Abstract This study evaluates the performance of the Consortium for Small‐scale Modeling (COSMO) and the Icosahedral Nonhydrostatic (ICON) numerical weather prediction models in forecasting a strong south foehn event in the Alpine Rhine Valley in November 2016. Comparisons with surface observations and Milan soundings indicate that ICON outperforms COSMO in forecasting near‐surface characteristics such as temperature, wind speed, foehn onset and cessation, and foehn spatial extent. Notably, ICON reduces the average cold bias from 3 K (in COSMO) to 1.5 K. The larger cold bias in COSMO is partly attributed to stronger evaporation, which enhances evaporative cooling and increases the downward surface sensible heat flux during foehn conditions. These larger surface fluxes are linked to COSMO's overestimation of near‐surface winds. The differences between the models, however, extend beyond local scales and also influence upstream flow conditions. ICON more accurately reproduces the observed Milan sounding profile, whereas COSMO simulates a stronger cold‐air pool in the western Po Valley. This disparity affects small‐scale gravity waves and lower level wind fields, and may alter the origins and pathways of the foehn air arriving in the Rhine Valley. Sensitivity experiments did not identify a single factor for the differences between the models. The findings suggest that, in addition to the physical parametrizations investigated, differences in the dynamical cores, orography representation, grid structures, and vertical coordinate systems may also influence foehn forecast quality.
Tian et al. (Wed,) studied this question.