Bo Huang- Nov 22

Scale-dependent-localization for global FV3GFS hybrid 4DEnVar data assimilation system   Bo Huang OU School of Meteorology November 22 2019, 3:30 PM National Weather Center, Room 5600 Successful application of ensemble-based data assimilation (DA) relies on localization to reduce sampling error due to running a small ensemble. In a simple form,

Start

November 22, 2019 - 3:30 pm

End

November 22, 2019 - 4:30 pm

Address

120 David L Boren Blvd, Norman, OK 73072   View map

Scale-dependent-localization for global FV3GFS hybrid 4DEnVar data assimilation system

 

Bo Huang

OU School of Meteorology

November 22 2019, 3:30 PM

National Weather Center, Room 5600

Successful application of ensemble-based data assimilation (DA) relies on localization to reduce sampling error due to running a small ensemble. In a simple form, the localization function is defined spatially homogeneous at all scales as implemented in the Gridpoint Statistical Interpolation (GSI)-based hybrid four-dimensional ensemble-variational (4DEnVar) DA system operationally at the NCEP. However, advanced localization methods should be aware of scale variation especially for high-resolution global DA system. In this study, the multi-scale covariance localization capability is implemented in the GSI-based global hybrid 4DEnVar system for the FV3-based GFS model by introducing the scale-dependent localization (SDL) methods with and without considering cross-band correlations (i.e., SDL-Cross and SDL-NoCross, respectively). In the SDL methods, different appropriate amount of localization is applied to different ranges of scales of ensemble background error covariances while performing a single-step simultaneous assimilation of all the available observations.

One-month cycled DA experiments using the FV3-based GFS model were performed to evaluate the impacts of the SDL methods with respect to the global forecasts and tropical cyclone track forecasts. An inter-comparison was conducted between two sets of SDL experiments adopting two and three wavebands, respectively. The results show that the SDL-NoCross and SDL-Cross improve the general global forecasts and storm track forecasts over applying the fixed localization for all scales. The SDL-Cross generally shows more improvement than the SDL-NoCross especially at longer lead times. In addition, the three-waveband SDL outperforms the two-waveband SDL counterpart within three days below 100 hPa. Additional diagnostics are planned to be discussed in the presentation.