Sunday, 30 October 2011

Wordplay - design

Samuel Johnson's dictionary 3rd edition published in 1766 defines 'design' as:-

To Design, v.

1. To purpose; to intend anything.
2. To form an order with a particular purpose.
3. To devote intentionally.
4. To plan; to project.
5. To mark out.

Computers have been used as design tools since the early days of discrete transistors. Rolls-Royce used computers to design turbine blades before they actually thought of using them to do payroll calculations.

Computer-based climate models are really imaginary climate designs which have yet to be tested systematically in the real world. This is not so much a scientific issue as a question of language, but climate scientists tend to dive into the science before the terminology has been sorted. Yet the language is important and here, climate scientists are not the experts.

So referring to climate models as models is misleading. Climate designs is a better way to describe imaginary computer-based climates which haven't yet made it into the real world. Climate scientists may well be designing climate models and that's fine, but we shouldn't omit the word design - it's crucially important.

After all, these climate designs haven't even passed even the most rudimentary testing stage and climate scientists certainly cannot skip physical testing by comparing one design with another as if the whole process is akin to Strictly Come Dancing. Yet it's what they are trying to do with their so-called ensembles, another furtive misuse of language.

Yet it's all really obvious. If an imaginary computer-based climate design makes a 30 year prediction, then that is it's main design parameter, the physical test it must pass before it can be taken seriously as a climate model. In 30 years time it may pass the test and become a tested design or it will fail and become a failed design. Until then it is merely a design.

At the moment, computer-based climate designs are rather less real than Shrek, who at least seems to understands his own language.


rogerh said...

Making a model is tricky, at the beginning it merely fits-where-it-touches and no big decisions can be made using it - overtly. The only test of a model is a good experiment.

As a lab rat long ago I suffered a few runaway experiments and was lucky. Let us hope humanity is lucky too, there being no senior lab rat around to say "er, I
wouldn't do that if I were you...". Finally, much excellent data is fabricated on a Friday afternoon.

A K Haart said...

rogerh - I was a lab rat too and I certainly remember Friday afternoon sessions "tidying up".

I also remember how problematic environmental data really is - uncertainty of measurement is just the start of the fun and games.

James Higham said...

Or to put it another way - they are hypotheses. Nothing wrong with a hypothesis, as long as it's treated as such and not as some god-given pronouncement from on high.

A K Haart said...

JH - that's another word they avoid by using the word 'model'.