(C204) <<< Previous | Back to Abstracts Page | Next >>> (C206)
Surface Drift Prediction Using Linear and Non-linear Hyper-ensembles of Atmospheric, Ocean and Wave Operational Models in the Adriatic
Michel Rixen, Emanuel Ferreira-Coelho
NURC NATO
rixen@nurc.nato.int(Abstract received 04/08/2005 for session D)
ABSTRACT
The Adriatic is an ideal natural laboratory and test bed to conduct oceanographic research, with a high concentration of observational and modeling operational efforts and expertise. The forecast of surface velocities, which usually results from a complex combination of ocean, atmospheric and wave forcing, has always been very challenging in the basin. However, our current monitoring and operational predictive capabilities do not allow yet inferring accurate surface drift velocities despite high- resolution modeling and extensive surface drifter data sets in the basin. It appears that an entirely different approach is warranted for a definitive significant improvement in surface drift forecast skills. Multimodel superensemble forecasts, which exploit the power of an optimal local combination of individual models usually show superior forecasting skills when compared to individual models because they allow for local correction and/or bias removal. Here we apply linear and non-linear statistical methods to generate hyper-ensembles, which are optimal combinations of models of different kinds, namely atmospheric, ocean and wave operational models, to investigate the Adriatic Sea Dynamics. Optimization methods are based on a training/forecast cycle and include simple least-square methods, neural networks and genetic algorithms. The performance and the limitations of the hyper-ensembles and standard surface drift methods are illustrated and discussed.
(C204) <<< Previous | Back to Abstracts Page | Next >>> (C206)
2005 LAPCOD Meeting, Lerici, Italy, June 13-17, 2005