總計畫暨子計畫:CReSS-NHOES高解析大氣海洋偶合模式之建置與梅雨防災應用(II)

The purpose of this three-year project is to adopt the Cloud-Resolving Storm simulator (CReSS)-Non-Hydrostatic Ocean model for Earth Simulator (NHOES) coupled model to perform mei-yu QPF, in a hope to further improve the heavy-rainfall QPF in the mei-yu season in Taiwan based upon our current results. In this year (second year), in addition to the implementation of the CReSS-NHOES model for routine forecasts, the current 2.5-km and 5-km CReSS forecasts during 2012-2017 mei-yu seasons are also evaluated and compared to assess the importance of cloud-resolving capability. Also, the daily eight-day forecasts results since 2012 are updated for mei-yu season, and also evaluated and discussed for typhoon QPFs. The implementation of the CReSS-NHOES coupled model has been going smoothly. The model has already been successfully installed and tested on our cluster computer, and we will adopt the JCOPE2 analysis/forecast data from JAMSTEC as the initial and boundary conditions (IC/BCs) for NHOES. The JCOPE2 data can be routinely downloaded automatically, and satisfy our need for the coupled forecasts. We have completed automating the procedure for daily routine forecasts, which will start soon. Some cases in recent seasons are also produced through hindcasts for future evaluation. For the mei-yu season, the 2.5-km operational CReSS (within 3 days) is clearly more capable to produce heavy rainfall and achieve higher threat scores (TSs) across high thresholds when compared to the 5-km model. The updated results show that such an advantage is more evident toward the higher thresholds and in bigger rainfall events. The once daily 2.5-km eight-day forecasts (at 0000 UTC) are also evaluated quite thoroughly for typhoons in 2012-2016 in this study, and this results within 72 h are compared with the three-day forecasts. Overall, the two forecasts are very similar, but the eight-day forecasts, with a larger domain, produced higher overall TS in the first day (0-24 h) at the extreme threshold of 750 mm compared to the three-day forecasts. Beyond the range of 72 h, the quality of the QPF as measured by the Fraction Skill Score (FSS) is negatively correlated to the track error, as expected, but it shown shows a tendency to be better in bigger events. In the appendices, we list the relevant SCI papers and conference papers in this year for reference.