臺灣梅雨期間降水之時空分析

本研究從綜觀氣候分類角度，將台灣梅雨季期間梅雨鋒面降水的空間分佈進行分類。資料以涵蓋1960-2016年的1km解析度日降水網格資料庫和經過標準化後的Zscore全台測站日降水資料庫為基礎。首先定義梅雨日和梅雨事件的日降雨門檻進行篩選，篩選出符合門檻的降雨日，接著使用經驗正交函數和旋轉經驗正交函數，分類出六種降水空間型態分佈(spatial pattern)，包括西南沿海地區為主、北部地區為主、西部沿海地區為主、中央山脈山區為主、東部沿海地區為主和西北部地區為主。將六種空間分佈型態分別對應的降雨日日雨量進行平均，得到六個空間型態分佈的實際降水平均分佈。利用歐洲中期天氣預報中心之重分析環流資料與空間型態的實際降水分佈做組合分析，分析出與六種空間型態分佈耦合的大尺度環流距平場，包括氣壓距平場、垂直速度距平場和流線距平場的發展特徵，以了解六種降水空間分佈型態的大尺度環流狀態。將六種空間分佈型態進行長期趨勢分析，以觀察六種降水間分佈型態的頻率和強度有無趨勢變化，並以拔靴法進行檢定。發現在頻率變化上，空間型態3（西部沿海地區為主）-1.94次、空間型態4（中央山脈山區為主）-6.4次和空間型態5（東部沿海地區為主）-2.57次的百年趨勢;在強度變化上，空間型態4（中央山脈山區為主）有29.42mm增強的百年趨勢，以及所有空間分佈型態未分類前的梅雨事件有11.71mm增強的百年趨勢。

關鍵字

梅雨；綜觀氣候分類；旋轉經驗正交函數；趨勢分析

並列摘要

From the perspective of climate classification, this study classifies the spatial distribution of rainfall in the Meiyu front during the rainy season in Taiwan. The data are based on the 1km resolution daily rainfall grid database from 1960-2016 and the daily rainfall database from the standardized Z-score stations all over Taiwan. First, the daily rainfall totals of the Meiyu days and Meiyu events are defined to select the rainy days that meet the amounts. Empirical orthogonal functions and rotating empirical orthogonal functions are then used to classify these Meiyu days into six rainfall spatial patterns, including the coastal areas in the southwest, the sections in the north, the western coastal areas, the mountainous areas in Central Mountain Range, the eastern coastal areas and the northwestern regions. The average daily rainfall distributions of the six spatial distribution patterns are averaged to obtain the average rainfall distribution of the six spatial distribution patterns. Based on the analysis of the circulation data and the rainfall distribution of spatial patterns using the European Reanalysis-Interim (ERA-Interim) of European centre for Medium-Range Weather Forecasts (ECMWF), the large-scale circulation anomaly fields coupled with the six spatial distributions are analyzed, including the pressure anomaly field, vertical velocity anomaly field, and the development characteristics of the streamlined anomalous field. These analyses are utilized to detect the large-scale circulation state of the six rainfall spatial patterns. The long-term trend of identified spatial distribution patterns are also analyzed to detect their possible frequency and intensity changes based on the bootstrap method. For the frequency changes, it is found that the spatial patterns 3, 4 and 5, mainly located in the western coastal areas, the Central Mountains Range and eastern coastal areas individually, showed a -1.94/100-yr, -6.4/100-yr and -2.57/100-yr reduction, respectively. As for the intensity changes, the spatial pattern 4 (i.e. Central Mountain Range) has a significant 29.42mm/100-yr rising trend, and all (i.e. the unsorted) spatial distribution patterns Meiyu events have an 11.71mm/100-yr rising trend.