LAUSR

Representing mesoscale convective systems (MCSs) and their multi-scale interaction with the large-scale atmospheric dynamics is still a major challenge in state-of-the-art global numerical weather prediction (NWP) models. This results in potentially defective forecasts of synoptic-scale dynamics in regions of high MCS activity. Here, we quantify this error by comparing simulations performed with a very large-domain, convection-permitting NWP model to two operational global NWP models relying on parameterized convection. We use one month’s worth of daily forecasts over Western Africa and focus on land regions only. The convection-permitting model matches remarkably well the statistics of westward-propagating MCSs compared to observations, while the convection-parameterizing NWP models misrepresent them. The difference in the representation of MCSs in the different models leads to measurably different synoptic-scale forecast evolution as visible in the wind fields at both 850 and 650 hPa, resulting in forecast differences compared to the operational global NWP models. This is quantified by computing the correlation between the differences and the number of MCSs: the larger the number of MCSs, the larger the difference. This fits the expectation from theory on MCS–mean flow interaction. Here, we show that this effect is strong enough to affect daily limited-area forecasts on very large domains.