The role of the mixing length in determining the size of convective storms in high resolution models

The Met Office UKV, running routinely since 2009, is the operational variable resolution version of the Unified Model (UM) and contains a fixed resolution region with a gridlength of 1.5km. In this model, convection is represented explicitly. In many situations the model shows a large improvement in the representation of showers compared to models where the convection is primarily parameterised. However it is clear that there are also deficiencies in this representation, the main ones being that the convective cells tend to be too large and intense and that convective initiation is delayed. In studies where the resolution of the UM is increased it is observed that as the gridlength decreases, there are a greater number of convective cells and these are narrower. In order to improve the forecast of convection in present and future operational models, it is of interest to investigate what processes influence the size and number of convective cells in the model at all scales both within and above the boundary layer and compare model results with observations of real convective cases.

In the UM, subgrid mixing is represented using a Smagorinsky-type turbulence scheme alongside a 1D boundary layer model. In the UKV model, the turbulence scheme is run only in the horizontal but in the higher resolution models, i.e. gridlength 500m or less, this is extended to the vertical. The choice of mixing length used in the turbulence scheme, which is the product of a specified constant and some representation of the gridlength, can be shown to influence the size of the convective cells. The cell statistics (e.g. size and number) from a higher resolution model run can be made to look more like a coarser resolution run if the mixing lengths are chosen to be the same in both models. However, an improvement in the number of small cells present in a run can often be at the expense in the number of large cells so a simple change to the magnitude of the constant mixing length is not the solution.

In this study, data collected during the DYMECS (Dynamical and Microphysical Evolution of Convective Storms) project is used. This project, a collaboration between The University of Reading and The Met Office, has obtained a large database of convective storm lifecycles by tracking storms with the Chilbolton Advanced Meteorological Radar. Individual storms were tracked on 40 days using a combination of scanning techniques to extract the dynamical and microphysical properties of the storm (such as storm size, vertical velocity, maximum surface rain-rate and hail intensity). This provides a wealth of observational data with which to compare and validate the modelling work.

A nested suite of UM models, with gridlengths ranging from 1.5km down to 100m, is used in this study to investigate how the size and number of convective cells changes with choice of mixing length in the subgrid mixing scheme. We explore the potential to improve the distribution of convective cell size and number by varying the mixing length with height and including the vertical resolution in the representation of gridlength. Statistics of cell size and number are shown along with other storm attributes with results informing the direction of future research in this area.