2021年6月4日臺北午後對流個案之系集預報研究

Ensemble Prediction of the Afternoon Thunderstorm System at Taipei on 4 June 2021

余世暘(Shih-Yang Yu)

指導教授 ： 楊明仁

國立臺灣大學/理學院/大氣科學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202201984

摘要

短延時強降雨為使臺北都會區淹水的主要因子之一。2021年6月4日的梅雨鋒前午後對流事件，於臺北盆地東南側降下破百毫米的時雨量，使都市排水系統宣洩不及，造成多處淹水。本研究透過系集預報模式進行分析，欲了解不同模式設定對於事件中各個重要物理過程的影響。系集預報系統由兩種初始邊界條件、四種積雲參數法、以及四種微物理參數法組合成32個系集成員，欲探討的對象則由綜觀環境至中尺度過程，最後探討定量降雨預報結果。
在綜觀環境方面，北側海面的梅雨鋒面非常接近臺灣，因此本個案並非弱綜觀環境下的午後對流事件，雖然觀測資料顯示當天臺北盆地的風場仍由局部環流主導，但梅雨鋒面於模式中的南移速度可能影響降雨型態。研究結果顯示，地面梅雨鋒面位置主要為初始場主導，ECMWF系集成員的鋒面移速偏慢，而NCEP成員的移速偏快，臺灣北部陸地周邊的水氣、溫度受鋒面位置的差異影響，使對流發展時的ECMWF成員環境較為暖濕、NCEP成員較為冷乾。
中尺度的物理過程除了探討日間輻射加熱之外，亦參考Jou (1994)及Miao and Yang (2020)的研究，著重分析日間淡水河海風帶來的水氣以及對流肇始後的冷池行為。日間短波輻射加熱對於盆地上空的高雲雲量敏感，因此主要由微物理參數法主導，高雲含量過多的Morrison參數法使該組系集成員的加熱時序較其他成員晚約一個小時，此系集差異向下傳遞至由海陸加熱對比造成的海風，Morrison參數法系集成員中的海風亦較其他系集成員延遲約一個小時。冷池方面，除了Morrison參數法的整體時間延遲之外，初始場所帶來的環境差異亦顯著影響。ECMWF成員的環境暖濕，有利於冷池發展，因此冷池於對流前緣沿著雪山山脈下山，並於臺北盆地內部與海風輻合，激發顯著的對流訊號；相對的，NCEP成員的環境冷乾，不利冷池發展，其訊號僅在山區強降雨區出現，因此對流胞發展侷限於雪山山脈上。
系集降雨預報的校驗則透過Fractions Skill Score (FSS; Roberts and Lean 2008) 和Method for Object-Based Diagnostic Evaluation (MODE; Davis et al. 2006) 兩種方法進行分析。FSS校驗結果顯示，雖然模式的水平解析度高達1公里，其能有效給予的降雨資訊僅介於1公里至21公里，為對流在空間上的不確定性所造成。FSS的三種衍生指數 (dFSS, eFSS, 及LFSS) 指出：系集在1公里解析度時的降雨區域發散程度不足，因大部分雨量都集中於雪山山脈山區，此現象亦透過MODE校驗可見，模式降雨中心由臺北盆地向東南側的山區偏離2至27公里。
最後由層次聚類法 (Hierarchical Clustering) 分析32組降雨結果的相關性，降雨結果主要由初始場主導，而積雲參數法及微物理參數法則在此框架下提供更高的多樣性，此分析結果與前述的物理過程相符。

