Improving Prediction of Typhoons Impinging Taiwan Using Global Model Mpas

GPS radio occultation (RO) observations offer abundant useful atmospheric information, which can be proven beneficial to improvement of model forecast for various weathers after assimilated into global models as well as regional models. Anomaly correlation of forecasted 500-mb height at 7th day by ECMWF global model has been remarkably improved by assimilation of the RO observations (mostly from FORMOSAT-3), however the RO impacts on typhoon track predictions by both CWB global model (GFS) and regional model (TWRF) are not as notable as expected. This indicates that we should further investigate how to enhance the RO data impacts on typhoon prediction.MPAS (Model for Prediction Across Scales) is a new-generation global nonhydrostatic model which allows a convenient higher mesoscale resolution for a specific domain of interest. We have employed the 60-15 km variable resolution MPAS to successfully simulate the 7-day track of Typhoon Haiyan (2013) which commences from an earlier spin-up stage to the final dissipation stage. We have also used the MPAS to predict the track and rainfall of Typhoon Morakot (2009) in comparison with WRF simulations with three nested domains (45 km, 15 km, 5 km). The primary task delivered by this project in this year phase will be on assimilation of RO observations into MPAS, through GSI 3DVAR (or hybrid) to assimilate FORMOSAT-3 refractivity and bending angle data, mainly for investigation of the differences in their impacts on typhoon prediction. More plentiful RO data will be available from FORMOSAT-7/COSMIC-2 in 2016 Q3 particularly over global tropical regions, and this timely project of two years (2015 and 2015 year phase) will develop RO data assimilation in global MPAS oriented for improving 60-15 km track prediction and 30-6 km high-resolution rainfall prediction for typhoons impinging Taiwan. The achievements of this project can be provided to related research and operation institutes for analysis and reference.