Abstract/Summary

A new approach is introduced to assimilate cloud information into a convection-permitting numerical weather prediction model with an ensemble Kalman filter. Cloud information is obtained from a satellite cloud-top height product based on Meteosat SEVIRI data. To assure quality and specify the observation error based on data reliability, the satellite cloud-top height is merged with cloud-top height derived from radiosonde humidity soundings. For that purpose, the radiosonde cloud-top information is assumed to be representative not only for the location of the ascent itself but also in spatially and temporally nearby areas which have the same cloud type according to a satellite cloud-type classification. From the cloud-top height information, different variables are derived to be used as ‘pseudo observations’ within the ensemble Kalman filter. At cloudy locations these are cloud-top height and 100% relative humidity at that cloud-top height. Also zero cloud cover amounts at up to three heights are assimilated at cloud-free observation locations and above low- and mid-level cloud observations. The approach has been applied in a number of single-observation experiments to assess the effect in detail of assimilating these variables on the atmospheric column. Three examples are shown of single-observation experiments for a stable wintertime high-pressure synoptic situation with low stratus clouds, where model background and observation have large discrepancies. The ensemble Kalman filter is able to draw the ensemble closer to the observations. The analysis shows improved cloud variables in both cloudy and cloud-free scenes. A reasonable effect on the temperature profiles through cross-correlations in the background ensemble is obtained. The approach represents a method to exploit high-resolution remote-sensing data for short-range meso-γ-scale weather forecast models with a view to applying this method in a rapid-update cycle, particularly useful for local-scale weather phenomena.