Stochastic Parametrisation of the Boundary Layer Fluxes in Cold-air Outbreaks

Ensemble forecasts, thought widely accepted as the state of art in numerical weather forecast (NWP), have a tendency to under-represent the spread on the longer timescales and might often fail to predict serious weather events in advance. Another new approach based on stochastic treatment of uncertainty is arising as well. Effect of various unresolved processes and variation in the initial conditions are represented by deliberate introduction of an uncertainty into models. Specific forms of stochastic parametrisation have been already implemented to some established NWP, however there are still number of issues surrounding the choice of introduced uncertainty.

With a goal to improve the quality of stochastic schemes, the underlying uncertainty in the atmospheric boundary layer fluxes of momentum, sensible heat and latent heat is analysed in a mesoscale phenomena of cold-air outbreak. Specific cases were observe during the DIAMET and ACCACIA field campaigns of the FAAM aircraft. Fluxes are estimated based on the measurements from airborne instruments and the variability is analysed. The Met Office Large Eddy Model (LEM) is then employed for fine simulations of of these cases, with outputs of European Centre for Medium-Range Weather Forecasts (ECMWF) used for initialisation and large scale forcing in the simulations. Time and spatial variation in the fluxes at chosen altitudes is then examined, leading to an approximation of the statistical distribution of uncertainty in the values of fluxes.

In addition to case simulations, a variety of idealised cases of developing cold-air outbreaks is modelled in LEM. Approximated mean values and variance of the fluxes are compared against similarity theory and the parametrisation used by ECMWF.

Results suggest possible adjustment to probability distributions of random numbers used in stochastic schemes.