Friday, March 4, 2011

Uncertainty and skill

I was reading about what the National Hurricane Center will be doing this year. It was quite interesting:
1. Watch lead time to 48 hours; warning lead time to 36 hours
2. The size of the cones will be based on 5 year running mean of 2/3 of track error.  The cone size radius varies in the Atlantic basin from 36 to 59 miles from 12 to 24 hours and grows about 35-40 miles per 24 hours thereafter.*

This means they have skill out to 48 hours in advance. They will be communicating actively their uncertainty via the graphical product (cones).

My foolish expectation would be to shy away from any sort of climatological cone of uncertainty and use ensemble guidance. This may not be the best option since it could provoke the so-called meteorological cancer i.e. over-dependence on models which have little value, not necessarily little skill. On the plus side it allows forecasts to be naturally consistent ... fairly certain in their ability to track where storms are, where they are going in the immediate future. To be fair the cone widens for a reason: tropical storms and hurricanes can encounter harsh or favorable environments quickly and these types of environments are hard to recognize over the ocean at longer lead times. Of course these environments can bring about changes to the inner workings of tropical storms in which case certain status quo forecasting rules may not work so well, and of course models also tend to not be spot on with hurricane intensity changes.

If nothing else, just seeing the products and how they are discussed should be interesting. It will be worth paying attention to see how the "public" reacts to be under "threat" for longer periods of time.

* I have not seen what this will look like but it will certainly be interesting!