LAUSR

We analyse how the representation of deep convection affects the characteristics of convective precipitation over tropical Africa in global storm‐resolving simulations with the European Centre for Medium‐Range Weather Forecasts (ECMWF) Integrated Forecasting System (IFS). The simulations consist in 48 hr forecasts initiated daily at 0000 UTC for a 40‐day period in August and September 2016, at ∼ 9 and 4 km resolution. We find that results agree best with the satellite retrieval Global Precipitation Measurement (GPM) Integrated Multi‐satellitE Retrievals for GPM (IMERG) when a term for the integrated and scaled total advective moisture tendency is added to the convective instability closure. This allows for a better coupling of the convective parametrization to the mesoscale dynamics and improves precipitation intensity (increased), precipitating area fraction (decreased), the size of individual deep convective systems (decreased) and their propagation (westward), even though the duration of rain events remains underestimated. With explicit deep convection, the duration of rain events is more realistic, but precipitation intensity is significantly overestimated, and extreme rain events (>400 mm · day −1 ) are three times more likely than in GPM IMERG, causing biases in the organization of deep convection. In particular, individual convective systems are too small, hindering the formation of mesoscale convective systems. These findings are also valid for two other regions: the Maritime Continent and the contiguous United States. Our results suggest that deep convection is not completely resolved at a resolution of 9 or even 4 km, and that some scale‐aware parametrization is still needed for global numerical weather prediction.