Abstract:When one makes short-range regional ensemble forecast by using the Breeding of Growing Modes (BGM) method,a critical problem first to be faced is what the typical characteristics of evolution of initial perturbations are in the short-range ensemble prediction system (EPS).Consequently,a short-range EPS based on the BGM method has been developed with WRF3.6.The regular rescaling scheme has also been incorporated into this system.Meanwhile,the short-range EPS that has covered the uncertainties of horizontal wind,vertical velocity,potential temperature,geopotential height and water vapor mixture ratio takes the large-range rainstorm in southern China in June 2016 as an example to recognize the evolving mechanism of perturbations.The results show that:(1)the perturbation growing process of physical quantities in the upper,middle and lower levels of model atmosphere can be divided into two stages,one of which is the rapid linear growth of perturbations and the perturbations quickly complete the total increase of themselves in this phase,another of which is the nonlinear stable phase of perturbations growing and the transition from the fast linearly growing phase to the nonlinear stable phase takes about 24 h.(2)The perturbations of physical quantities take approximately the same length of time to enter the nonlinear stable phase through the temporal evolution features of perturbation growth rate,correlation coefficient and perturbation growing modes.Nonetheless,when the perturbations come into the nonlinear stable stage,the numerical values and evolving characteristics of each assessment parameter are different for the same pressure level with different physical quantities or the same physical quantity at different pressure levels.Moreover,there is a diurnal oscillation phenomenon with time for each assessment parameter at the nonlinear stable stage.(3)For the initial ensemble of different random patterns with the same size of perturbation amplitude,the impacts of different random patterns on perturbations breeding mainly yield differences in the nonlinear stable stage while the differences between each pattern are too small to distinguish in the fast linearly growing stage.(4)For the initial ensemble of the same random pattern but with different sizes of perturbation amplitudes,the influences of different amplitudes on the evolution of perturbations mainly occurs in the fast linear growth phase,while the differences between each amplitude are quite small in the nonlinear stable phase.Additionally,the different sizes of initial amplitudes have no influence on the characteristic time scale of the perturbations getting into the nonlinear stable phase.