Marek Wlasak, Met Office
An examination of the short-range forecast error characteristics of perturbations from an ensemble of 4DEnVars when using different inflation schemes
Coauthors
Neill Bowler, Mohamed JardakAbstract:
At the Met Office an ensemble of 4DEnVars has been developed as a potential replacement for the current ETKF with the aim of providing an improved representation of the flow-dependent forecast error for use in hybrid data assimilation. In addition to standard verification approaches one can use a qualitative approach by looking at the error covariance structure implied by the ensemble background perturbations. This presentation demonstrates the power of covariance statistics by examining the effect of different inflation schemes on the structure of the perturbations.
One typically expects the forecast error to be substantially less balanced than the full atmospheric state since the large scales are well observed and thus short-range forecast errors will be mostly small-scale and unbalanced . When using the relaxation to prior perturbation (RTPP) inflation scheme the perturbations were seen to be large-scale and well balanced, which we believe to be incorrect. An ensemble which uses only static covariances in the ensemble update and a low value of RTPP produces covariance structures similar to the scaled random analysis increments which are used as model forcing. Relaxation to prior spread (RTPS) inflation scheme produces noticeably tight vertical correlations for pressure and generally longer horizontal length scale for winds, potential temperature and total water perturbations.