Letter: The Risk of Modelling – It’s the Gusts, Not the Average

Published on: 7 Jul, 2017
Updated on: 7 Jul, 2017

From Harry Eve

We are used to being wary of propaganda, fake news and the abuse of statistics. In the computer age we should also be wary of the use of models. We can trust our weather forecasters to do their best, using models, to produce a reliable forecast.

They will sometimes state, openly, that they have compared the predictions of more than one model and found that they differ – so there is no great certainty whether you will need sun cream or an umbrella.

The indications from a modelling exercise depend on the quality of data fed into it and the assumptions chosen, as well as the refinement of its functionality. The reliability of the indications also depends on sensitivity to the assumptions, and reduces as a model looks further into the future.

Examples of assumptions in modelling include future mortality rates for population projection and the capacity of a road or junction for traffic modelling. Where there is uncertainty, a range of assumptions can be tested to see what effect they have on the modelled result. The end result can be highly misleading if all or most of the assumptions are slanted towards a particular outcome.

Where modelling is used as part of the evidence to support an agenda that may involve politics or profit we should do our best to give it close scrutiny and alarm bells should start ringing if the information needed for scrutiny is withheld.

We should not accept, blindly, the figures derived from modelling just because they are the headlines in complex, and lengthy, technical documents that state that they follow industry standards and are produced by experts.

It is all too easy to “model away” a problem – wittingly or unwittingly. A simple example is the use of averaging techniques. Average wind speeds may be of some interest but it is the strength of the gusts that does the damage.