Satellite bias correction in the regional model ALADIN/CZ: Comparison of different VarBC approaches

Abstract

The bias correction of satellite radiances is an essential component of data assimilation system in numericalweather prediction (NWP). The variational bias correction (VarBC) scheme is widely used by global NWPcenters, but there are still open questions regarding its use in limited-area models (LAMs). We present a studyof key VarBC aspects in the limited-area 3D-Var system using the state-of-the-art NWP system ALADIN. Two basic VarBC applications are tested, specifically adopting bias coefficients from the global modelARPEGE and cycling bias coefficients independently in the LAM ALADIN (VarBC-LAM). The latterapplication is studied using daily update of bias coefficients with regards to static and dynamic settings of theVarBC stiffness. Extensive testing shows that the VarBC-LAM methods outperform the use of global coefficients from ARPEGE providing the better quality of the model first guess (3-h forecast), in the assimilation cycle with the largest normalized impact of 2%-3% for temperature and wind components in themidtroposphere. Compared to the global coefficients, there was little forecast impact between 24 and 48 hfrom using the VarBC-LAM coefficients. The various VarBC-LAM methods were comparable, but the CAMmethod may be most useful when an unexpected bias shows up.