B Questions about climate and physics

[DrClaude]

Thread closed pending moderation.

 

[bhobba]

The second problem is how much tuning should be done on the data. If you had a model good enough to predict 5 hurricanes ina time window and you saw 4 or 6 would this be evidence in favor or against? A supercomputer is not going to tell you the answer to that.

His point is it is inherently probabilistic, as the climate is chaotic. A better supercomputer will give us better probabilities. He knows this from his work on long-term climate change and medium-term climate prediction. When a hurricane is forming, it is impossible to predict what it will do - all you can do is give probabilities of what will happen. Better computers will give better predictions. It is of great practical value to those tasked with preparing for emergencies who are very interested in the probability of if, for example, it will hit Brisbane rather than miss Australia completely.

Just a note about how he does it. The models all require parameters. What he does is change the parameters slightly and run it again. The model must be run many times to get good probabilities - the more times, the better. Faster computers mean better models that can be run more times and get better probabilities. I think further discussion should be in a thread on the computational modelling of chaotic systems.

I don't think there is anything more to discuss, so I have permanently shut the thread.

Thanks
Bill