臺灣梅雨季降雨之物理-經驗預報模式分析

過去研究指出，能夠藉由多種影響臺灣梅雨季(5-6月)降雨的機制，選取預報因子，並建立出具有物理意義的預報方程式。Yim et al. (2015)以海表與兩米溫度變化趨勢作為預報因子，並且建立了三個預報領先時間(0-month lead、1-month lead和2-month lead)之預報方程式，其中以0-month lead所建立的預報方程式對於臺灣梅雨季降雨之年際預報結果最佳。之後，張等(2017)針對Yim et al. (2015)提出的物理-經驗預報方程式對臺灣梅雨季降雨預報能力進行改善。然而這些研究僅針對2015年前的資料進行分析，其對於2016-2020年臺灣梅雨季的降雨預報能力尚不可知。在本研究中，我們以中央氣象局(簡稱CWB) 測站資料作為臺灣梅雨季降雨資料，針對以下三個議題進行探討：(1) Yim et al. (2015)與張等(2017)所建立的物理-經驗預報方程式對於2016-2020年臺灣梅雨季降雨預報是否適用？ (2)以2008-2015年為訓練期，重新建立的預報方程式是否能改善對於2016-2020年臺灣梅雨季降雨預報結果？ (3)以風場相關的氣象參數做為預報因子，重新建立的預報方程式，是否能有效的掌握2016-2020年臺灣梅雨季降雨？針對上述議題，本研究發現前人所建立的預報方程式，對於2016-2020年臺灣梅雨季降雨的預報結果不盡理想。而我們以2008-2015年為訓練期或是調整預報因子選取範圍重新建立的預報方程式對於2016-2020年臺灣梅雨季降雨預報也無明顯改善。對於此結果，我們藉由改變預報因子參數，由原本的溫度趨勢改為與風場相關之參數趨勢，結果發現重新建立的預報方程式中以850hPa速度位趨勢作為預報因子最能夠有效的提升預報能力。此研究成果對「預報因子從溫度改為風場相關參數」的可能性提供了新的見解，有利於提高臺灣梅雨季降雨的物理-經驗預報模式的能力。

關鍵字

梅雨季降雨；預報方程式；預報因子

並列摘要

Past studies have pointed out that various rainfall mechanisms of Meiyu season (May to June) can be used as prediction factors to help establish prediction equation with physical meaning. Yim et al. (2015) used trends change of sea surface temperature and 2 meters temperature as prediction factors to establish three forcast equations at different lead time (0-month lead, 1-month lead and 2-month lead), and found that the forecast equation established by the 0-month lead is the best for rainfall forecast in Taiwan during the Meiyu season on the interannual timescale. Afterward, Chang et al. (2017) modified Yim et al. (2015)’s equation to improve the Meiyu season rainfall forcast. However, both these studies have focused on the data prior to 2015, and the ability of the forcast equation in prediting 2016-2020’s Meiyu rainfall in Taiwan is unknown. In this study, we used the Central Weather Bureau (CWB) station data as the rainfall data for the Meiyu season in Taiwan, and discussed the following three issues: (1) Do the physical-empirical forecast equation established by Yim et al. (2015) and Chang et al. (2017) suitable for rainfall forecast in Taiwan during the Meiyu season from 2016 to 2020? (2) If we use 2008-2015 as the training period to re-establish the forecast equations, can we improve the results of the rainfall forecasts in Taiwan for the Meiyu season from 2016 to 2020? (3) If we use wind field-related meteorological parameters as predictors to re-establish the forecast equations, can we improve the results of the rainfall forecasts in Taiwan for the Meiyu season from 2016 to 2020? Our results show that both the forecast equations established by Yim et al. (2015) and Chang et al. (2017) are not capable to predict the Meiyu season in Taiwan during 2016 to 2020. In addition, the re-established forecast equation that use 2008-2015 as training period has not obviously improve the rainfall forecast for the Meiyu season during 2016 to 2020. In contrast, the re-established forecast equation that use the 850hPa velocity potential trend as the predictor can efficiently improve the ability of rainfall forecast. This finding provides new insight for the possibility of changing the predictor from temperature to wind-related factor to improve the ability of physically based empirical forecast model of Meiyu season rainfall in Taiwan.