1/07/2015

What Good Are Climate Models When They Get The Math Wrong?

Considering the temperatures outside The Manse are well below zero at the moment, it seems almost contradictory to be writing about global warming. But I'll make an exception in this case.

And “this case” deals with the simplistic nonsense that is the climate modeling used by the warmists to 'prove' we're all doomed unless we take drastic action. (More often than not the actions they want us to take will impoverish everyone...except them, and force us to live under a green dictatorship.)

As an engineer, my first experience of a computer model taught me nearly all I needed to know about models.

Some people suggested I make use of the acoustic coupler to input my design and optimise it with one of the circuit modeling programs they had devised. The results were encouraging, so I built it. The circuit itself was a dismal failure.

Investigation revealed the reason instantly: the model parametrised parasitic capacitance into a simple single value: the reality of semiconductors is that the capacitance varies with applied voltage – an effect made use of in every radio today as the ‘varicap diode’. for small signals this is an acceptable compromise. Over large voltage swings the effect is massively non linear. The model was simply inadequate.

I have used and continue to use modeling software much like that to simulate the performance of circuits we use in our designs. It is always understood the models are an approximation and may not reflect the actual performance of the circuit. That's why we build prototypes to test the design under real world conditions.

While the modeling software we use is very sophisticated, the components in the models are represented by a limited number of parameters, far less than are used in climate models. The electronic parameters are well understood and have been refined over the years, and still the modeling software will get it wrong (though not grossly wrong).

The climate model software should take into account many more parameters, many of which are poorly understood, yet we are supposed to believe they're right. The biggest problem with the models?

They get the math wrong.

Most of engineering is to design things so that small unpredictable effects are swamped by large predictable ones. Any stable design has to work like that. If it doesn’t, it ain’t stable. Or reproducible.

That leads to a direct piece of engineering wisdom: If a system is not dominated by a few major feedback factors, it ain’t stable. And if it has a regions of stability then perturbing it outside those regions will result in gross instability, and the system will be short lived.

So called ‘Climate science’ relies on net positive feedback to create alarmist views – and that positive feedback is nothing to do with CO2 allegedly: on the contrary it is a temperature change amplifier pure and simple.

If such a feedback existed, any driver of temperature, from a minor change in the suns output, to a volcanic eruption must inevitably trigger massive temperature changes. But it simply never has. Or we wouldn’t be here to spout such nonsense.

[Climate is] a massive non linear hugely time delayed negative feedback system. And that’s just water and ice. Before we toss in volcanic action, meteor strikes, continental drift. solar variability, and Milankovitch cycles…

The miracle of AGW is that all this has been simply tossed aside, or considered some kind of constant, or a multiplier of the only driver in town, CO2.

When all you know is linear systems analysis everything looks like a linear system perturbed by an external driver.

When the only driver you have come up with is CO2, everything looks like CO2.

Engineers who have done control system theory are not so arrogant. And can recognise in the irregular sawtooth of ice age temperature record a system that looks remarkably like a nasty multiple (negative) feed back time delayed relaxation oscillator.

It also doesn't help that our climate is a semi-chaotic system, something that is extremely difficult to model. More often than not many of these models average the hell out some of the data inputs which has the (supposedly) inadvertent effect of minimizing some of the feedback mechanisms, particularly the negative feedback mechanisms. This can result overestimating the effects of the positive feedback or the sensitivity of some factors, like CO2. Hence you come up with model results that do not match the real world data. What should that tell us?

The models are seriously defective and should not, under any circumstances, be used to predict the future climate. This also implies they should also not be used to create climate policy because the predictions are crap. In other words, they got the math wrong. If my math was as wrong as theirs, I would be fired. Too bad they haven't been.