台灣地區最大機率雨量分析之最適機率分佈研究

Study on the Best Probability Distribution for Analyzing Probable Maximum Rainfall in Taiwan

陳明杰(Ming-Jei Chen) ； 陳信雄(Hsin·Hsiung Chen)

《中華水土保持學報》 19卷2期 (1988/11) Pp. 49-62

https://doi.org/10.29417/JCSWC.198811_19(2).0005

摘要

蒐集中央氣象局所屬之測候站的最大一日雨量和最大一小時雨量等資料，就雨量資料之頻度分佈情況套配國內常用之四種理論機率分佈，經由統計上之適合性檢定，尋找各個測候站的雨量資料之最適理論機率分佈。其次，利用經頻率分析所得之各個迴歸週期最大一日和最大一小時雨量資料，以特性係數法探討物部公式之係數值K的變化情形。得結論如下：1.依照各個測候站的雨量資料頻度分佈的特性，選擇最適合的理論機率分佈。2.依據適合性檢定結果，大致上以對數常態分佈的適合性較佳。3.K-S檢定法爲三種統計檢定方法中，敏銳性最差者。4.物部公式中的係數値K，因各個測候站和不同廻歸週期條件而異，不能一槪訂定爲0.6。

關鍵字

機率分佈 ； 適合性檢定 ； 迴歸週期 ； 特性係數法

並列摘要

Maximum one-day and one-hour rainfall data of meteorological stations of Central Weather Bureau were collected, and their frequency distribution that will be suitted to the four type probability distribution in common use in Taiwan were analyzed. According to the testing for goodness of fit, the best probability distribution type suit to the frequency distribution of rainfall data of each meteorological station were selected. Besides, on the basis of maximum probable one-day and one-hour rainfall data in each return period that were calculated by frequency analysis of choosing probability distribution, the change of coefficient K in Monobe long-term rainfall intensity formula was investigated by specific coefficient method. Four conclusions stated as follow: 1. The best probability distribution is chosen by considering the frequency distribution characteristics of rainfall data of each meteorological station. 2. According to the testing for goodness of fit, generally, the lognormal distribution is better than other probability distribution. 3. Among the three nonparametric statistics test method, the K-S test method is the least acute one. 4. The coefficient K in Monobe long-term rainfall intensity formula will change under the condition of different station and return period, and K is not equal to 0.6 which is in common use in Taiwan.

並列關鍵字

Probability distribution ； Testing for goodness of fit ； Return period ； Specific coefficient method