Why the Climate Models will never work
Mike Jonas
The climate models are among the most sophisticated computer models ever developed. Billions of dollars and countless man-hours have been spent on their development. The IPCC references about 70 computer models in its regular climate reports. So the idea that the models will never work may sound absurd. But it’s true, and this article shows why, and to do that it bypasses all the complexity and just goes to the heart of the matter – which is surprisingly simple.
I should point out that I am not the first person to say that the climate models will not work. Many people have done that, for example Robert L Bradley Jr wrote climate models can never work, published in AEIR (American Institute for Economic Research), but somehow the gatekeepers manage to prevent the message from getting through to the general public. The problem is that the proponents of the climate models can use climate’s complexities to obfuscate – for ever.
Cambridge Dictionary: Obfuscate – to make something less clear and harder to understand, especially intentionally.
So, how can I simplify the picture? Let’s start with the statement by the IPCC way back in 2001: “The climate system is a coupled non-linear chaotic system, and therefore the long-term prediction of future climate states is not possible.“. If we could look at another coupled non-linear chaotic system, and understand how the structure used in the climate models could never work for that system, then maybe we could understand why the IPCC made that statement, and why the current set of climate models can never work.
Let’s look at road traffic.
A road traffic controller can tell pretty well how the next few minutes of traffic will go at a particular set of traffic lights or roundabout by seeing what traffic is on the approach roads. But to predict even the next few hours, they must know much more – will people be leaving work, is there a football match starting. Even to look a few hours ahead, knowledge of how many cars are currently on the roundabout a block away is already useless. Larger factors are at work.
What about the next few days – are roadworks scheduled, are school holidays starting. You can see that a model that starts with the amount of traffic on the road at this minute, and then works forward minute by minute calculating how the traffic changes, is a complete waste of time. Sure, you can feed in data about when people will leave work, when football matches are scheduled, when roadworks are scheduled, when the school holidays start, and your model may give some respectable answers. But the reality then is that all the minute-by-minute calculations are meaningless, what really matters is whether you get the bigger picture items right. And if you get the bigger picture items right, you can give a reasonable overall forecast of next month’s or next year’s traffic without going through all the intervening days minute by minute. In other words, in order to forecast next year’s traffic at Thanksgiving you won’t need to work out along the way how many cars will be on each road at 11:30am next Paddy’s Day.
Well, that’s what climate modellers need to understand. They march their models through time, typically twenty minutes at a time, using a single number for the climate equivalent of the speed of all traffic at a point in time between the set of traffic lights in the middle of town and the roundabout several blocks away on the edge of town, and think that they can produce an accurate forecast for the next few decades for the whole planet. (There’s a description here). Even if they had 10,000 numbers instead of a single number, it wouldn’t help. The reality is that they need to know what the sun will be doing over the next few decades, and the clouds, and the Pacific ocean, and so on, and then they will be able to give a reasonable climate forecast without a climate model, ie, without calculating what the temperature will be in the mid troposphere over a particular patch of the Atlantic next Wednesday at 3:20pm.
If they don’t know what the sun will be doing a few decades from now, and the clouds, and the Pacific ocean, and so on, then they can’t make a credible forecast, with or without a climate model. And I shouldn’t have to add that if they do know, then they don’t need a 20-minute climate model.
Postscript
I have kept the above article deliberately simple, so that it is easy to follow. It uses analogy, so it has limits, and given the complexity of Earth’s climate there are necessarily some gaps. In this postscript I will try to address some of those limits and fill in some of those gaps.
1. This article is loosely based on Edward Lorenz’s Chaos Theory. But you don’t need to understand chaos theory in order to understand this article. You just need to understand one fact that comes from Chaos Theory: In a chaotic system, a tiny error will relentlessly increase in size until it has completely swamped the predictions. See Explainer: what is Chaos Theory?
