Monday, December 14, 2009

Climate Models

Peter Newnam and I have had a long-running discussion about how models are used in Climatology as compared to in engineering, physics, and astronomy. We concluded that there is are several significant differences, and we have done our best to describe these differences in Climate Models, a new section of our analysis page. It's more of a philosophical piece than a scientific one, but we hope it will at least serve to stimulate debate upon the subject of modeling in science.

UPDATE: Clive James talks about making too many predictions.


  1. You are saying is that we cannot have much confidence in the results of the climate models and it seems to me that you are probably right. But I would be interested to see your discussion broken down to address each of the different types of models that compose the climate model.

    Let's take the hydrodynamic models for example. They are based on the Navier-Stokes equation which has solid physical foundation and plenty of experimental evidence to support it. These models have well known inputs such as bathmetry, rotational speed of the Earth, etc. They are usually solved using a finite element method, but I wonder if the grid cell size is too large to be able to capture smaller scale features which could have a cumulative impact on oceanic circulation (not to mention heat transfer).

    There might be other models which are more of a "black box" type model which are used to simulate biological or chemical processes. Presumably their coefficients have been investigated through controlled experiments. To have confidence in these models, you usually have to a good validation process which seems like it would be difficult with a climate model.

    With all this said, being a modeler, I like the climate models. I think there are a lot of very smart people out there doing their best to understand our world's climate. And by studying the subject we learn a lot. But, while I think this is a worthy endeavor, the forecasts from the models should not be used to set policy. Or make dire predictions for the future. I suspect that the way the predictions are used is one of the things that concerns you.

  2. I agree that climate models would be useful in helping us think up new experiments to perform in the laboratory, and to suggest new observations we might make to improve our understanding. I agree that many of the physical processes to which climatologists refer are well-proven. But its the "CO2 doubling temperature" and things like that which are unique to climatology that have no empirical basis, and are calibrated by the models.

    The CO2 doubling temperature is the amount by which the world is supposed to warm up if the CO2 concentration doubles. There is no empirical basis for this theory. It sounds reasonable because CO2 does absorb heat transmitted by the earth. The climate models assume there is some temperature, T, by which the world warms up if you double CO2 concentration, and the models choose T to suit the climate history.

    As a result of many models, climatologists now state that a value of T around 3°C is "well-established". But there is no empirical basis whatsoever for this conclusion. It's just a hypothesis.