The impact of surface heterogeneity on the diurnal cycle of deep convection

Despite some recent improvements, major deficiencies remain in model simulations using parameterised convection in capturing both the phase and amplitude of the daily cycle of precipitation in tropical regions. The difficulties are particularly acute in regions of heterogeneous surface conditions, since the simulations need not only to respond appropriately to the local forcing from surface fluxes but also to capture the interactions with near-surface mesoscale circulations. Here, we examine such a situation by means of idealised cloud-resolving simulations of deep convection over a heterogeneous surface, performed using the cloud-resolving simulation known as the Met Office NERC Cloud model (MONC). In these simulations, we show that precipitation forms preferentially over dry and warm patches ("DRY") compared with wet and cold patches ("WET"), with the peak precipitation rates differing by a factor of approximately 4. The initiation of precipitation occurs approximately 1.5 hr earlier in the DRY patches compared with the WET patches. Moreover, within the WET and DRY patches there are marked differences in the spatial distribution of the precipitation. These cloud-resolving simulations are then used as a benchmark to assess the behaviour of simulations using parameterised convection, performed using the idealised configuration of the Met Office Unified Model (MetUM). The MetUM simulations produce a response with some qualitative similarities to the cloud-resolving simulations. In particular, although the simulations with parameterised convection initiate precipitation too early, they are capable of capturing the relative amounts of daily-mean precipitation in the DRY and WET patches. We propose that the cloud-resolving simulations could be used further to investigate the impact of fully interactive surface schemes and as benchmark simulations to evaluate new parameterisation schemes.