台灣地區廿四小時延時降雨事件之特性研究

Characteristics of 24 Hours Rainfall Events in Taiwan

張守陽(Shou-young Chang)

《中華水土保持學報》 28卷3期 (1997/09) Pp. 231-245

https://doi.org/10.29417/JCSWC.199709_28(3).0006

摘要

本研究以台灣地區逐時雨量記錄分割出24小時延時降雨事件，探討降雨事件之降雨特性，並以區域化變數理論對降雨特性參數進行空間分析。主要成果分述如后：各主要雨量站降雨事件之降雨重心發生機率呈現左右對稱之鐘型分佈，可以以常態分佈描述之；而尖峰降雨機率分佈略爲向右偏斜；另外尖峰降雨發生時間之機率分佈除降雨初期與末期外，其餘時段發生機率呈均匀分佈，此說明實際降雨事件之尖峰時間離散度頗大，當模擬設計長延時降雨分佈時，以尖峰時間最具不易掌握。台灣地區多數雨量站逐時雨量之逢機性較持續性明顯。以地區而言，東部地區逐時雨量通過持續性檢定比率普遍較低，南部地區涵蓋雨量站較多，通過檢定比率並不一致。降雨事件降雨量之機率分佈經以柯-司（K-S）檢定，可以以常態分佈或對數常態分佈加以套配。分別以逆距法及克利金法進行區域變數分析可發現，不同雨量站之逐時雨量有其區域特性，其中除平均尖峰降雨以阿里山地區爲一最小值外，其餘參數皆以阿里山地區爲一最大值。

關鍵字

逐時雨量 ； 降雨事件 ； 機率分佈函數 ； 持續性檢定 ； 區域變數分析

並列摘要

This study is conducted to investigate the characteristic rainfall events concerning about 24 hours duration in Taiwan area theory of regionalized variables was applied to analyze the spatial changes of characteristic parameters of rainfall. The primary conclusions may be staled as follows The probability density function (p. d. f.) for the weighting position of rainfall events is bell-shaped and symmetrical about 45%. Accordingly, this function can be undoubtedly fitted as a 'Gaussian distribution' The p.d.f. of the rainfall peak were slightly positively-skewed In the meantime, the p.d.f. of 'time to peak' tend to distribute uniformly over the whole occurring period except for the initial and final stages. This may be explained by the fact that the scatters of 'time to peak' were significant. Therefore, the 'time to peak' were not easily determined when rainfall events were simulated From persistence analyses, it is found that the time series of hourly rainfalls in an event in general, tends to be a random process rather than a persistence process. It is also found that the persistence for the rainfall events is negligible in eastern area. Contrarily, the persistence for the rainfall events were complex in southem area due to the existence of large amount of rainfall stations in those area. Via the 'Chi-square goodness of fit test', the rainfall events were also found to follow either a normal or a long-normal distribution. Two regionalized variables analyses, verse distance method and Kringing method, were conducted to have a better understanding of spatial structure on parameters of rainfall events Regional analyses of events indicate that all the parameters of the A-i-san station were extreme cases among those stations.

並列關鍵字

Hourly Rainfall ； Rainfall Event ； Probability Density Function ； Persistence Analyses ； Regionalized Variables Analyses