AlaninWales wrote:A prediction that shows (say) 30% probability of x; 70% probability of y may be completely accurate.
But you can only know if it is accurate if, over a long period, the things you say will turn up 30% of the time do indeed turn up 30% of the time.
To say a forecast is 10% better than climatology means that in comparing the specific forecast and the alternative "climatological" forecast that average conditions will prevail, the specific forecast is on average better 55% of the time (ie 10% more than the 45% that "climatology" was better). Clearly this only makes sense in relation to a forecast for an extended period such as a month. To tell a forecaster his forecast is only as good as climatology is to say that it is in effect worthless.
Forecasts like 30% of x, 70% of y, if accurate, are very useful for insurance companies, traders and the like, who can deal with short term uncertainty and long term averages, it's their job. Normal people, as you say, like certainty, which can rarely be given. Days like yesterday, where there unforecast heavy downpours in the SE - at least not forecast the previous evening - have become rarer.
I think we were a bit lucky last winter. The blocking formations that have given us the cold snaps in many recent winters all happened over the America and Asia, not Europe, and they got all the cold. The state of the North Atlantic Oscillation,* which tends to get stuck for periods of a decade or two, and flipped a few years back, means blocking formations are more likely in winter in the north. We are stuck with this risk of colder winters for some time to come. Then it will flip back and there will be few cold winters for a decade or so.
*Even this is a bit statistically troublesome. The NAO is often described as varying over the course of a couple of weeks or so. But it is its annual average that seems to flip between a high state and a low state with a frequency that is around decadal or so. Even that is hard to describe, sometimes it has stayed one side for 30 years or so. And who are we to say that what happened in the past will continue in the future. What use are a mere hundred years of good weather records (though we have other methods of looking further back, though those tend to be disputed) when looking at things that can persist for 30 years?
