sonRAIL is a new Swiss simulation model for railway noise. Its propagation model is applying a ray tracing approach to take meteorological influences on sound propagation into account. The ray tracing algorithm is designed to allow for arbitrary sound speed profiles, which additionally can vary along the propagation path, for example as a consequence of changing surface properties. In order to do noise mapping for long-term averages, weather statistics for entire Switzerland were derived on the basis of model data of COSMO-2, the most recent version of the operational numerical weather prediction model of the Swiss Federal Office of Meteorology and Climatology, MeteoSwiss, with a grid resolution of 2.2 x 2.2 km². Especially in regions with complex topography, as it is the case in Switzerland, this high spatial resolution is a major advantage. Additionally, a profile generator was devised, which allows to calculate vertical profiles of wind speed, temperature and humidity based on input data describing the general meteorological conditions and the local surface properties. Based on this profile generator a classification scheme for meteorological conditions was developed and a set of standard profile functions for wind speed and temperature is given for each class. The validation of the profile generator in comparison with profiles directly taken from COSMO-2 showed good agreement in general, but with a tendency to overes- timate temperature gradients under highly stable conditions. A comparison of the probability of occurrence of the meteorological classes derived from COSMO-2 with observation data of conventional stations also yielded a high compliance. However nighttime situations with hardly any wind were underpredicted by COSMO-2. These findings confirm the fact that meteorological modeling is generally most difficult for low wind stable conditions. Yet the advantage of having meteorological statistics available in a small spatial resolution clearly outweighs this restraint.