Deliverable No. 2.3 Recommendations on the inclusion of APPLICATE model enhancements in NWP models

This deliverable provides an assessment of the impact of new and/or improved model components developed within APPLICATE WP2 into numerical weather prediction (NWP) systems. Based on these conclusions, it provides recommendations about priorities to pursue for further model developments and components to include in the next generation of NWP systems.

The APPLICATE developments considered in this deliverable are:

• the inclusion of dynamical and/or thermodynamic sea ice models.

• the extension of single-layer snow schemes on land to multi-layer snow schemes.

• the inclusion of snow schemes on top of sea ice.

• an improved version of the eddy-diffusivity mass-flux scheme which detects decoupled stratocumulus.

  Introducing an interactive sea ice model into numerical weather predictions leads to a robust increase in forecast quality for near-surface temperature close to the sea ice edge in the ARPEGE/AROME prediction system. Both bias and standard deviation are substantially reduced. The temperature gradient across the sea ice edge is more realistic. Exploiting both observed sea surface temperatures and sea ice fraction to initialize the model is essential to optimize the benefits from the interactive sea ice model. Similarly, extracting boundary conditions from a global model which also uses an interactive sea ice model improves substantially the performance of the regional prediction system. The dynamic coupling of the ice cover in IFS allows for capturing accurately rapid changes in sea ice concentration, especially along the sea ice edge as illustrated on a case study. A substantial impact is seen in the near surface temperature (up to 6°C) when the large-scale meteorological situation is favourable.

  At high latitudes, near-surface temperature tends to be overestimated during cold events occurring in clear-sky conditions. Introducing a multi-layer snow scheme over land into the IFS prediction systems leads to a better representation of these cold events at forecast day 2 over the Sodankylä observational supersite. The diurnal cycle in near surface temperature is also better captured with the multi-layer snow scheme than with the previous single-layer scheme, with a reduced warm bias at the daily minimum and cold bias at the daily maximum.

  Including this multi-layer snow scheme over sea ice allows for a marginally better simulated temperature in the lower atmosphere for a case study from the SHEBA campaign. However, during periods of liquid-phase cloud cover, systematic errors in the representation of such clouds leads to the increase of the errors of skin temperature in the simulation using the multi-layer snow scheme. These results highlight the key role of compensating errors in the surface energy balance and the need for improved representation of Arctic boundary layer processes.

  A revised version of the eddy-diffusivity mass-flux scheme was proposed to identify decoupled stratocumulus. This scheme detects and simulates cloud layers that would not have been represented before, as seen in an ASCOS Arctic stratocumulus case. The root mean square error for clouds and radiative properties are drastically reduced with this improved scheme.