Research On Simulation Of South China Precipitation With Different Cumulus Parameterization Schemes

Since cumulus parameterization is the key problem to enhance the performanceof numerical model, this paper is to deeply analyze the feature of schemes in modelGRAPES-Meso, their ability to present convection and their compatibility with themodel through simulation which is based on detailed analysis of a continuous rainprocess lasting from June 17th, 2005 to June 24th in south China.The occurrence of this rain process is the outcome of both large scale circulationand some meso-scale weather systems: strong southwest monsoon in Bay of Bengal,Indo-China and South China Sea bring consistent water vapor to Guangdong;maintenance of 850hpa shear line between southern part of South yanze and northernSouth China; existence of 500hpa southern trough over west Guan.gdong andGuangxi for a long period and location of surface shear line matched torrential rainarea very well.In the 24 hours rain simulation of June 21st using the three schemes, the rain beltis very close to observation, but the simulated meso-scale torrential centre differ a lotwith each parameterization scheme. Initial provocation status and time change ofconvection in three schemes are quit different and relate closely to their 24 hoursprecipitation simulations. Convection provocation time is not the same as weli withthree schemes. Generally speaking, SAS scheme predicts the convections, better, butthe shortcoming is the triggered rain mount is nearly the same while in theobservation rain mount differs a lot. In the simulation with KFETA scheme, thevertical velocity in the centre of troposphere tends to be smaller than normal is onereason for the less predicted rain.Computation of variables and parameters in schemes should be adjustedaccording to the numerical model and resolution. Vertical speed, relax time andprecipitation efficience in KFETA, BMJ and SAS schemes have strong influence on simulation results, so change these parameters can enhance model performance.Plant a new parameterization scheme——Grell-Devenyi scheme in GRAPESand use it to perform a 24 hours rain simulation. The result shows that this veryscheme that combine ensemble and data assimilation techniques has better forecastwhich is close to observation.