Real-time mesoscale analysis and prediction with NCAR 4D-REKF

A Four-Dimensional Relaxation Ensemble Kalman Filter (4D-REKF) mesoscale analysis and forecasting system has been developed jointly by NCAR and ATEC (Amy Test and Evaluation Command). 4D-REKF is built upon the multi-model (MM5 and WRF), multi-approach (perturbations), and multi-scale (nested-grid) E-RTFDDA (Ensemble Real-Time Four-Dimensional Data Assimilation and forecasting system) that has been running operationally at the Army Dugway Proving Ground. E-RTFDDA employs Newtonian-relaxation 4D data assimilation algorithms to achieve continuous 4D analysis and rapid-update forecast cycles. 4D-REKF computes the background error covariance using the multi-model E-RTFDDA forecasts and uses it to construct the spatial weight functions for assimilating each observation (i.e. Kalman gain). With 4D-REKF, both observations the their Kalman gains are ingested into the WRF models and the observation information is spread in the model space and time according to the time-dependent Kalman gains. A Local Ensemble Kalman Filter (LEKF) approach is employed to take account of multi-observations. 4D-REKF retains and leverages the advantages of both traditional Newtonian-relaxation and Ensemble Kalman Filter data assimilation schemes. 4D-REKF represents a new way to extend the traditional (intermittent) EnKF data assimilation method to a 4D continuous data assimilation paradigm that avoids the dynamic shocks associated with the intermittent EnKF processes. In this paper, we will describe the science and initial implementation of 4D-REKF, present the validation experiments and comparison studies with NCAR DART-ENKF, WRF-3DVAR, WRF-4DVAR and transitional Newtonian nudging schemes. Verifications statistics of the model output from real-time parallel data assimilation and forecasting experiments at Dugway Proving Ground will be also discussed.