Atmospheric data assimilation using the Ensemble Kalman Filter at BMRC

Abstract

In 2003, BMRC commenced a research program aimed at exploring the use of an Ensemble Kalman Filter (EnKF) as a progression from our current GenSI system. The aim was to support both deterministic NWP and ensemble prediction systems (EPS) at various scales. The advent of ACCESS has meant that these aims have been superseded, with the Met Office VAR assimilation system to become the core algorithm. However, the various flavours of EnKF have their attractions, including for EPS, and there remains a significant possibility that a hybrid VAR-EnKF system could outperform both in some circumstances. In addition, the EnKF remains the preferred BMRC strategy for oceanic data assimilation (DA) and has demonstrated utility in applications such as land surface DA. Thus an EnKF research effort will continue. Even a casual survey of the EnKF literature will reveal a lack of consensus as to the advantages of the competing algorithms. Further, it is our perception that while VAR only became operationally feasible following a couple of crucial insights into the detail of the formulation, the EnKF has not to date received sufficient similar attention. Thus our strategy was to investigate the properties of the various EnKF algorithms in simple settings, in parallel with the development of a prototype atmospheric EnKF built around the GenSI infrastructure. This paper reviews a couple of specific investigations into EnKF properties, and summarises our progress in developing a prototype atmospheric EnKF.