基于集合变换方法的华北冬季降雪过程的高分辨率集合预报试验比较研究OA

Improving initial perturbation techniques for convection-scale ensemble forecasting is essential for ad-vancing operational numerical weather prediction in China.To explore effective approaches for constructing initial perturbations in high-resolution ensemble systems,this study implements the ensemble transform(ET)method within the China Meteorological Administration Mesoscale Model(CMA-MESO),a regional numerical prediction system independently developed by the China Meteorological Administration.A series of continuous ex-periments was conducted for winter high-impact snowfall events over North China,and the results were compared with those obtained using the multi-scale blending(MSB)perturbation method.The results reveal distinct physical characteristics of convection-scale perturbation fields generated by the two methods.Analyses of error growth indicate that both schemes effectively capture multiscale perturbation information,with reasonable struc-tures and amplitudes of dominant modes,demonstrating the overall reliability of the ensemble configurations.The perturbation fields exhibit enhanced kinetic energy at meso-and small-scales,and the spectral energy of upper-level perturbations increases with forecast lead time.In both schemes,total perturbation energy first increases and then decreases from the lower to the upper troposphere.The ET method produces higher initial perturbation energy than the MSB method,while the two methods show comparable energy levels at later forecast times.This differ-ence is likely attributable to the ET method's ability to incorporate more localized small-scale perturbations,there-by enhancing the perturbation energy spectrum and improving the representation of the initial dynamical field for short-range ensemble forecasts.However,the lack of spatial filtering in the ET method introduces additional noise,leading to excessive initial perturbation energy,which warrants further refinement.Statistical verification demon-strates that the ET method generally outperforms the MSB method in forecasting meteorological variables at mul-tiple pressure levels.Within a 24 h forecast period,the ET scheme yields lower continuous ranked probability scores(CRPS)and reduced forecast errors for geopotential height and temperature in the middle and upper tropo-sphere as well as for surface variables,indicating improved probabilistic forecast skill.Unlike previous studies fo-cused primarily on warm-season precipitation,this work specifically evaluates model performance for winter snowfall events.Synoptic verification shows that both schemes perform comparably in predicting snowfall distribu-tion and intensity over North China.Notably,the ET method provides more reliable forecasts of surface tempera-ture and wind at mountainous stations,highlighting its advantages in complex terrain.Additionally,the ET method demonstrates higher computational efficiency,supporting its potential application in rapid-update cycling and stable operation within future high-resolution operational ensemble forecasting systems.