關鍵字

臺北午後對流 ； 系集預報 ； 淡水河谷海風 ； 冷池 ； 降雨校驗

並列摘要

The short-duration high-intensity precipitation is one of the factors that can cause the inundation in the Taipei urban area. On 4 June 2021, an afternoon thunderstorm happened, and the rainfall intensity exceeded 100 mm/hour in the southeast corner of the city, causing severe flooding in the urban area. In this study, an ensemble prediction was conducted to figure out the factors in the model that affected the physical processes in this event. The ensemble members were established by the variations of two initial conditions, four cumulus parameterizations, and four microphysics schemes. The physical processes analyzed included the synoptic environment, mesoscale processes, and the resulting precipitation.
For the synoptic conditions, the Mei-Yu front was about 50 km to 100 km offshore of northern Taiwan, so this event cannot be considered “weak synoptic.” Instead, the Mei-Yu front in the numerical model might influence the event even though the local circulation dominated the area based on the observation. Results showed that the initial conditions mainly determined the location of the surface Mei-Yu front in the model. The fronts in the ECMWF members moved southward slower, while those in the NCEP members moved faster. Therefore, the surface environment near northern Taiwan differed between these two groups. The ECMWF members tended to be warmer and wetter, but the NCEP members were cooler and drier during the development of the thunderstorms.
According to the previous studies (Jou 1994; Miao and Yang 2020), the main focuses of the mesoscale processes included the solar heating in the morning, the sea breeze along the Tamsui River Valley (TRV), and the thunderstorm cold pool. The solar heating in the morning was sensitive to the high cloud pattern. Therefore, the microphysics schemes mainly dominated this process. The Morrison scheme tended to produce too many upper-level hydrometeors, so the heating timing was about one hour later than other schemes. This ensemble spread would be transported to the following physical processes, such as the sea breeze induced by the land-sea heating contrast and the thunderstorm initiation time. The detection time of the resulting thunderstorm cold pool was also delayed in the Morrison members. However, in addition to the microphysics schemes, the diversities of initial conditions influenced the characteristics of the thunderstorm cold pool. The synoptic environment in ECMWF members was warmer and wetter, which was favorable for developing cold pools. The cold pool in these members propagated downslope from the Snow Mountain Range (SMR) to the Taipei Basin, converging with the sea breeze and causing heavy rainfall in the plain area. In contrast, the synoptic environment of NCEP members was cooler and drier, which inhibited the intensity of the cold pool and was unfavorable for triggering new convection in the basin. The resulting rainfall was therefore concentrated in the mountainous area.
The verification of the ensemble quantitative precipitation forecast (QPF) was evaluated by the fractions skill score (FSS; Roberts and Lean 2008) and the method for object-based diagnostic evaluation (MODE; Davis et al. 2006). The results of FSS showed that although the finest horizontal resolution was up to 1 km, the informative spatial scale could range from 1 km to 21 km, which was brought about by the intrinsic spatial uncertainty of the afternoon thunderstorm in the numerical model. The three FSS derivatives (dFSS, eFSS, and LFSS) indicated that the spatial distribution of the precipitation area was underspread under the highest resolution (i.e., 1 km). The rainfall areas were mainly concentrated on the SMR, and the MODE method also pointed out this systematic bias with the distance from 2 km to 27 km.
Last, the hierarchical clustering technique was applied to the 6-hour precipitation between 12 LST and 18 LST to understand the relationship between the ensemble members. The initial conditions mainly dominated the rainfall pattern, while the cumulus schemes and the microphysics parameterizations further added the diversities based on this background. These results were consistent with what was found in the physical processes analyses.

並列關鍵字

Taipei afternoon thunderstorm ； ensemble prediction ； Tamsui River Valley sea breeze ； thunderstorm cold pool ； QPF verification

參考文獻

陳白榆，李志昕，葉天降，2018：物件基礎診斷校驗技術於中央氣象局區域模式定量降水預報能力之初步評估結果。中央氣象局107年天氣分析與預報研討會，A2-8。

Davis, C. A., B. G. Brown, and R. G. Bullock, 2006: Object-based verification of precipitation forecasts. Part I: Methodology and application to mesoscale rain areas. Mon. Wea. Rev., 134, 1772–1784.

Dey, S. R. A., G. Leoncini, N. M. Roberts, R. S. Plant, and S. Migliorini, 2014: A spatial view of ensemble spread in convection permitting ensembles. Mon. Wea. Rev., 142, 4091–4107.

Epstein, E.S., 1969: Stochastic dynamic prediction. Tellus, 21, 739–759.

Ferrett, S., T. H. A. Frame, J. Methven, C. E. Holloway, S. Webster, T. H. M. Stein, and C. Cafaro, 2021: Evaluating convection-permitting ensemble forecasts of precipitation over southeast Asia. Wea. Forecasting, 36, 1199–1217.