Blending of global and regional ensembles in cycling hybrid ensemble-variational data assimilation for the convection-permitting prediction of typhoon Merbok (2017)

Abstract

Several blending approaches have been implemented in regional cycling data assimilation to introduce the large-scale weather features of the global model forecast or analysis effectively into regional model forecasts. While in previous studies the blending schemes for data assimilation have generally been applied in stand-alone variational data assimilation systems, here we incorporate the new large-scale ensemble blending schemes into the regional hybrid ensemble-variational (EnVar) data assimilation system. It uses a low-pass Raymond tangent implicit filter to introduce the large-scale background or analysis information from global ensembles. Five parallel cycling hybrid ensemble-variational data assimilation experiments for the convection-permitting prediction of typhoon Merbok in 2017 show that the large-scale ensemble information from Global Ensemble Forecast System (GEFS) can suppress cumulative error growth in data assimilation cycles. On the other hand, the large-scale part of the global ensemble improves the typhoon track and intensity forecasts, possibly due to a better description of the horizontal and vertical structure of the typhoon. In addition, the blending scheme using global ensemble analysis provides more accurate rainfall forecasts of typhoon than the other blending schemes.