In this study, we investigate the role of convective parameterization (CP) in simulations of a convection band over the mid-Korean Peninsula at grey-zone resolutions, at which convection is partially resolved, partially subgrid. An approach similar to that used in ‘observing system simulation experiment’ is adopted. Simulations with a 500-m grid size serve as benchmark simulations. The impacts of resolution and convective parameterization at grey-zone resolutions (i.e. 3, 6 and 9 km) are then investigated. Results indicate that a grid size of 3 km is sufficient to resolve the convection band and CP for this size grid is not necessary. With 6 and 9 km grids, explicit simulations or those based on a Kain–Fritsch CP scheme do not simulate the atmospheric structure surrounding the band accurately.

A major problem with CP is excessive triggering of parameterized convection. False triggers of CP in the band adjacent area suppress evolution of the resolved convection band through excessive stabilization of inflow air. We obtain significant improvements by using a modified trigger function, resulting in reduction of the area of parameterized convection, which in turn leads to stronger development of a resolved convection band. Furthermore, our approach reduces bias in the domain-averaged vertical thermodynamic structure.