Outer Loops and Scale-Dependence in the NCEP GDAS Hybrid 4D EnVar Assimilation Scheme

NCEP is planning to implement a hybrid 4D EnVar scheme for use in the operational global data assimilation system (GDAS) and global forecast system (GFS) model in early 2016. Beyond the initial implementation, several enhancements are being pursued for impact testing. Outer loops have been demonstrated to improve analysis quality within the context of 4DVAR and EnKF by helping to address nonlinearities. Within the context of 4D EnVar, an outer loop can be run by re-initializing the full-resolution nonlinear model at the beginning of the assimilation window and performing an addition inner loop minimization. This may prove to be particularly important for dual-resolution configurations that use low-resolution ensemble members within high-resolution deterministic assimilation. Another potential improvement that is being pursued is the application of scale-dependence in both the hybrid weighting parameters as well as the localization of the ensemble control variable. Results from toy model and GFS impact experiments will be presented to demonstrate the utility of the use of outer loops and scale dependence within the context of 4D EnVar data assimilation.