臺北地區2002-2005年溫度預報之分析

陳白榆(Bai-Yu Chen)

指導教授 ： 王國英

國立中央大學/地球科學學院/大氣物理研究所/碩士(2006年)

摘要

為了瞭解大都會地區的天氣預報能力，本研究以中央氣象局對臺北市三年的一週溫度預報為主進行分析，同時也搜集國外其他單位（CNN、JWA、WMO和WN）的預報資料互相作比較。研究中詳細地計算各月預報的技術得分，以及預報誤差的分佈情形。此外，為了進一步探討各季節溫度預報是否有系統性的誤差存在，本研究使用三維空間氣流軌跡模式（IMS-Lagrangian）分析在不同預報狀況下的軌跡來源及其分佈情形，希望藉此瞭解氣流來源和溫度預報能力間是否有相關性。
結果顯示，一天溫度的預報技術得分大部分比persistence高，表示一天預報有技術性。各季節的溫度預報技術得分比較，顯示在冬季的技術得分較高，夏季較低，尤其是6月份最低。且發現愈短時間的預報，最高溫的預報技術較佳；愈長時間的預報，則是最低溫的預報技術較佳。與國外其他單位的預報技術比較，以中央氣象局預報的技術能力最優秀。因為預報的誤差會隨預報時間加長而增大，故誤差頻率分佈圖亦隨時間加長分佈愈廣，且與各季節的氣流來源一致性有相關。
由三維空間氣流軌跡模式計算高低溫氣團的軌跡分佈情形，發現氣團之來源與特性有密切的關係。若實際上的空氣為來自西北方較冷的氣團，但預報時判定空氣來源為非北方且溫暖海洋的氣團，則有高報的現象；若實際上的空氣為來自東方、東南方或南方海洋上的高溫氣團，但預報認為來源為西北方的冷空氣，則會有低報的現象。西北大陸南下的大尺度天氣系統或是滯留的空氣來源，因為氣流特性與變化較易掌握，氣溫的預報也會較準確。
本研究希望藉由中央氣象局對臺北市的溫度預報，瞭解天氣預報的特性。並且使用氣流軌跡模式，探討氣溫預報誤差產生的可能因素，以期能作為預報改進之參考，相信此為另一個研究的新方向。

關鍵字

統計預報

並列摘要

In this analysis, we assess the accuracy of temperature forecasts for lead times of 0 to 7 days, issued daily by the Central Weather Bureau (CWB) for Taipei city during 2002-2005. The main purpose of this study is to understand the ability of weather forecasts for metropolis. Furthermore, we also collected weather predictions of other services in foreign countries for comparison, such as CNN, JWA, WMO and WN etc. The detail of skill score of temperature forecasts each month and the distribution of forecast errors were calculated. Besides, in order to find out if there is any correlation between the source of airstream and the ability of temperature forecasts for further evaluation of the existence of systematic errors in each season, we used a three-dimensional airstream IMS-Lagrangian model to analyze the origin of airstream’s trajectory and its distribution pattern in different cases of forecasts.
The result shows that the skill score of 1day forecasts of CWB are better than persistence at most, which means the 1day forecasts have skill. Comparison of skill score of each season shows that the skill score during winter time is the highest. During summer time the skill score is the lowest, especially in June. We also find that the skill score of maximum temperature forecasts is better for less lead times and the skill score of minimum temperature forecasts is better for longer lead times. The skill of forecasts of CWB is the best when comparing with other countries. Since the mean absolute forecast errors increase with lead time, therefore the range of the distributions of forecast errors are narrower for the early forecast days and broader for the later forecast days. The forecast errors were affected by the consistence of airstream’s origin.
After analyzing the trajectory distributions of higher and lower temperature air parcels, we discover that the origin and characteristic of air parcels are closely related to each other. If air parcels actually came from the colder region in the northwest of Mainland China, but forecasts may consider that its origin is from a warmer region instead, then the temperature will be overestimated; if air parcels actually came from east, southeast or south warmer air, but forecasts may consider that its origin is from the northwest colder region instead, then the temperature will be underestimated. And when the air parcels are rather stationary in the vicinity of Taiwan, or brought along by a southbound large scale weather system from the northwest, the characteristics of which can be easily to know well so the forecasts will be more accurate.

並列關鍵字

IMS-Lagrangian model ； persistence

參考文獻

鄭明典，2001：數值天氣預報近期的發展趨勢。物理雙月刊，23，422- 426。

Hack, J.J., B.A. Boville, B.P. Briegleb, J.T. Kiehl, P.J. Rasch, and D.L. Williamson,

Anthes, R. A. and D. P. Baumhefner, 1984: A diagram depicting forecast skill and

Gilchrist, A., 1986: Long-range forecasting. Q. J. R. Meteor. Soc., 112, 567-592.

1993: Description of the NCAR Community Climate Model (CCM2). NCAR