以事件角度分析台灣極端降雨變化

Abstract

極端降雨在氣候風險評估中屬於主要危害之一，過去因極端降雨所引發之自然災害造成許多損失，因此極端降雨過去及未來如何變化逐漸受到重視。台灣的極端降雨往往發生在特定的環境條件，如颱風或梅雨鋒面等。然而，過去相關的極端降雨研究多使用極端降雨指標來做分析，且極端指標主要是以各格點的資料做計算，而非獨立之極端降雨事件。因此，本研究以事件角度分析極端降雨變化提供了該領域研究下新的觀點。本研究利用高解析度網格化觀測資料，以深度優先追蹤演算法偵測1960年至2019年間的台灣極端降雨事件，並設定兩種門檻（80mm和350mm）來篩選出大雨事件和大豪雨事件。本研究利用極端降雨事件之特性進行分析，其中事件之總降雨體積可分為平均降雨強度、平均影響面積，以及持續時間之貢獻，因此透過此關係除了可以了解事件間的差異，也可以取得各年或各區域極端降雨的主要貢獻事件。而台灣極端降雨事件前十名之特性也一併在本文中進行討論。

本研究發現兩種強度之台灣極端降雨事件在過去60年中發生頻率及總強度皆呈顯著增加，其中大豪雨事件在後30年的發生頻率較前30年增加76%，大雨事件則增加23%。對於總降雨體積變化之貢獻，大豪雨事件之平均降雨強度、平均影響面積及持續時間變化皆呈增加，但持續時間之變化不顯著。而大雨事件之趨勢也相同，但持續時間變化在大雨事件中則顯示減少趨勢。除此之外，平均影響面積之變化為總降雨體積變化的主要貢獻。Understanding the changes in extreme rainfall has attracted a lot of attention since it is one of the major exposures in terms of climate risk. In Taiwan, extreme rainfall typically occurs under unique environmental conditions, such as tropical cyclones or Mei-yu fronts. While Extreme indices has been widely used for analyzing extreme rainfall, the data is sampled based on each grid instead of the extreme rainfall event itself. Therefore, investigating the extreme rainfall from the event perspective can provide a new insight for the changing extremes. Here we apply the event-tracking method by using Depth-First Search algorithm on high-resolution gridded observation data to track the extreme rainfall events from 1960 to 2019 in Taiwan. Two different thresholds (80mm and 350mm) are then selected for analysis of extreme event and the most extreme event.In this study, the characteristics of extreme rainfall event in Taiwan is being examined first. Since the total rainfall volume is contributed by mean rainfall intensity, mean affected area, and duration, the correlation shows the different structure between each extreme event. With the uniqueness of each event, the contribution of extreme rainfall of a year can be determined. Furthermore, Taiwan’s top10 event over the past 60 year have been discussed.

Our results show that the extreme rainfall events have increased significantly in both frequency and intensity, as of the 60-year long-term trend and the relative change between the period of 1960-1989 and the period of 1990-2019. Frequency changes in extreme rainfall events indicate a 76% increase for the most extreme rainfall event and a 23% increase for the extreme rainfall event. For the changes in intensity, we analyze the total rainfall volume and its main contributors with the selected top events in each period. Among all the contributors, mean affected area and mean rainfall intensity show an increase trend for both kinds of event, while duration increases for the most extreme rainfall event and decreases for the extreme rainfall event. Moreover, change in mean affected area contributes the most of the change in total rainfall volume.