NZ CLIMATE TRUTH NO 35
LIES, DAMN LIES AND... CORRELATION
It is a fundamental principle of formal
logic that a correlation, however convincing, is not, of itself,
evidence of a cause and effect relationship.
It is surprising how frequently this
principle is ignored. We are for ever being told that there is a "link"
between one factor and another, followed by an assumption that this proves
a relationship even in the absence of other evidence.
Of course, a correlation is a stimulus
to further investigation, which may confirm or deny the value of the correlation.
My statistics textbook ("Modern
Elementary Statistics" by John A Freund, 1967) Gives two examples.
It was found that there was a correlation
between the number of storks in Holland and the number of babies delivered.
Does this prove that the storks deliver the babies? Further investigation
would show that there are no observations supporting the hypothesis.
The other example given was the correlation
between the salaries of schoolteachers and the consumption of alcohol. Does
this prove that schoolteachers are the main drinkers? Further study would
show that alcohol consumption correlated with general salary increases.
The procedure for checking the reliability
of a correlation is called Validation.
Validation might include some or all
of the following
-
Check whether there are other, more
plausible correlations
-
Develop a plausible theory which
might explain the correlation
-
Carry out an experimental programme
to try and prove the correlation
-
See whether the correlation fits
a variety of circumstances
-
See whether the correlation works
for a wide range of variables
-
See whether the correlation can predict
the future.
The Intergovernmental Panel on Climate
Change has featured the use of compute-generated climate models and has
argued that since they correlate, with climate observations, to a greater
or lesser degree, they are a correct representation of climate behaviour,
and are an effective way to predict future climate.
They have, however, completely failed
to validate any model.
The first IPCC Report (1990) had a Chapter
entitled "Validation of Climate Models"
The "Executive Summary" begins with
"The validation of the present day climate simulated by atmospheric general
circulation models shows that there is considerable skill in the portrayal
of the large scale distribution of the pressure, temperature wind and precipitation"
In other words, there is a correlation.
If you take a look, it doesn't seem so
convincing, and in the "Policymakers Summary" they admit, for temperature,.that
"The size of this warming is broadly
consistent with predictions of climate models". But since the models show
a steady increase of temperature over time and the measurements do not,
this conclusion is dubious.
The Chapter on "Validation of Models"
nowhere defines what "validation" implies, and it carries out none of the
"validation" exercises suggested above. The conclusions of the Report are
based entirely on correlation, without any supporting information that could
legitimately be called "validation"
The Second IPCC Report (1995) circulated
its first draft for comment to the expert reviewers, including myself. It
included Chapter 5 entitled "Validation of Climate Models". I made
the comment that there was no discussion in the Chapter on what should be
done to validate the models, and no actual measures taken to carry out such
an exercise. I therefore considered that the term "validation" was not appropriate,
To my surprise, they accepted this criticism.
Not only did they change the title of the Chapter to "Climate Models
- Evaluation", they altered the words "validation" and "validate"
to "evaluation" and "evaluate" no less than fifty times, throughout the
Chapter. Since then, "validation" has been a dirty word, never
to be mentioned. This applies to the whole of "Climate Change 2001". although.
It slipped through once or twice.
Chapter 5 of "Climate Change 2001", "Climate
Models- Evaluation" starts with a heading "What is Model Evaluation
and Why is it Important?. They state
"we are trying to demonstrate the degree
of correspondence between models and the real world they represent. We will
therefore avoid the use of the word "validation", although this term is
commonly used in the sense of evaluation in the climate modelling community"
In the "Conclusion" they say
"Confidence in climate models depends
partly upon their ability to simulate the current climate and recent climate
changes, and partly upon the realistic representation of the physical processes
that are important in the climate ate system"
This makes it plain that they are not
interested in finding out whether the models can predict climate changes
as they occur. They are concerned exclusively with correlation, and
do not intend to validate any model.
The Chapter contains no discussion
of what measures need to be taken to provide a climate model that is sufficiently
reliable to be used to forecast future climate. I have consistently argued
that until a procedure has been agreed and applied the models should not
be used for this purpose.
All there is, is correlation. Since the
parameters in climate models are uncertain, and sometimes extremely uncertain,
it is a relatively simple matter to adjust them so that the model fits almost
any sequence of climate observations. We used to call this procedure "fudging",
and it usually called for low marks. Even with the most desperate fudging
the models cannot explain why the Northern Hemisphere warms up faster than
the Southern Hemisphere, or why the poles do not heat more rapidly than
the equator.
A crucial validation test of climate
models would be a proven capability of predicting future climate. All
users of computer generated models except those connected with the IPCC
automatically check whether their models are in line with what is actually
happening, and adjust the model accordingly.The IPCC refuse even to consider
this.
Computer models of the climate have
now been going for a number of years. Comparisons between the predictions
of models and actual climate observations, such as those published by myself
(see my book "The Greenhouse Delusion") invariably show that the models
exaggerate. But IPCC scientists are just not interested in correcting their models
to agree with actual climate trends.
The IPCC is about to launch its
fourth Report. As before, they have a Chapter on " Climate Models and their
Evaluation" which includes no paragraph dealing with a genuine validation,
and no mention of possible success in prediction of climate trends.
Vincent Gray
75 Silverstream Road
Crofton Downs
Wellington 6004
New Zealand
Phone/Fax (064) 4 9735939
Email vinmary.gray@paradise.net.nz
"It's not the things you don't know that fool you.
It's the things you do know that ain't so"
Josh Billings