The Impact of a Simple Physically-Based Stochastic Boundary Layer Scheme on the Initiation and Development of Convection in Convection-Permitting Versions of the Met Office Unified Model.

Atmospheric models are now routinely run at convection-permitting resolutions with convection represented explicitly. At these resolutions, however, many aspects of convection will be under-resolved resulting in deficiencies in the representation of convection in these models. In convective regimes, it is common to see a delay in the initiation and development of convection in the model along with convective cells which are too large and intense when compared with observations. This can be as a result of the supression of variability from BL schemes which are designed to represent the ensemble mean of the turbulence.

The missing variability can be introduced via the application of random perturbations to the temperature and moisture fields and this method has shown to improve, in particular, the timing of the convection in models. Often, however, these perturbations are arbitrary and, although the addition of the perturbations in models can have a positive impact on forecasts, they are not always physically justifiable.

This contribution describes a simple, physically-based, stochastic modification to the Met Office boundary layer scheme and reports the impact that this has on the timing and development of convection in convection-permitting versions of the Met Office Unified Model (UM). The scheme is based on random sampling of turbulence governed by standard convective scaling laws with perturbations added to the temperature and moisture fields at each timestep. A time correlation is also imposed on the perturbations. The performance of the scheme in both the full and idealised versions of the UM is summarised and future application discussed.