Based on the GRAPES（Global/Regional Assimilation and Prediction System） regional ensemble prediction system and 3DVAR（three-dimensional variational） data assimilation system,which are implemented operationally at ...Based on the GRAPES（Global/Regional Assimilation and Prediction System） regional ensemble prediction system and 3DVAR（three-dimensional variational） data assimilation system,which are implemented operationally at the Numerical Weather Prediction Center of the China Meteorological Administration,an ensemble-based 3DVAR（En-3DVAR） hybrid data assimilation system for GRAPES-Meso（the regional mesoscale numerical prediction system of GRAPES） was developed by using the extended control variable technique to implement a hybrid background error covariance that combines the climatological covariance and ensemble-estimated covariance.Considering the problems of the ensemble-based data assimilation part of the system,including the reduction in the degree of geostrophic balance between variables,and the non-smooth analysis increment and its obviously smaller size compared with the 3DVAR data assimilation,corresponding measures were taken to optimize and ameliorate the system.Accordingly,a single pressure observation ensemble-based data assimilation experiment was conducted to ensure that the ensemble-based data assimilation part of the system is correct and reasonable.A number of localization-scale sensitivity tests of the ensemble-based data assimilation were also conducted to determine the most appropriate localization scale.Then,a number of hybrid data assimilation experiments were carried out.The results showed that it was most appropriate to set the weight factor of the ensemble-estimated covariance in the experiments to be 0.8.Compared with the 3DVAR data assimilation,the geopotential height forecast of the hybrid data assimilation experiments improved very little,but the wind forecast improved slightly at each forecast time,especially over 300 hPa.Overall,the hybrid data assimilation demonstrates some advantages over the3 DVAR data assimilation.展开更多
基金Supported by the National Natural Science Foundation of China(91437113 and 41275111)China Meteorological Administration Special Public Welfare Research Fund(GYHY201506005)
文摘Based on the GRAPES（Global/Regional Assimilation and Prediction System） regional ensemble prediction system and 3DVAR（three-dimensional variational） data assimilation system,which are implemented operationally at the Numerical Weather Prediction Center of the China Meteorological Administration,an ensemble-based 3DVAR（En-3DVAR） hybrid data assimilation system for GRAPES-Meso（the regional mesoscale numerical prediction system of GRAPES） was developed by using the extended control variable technique to implement a hybrid background error covariance that combines the climatological covariance and ensemble-estimated covariance.Considering the problems of the ensemble-based data assimilation part of the system,including the reduction in the degree of geostrophic balance between variables,and the non-smooth analysis increment and its obviously smaller size compared with the 3DVAR data assimilation,corresponding measures were taken to optimize and ameliorate the system.Accordingly,a single pressure observation ensemble-based data assimilation experiment was conducted to ensure that the ensemble-based data assimilation part of the system is correct and reasonable.A number of localization-scale sensitivity tests of the ensemble-based data assimilation were also conducted to determine the most appropriate localization scale.Then,a number of hybrid data assimilation experiments were carried out.The results showed that it was most appropriate to set the weight factor of the ensemble-estimated covariance in the experiments to be 0.8.Compared with the 3DVAR data assimilation,the geopotential height forecast of the hybrid data assimilation experiments improved very little,but the wind forecast improved slightly at each forecast time,especially over 300 hPa.Overall,the hybrid data assimilation demonstrates some advantages over the3 DVAR data assimilation.
