Abstract/Summary

Mesoscale convective systems (MCS) are difficult to forecast due to their inherent unpredictability and small scale. Here the impact of model resolution on MCS structure and downstream forecast evolution are determined. Four case analyses of European MCSs using output from a 12-km grid-spacing model reveal a consistent potential vorticity (PV) structure with a positive PV anomaly in the mid troposphere (5 PVU) and negative PV anomalies above and to either side of it (-1 PVU). Convection-permitting models produce stronger MCS PV anomalies than convection-parametrizing models in a case study from July 2012. These differences persist after coarse graining to 100-km grid spacing and are largest in the upper troposphere. The effect of poor representation of MCSs on forecasts is investigated by adding MCS perturbations, calculated as differences from coarse-grained convection-permitting (4.4- km grid spacing) model output, to convection-parametrizing (25-km grid spacing) deterministic and six-member ensemble (with operational initial condition perturbations) model forecasts. Upper-level MCS perturbations have more impact than those at middle levels, though using all levels yields the greatest impact. For the first 30 hours differences grow on the convective scale related to the MCS and a developing UK cyclone, despite perturbation damping. Subsequently, differences grow rapidly onto the synoptic-scale and by five days impact the entire northern hemisphere. The MCS perturbations systematically affect the ensemble forecasts though the differences are smaller than those generated by the initial condition perturbations. MCS perturbations slow the eastward movement of Rossby waves due to ridge amplification. A downstream cyclone deepens by up to 3 hPa after five days and forecast errors (compared to analysis) in mean-sealevel pressure are reduced in most members (by up to 2%). Thus, perturbing convection-parametrizing models to include PV anomalies associated with MCSs produces alternative realisations to those generated by initial condition perturbations and so could be useful operationally.