A Three-Dimensional Variational Data Assimilation System for MM5: Implementation and Initial Results

Abstract

A limited-area three-dimensional variational data assimilation (3DVAR) system applicable to both synoptic and mesoscale numerical weather prediction is described. The system is designed for use in time-critical real- time applications and is freely available to the data assimilation community for general research. The unique features of this implementation of 3DVAR include (a) an analysis space represented by recursive filters and truncated eigenmodes of the background error covariance matrix, (b) the inclusion of a cyclostrophic term in 3DVAR's explicit mass?wind balance equation, and (c) the use of the software architecture of the Weather Research and Forecast (WRF) model to permit efficient performance on distributed-memory platforms. The 3DVAR system is applied to a multiresolution, nested-domain forecast system. Resolution and seasonal- dependent background error statistics are presented. A typhoon bogusing case study is performed to illustrate the 3DVAR response to a single surface pressure observation and its subsequent impact on numerical forecasts of the fifth-generation Pennsylvania State University?National Center for Atmospheric Research Mesoscale Model (MM5). Results are also presented from an initial real-time MM5-based application of 3DVAR.