Scale-Dependent Localization and Multiscale Assimilation for NCEP Data Assimilation Systems

Most of the operational data assimilation applications at NCEP utilize ensemble-based covariance estimates through integration into a hybrid EnVar update using the Gridpoint Statistical Interpolation (GSI). The GSI-based hybrid system has recently been updated to allow for scale-dependent localization and multi-resolution applications. The use of scale-dependent localization (SDL) is well documented in the literature and has shown to improve the quality of analysis and subsequent forecast skill. In this presentation, an overview of the GSI-based multiscale system will be given. The use of SDL in GSI is currently being considered as part of a future upgrade to the Global Forecast System (GFS) as well as initial implementation of the Rapid Refresh Forecast System (RRFS). Results from experiments with prototype versions of these systems will be presented. Plans for integration of these capabilities into the Joint Effort for Data assimilation Integration (JEDI) will be discussed.