Direct Radar Reflectivity Assimilation within MPAS-JEDI using Reflectivity Analysis Variable and Multivariate Background Error Covariance

Abstract. This study implements direct reflectivity assimilation within the Joint Effort for Data assimilation Integration (JEDI) with the Model for Prediction Across Scales–Atmosphere (MPAS-A) (i.e., MPAS-JEDI) and evaluates its performance for hourly cycled radar assimilation in heavy rainfall forecasts over the deep tropics. Radar reflectivity observations are directly assimilated using the hybrid 3DEnVar method, in which reflectivity is treated as an analysis variable. Multivariate correlations between reflectivity and temperature, humidity, and hydrometeors are incorporated into the static component of the background error covariance (BEC), allowing reflectivity information to propagate to the model state variables. In addition, reflectivity states are updated from the analyzed hydrometeors across successive outer loops, seeking improved consistency between reflectivity and hydrometeor fields and a better fit to the reflectivity observations. Diagnosis of the multivariate BEC reveals physically consistent cross-variable correlations among thermodynamic, microphysical, and reflectivity fields. Single observation assimilation tests demonstrate that direct reflectivity assimilation effectively propagates reflectivity increments to both hydrometeor and thermodynamic variables. Results from a Sumatra squall line case indicate that updating reflectivity from analyzed hydrometeors across successive outer loops produces a closer fit to observed reflectivity and improves the forecast accuracy of the squall-line system. Furthermore, the hybrid multivariate BEC outperforms the ensemble-based BEC in reflectivity assimilation by substantially improving the analyses of dynamical and microphysical states, leading to better predictions of squall-line intensity, orientation, and propagation. The multi-case quantitative evaluation further demonstrates the superiority of hybrid multivariate BEC over the ensemble-based BEC in improving both composite reflectivity and 3-h accumulated precipitation forecasts over the Singapore region. Overall, the successful implementation of direct reflectivity assimilation in MPAS-JEDI highlights the added value of incorporating a multivariate BEC for improving heavy rainfall prediction in the deep tropics.