The Breeding of Growing Modes with Dynamic Rescaling in Ensemble Prediction and Improvement on Free Breeding
集合预报中动态调整的增长模繁殖及其对自由繁殖的改进

To the ensemble prediction of initial conditions, it is very important how to generate initial perturbation.Of all the methods to create perturbations for ensemble forecast,the "breeding of growing modes"(BGM) method has obtained more and more favor for its good performance and almost "free" cost of computation.The breeding method simulates the development of growing errors in the analysis cycle.When a breeding cycle is started,an arbitrary initial perturbation field is added upon the control analysis.After some time of breeding,in some sense, the perturbation in the breeding cycle reaches "saturation".This saturated status of perturbation is regarded as the estimate of the fast-growing modes in the realistic analysis error,and then is applied as the initial perturbation of ensemble prediction.Using the global spectral model T63L9 and NCEP re-analysis data,the experiments with "free breeding" have been designed(Yu and Zhang, 2005) to view the evolving of bred modes in the breeding cycles.Results have shown that the breeding of growing modes has a clear saturation character in both the norm and the form after 34-day breeding.However,the growth rate and the form evolving of the perturbations are associated with the size of the initial modes.And more importantly,the growing modes in the breeding cycles need to be rescaled timely to accord with the size of initial analysis error.In this paper,the experiments with "dynamic rescaling breeding" are designed to view what effects will be made on the breeding cycles.The method of "dynamic rescaling" used here is defined as: when an integration cycle ends,the output modes(perturbation) are obtained.The cathors will scale down the perturbation before it is inputted to the next cycle.The purpose of rescaling is just to make the magnitude of growing modes comparable with the initial analysis error in RMSE(root-mean-square error) sense.Results show that the "dynamic rescaling breeding" can make a larger growing rate and shorter saturated time.To the form saturation of growing modes,the dynamic rescaling breeding can to a great extent overcome the lagging of reaching saturation at low levels.Besides,the saturation character appears to be more obvious by using the dynamic rescaling,To dynamic rescaling breeding,the size of initial modes must be appropriate,which is the same as free breeding.Conclusions obtained from our experiments show that,to the T63L9 model,if the dynamic rescaling breeding is adopted,the initial size of geopotential height and wind perturbations can be respectively chosen as 1015 gpm and 1.22 m/s at 500 hPa,and larger sizes are chosen at high levels.To the relative humidity perturbation,about 5% can be chosen below 500 hPa.These results also have important references to other models.