Science and Technology Seminar

MOS Probability Forecasts

Part 1 - Event Probabilities Kathy Gilbert and Judy Ghirardelli

Part 2 Probability Distributions of Continuous Variables Matt Peroutka

Meteorological Development Laboratory

Office of Science and Technology

Probabilistic weather forecasting may finally be on the upswing. Contributing to the renewed interest is the 2006 National Research Council report “Completing the Forecast.” Probability of precipitation forecasts have been issued to the public since 1966, but most other forecasts are single-value or of a descriptive nature. The Meteorological Development Laboratory (MDL) has been providing probabilistic guidance forecasts of several weather elements in various formats for many years. These forecasts are of specific events, such as the probability of thunderstorms, or of categories of variables that have very skewed distributions and do not lend themselves to statistical treatment as continuous variables. Examples of these latter forecasts include the probability of ceiling height less than 500 ft. and the probability of greater than one inch of rain. These are essentially “event” probabilities.

More recently, we have been developing methods to deal with probability distributions of continuous variables such as maximum temperature. All of these forecasts are, of course, based on output of numerical models. With the latest development, ensembles produced by the National Centers for Environmental Prediction have been used.

The first seminar will describe MDL’s probabilistic forecasts of weather events and lay the groundwork for the next seminar which will describe the techniques and results of producing probabilistic forecasts of continuous variables.

Part 1 - Wednesday,

September 19, 2007

2:00 3:00 p.m. ET

Part 2 - Wednesday,

October 3, 2007

2:00 3:00 p.m. ET

SSMC#2, Room 2358

(Contact: Bob Glahn at (301) 713-1768)