HR: 0830h
AN: A41A-06    [Abstracts]
TI: Combining Multiple Atmospheric GCM Ensembles for Seasonal Prediction with a Bayesian Method
AU: * Robertson, A W
EM: awr@iri.columbia.edu
AF: International Research Institute for Climate Prediction, 61 Route 9W, Palisades, NY 10960 United States
AU: Lall, U
EM: ula2@columbia.edu
AF: International Research Institute for Climate Prediction, 61 Route 9W, Palisades, NY 10960 United States
AU: Zebiak, S
EM: steve@iri.columbia.edu
AF: International Research Institute for Climate Prediction, 61 Route 9W, Palisades, NY 10960 United States
AU: Goddard, L
EM: goddard@iri.columbia.edu
AF: International Research Institute for Climate Prediction, 61 Route 9W, Palisades, NY 10960 United States
AB: The skill of seasonal climate predictions can be enhanced by combining together the predictions made with different models, and it is desirable to be able to combine them in an optimal way. Here we develop an improved Bayesian optimal weighting scheme to combine six atmospheric general circulation model (GCM) seasonal hindcast ensembles. The approach is based on the prior belief that the forecast tercile-category probabilities are equal to the climatological ones. The six GCMs are integrated over the 1950-97 period with observed monthly SST prescribed at the lower boundary, and the scheme is applied to seasonal-mean simulations of precipitation as well as near-surface temperature. A key ingredient of the scheme is the climatological equal-odds forecast, which is included as one of the models in the multi-model combination. The weights of the individual models are determined by maximizing the log-likelihood of the combination by season over the integration period. Refinements are made to the original Bayesian scheme of Rajagopalan, Lall and Zebiak (2002), by reducing the dimensionality of the numerical optimization, averaging across data sub-samples, and including spatial smoothing of the likelihood function. These modifications are shown to yield increases in cross-validated Ranked Probability Skill Score (RPSS) skills. The Bayesian optimal weighting scheme is shown to outperform a simple unweighted pooling together of the models, which in turn outperforms the individual models. In the extratropics, the main benefit is to bring much of the large area of negative precipitation RPSS up to near-zero values. The skill of the optimal combination is almost always found to increase when the number of models in the combination is increased from 3 to 6, regardless of which models are included in the 3-model combination.
DE: 3319 General circulation
DE: 3337 Numerical modeling and data assimilation
DE: 3339 Ocean/atmosphere interactions (0312, 4504)
DE: 3354 Precipitation (1854)
DE: 3367 Theoretical modeling
SC: Atmospheric Sciences [A]
MN: 2005 Joint Assembly