15A.1 Experiments with anisotropic background error correlations in the Rapid Refresh system

Thursday, 4 June 2009: 1:30 PM
Grand Ballroom East (DoubleTree Hotel & EMC - Downtown, Omaha)
Dezso Devenyi, NOAA/ESRL/GSD and CIRES/University of Colorado, Boulder, CO; and S. S. Weygandt, M. Hu, and S. G. Benjamin

19th Conference on Numerical Weather Prediction

Abstract

Experiments with anisotropic background error correlations in the Rapid Refresh system

The Rapid Refresh (RR), a new hourly updated data assimilation and forecast system running in real-time at NOAA/ESRL/GSD, is slated to replace the Rapid Update Cycle (RR) in operations at NCEP in 2010. The main components of the RR are the Gridpoint Statistical Interpolation (GSI) 3dvar analysis and the WRF ARW core model.

The current formulation of GSI for RR uses isotropic background

error correlations applied by digital filters. GSI has also has an option to allow for the use of anisotropic background error correlations. Following the Riishojgaard method, this option is based on potential temperature gradients. Anisotropy is introduced by distortion of

original isotropic correlations using the outer product of the gradient

with itself.

In the poster, results of experiments with anisotropic option in GSI

will be presented with special attention to the atmospheric boundary

layer and the impacts on near surface fields. A possible generalization of anisotropy will be introduced using higher order local approximation to the correlation tensor. Results from real time RR cycle will be presented to illustrate the utility of anisotropy approach.

Poster presentation

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner