Scale-Dependent Weighting and Localization for Global Numerical Weather Prediction

Various forms of hybrid (ensemble/variational) data assimilation have become standard practice for operational global numerical weather prediction. NCEP has been utiliziling hybrid 4DEnVar since 2016 and will continue to do so with the initial implementation of the FV3-based Next Generation Global Prediction System (NGGPS) model in early 2019. Taiwan’s Central Weather Bureau (CWB) has been utilizing a GSI-based hybrid 3DEnVar for the past several years.

Many hybrid schemes utilize single global weighting parameters to prescribe the contributions from the ensemble and climatological error covariances, respectively. Furthermore, the implemented versions of the GSI-based hybrid utilize localization for the ensemble part of the hybrid increment that is assumed to be Gaussian, with the horizontal localization varying only by vertical level. Several studies have already shown that spectral and scale-dependent localization can be more effective within the context of EnVar (Buehner 2012, Buehner and Shlyaeva 2015, Lorenc 2017). Early work with toy models has shown that a similar idea of scale-dependence can be extended to the hybrid weights.

Here, we will describe an effort toward applying waveband-dependent localization as well as scale-dependent weighting between the static and ensemble contributions within a hybrid assimilation paradigm. Results will be shown for low resolution prototype versions of the NGGPS FV3-based modeling system at NCEP as well as the spectral model used at Taiwan’s CWB. Future work regarding the use of mechanisms to quantitatively estimate the scale-dependent weighting and localization parameters will be discussed.