2. I used traffic flow as an analogy for climate, because it is easier to understand how to predict traffic – and how not to predict it. Traffic is a legitimate analogy, because both traffic and climate are chaotic systems. From Chaos Theory: “Chaotic behavior exists in many natural systems, including fluid flow, heartbeat irregularities, weather and climate. It also occurs spontaneously in some systems with artificial components, such as road traffic.“. Obviously the actual data and equations used for traffic are completely different to those for climate, but the basic principles of chaotic behaviour still apply. It is pointless for the climate models to look at the climate equivalent of traffic lights and roundabouts, ie. trying to compute climate on a micro scale (see 5. below).
3. One of the features of chaotic systems is that models can predict behaviour reasonably well for a short period, and then they rapidly deviate. From Butterfly effect: “complex systems, such as the weather, [are] difficult to predict past a certain time range (approximately a week in the case of weather) since it is impossible to measure the starting atmospheric conditions completely accurately.”. I would argue that a lot more than just accurate starting conditions are involved, because the predictions for the end of the first day are the starting conditions for the second day. In other words, even if you get your starting conditions absolutely perfectly, you will still be a long way out in the second week. There’s more at What is chaos? A complex systems scientist explains, eg. “A hallmark of chaotic systems is predictability in the short term that breaks down quickly over time, as in river rapids or ecosystems.“.
5. The statements I made about how climate modellers “march their models through time, typically twenty minutes at a time, using a single number for the climate equivalent of the speed of all traffic at a point in time between the set of traffic lights in the middle of town and the roundabout several blocks away on the edge of town” is correct. Climate models really do that. The way they operate was described in Climate Models: “Climate models are a mathematical representation of the climate. In order to be able to do this, the models divide the earth, ocean and atmosphere into a grid. The values of the predicted variables, such as surface pressure, wind, temperature, humidity and rainfall are calculated at each grid point over time, to predict their future values.“.
6. Some modellers claim that they have ‘moved on’ from the difficulties of chaos theory, and that certain factors like viscous effects would “tend to damp out small perturbations” – from Butterfly effect again. Well, until they can prove it by forecasting weather a month ahead, say, that to my mind is just magical thinking. Like the IPCC said: “The climate system is a coupled non-linear chaotic system, and therefore the long-term prediction of future climate states is not possible.“, and the same applies to weather, where long-term is just a week.
7. There is even some argument about whether the climate is deterministic. Climate modellers are reportedly getting more confident that climate is deterministic, but don’t be fooled into thinking that means that their climate models can predict it. Edward Lorenz demonstrated that a deterministic system could be “observationally indistinguishable” from a non-deterministic one in terms of predictability. (That’s in Butterfly effect too). For example, the behaviour of a snooker ball is deterministic, but the outcome of the first stroke in a snooker game can’t be predicted. There’s a nice demo of the unpredictability of a deterministc system at demonstration of the butterfly effect.
Summary
In summary, a climate model that works in tiny time steps on a coarse grid can never work for more than a very short time. It can be useful for helping people to understand climate, but it is useless for predicting future climate. And it doesn’t matter how fine a grid is used, or how sophisticated the partial differential formulae are, or how carefully the boundary conditions and chaotic attractors are managed, etc, etc, it still can’t predict more than a short time ahead. Certainly, it might get lucky sometimes, but that’s not the same as being reliable. And even if all the models get their forecasts right for a short period, there is still no reason to suppose that they will continue to be right – remember, the nature of chaotic systems is that they are predictable only for a short time.
Coming back to the last paragraph of the body of this article, clearly what we need in order to be able to make reasonable predictions of climate is an understanding of the longer term factors like solar activity, ocean oscillations, clouds, storms, etc, and greenhouse gases too, of course. We still don’t know what the sun will do next, and we still don’t know how the sun affects our climate. We still don’t know exactly how all the other factors work and interact. And if we did know all of those things, we wouldn’t put them into a 20-minute model, we would say that in x decades time the sun would be doing this and the oceans and clouds and so on would be doing that, so the climate will be doing such-and-such. Approximately.
Related
