Response of moist convection to multi‐scale surface flux heterogeneity

We investigate the response of moist convection to the spatial variation of surface sensible heat flux (SHF) in a mesoscale domain during the evolution of the afternoon convective boundary layer (CBL), using large‐eddy simulation. The surface SHF heterogeneity in the domain is analytically created as a function of the spectral slope in the wavelength range from a few tens of kilometres to a few hundreds of metres in the SHF spectrum on a log–log scale. Assuming surface energy balance and spatially uniform available energy, the prescribed SHF has a phase lag of 180° with respect to the latent heat flux (LHF) in the domain. Two sets of three simulations are forced by heterogeneous surface SHF fields, which are characterized by similar statistics. One set, however, is created with a spectral slope of k−3 (where k is wave number) and the other with a slope of k−2. All of the simulations are integrated with the same observation‐based initial sounding favourable for moist convection. In all of the k−3‐slope cases, early non‐precipitating shallow clouds further develop into deep thunderstorms. But in all of the k−2‐slope cases, only shallow clouds develop. A key process in the transition to deep convection is the formation of a mesoscale pool of cool and moist air just above the top of the CBL. This high relative humidity (RH) pool is formed by repeated deep penetrations of turbulent plumes into the free atmosphere over a mesoscale surface of high SHF. These cross‐scale fluxes, the vertical transports of mesoscale moisture and heat fluctuations by turbulent updraughts, are critical for the formation of the mesoscale pool of high RH. However, these cross‐scale fluxes are cancelled out in the process of averaging, and thus appear negligible in the vertical profiles of domain‐averaged moisture and heat fluxes.