Application of WRFVar (3DVar) to a High Resolution (3-km) Model over Beijing Area

Assimilation of mesoscale observations, such as AWS, GPSPW, etc., and initialization of the high resolution (3km or less) model with the 3DVar approach are still challenge problems. To improve the weather forecast over the Beijing area for the 2008 Olympic game, a 3 nests (27/9/3-km) WRFVar/WRF system with 3-h update cycle is established. The background error statistics (BES) files for each nested domain were derived based on the differences between 24-h and 12-h WRF model forecasts valid at the same time. The BES for the fine domain (3-km) obtained by interpolation from its parent domain (9-km) was tested as well. This was shown to work well. User can now save much work related to the derivation of the fine mesh BES from the forecast over a period of time. To keep the dynamic balance existed in the first guess as much as possible for the high resolution frequent cycling run, the experiments with the tuned background error variance using different factors in the 3-km resolution WRFVar analysis are conducted. The adaptive tuning of observation error parameters (Desroziers and Ivanov 2001) was also tested. For a convective event occurred on August 1, 2006, the results showed that the background and observation error statistics tuning in WRFVar improved the skill of the precipitation forecast.