For management and planning valuable groundwater resources, it is very important to predict groundwater level and have a correct understanding about aquifer changes. In this paper for the first time, the wavelet Halt-Winters hybrid models (WHW) were used and tested for
More
For management and planning valuable groundwater resources, it is very important to predict groundwater level and have a correct understanding about aquifer changes. In this paper for the first time, the wavelet Halt-Winters hybrid models (WHW) were used and tested for groundwater forecasting. A monthly data set of 16 years consisting of groundwater level fluctuations was used in two observation wells of Urmieh coastal aquifer. In the WHW, the dataset was converted into several sub-dataset with different time scales. Then, the sub-series were used in the HW model as inputs. Subsequently, the performance of the WHW model was compared with ARIMA, HW, and SARIMA as linear models and neural network models (ANN) and Support Vector Regression (SVR) as nonlinear models. The results showed that the NSE and RMSE values of the WHW model were upgraded up to 30% and 60% respectively, in comparison with linear models. The WHW hybrid model also has the same performance compared to nonlinear models. This research reflects that if there are multiple seasonal fluctuations in the groundwater time series, the performance of the WHW model compared with linear models will be more accurate.
Manuscript profile
