A new observational dataset separating stratiform and convective precipitation at a one-hour temporal resolution is produced for Germany to evalutate the COSMO-EU and DE model. The underlaying idea of this work is to connect rain producing cloud types taken from synoptic observations and derived cloud types from Meteosat data by a interpolation scheme. The accuracy of the cloud types derived from Meteosat data are compared with observed cloud types by an error matrix. The distinction between convective and stratiform precipitation was made by using hourly observations of clouds, ‘weather watch’ and the amount of precipitation. In summary the clouds of the convective scale are cumulus, cumulus congestus as well as cumulo-nimbus/nimbostratus and for stratiform precipitation events stratus fractus, stratocumulus as soon as altocumulus with altostratus/nimbostratus. In order to be able to use cloud types for the interpolation, these data are normalized to yield values of frequency of rainfall. A precipitation analysis is created, first without satellite data and in a second step using satellite data.