Date and Time：27th, May,（Tue）　16:30〜　（This seminar will start after the lab-seminar.）
Location：Ew-305, IIS, University of Tokyo
Invited Speaker：　Dr. Devin Kilminster　（Senior Scientist, MetService Adjunct
Lecturer, Victoria University)
TITLE:  Probabilistic weather forecasting and ignorance tricks.
ABSTRACT:
At MetService, we are developing statistical methods for combining
multiple sources of information into unified probabilistic forecasts
of the weather.  These sources include different available runs from
both global and limited area numerical weather prediction models,
ensembles, and even raw observations.

An important question is how the performance of such a probabilistic
forecast can be measured.  Perhaps the simplest method is to score
the negative logarithm of the forecast probability of the actual
outcome
--- a measure of the information required to transmit the outcome to a
receiver already in possession of the forecast, that is, the receiver's
ignorance.

I will show some examples of the unified probabilistic weather
forecast products being developed at MetService.  I will then show
how the viewpoint of ignorance supplies some "tricks" for attacking
the problem of combining forecast information.