Non-Stationarity Identification in Flow Time Series Using Wavelets Transform Technique

The current study explored the possibility of using Discrete Wavelets Transform technique (DWT) in diagnosing the non-stationarity in hydrologic time series, which typically masks the real characteristics of those series. This helps in diagnosing the appropriate model and using it for prediction purposes.Basically, this manuscript divided into two phases: in the first phase, a defined stochastic linear model parameter, i.e. (ARMA (1,1)) was developed with known parameters 1 and of (0.8 and 0.4) respectively. The ACF and PACF analyses before and after intentionally adding some defined deterministic components (such as trend, periodicity, etc.) confirm the capability of (DWT) in diagnosing those non-stationarity sources. While phase two makes use of (DWT) technique in diagnozing the non-statioarity in an observed flow time series of al-Khabor river, Kurdistan region-Iraq, where 24 years of flow time series is available. After removing the source of the non-stainarity diagnozed by the proposed method in the data, a stationary model (ARMA (2,1)) has been fitted. The study indicated that the proposed model was distinguished by its capabilities to work in real time, thus, the outcomes of the model is almost following the same pattern of the observed outcomes of the process under study.