Abstract
The authors demonstrate that an appropriate treatment of convective entrainment is essential for determining spatial distributions of and temporal variations in precipitation. Four numerical experiments are performed using atmospheric models with different entrainment characteristics: a control experiment (Ctl), a no-entrainment experiment (NoEnt), an original Arakawa–Schubert experiment (AS), and an AS experiment with a simple empirical suppression of convection depending on cloud-layer humidity (ASRH). The fractional entrainment rates of AS and ASRH are constant for each cloud type and are very small in the lower troposphere compared with those in the Ctl, in which half of the buoyancy-generated energy is consumed by entrainment. Spatial and temporal variations in the observed precipitation are satisfactorily reproduced in the Ctl, but their amplitudes are underestimated with a so-called double intertropical convergence zone bias in the NoEnt and AS. The spatial variation is larger in the Ctl because convection is more active over humid ascending regions and more suppressed over dry subsidence regions. Feedback processes involving convection, the large-scale circulation, free tropospheric moistening by congestus, and radiation enhance the variations. The temporal evolution of precipitation events is also more realistic in the Ctl, because congestus moistens the midtroposphere, and large precipitation events occur once sufficient moisture is available. The large entrainment in the lower troposphere, increasing free tropospheric moistening by congestus and enhancing the coupling of convection to free tropospheric humidity, is suggested to be important for the realistic spatial and temporal variations.
