非線性RCA模式應用於河川月流量之模擬與預測

虞氏（1986）曾使用雙波譜證明水文時序通常爲非線性，本研究主要探討非線性時間序列模式中之隨機係數自迴歸模式，簡稱RCA模式，及其應用於臺灣河川月流量資料之適用性，並與傳統線性自迴歸模式，做一比較。本研究所使用RCA模式之參數推估法，主要採用Nicholls及Quinn（1982）所提出之最小二乘方法及最大概似法。研究結果顯示，於合成資料參數推估時RCA模式受推估法及樣本數之影響較大；於實測資料分析上，就統計特性保存能力方面，RCA模式於平均值及變異數之保存能力則優於AR模式，另於預測能力表現上，於臺灣南部地區，除勢後之RCA模式則皆具較除勢後之AR模式爲佳。總體而言，除勢後之RCA模式於預測能力與統計特性保存能力上表現均優於除勢後AR模式。

關鍵字

隨機係數自迴歸模式；月流量

並列摘要

Yu (1986) used bispectrum to show that the hydrological series are often nonlinear. The major objective of this present study is to investigate the Random Coefficient Autoregressive Model, one of the nonlinear time series models denoted RCA model in brief. The linear autoregressive model is also compared with nonlinear models. In this research, the parameters estimation of the nonlinear model are studied. The monthly riverflow data in Taiwan are employed to investigate the aptness of these nonlinear time series models. In this study, least squares method and maximum likelihood method that, proposed by Nicholls and Quinn (1982), are used to estimate RCA model. The results indicate that the estimation of parameter for RCA model is affected by sample size and estimation method. The RCA model has better forecasting ability than AR model when the data are detrended in south area of Taiwan. In conclusion, the nonlinear time series models are appropriate for riverflow in Taiwan.

並列關鍵字

Random coefficient autoregressive model ； Monthly riverflow

主題瀏覽