Ensemble Square Root Filters*

Ensemble data assimilation methods assimilate observations using state-space estimation methods and lowrank representations of forecast and analysis error covariances. A key element of such methods is the transformation of the forecast ensemble into an analysis ensemble with appropriate statistics. This transformation may be performed stochastically by treating observations as random variables, or deterministically by requiring that the updated analysis perturbations satisfy the Kalman filter analysis error covariance equation. Deterministic analysis ensemble updates are implementations of Kalman square root filters. The nonuniqueness of the deterministic transformation used in square root Kalman filters…

• NA
  National Aeronautics and Space Administration
• NO
  National Oceanic and Atmospheric Administration
• OO
  Office of Naval Research

  Awards: 0601153N, N00014