Evaluation of extended intraseasonal forecasting using slow manifold modeling

Simulation and extended forecasting of the observed high amplitude tropical intraseasonal activity, a very strong and coherent signal and arguably the most useful time scale from the point of view of the user community, has proven to be a difficult task for the state of the art climate models.

In order to improve predictability of the Intraseasonal Oscillation (ISO) activity, a Bayesian physically based empirical scheme was designed. The scheme uses wavelet banding to separate the selected predictand (e.g., regional precipitation in a sector of South Asia, Brahmaputra river discharge) and the set of predictors (e.g. OLR, 200mb and 925mb zonal wind over the equatorial Indian Ocean) into physically significant bands where linear regression followed by recombination of the bands is used to generate the forecast. Results of the empirical scheme (Webster and Hoyos 2004: BAMS) suggest that isolating the evolution of the intraseasonal signal from higher frequency variability and noise improve the skill of the extended prediction. We hypothesize that a similar phenomenon occurs in numerical models: Specifically, the strong intraseasonal signal observed in nature is eroded by high frequency errors due to the model parameterizations, especially in convection.

To evaluate the hypothesis, 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, allowing the examination of the skill of the model relative to the phase of the oscillation. Two different cases were studied: December of 1992/93 during TOGA COARE and the onset of the monsoon in 2004. The results represent well the observations for about 7 days after which the magnitude of the errors is greater than the signal itself. To minimize the random error growth associated with poorly modeled convection, we have developed a model modification scheme for the coupled ocean-atmosphere general circulation model that tries to mimic the philosophy of the banded wavelet empirical scheme. The modified model (referred to as the Slow Manifold Model: SMM) is used for the two series of 30-day forecasts described above. The propagation features associated to ISO activity are greatly improved. More experiments are underway to confirm the earlier success of the Slow Manifold Model. The Slow Manifold integration scheme has been designed to be easily integrated into operational systems, providing global forecasts in the intraseasonal time scale.