How do weather forecasters improve the accuracy of their predictions?
You can predict the weather months in advance, but sometimes the weather forecasters don't even get it right next week. Do you look back on your predictions and work out how accurate you are and refine your models?
That's a very good questions actually because weather forecasts do go wrong and we've all seen that. We have a scientific basis for understanding that and it's a phenomenon called chaos theory. Things like the tides, for example, are very predictable. We can predict tides years or decades ahead. We can predict eclipses hundreds of years ahead. But these are not chaotic phenomena. The weather is a chaotic phenomena. What this means is that we can never when we go out to these time scales of years and months or seasons ahead, we can never make absolute definite predictions. What we make are what we call probabilistic predictions. What we will do is say that the chance of it being wetter than normal or warmer than normal would be maybe 80% or 90% depending on the confidence we have in the prediction and this will be useful information because obviously without that information all you could say is that there's a 50% chance of it being warmer than normal or colder than normal. So this is how we deal with chaos theory in the weather by going towards a more probabilistic prediction. But nevertheless, in many applications these turn out to be very useful types of forecast.