•For lighter, more
frequent events.
–MOS-type non-linear statistical approaches MAKE SENSE.
•Provide
WELL CALIBRATED probabilities.
–Calibrated ensemble methods also work well
•may be
more computationally expensive in long run.
–Statistical-man mix may be an option. Since a forecaster
might be able to take into account the predictability
of the pattern.
• a person might be
able to combine information from ensemble and statistical methods to adjust POPS (this is
already being done at HPC in the 3-7 day range).