Evaluating Convective Initiation in the Met Office Unified Model over South Africa against Radar Observations.

Since 2016, the South African Weather Service (SAWS) has routinely run convective-scale simulations to assist with forecast operations across southern Africa. These simulations are run with a tropical configuration of the Met Office Unified Model (MetUM), nested in the Met Office global model, but without data assimilation. Consistent with previous studies of convective-scale simulations, we find that the onset of convection occurs earlier when model grid length is reduced from 4 km to 1.5 km. However, compared against satellite and ground-based observations, the model does not have a consistent bias for the timing of convective initiation, ranging from 2 hours too early to 2 hours late. Here, we combine storm-tracking techniques in radar data with analysis of the vertical structure of the atmosphere to investigate the model bias in convective initiation.

Using the TITAN software, a 3D composite is derived from the SAWS radars, which perform volume scans every 6 minutes. We detect convective initiation in the SAWS radar data when the 10-dBZ echo top height of individual storm features reaches above 8 km. The same method is applied to model simulated radar fields. When relating the storm life cycle to the storm three-dimensional structure, the simulated storms appear to intensify too quickly compared to the radar-observed storms.

Analysis of the vertical profiles indicates that an error in timing of convective initiation can be due to various biases – typically inherited from the driving global model – including a lack of convective inhibition and a horizontal displacement of synoptic systems. We perform a composite analysis of our cases to study if biases of convective initiation are consistent when conditional on their synoptic set up.