An improved snow cover scheme for high-resolution numerical weather prediction models.

Abstract:

Numerical weather prediction (NWP) is the core of any operational weather service. The horizontal and vertical resolution of numerical weather prediction models strongly increased during the last decades. However, numerical weather prediction in complex terrain is still challenging, because the underlying physics in the majority of subgrid-scale parameterizations have been developed for flat or idealized terrain. Weather prediction in alpine countries – such as Austria or Switzerland – is not only challenged by complex topography, furthermore, for a good part of the year the ground is snow covered influencing boundary layer processes such as turbulence and radiation. Currently, most NWP models predict the formation and evolution of the seasonal mountain snow cover in a simplified way, i.e. often a single layer model. We validated the performance of the currently implemented snow cover scheme of the COSMO model (Consortium for Small-scale Modelling) in terms of the snow surface temperature, a key parameter for the evolution of the snow cover, as well as snow height. Snow surface temperature and snow height from 120 alpine weather stations located across the Swiss Alps were compared to the corresponding COSMO output. Surface temperature was found to be overestimated especially during the night (up to 10 °C, RMSE = 6.0 °C). Snow height tends to be underestimated during the ablation phase, i.e. the COSMO model becomes snow-free too early. A new multi-layer snow module, which minimizes the energy balance equation with regard to snow surface temperature and then iteratively solves the heat equation has been implemented, predicting the daily cycle of the snow surface temperature accurately (RMSE = 1.8 °C). Furthermore, by implementing densification, melt-freeze processes and water transport snow height, especially during the ablation phase, was found to be in good agreement with the observations. Our suggested snow scheme shows promising potential not only for accurately modeled snow surface temperature and snow height it has also the potential to improve the NWP performance, e.g. in predicting the near surface air temperature during snow covered periods.