parlance, doubling down. They tell the Daily Mail that not only are the models to be trusted, “The long-term trend points to severe warming of the climate”.
How could they not see that this is silly?
I don’t know the answer. I’m too depressed to continue, except to remind us we talked about this twice before.
Look, sisters and brothers, if we (as in climate scientists) knew what the temperature was going to be, we would have been able to skillfully forecast it. We were not able to skillfully forecast it, so we did not know what the temperature was going to be.
To speak of a “hiatus” or “pause” logically implies we knew the “hiatus” or “pause” was going to be there, that it was expected, that we knew in advance its causes. We did not know. If we did know, we would have predicted it. Which we didn’t.
Natural variability, sisters and brothers, is what the models said they could predict skillfully. The models did not skillfully predict natural variability. Natural variability just is, in this sense, what the temperature does.
There is another sense of the phrase, though, a kind of enviro-religious sense that people might be using, which is, “What the temperature would do in absence of humans”. Now that is a valid thing to study. Only trouble is, it’s counterfactual. We can produce answers by the grant-load, but we’ll never know, or that is, we can never verify, whether any of them are true.
Update Reader Gary (see below) suggested writing a letter to Nature. I submitted one. Here it is.
Models Which Make Bad Forecasts Are Wrong
You create a model which predicts the sun will rise in the west. The sun fails to cooperate and rises in the east. Do you admit failure or write a paper to Nature which explains your failure as a success because the sun was exhibiting “natural variability”, a phenomenon of which your model cannot be expected to capture fully?
This scenario is no different than attempts to explain away climate model failures, like that found in the paper “Forcing, feedback and internal variability in global temperature trends” by Jochem Marotzke1, Piers M. Forster, in Nature, 517, 565–570 (29 January 2015). Those authors write “The difference between GMST observations and simulations is caused in part by quasi-random internal climate variability, which arises because of chaotic processes in the climate system…”
This is false. A GMST model is built to predict GMST. Any departure of the model’s predictions from the GMST observations is due to a fault or faults in the model and nothing else. It might be that GMST is chaotic, or again it might not; either way, it is the duty of the model to capture the essence of GMST, whatever that essence is. If that essence cannot be captured, for whatever reason, the model has failed. Why the model failed is interesting to the model creators, but not necessarily to anybody else. Further, identifying the reasons for the failure belongs solely to the model creators. A person showing the model has failed has no responsibility whatsoever to offer reasons for the failure.
A fortiori, if the model has loudly and repeatedly make promises it has not kept, it is irrational to trust the model. Now this used to be a well known scientific principle. Once, it was known, believed, and acted upon that models which produced rotten forecasts were admitted to be wrong, and were fixed or abandoned. Never were the predictands—the things the model forecasted—blamed for error. Never were mistakes waved aside and statements about the model’s intrinsic goodness given out. Why this reversal from common sense to the state we are in now is a question I leave the reader to answer.
The bulk of the climate signal must be natural variability. This is so unless somebody has proof that man drives most of the (entire) climate. Any climate model worthy of respect thus must be able to explain and predict natural variability. If model cannot capture this natural variability, the model is in error. Global climate models promised good predictions of actual GMST. In this they have failed. Therefore they are in error. There is no escaping this logical necessity.
Update An editor at Nature kindly showed me that I clicked the wrong button (or whatever). The letter has been resubmitted as of 1:30 PM, 6 February.
Thanks to Al Perrella and others who pointed out this newest attempt to avoid responsibility.