| Summary: | 碩士 === 淡江大學 === 水資源及環境工程學系 === 89 === Model uncertainty is one of the difficulties in generating synthetic time series. According to the time series decomposition method, a time series can be composed into the data of the marginal distribution and the order series. Therefore, it can to used to generate synthetic time series without identifying the model for the time series. In this study, instantaneous transformation method is used to generate non-Gaussian time series as the original data to verity the time series decomposition method. According to the results of synthetic data, the time series decomposition method preserves the statistic properties and autocorrelation structure of original time series in synthetic time series. Besides, the time series decomposition method provides a possibility to rectify the attenuated autocorrelation structure of non-Gaussian time series. Furthermore, the results of observed hydrological, indicate that the time series decomposition method is better than detrended model in preserving Hurst coefficient in the synthetic time series.
|
|---|
