The impact of diabatic initialization on stratospheric analyses, forecasts, and transport experiments

An efficient diabatic initialization procedure has been applied to the entire tropospheric and stratospheric domain within the Goddard Earth Observing System (GEOS) 4-D data assimilation system. The initialization, or initial balancing and noise control technique, is based on the application of the iterative Euler scheme as a highly selective, efficient filter. The GEOS 46-layer General Circulation Model extending to 0.1 hPa was used to produce the first-guess fields within the data assimilation system. This model is also used to calculate medium-range stratospheric forecasts. Within the diabatic initialization approach, the model itself is used to balance the resulting analysis, or initial fields. The diabatic initialization technique appears to be a useful tool which improves stratospheric analyses and forecasts by controlling small-scale high-frequency non-meteorological oscillations and shocks, or spin-up effects resulting from initial imbalances. Changes due to initialization are cumulative with time, hence initialization affects positively the slowly varying components of stratospheric flows, resulting in better representation of climate characteristics. Experiments performed to assess the impact of diabatic initialization on monthly mean analyses and diagnostic fields produced by the data assimilation system show that it provides an efficient noise control, especially for the upper stratospheric levels and in the polar regions. Stratospheric analysis errors have been significantly reduced by applying diabatic initialization. The major characteristics of stratospheric circulation, such as the residual mean meridional circulation, potential vorticity and Eliassen-Palm or EP-flux distributions, are improved when using diabatic initialization. The ozone transport experiments show improvements due to using initialized winds in terms of better agreement with data. Ten-day stratospheric forecast scores have been improved when using initialized, or balanced initial conditions. The initialization method used in the study is easy to implement and is computationally efficient. It appears to be a useful component of a stratospheric data assimilation system.