本研究主要是評估同化地基式GPS天頂總延遲（zenith total delay，ZTD）資料對臺灣地區夏季午後
對流之模式定量降水預報的影響。研究結果顯示，同化地基式GPS ZTD可明顯增加分析場中距離臺灣半徑約400公里範圍之水氣，同時可以改善受西南季風影響下，分析場水氣由南向北遞減的現象。
在模式預報方面，同化地基式GPS ZTD可改善模式在臺灣地區及周圍海域偏乾及降水預報不足的現象。在定量降水預報校驗方面，同化地基式GPS ZTD後，0～12小時預報之ETS得分在各個降水門檻都大幅改善，能夠有效提升模式短期定量降水預報能力。

To improve the model quantitative precipitation forecast (QPF), the better model moisture analysis in the model initial condition is the most important. Taiwan, an island with complicate terrain, always exist a great challenge on the model QPF due to the limit atmospheric moisture observation. The ground base global positioning system (GPS) observation is able to provide useful moisture information on nowcasting and data assimilation purpose with the high density both in time and space. This paper is to study the impact of GPS zenith total delay (ZTD) observations on the data assimilation and model QPF performance. A total of 22 afternoon thunderstorm cases were selected to assess the impact of the GPS ZTD data
assimilation the model QPF performance. The results show that the moisture filed in the initial condition is increased around the Taiwan area with radius about 400-km as assimilating the GPS ZTD. In particular, the assimilation of the GPS ZTD results in the increase of the moisture more in southern Taiwan. Due to the moisture was supplied by the southwest monsoon during summertime, therefore, the assimilation of the GPS ZTD does not only increase the amount but also the distribution of the moisture filed in the initial condition in a more reasonable way.
The impact of the GPS ZTD data assimilation shows the improvement of the model QPF does not only over island but also the ocean area in the cases study. The statistics over the 22 afternoon thunderstorm cases show that, assimilating the ZTD GPS can reduce the model spin-up problem significantly. The verification results indicate model QPF performance is also improved as revealed by the equitable threat score in both 06-12 and 30-36 forecasts. This study demonstrates that the GPS ZTD data assimilation is able to provide useful moisture information for the numerical weather prediction application.