A hybrid nudging-ensemble Kalman filter approach to data assimilation. Part I: Application in the Lorenz system

Lili Lei, David R. Stauffer, Sue Ellen Haupt, George S. Young

Research output: Contribution to journalArticlepeer-review

10 Scopus citations

Abstract

A hybrid data assimilation approach combining nudging and the ensemble Kalman filter (EnKF) for dynamic analysis and numerical weather prediction is explored here using the non-linear Lorenz three-variable model system with the goal of a smooth, continuous and accurate data assimilation. The hybrid nudging-EnKF (HNEnKF) computes the hybrid nudging coefficients from the flow-dependent, time-varying error covariance matrix from the EnKF's ensemble forecasts. It extends the standard diagonal nudging terms to additional off-diagonal statistical correlation terms for greater inter-variable influence of the innovations in the model's predictive equations to assist in the data assimilation process. The HNEnKF promotes a better fit of an analysis to data compared to that achieved by either nudging or incremental analysis update (IAU). When model error is introduced, it produces similar or better root mean square errors compared to the EnKF while minimising the error spikes/discontinuities created by the intermittent EnKF. It provides a continuous data assimilation with better inter-variable consistency and improved temporal smoothness than that of the EnKF. Data assimilation experiments are also compared to the ensemble Kalman smoother (EnKS). The HNEnKF has similar or better temporal smoothness than that of the EnKS, and with much smaller central processing unit (CPU) time and data storage requirements.

Original languageEnglish (US)
Article number18484
JournalTellus, Series A: Dynamic Meteorology and Oceanography
Volume64
Issue number1
DOIs
StatePublished - 2012

All Science Journal Classification (ASJC) codes

  • Oceanography
  • Atmospheric Science

Fingerprint Dive into the research topics of 'A hybrid nudging-ensemble Kalman filter approach to data assimilation. Part I: Application in the Lorenz system'. Together they form a unique fingerprint.

Cite this