Recent Development of Multiscale and Multi-resolution Data Assimilation in Hybrid EnVar for Global and Regional Numerical Weather Prediction

The most recent development of the multiscale and multi-resolution data assimilation in hybrid 4DEnVar for global NWP is first discussed. A multi-resolution ensemble (MR-ENS) method is developed to resolve a wider range of scales of the background error covariance (BEC) in the NCEP hybrid 4DEnVar to improve global forecast while saving the computational cost. Experiments show that MR-ENS improves global and tropical cyclone track forecasts compared to the operational Single-Low-Resolution configuration (SR-Low). Compared to the expensive configuration where background ensemble members are at high resolution (SR-High), MR-ENS decreases the overall cost by about 40% and shows comparable global and tropical cyclone track forecast performances. Various diagnostics to facilitate the understanding of the difference of MR-ENS relative to these other configurations were performed and will be shown and discussed.

In addition, a scale-dependent covariance localization (SDL) method is implemented and integrated with MR-ENS. The integration is termed as SDL-MR-ENS. Experiments suggested that the SDL approach not only improved the global forecasts relative to using a fixed localization for all scales, it also improves relative to well-tuned operational level dependent localization. Results of SDL with vertical extension and various SDL-MR-ENS experiments with the same computational cost constrained are planned to be discussed together with the initial SDL implementation for regional convection allowing application.