Impact of a Multi-Layer Snow Model in the ECMWF Integrated Forecasting System.

Due to its unique physical properties, snow has a significant impact on the exchanges of heat and moisture between the surface and the atmosphere, modifying the distribution of surface energy fluxes. Cold processes associated with snow have implications spanning from short-range weather predictions to seasonal and climate projections. To resolve a range of time-scales associated with snow from the sub-diurnal to seasonal, physically-based multi-layer snow models are required, which enable a detailed description of internal snowpack processes. In this work the impact of a new multi-layer snow scheme in the ECMWF Integrated Forecasting System (IFS) is evaluated. Simulations with the multi-layer snow scheme and with the current single-layer snow scheme are compared. 10-day forecasts are initialized every day at 00UTC for the wintertime 2016/2017 and 2017/2018. To produce equilibrated snow initial conditions for each of the forecasts, surface initial conditions are derived from long integrations using the land-surface model only. The evaluation in the northern hemisphere of the coupled forecasts using standard observations (synop) shows an improved representation of the snow in the multi-layer snow scheme. The root-mean-square error of snow depth decreases by 17% at day 1 (mainly because of improved initial conditions) and by 19% at day 5. The single-layer snow scheme generally overestimates snow depth compared to the observations, which is reduced using the multi-layer snow scheme. The multi-layer snow scheme leads to the increase of the amplitude of the diurnal cycle of the near-surface temperature over snow-covered area, mainly because of the reduced thermal inertia of the surface snow layer. The impact on 2m-temperature biases and the relationships with other sources of error, like cloud cover, are discussed. These results show the added value of increased physical complexity in the representation of snow in numerical weather prediction applications.