Özet:

The development of different models for prediction of hydrological data such as rain and flow is important in order to be able to cope with water-related problems in the future. This study examines the monthly flow forecast performance of the Models of the Artificial Neural Network (ANN), the Otoregressive Integrated Moving Medium (ARIMA), the Wave-ARIMA (WARIMA) and the WARIMA-ANN. These models were applied at two stations in the Susurluk pool of Turkey. In this context, the flow data is first allocated to the components with wave conversion for WARIMA and WARIMA-ANN models. Then, flow forecasts for each model were made. As a comparison measure, the average ratio of errors (RMSE), Kling-Gupta efficiency (KGE) and Nash Sutcliffe efficiency (NSE) were taken into account. As a result, the WARIMA and WARIMA-ANN models have been found to perform better, especially compared to the ARIMA and ANN models.In addition, the wave transformation has been clearly shown to improve the performance of the ARIMA and ARIMA-ANN models.

Anahtar Kelimeler:

Özet:

The development of various models to estimate hydrological variables, such as precipitation and runoff is significant regarding handling the water-related problems in the future. This study investigates the performances of Artificial Neural Network (ANN), Auto-Regressive Integrated Moving Average (ARIMA), Wavelet-ARIMA (WARIMA), and WARIMA-ANN models for monthly streamflow forecasting. These models were utilized in two stations of the Susurluk basin in Turkey. In this regard, first, the streamflow data were decomposed into components by wavelet transformation for the WARIMA and WARIMA-ANN models. After that, runoff predictions were performed for each model. As comparison criteria, Root Mean Square Error (RMSE), Kling Gupta Efficiency (KGE), and Nash Sutcliffe Efficiency (NSE) were taken into consideration. As a result, it was obtained that WARIMA and WARIMA-ANN models performed better than the ARIMA and ANN models, particularly. In addition, it was seen that wavelet transformation improved the performance of ARIMA and ARIMA-ANN models, obviously.