Application of a Bayesian inflation approach to EnSRF radar data assimilation to improve the analysis and forecasting of an MCS

Abstract

Abstract A spatially and temporally varying Bayesian inflation algorithm was applied to the assimilation of the real radar data of a mesoscale convective system (MCS), which occurred on May 8–9, 2007 in Oklahoma and Texas (USA), using the ensemble square root filter (EnSRF) within the advanced regional prediction system (ARPS). For comparison purposes, two radar data assimilation and forecast experiments using Bayesian inflation and regular multiplicative inflation algorithms were performed. Results indicated that the analysis reflectivity structure of the MCS using the Bayesian inflation method appeared better than when using the multiplicative inflation method. The Bayesian inflation method also reduced the root mean square innovations (RMSIs) and increased the ensemble spreads for reflectivity and radial velocity. Furthermore, the reduction of RMSIs was related to the higher Bayesian inflation parameter assigned to the background. The flow‐dependent characteristic of the Bayesian inflation parameter means it varies with space and model state variables. This successful analysis helped improve subsequent forecast reflectivity and precipitation, especially for stratiform precipitation. It also promoted the development of the MCS with a strong surface cold pool and outflow in the convective regions. Quantitative reflectivity and precipitation forecast skills were also improved in the Bayesian inflation experiment. These encouraging results demonstrated the greater potential of using the Bayesian inflation algorithm for EnSRF radar data assimilation compared with multiplicative inflation.

Keywords

• Bayesian inflation algorithm
• ensemble square root filter (EnSRF)
• mesoscale convective system (MCS)
• multiplicative inflation algorithm
• radar data assimilation