運用自組織映射圖分類法建構之春夏轉換期東亞季風區低層風場天氣曆分析聖嬰現象與台灣及菲律賓區域的颱風和極端降雨特徵

Abstract

本研究著重於東亞季風區的春夏季節轉換期，以客觀方法對逐日低層風場分群所辨識出的天氣類型為基礎，建構春夏轉換期東亞季風區之天氣曆，並藉此呈現聖嬰現象、西北太平洋颱風活動、台灣及菲律賓區域極端降雨與東亞季風間的關係。 東亞季風區低層風場與降雨特徵於春夏季節轉換期有顯著的轉變，包含西北太平洋副熱帶高壓的移動、南海夏季季風肇始、台灣梅雨與長江梅雨的雨帶形成與持續、季風槽的發展等。此外，東亞季風的年際變異明顯受到聖嬰現象影響，對於低層風場、降雨、乃至於颱風活動都具有一定的影響。本研究應用自組織映射圖分類法(self-organizing map, SOM)，將東亞季風區(0°-37.5°N, 90°-140°E) 1979-2023年的四到七月逐日低層風場分為九種天氣類型，發現低層風場與對應的降雨特徵以及偏好發生的時期有明顯的季風樣貌。將天氣類型依照發生時間前後編號排序建構的東亞季風區四到七月的季風天氣曆，呈現出東亞季風春夏季節演變的過程。根據各天氣類型發生頻率和組合，可將四至七月區分為春季晚期（第19至24侯）、轉換期（第25至33侯）與夏季初期（第34至42侯）三個季風發展階段。不同階段分別由不同的天氣類型所主導，顯示出季風發展過程氣候狀態的階段性演變和每日風場的關係。 我們根據不同聖嬰相位組成的春夏季風天氣曆來呈現氣候與天氣的關聯，顯示不同聖嬰相位的氣候背景對於逐日環流類型、極端降雨與颱風活動的影響。在聖嬰相位後一年的春夏轉換期，相比反聖嬰相位，主導的天氣類型凸顯出春季晚期菲律賓反氣旋與初夏西北太平洋颱風活動減少的特徵，同時反映出菲律賓雨季前降雨明顯受聖嬰現象所主導。我們進一步鎖定台灣梅雨季時期（4月26日至7月4日），2011與2009年兩個極端年份顯示季風天氣曆能夠反映出反聖嬰影響下2011年菲律賓的極端多雨以及2009年台灣偏乾的事件。然而，對於非極端少雨年份天氣曆並不能區分台灣梅雨季雨量的偏多或偏少，顯示聖嬰和季風等大尺度環境尚不足以解釋台灣梅雨季的年際變化，但對於雨量極端偏少的年份有值得繼續開發的應用價值。最後，我們運用東亞季風天氣曆討論聖嬰現象、東亞季風與颱風的關係，發現轉換期與夏季初期西北太平洋的颱風活動能夠從天氣類型呈現颱風活躍年與不活躍年的差異，呈現以客觀方法辨識的天氣類型具有的另一個應用面向。 本研究提出了以天氣類型為基礎的東亞季風天氣曆並說明如何藉以顯示整體大環境季風系統在春夏季時期發展演變的過程和年際間變異。建議未來可在此基礎上進一步擴大分析區域與時間範圍，完成全年季風天氣曆的建構，同時也可嘗試應用於次季節至季節尺度(sub-seasonal to seasonal, S2S)模式預報表現的評估以及季節演變過程的即時監測等實務面向。至於區域性極端多雨的天氣類型，則可參考本研究提出的季風發展階段，嘗試縮小分析區域至足以解析地形影響程度進行分類，了解動力場對區域極端氣候的影響。

The aim of this study is to build up an East Asian monsoon (EAM) weather calendar based on the weather types (WTs) identified using daily low-level wind, and to demonstrate that the calendar can reflect different large-scale background climate state, such as ENSO, influence on regional extreme weather, such as the extremely wet and dry cases and tropical cyclones (TCs). EAM exhibits substantial changes during the months of April-July. Transition from spring to summer regime is accompanied by the movement of western North Pacific (WNP) subtropical high, onset of South China Sea (SCS) summer monsoon, and development of the monsoon trough. The interannual variability of EAM can be modulated by ENSO, which also influences the TC activity over the WNP. Three monsoonal stages are objectively identified in this study using self-organizing map (SOM) method based on the WTs of the April-July daily 850-hPa wind data from 1979 to 2023 over the EAM region (0°-37.5°N, 90°-140°E). After arranging the WTs in chronological order according to their occurrence time, a EAM weather calendar is established. According to the WT occurrence frequency and types, three stages of the EAM during April-July is objectively identified: late spring (pentad 19-24), transition (pentad 25-33), and early summer (pentad 34-42). On this basis we demonstrated that EAM weather calendar can reflect the influence of distinctly different background climate conditions of opposite ENSO phases on daily low-level circulation patterns and the associated precipitation extremes and TC activity. During the post-El Niño spring-to-summer transition season the frequent weather types result in strong late spring anticyclonic circulation over East Asia and less early summer TC genesis in comparison to the post-La Niña transition season. The WTs well reflect the ENSO modulation on the pre-rainy season rainfall (wetter during La Niña) over the Philippines. To understand the ENSO influence on the WTs during the Taiwan Mei-yu season (April 26 – July 4), two extreme years (2011, 2009) are selected for detecting the relationship between WTs and the successive extremely wet/dry events. Results suggest that EAM weather calendar can reflect 2011 extremely wet condition over the Philippines and 2009 extremely dry condition over Taiwan area. However, no clear signals are detected for Taiwan wet Mei-yu years. It implies that the ENSO and monsoon large-scale conditions are not sufficient for explaining the wet conditions in Taiwan. Finally, we use the EAM weather calendar to identify the ENSO-EAM-TC relationship and find that the active and inactive TC signals can be detected in different WTs during the transition and early summer stages. Therefore, it is important to objectively identify monsoonal stages based on data-driven concept. It is demonstrated that the monsoon weather calendar can manifest the concept of “weather in climate” during spring-to-summer transition of EAM and interannual variability of ENSO modulations. In future studies, the research domain and study period can be extended to build up the calendar for a whole year. The potential of using it for evaluating the sub-seasonal to seasonal forecast model performance and real-time climate monitoring should be studied in the future.