﻿

Climate Modeling Pitfalls

INTRODUCTION

Alarmists whose sole evidence for catastrophic human caused climate-change are computer models suffer from a number of delusions.   Not the least of these is the claim that more and more computer simulations, each of which always fails to predict the future, when taken together will somehow be more accurate.

In science, but apparently not in politics, models which cannot predict real world results are generally considered wrong.  And more calculations will not fix the inherent biases and inaccuracies in the model that caused it to fail in the first place, but can only magnify these errors in mostly unpredictable ways.   Errors whose cause and exact nature is not known will not tend to cancel out but will rather accumulate in a seemingly random manner.

CHAOTIC SYSTEMS

One problem is that the short term weather is chaotic.  This means we cannot predict the weather beyond about 14 days.  The problem is not that we know too little but rather that we know too much.  Much like we have a mathematical proof that Fermat’s “Little Theorem” has no solutions for exponents greater than two, we have a similar mathematical proof that any simulation of the weather will be grossly inaccurate after two weeks.   The difficulty is that errors accumulate exponentially.  So if we improve our knowledge of the starting conditions by a factor of ten, we will only extend the accuracy of our predictions by perhaps 10%.  And so we will asymptotically approach a time in the future beyond which no accurate prediction is possible.

This means that while the atmosphere is governed by the deterministic Navier-Stokes equation, the weather is nevertheless unpredictable into the indefinite future.  In practice this means we will never be able to predict if it will be sunny or cloudy, or if it will rain or not, or whether it will be hotter or colder more than about two weeks from now.

PARAMETERIZATION

There is a difference between models which simulate real life behavior and curve fits called parameterization.

Another problem is that a calculation of the weather which depends only on an evolution from first principles is unstable.  And so alarmists substitute deterministic equations for smoothing curve fits which TOTALLY eliminates any dependence on the physics or physical behavior.  And of course, a curve fit is theoretically guaranteed to be an incompetent and inaccurate and stupid approximation of the real world if any of the physical inputs change.  But these curve fit parameterizations by the hundreds to thousands to tens of thousands are what climate alarmists depend upon.

ARITHMETIC AVERAGING OF EXPONENTIALS

Another problem is that the weather is strongly non-linear with exponentially increasing effects.  Simple arithmetic averages are simply wrong.

CONCLUSIONS

You just cannot make this stuff up.