Hydroclimatic Phenomena Modeling Using Wavelet-Holtwinters Model

Doubtlessly hydroclimatic models play important role in the management of water resources. The hydroclimatic time series have three principle components (autoregressive, seasonality and stochastic) and the performance of the models related to these components, In the current research, the wavelet transform was linked to the Holt-Winters (HW) model for prediction of Lighvanchai, Trinity, West Nishnabotna watersheds monthly runoff and minimum temperature of Tabriz. The obtained results were compared with autoregressive and seasonal models such as ARIMA, seasonal ARIMA (SARIMA) and HW. For this purpose, the main time series were decomposed to some multi-frequency time series by wavelet transform. Then due to the univariated nature of the HW model, these subseries were imposed as input data to the HW models with two considered scenarios. In the first scenario only approximation subseries and one detail subseries (resulting from the accumulation of all details subseries) and in the second scenario all subseries were used as input to HW models .The obtained results show the second scenario of hybrid wavelet-holtwinters model (WHW2) could lead to considerably increased accuracy of both runoff and temperature monthly modeling because of multiscale analysis and considering all multi-frequency subseries.