[3-2]

Climatological Structure Functions in HARMONIE Data Assimilation

Jelena Bojarova (MET Norway), Nils Gustafsson (SMHI), and Magnus Lindskog (SMHI)

Data assimilation background error statistics for cloud permitting numerical weather prediction (NWP) on a regional domain are derived by considering forecast differences between ensemble prediction members as a proxy for the background forecast errors for HARMONIE forecating system. In a first trial, simple downscaling of forecasts (ECMWF) from a global Ensemble Prediction System (EPS) was applied for this purpose. It is shown that forecasts from the cloud-permitting HARMONIE NWP model with global initial and lateral boundary data are seriously affected by a model spinup over 9-12 hours, that prohibits the use of derived background error spectra, for example, for data assimimilation purposes. In a second trial, Ensemble Data Assimilation (EDA), based on perturbation of observation errors, was applied with the cloud permitting model. Although the derived background error spectra appeared to be more realistic, the ensemble generation forced by perturbation of observation errors turned out to produce an un-balanced perturbations, triggering adjustment processes with fast ascillations moving quickly out of the domain. The ensemble of perturbation generated from a randomized background error covariance is used to investigate the error propagation in the cloud-permitting HARMONIE model and what assumed structure functions impliy for data assimilation. From the results of our work, we may question the present approach to derive climatological background error statistics for cloud-permitting model data assimilation based spatial and temporal averaging utilizing the assumption of homogeneity and isotropy.