none

• Main Authors: Yung-lin Teng, 鄧詠霖
• Other Authors: Yu-chieng Liou
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/51483927715553833458

碩士 === 國立中央大學 === 大氣物理研究所 === 104 === There are high special and temporal characteristics in radar observations. Thus, it is important in weather survillence around the world. Recently, radar observations are also usually used in data assimilation for convective-scale systems. In the research, we attempt to use the Doppler radar data in Taiwan, then acquire the model initial condition to improve the Quantitative Precipitation Forecast (QPF). In this paper, the radar data assimilation mainly involves in the multiple-Doppler radar wind synthesis system, thermodynamic retrieval method, and temperature/moisture adjustment. A real case of 1200 UTC on 14 June 2008 during Southwest Monsoon Experiment (SoWMEX) is selected. Observational data from 5 radars are utilized, including S-POL (operated by NCAR), RCWF, RCCG and RCKT (operated by the Central Weather Bureau of Taiwan), and National Central University C-POL. The ECMWF reanalysis data, radiosondes, and surface mesonet stations are prepared to simulate a background field, and they are instantaneously applied to update the three-dimentional wind field, thermodynamic fields (pressure and temperature), and moisture fields. Therefore, we can regard them as initial fields for 6-hour model forecast. There are two tests in our experiments. The one is trying to classify the missing value of radar reflectivity. In this way, we can increase the data usage of radar observations. The other is using the conditional instability to modulate the saturation thresholds. It allows the model to forecast the precipitation much accuratly over the southwestern area in Taiwan. Similarly, another real case of 0600 UTC on 16 June 2008 is selected to use this modified saturation thresholds. The results show that the strategy in temperature/moisture adjustment, to a certain extent, improve the model QPF skill. To conclude, the above-mentioned experimental results imply that the model QPF skill in the convective-scale can be significantly improved about 2-3 hours using fewer radar data in these real case studies.