4.3
A Bayesian hierarchical approach to model output statistics
David J. Nott, University of New South Wales, Sydney, NSW, Australia; and W. T. M. Dunsmuir, R. Kohn, and F. Woodcock
A Bayesian hierachical statistical model for generating model output statistics forecasts at a collection of stations in the Sydney region is developed. Our statistical model can allow for correlations between observations at different stations and at different times. An additiional attractive feature is the way that the hierarchical structure of the model can stabilize estimates of regression parameters at stations with only a short observational record. This was useful for the set of stations in our study, since some of the stations in the network have only recently been installed for the year 2000 Olympic games.
Session 4, Short-range forecasting
Wednesday, 10 May 2000, 10:40 AM-2:30 PM
Previous paper Next paper