極端降雨量其空間分布之群集特性分析

A Study on Clustering Characteristics of Spatial Distribution of Extreme Rainfall

陳嘉榮(Chia-Jung Chen)

《技術學刊》 20卷4期 (2005/12) Pp. 377-386

https://doi.org/10.29507/JT.200512.0006

摘要

本文旨在探討降雨型態空間分布之群集特性，以定義空間屬性均一區，做爲區域雨量分析之依據。以臺灣北部地區38個雨量站樣本作爲研究對象，選擇主要影響降雨空間分布之因子；如測站之空間位置、年雨量變異性、短延時降雨型態等物理特性做爲分析變量。在主成份分析結果中，顯示前三個主成份可說明原始29個變量之92.9%訊息，其中第一主成份佔82.7%爲降雨型態因子、第二主成份佔6.1%爲測站空間位置因子、第三主成份佔4.1%爲年降雨變化因子。最後，在主成份空間中，經比較兩階段群集分析與模糊聚類分析之結果，將研究區劃分爲三個降雨均一區，其分區結果除能結合地理位置之相鄰性外，又與氣候分區結果相吻合，顯現分區之合理性，將有助於雨量區域分析工作之進行。

關鍵字

區域分析 ； 主成份分析 ； 群集分析 ； 均一性

並列摘要

The aim of this study is to identify the spatial distribution of extreme rainfall in Northern Taiwan and further to divide this area into three homogeneous regions as the basis of regional analysis. Some physical characteristics of 38 recording rain gages, representing typical spacing, annual rainfall variations, and short period rainfall patterns, were chosen to be analytic variables. By principal component analysis, these 29 variables were reduced to three independent principal components as clustering variables. The similarities of clustering variables between rain gages were calculated by using Euclidean distance in the principal component space. The results of two-stage cluster analysis are better than fuzzy cluster analysis, and three homogeneous regions were determined. The first subregion lies in Taipei and Tao-Yuan. The second includes Hsin-Chu and Ta-Tung and Nan-Ao in I-Lan. The third encompasses the northern coast, Kee-Lung and I-Lan. These three regions not only have well-defined geographic boundaries, but also correspond to climatic characteristics. The first three principal components account for 92.96% of original variation, and depict the rainfall patterns, positions, and annual variation respectively. The study reveals that there was a reasonable result from grouping 38 rain gages into three homogeneous regions and the results could be an aid in regional analysis of Northern Taiwan.

並列關鍵字

regional analysis ； principal component analysis ； cluster analysis ； homogeneity