HR: 10:30h
AN: A32A-01 INVITED     [Abstracts]
TI: Extended intraseasonal predictions using Bayesian empirical methods and slow manifold climate modeling
AU: * Peter, W
EM: pjw@eas.gatech.edu
AF: School of Earth and Atmospheric Sciences, Georgia Institute of Technology, 311 Ferst Drive,, Atlanta,, ga 30332 United States
AU: Hoyos, C
EM: choyos@eas.gatech.edu
AF: School of Earth and Atmospheric Sciences, Georgia Institute of Technology, 311 Ferst Drive,, Atlanta,, ga 30332 United States
AU: Vitart, F
EM: Frederic.Vitart@ecmwf.int
AF: European Center for Medium Range Weather Forecasts, Shinfield Park, Reading, RG2 9AX United Kingdom
AU: Miller, M
EM: Martin.Miller@ecmwf.int
AF: European Center for Medium Range Weather Forecasts, Shinfield Park, Reading, RG2 9AX United Kingdom
AU: Palmer, T
EM: Tim.Palmer@ecmwf.int
AF: European Center for Medium Range Weather Forecasts, Shinfield Park, Reading, RG2 9AX United Kingdom
AU: Hortal, M
EM: Mariano.Hortal@ecmwf.int
AF: European Center for Medium Range Weather Forecasts, Shinfield Park, Reading, RG2 9AX United Kingdom
AB: A major goal of GEWEX is the prediction of precipitation from daily to seasonal time scales. Of all of these scales, prediction of intraseasonal variations stands as probably the most useful for agriculture, water resource management and disaster mitigation and relief planning. Unfortunately, it is the time scale that has proven most difficult to either simulate or predict even though the intraseasonal signal is strong, and the phenomena well described, especially in the tropics. We have approached the problem of intraseasonal prediction from a hypothesis that the strong intraseasonal signal observed in nature is eroded in weather and climate models by errors in high frequency convective parameterization. The following procedures were adopted:  We develop a Bayesian physically based empirical scheme that uses wavelet banding to separate physically significant bands. This procedure was adopted for the predictor (e.g., regional precipitation in a sector of India, Brahmaputra river discharge.) and a set of predictors. Linear regression and recombination of the bands provides pentad forecasts at 20 days (4 lags) with correlation coefficients of > 0.8.  The ECMWF coupled ocean-atmosphere model was run in ensemble mode for 30 day periods initialized daily for 15 days before to 15 days after major intraseasonal oscillations thus allowing the examination of the success and failure of a climate model relative to the phase of the oscillation. Two cases were chosen: the December period of 1992/93 during TOGA COARE and the onset of the monsoon in 2004. The results compare well with observations for about 10 days after which the fields rapidly diverge.  We develop a model that applies the philosophy of the banded wavelet empirical scheme to the coupled ocean-atmosphere general circulation model and use for 30-day forecasts for the two cases mentioned above. Adherence with observations is greatly improved and a full evolution of the monsoon ISO predicted throughout June 2004. The new Slow Manifold Model appears to show great promise and has been designed with use as an operational system in mind. As distinct from the empirical Bayesian model which is regionally bound and for which new predictors would have to be found for forecasting different intraseasonal system, the SMM model provides consistent global forecasts on the intraseasonal time scale.
DE: 3337 Numerical modeling and data assimilation
SC: Atmospheric Sciences [A]
MN: 2005 Joint Assembly