I had an interesting exchange of views on Twitter last night with the Guardian's Leo Hickman and the Met Office's Doug McNeall. I was pressing them on the issue of climate models and their reliability and more specifically the fact that their ability to predict temperature is unvalidated.
Leo rather ducked the issue, noting that he wasn't qualified to comment, but I think that Doug and I came to a meeting of minds. Doug draws some comfort from the fact that other aspects of the global climate can be validated - we mentioned tropospheric fingerprints and I'm assuming ENSO would be another one. We didn't go into the question of just how much of a model/data match there is in these areas and I realise the fingerprints are hotly disputed. Let's therefore park these questions for the minute.
What was more interesting was that Doug and I agreed that unknown unknowns are a big problem for climate models. I think this recognition does take the debate forward so I'm gratified for Doug's input. Now I think we have to ask ourselves whether this ignorance has been properly relayed to the public and to policymakers.
On a related question, Doug and Leo asked what it would take to convince me of the validity of the models. Scientifically, this is an easy question. You need a lot of out-of-sample data. This will clearly take time, but it's the only way to do it. Pretending otherwise is simply to fool yourself. That said, I think if we had had some sort of a match between data and models since the IPCC projections in 2001, I would probably have moved a long way towards Leo's "convinced" position. When you think about it, it's remarkable that this failure of the models doesn't appear to have dented Leo's confidence in them at all.
Bishop Hill
Leo tweets:
I didn't say I was confident in models or not. I was talking about about basic principle tht rising CO2 causes warmng and that we'd b foolish to ignore that risk.
Interesting stuff.