Impact of Background Error Statistics on Variational Assimilation Systems

Assimilation of high-resolution Doppler Weather Radar (DWR) data is a powerful technique for forecasting the evolution of atmospheric conditions and provide critical information in the case of severe weather. The quality of assimilation in a variational assimilation system is dictated by the background error (BE) which provides information about the location and pattern of convection by spreading DWR observations in space and filtering analysis increments. Traditionally, BE is determined for a limited set of control variables (CV) which are calculated from model variables. Hence, the choice of control variables in modelling BE has a major influence on the atmospheric state variables represented in the analysis. In the present study, we have analysed the impact of different unconventional momentum control variables on BE statistics in a limited area variational assimilation systems. We have constructed BE statistics with two pairs of control variables namely stream function velocity potential (ψχ) and horizontal wind components (uv). Five experiments are carried out using WRF 10 km model namely, Control (without data assimilation), 3DVAR-uv, 3DVAR-ψχ, 4DVAR-uv, 4DVAR-ψχ using uv and ψχ BEs in three dimensional and four-dimensional assimilation systems respectively. The short range forecast results are presented for an intense convective event which occurred in Chennai, India in year 2015. Comparative analysis with radiosonde and synoptic station observations highlight a significant reduction in RMSE during the use of uv control variables in both 3dvar and 4dvar assimilation system. Results indicate an improved low-level wind (850 hPa) magnitude and convergence of vertically integrated moisture transport in 3dvar-uv experiment resulting in a spatial precipitation pattern similar to observed GPM accumulated precipitation. Statistical verification of precipitation forecasts suggest that uv experiments in 3dvar and 4dvar showed 80% and 50% comparative increase in average hit rate and a significant increase in mean Forecast Skill Scores.