Data assimilation of a dense wind profiler network and its impact on convective forecasting

Abstract

Two momentum control variable schemes are typically used in most data assimilation systems: stream function and unbalanced velocity potential (ψ/χu scheme) and eastward and northward velocity (U/V scheme). The wind profiler radar (profiler) plays an important role in the expansion of meteorological observations networks. In this study, the impacts of two momentum control variable schemes on assimilating a dense wind profiler network are discussed based on the single profiler station observation tests and six convective rainfall events cycling experiments. Single profiler station observation tests indicate that profiler data assimilation is sensitive to control variables. The dynamical increments using U/V scheme contain more elaborate structure, which contributes to valuing the straightforward effects of profiler observations objectively. By contrast, unrealistic increments occur around the observation station in the experiment using ψ/χu scheme. Six continuous cycling experiments further demonstrate that, for convective rainfall events, experiment using U/V scheme leads to more skillful quantitative precipitation forecasts of convective rainfall. Diagnostic analysis results show that analysis using ψ/χu scheme prevents a good fitting to the profiler data, and accurate dynamical analysis containing abundant small-scale disturbances are the main reason for improving precipitation prediction of convective rainfall when U/V scheme is adopted.