•Advantages
–Optimal predictor selection
–Recognition of model predictability,
but not pattern predictability
–Removal of some systematic bias
–Reliable probabilities
•Disadvantages
–can be over trained (too many predictors can cause
problems)
–Short samples
•HOW DO YOU HANDLE EXTREME EVENTS? They are often not in the statistical sample
– Changing NWP models
–Does not distinguish between more
predictable pattern and less predictable
ones.
•