台灣擁有複雜地形使得預報對流初始、發展、增強及傳遞是更具有挑戰性，本篇研究選擇2014年8月19日北台灣夏季午後對流個案，分析方法分為兩部分。第一部分為地面測站資料分析，第二部分使用最新研發的四維變分都卜勒雷達資料同化系統(IBM_VDRAS)，其運用沉浸邊界法(Immersed Boundary Method)，因此具有解析地形的能力，並且能同化雷達及地面觀測資料，產出每17.5分鐘更新的完整三維高時空解析分析場，本研究共產生8個分析場來分析。 在對流產生降雨前，地面測站資料可分析出水平輻合帶由山區往平地移動，溫度場顯示對流蒸發冷卻效應，降雨觀測可看出被地形隔離的兩個降雨帶。從IBM_VDRAS分析場可以看出，此降雨事件主要是由兩個獨立對流胞成長開始，其中一個對流胞的外流邊界與另一個對流胞合併，使得後者增強，並且往台北市移動，產生80mm的觀測雨量。從近地表輻合和相對溼度場顯示，外流邊界的合併提供動力輻合增強，以及平流潮濕的環境有利對流發展。之後進行移除地形及地面資料同化變數的敏感度實驗，探討地形及同化變數對於定量降水預報的影響。結果顯示，雪山山脈在此事件中扮演的角色為阻礙外流邊界向南傳遞，而陽明山及林口台地增加外流邊界移動速度，此外，同化地面風場可修正地面風速偏差量及地表輻合帶強度並改善地量降水預報結果。 ;The complex terrain in Taiwan area makes it more challenging to forecast convection initiation, intensification, and propagation. In this research, the heavy rainfall event occurring on 19 August 2014 in northern Taiwan is selected. We use a newly-developed four-dimensional variational Doppler radar assimilation system (IBM_VDRAS), which is capable of simulating the topographic effect by adopting the so-called Immersed Boundary Method, and assimilating radar observations and surface station data. The products of IBM_VDRAS are a series of frequently-updated three-dimensional analysis fields over the complex terrain. In this case study, a total of eight analysis fields times are generated with a temporal interval of 17.5 min over a period of 2.5 h. From the surface observations and the high temporal/spatial resolution analysis fields generated by IBM_VDRAS, it is found that the rainfall process started with the initiation of individual convective cells. The outflow of one of the convective cells merged with another convective system and helped to intensify the latter. The intensified major convective cell then moved into the Taipei metropolitan area and produced 80 mm of heavy precipitation within 2.5 h. The role played by the topographic forcing on the development of the convective system is investigated. A series of experiments are also designed and conducted by moving out terrain or surface assimilated variables to examine the performance of IBM_VDRAS in short-term rainfall forecasts. The result shows that SMR prevents the outflow from propagating southward, and LKHL and MTYM increase the outflow propagation speed. The surface wind assimilation improves the QPF skill by correcting the wind speed bias and controlling the magnitude of low-level convergence.
