運用大尺度西南氣流指數法預報臺灣梅雨季極端降雨頻率年際變化

A downscaling Method for Predicting Taiwan Mei-yu Seasonal Extreme Rainfall Event Frequency Using a Large-Scale Southwest Flow Index

卓盈旻(Yin-Min Cho) ； 盧孟明(Mong-Ming Lu)

《大氣科學》 Volume 45 Issue 2 (2017/06) Pp. 83-99

https://doi.org/10.3966/025400022017064502001

Abstracts

梅雨季極端降雨事件是臺灣的重要災害天氣，然而事件發生趨勢的季度預報研究卻尚未見文獻探討。本文提出大尺度環流指數降尺度預報法，以中央氣象局月與季短期氣候預報系統(TCWB2T2)產品應用為例，說明臺灣梅雨季極端降雨頻率二分類機率預報及準確度評估結果。極端降雨事件的判斷標準乃是根據臺灣西部10個測站的各站50年(1951～2000)五、六月發生最大日雨量的中位數為各站的門檻值，收集至少有一站事件發生的日期與資料研判利於極端事件發生的大尺度條件。極端事件發生的177天中，有70%的個案出現正渦度距平低壓氣旋環流從華南經臺灣通過東海到日本南方洋面，並有從西南往東北方向的負渦度距平高壓反氣旋環流從南海東北部經菲律賓往西北太平洋延伸，而在南海北部到琉球附近的正、負渦度距平之間有強勁的西南氣流等特徵，據此設計了西南氣流指數(Southwesterly Flow Index, SWFI)作為預報因子。預報對象為梅雨季10個測站加總的事件數，定義為梅雨季極端降雨事件發生頻率(Taiwan Mei-yu Extreme Rainfall Frequency, TMERF)，TMERF與SWFI的50年相關係數達0.62，在16年驗證期(2001～2016)的相關係數有0.75。由於SWFI的預報結果可從全球數值預報系統產品計算獲得，因此能用SWFI預測值進一步預測TMERF的變化，也就是將全球模式的預報結果降尺度到臺灣應用。以氣象局TCWB2T2作業性預報為例，以每年四月的每日觀測分析場作為預測初始值得到120組SWFI的預測值，藉此即能預測TMERF的變化傾向。TMERF在回溯預報期(1982～2011)的二分類機率預測有56.7%命中正確類別，5年實時預測(2012～2016)有3年預測正確。SWFI降尺度方法的物理基礎在於它的變化反應出與臺灣息息相關的東亞與西北太平洋季風氣候系統和太平洋與印度洋海溫變異的關係。本預報方法未來可廣泛運用到其他全球預報系統產出的季內至年際尺度預報資料，能改善臺灣梅雨季極端降雨事件發生趨勢的季度預測。

Keywords

梅雨 ； 極端降雨預報 ； 長期天氣預報 ； 月與季預報 ； 臺灣氣候

Parallel abstracts

The extreme rainfall events during Taiwan Mei-yu season (May and June) is one of the major hazardous weather phenomena in Taiwan. However, to the authors' best knowledge up to this moment no research on Taiwan Mei-yu extreme rainfall frequency seasonal outlook has ever been published. This paper proposes a large-scale index method applicable to downscale global monthly and seasonal forecast model product to outlook Taiwan Mei-yu season seasonal extreme rainfall frequency. The global forecast model output used in this study is generated at the Central Weather Bureau (CWB) of Taiwan by Taiwan CWB 2-Tier monthly and seasonal forecast system version 2 (TCWB2T2). The extreme rainfall events identified on the station basis are based on the data at 10 stations to the west of the central mountain range. During 50 Mei-yu seasons from 1951- 2000, 177 days are identified with rainfall extremes, which means at least one in ten stations received above-threshold daily rainfall amount. The predictand of study is Taiwan Mei-yu Extreme Rainfall Frequency (TMERF), which is the sum of the total extreme events of 10 stations during May and June. The predictor is a southwesterly flow index (SWFI), which is defined according to the large-scale circulation pattern and strong southwesterly wind condition satisfied by at least 70% of the 177 days with the extreme events. The 50-year correlation between SWFI and TMERF during the statistical relationship identification period (1951-2000) is 0.62. The correlation during the verification period (2001-2016) is 0.75. Applying the SWFI and TMERF relationship to TCWB2T2 for two-category downscaled forecast, during the hindcast period (1982-2011) the hit rate is 56.7%. For the experimental forecast years (2012-2016), three out of five of which the TMERF category is correctly forecasted. The downscaling method presented in this study can be easily applied to the blooming forecast database generated by WMO/GFCS/Global Producing Centers for Long-range Forecasts and research institutes.

Parallel keywords

Mei-yu (Baiyu) ； Extreme rainfall forecast ； Long-range forecasts ； Monthly and seasonal prediction ； Taiwan climate