Validation of mesoscale numerical weather prediction models calls for high-resolution observations. Conventional data sources, such as radiosonde data, provide a reliable benchmark but their spatial resolution is insufficient for model validation in the kilometric scale. Doppler radar radial wind observations are used in this study as an alternative data source for model validation. Data assimilation system in general, and observation modelling in particular, are used as a toolbox for validation.
Two versions of the High Resolution Limited Area Model (HIRLAM), which differ only in the formulation of the surface stress direction, are validated over a period of 1 month. The observation minus background (OmB) statistics advocate the merits of using high-resolution radar data as model validation material. It is demonstrated that subtle differences in model versions due to different parameterization details can be distinguished with high statistical confidence. Copyright © 2008 Royal Meteorological Society