Crytic period analysis model of hydrological process and its application

It is difficult to analyse the crytic period of the hydrological process, because hydrological time series is probably characterized by heteroscedasticity. To find out the crytic period, a model is constructed as follows: (1) after using zero-mean transformation for the data, to do Augmented Dickey-Fuller stationary test for the sequence, to build the corresponding AR(p) model and then to do ARCH effects test and white noise test for residual series; (2) for those time series that cannot pass through ARCH test, using logarithm transformation to reduce the heteroscedasticity, and then to redo step (1) until they pass through ARCH test and stationary test; (3) using periodogram analysis to determine all the possible the prime periods and further to put forward three kinds of tests to determine significance level of those prime periods. As examples, the hydrological processes of streamflow from 1784 to 1997 for the gauging stations of Alaer and Xinquman along Tarim River are analysed. After reducing their heteroscedasticity, AR(4) and AR(2) models are developed, respectively. Our results show that the streamflows from the two gauging stations have the same cryptic period of 42.7 years. Furthermore, the reliability for the crytic period model is testified by variance analysis, which shows that the crytic period model is useful and reliable. Copyright (C) 2009 John Wiley & Sons, Ltd.