Hybrid DARIMA-NARX model for forecasting long-term daily inflow to Dez reservoir using the North Atlantic Oscillation (NAO) and rainfall data

Abstract Proper water resources management cannot be achieved without accessing comprehensive data, suitable resources exploitation programs, and quantified forecasts of water resources. Thus, it is necessary to develop new forecasting models of water resources. Autoregressive integrated moving average (ARIMA) models (classified as time series models) and artificial neural network models have performed well in forecasting linear and non-linear stream flow, respectively. In this paper, a hybrid method was used to evaluate the accuracy of daily flow forecasts through using the capabilities of ARIMA model and nonlinear auto regressive model with exogenous inputs (NARX). Moreover, the efficiency of forecasters such as North Atlantic oscillation (NAO) (as a large scale climate signal) was analyzed for flow forecasts. The forecasting results which compared using proposed error index (IIFFE) to assess mean absolute relative error (MARE), time to peak, and peak flow of forecasted flow. The results showed that forecasting accuracy was enhanced by using the hybrid model. It also displays that using rainfall as a forecaster has the most prominent influence on the increasing forecasting accuracy, while the accuracy is not achieved by using NAO singular or together with rainfall data. Finally, the proposed hybrid model decreased the IIFFE index from 1.25 (achieved by the best ARIMA forecast) to 0.36 and improved the accuracy daily flow forecasting considerably which enhance real time optimal operation of reservoirs.

[1]  Using MODIS snow cover and precipitation data to model water runoff for the Mokelumne River Basin in the Sierra Nevada, California (2000–2009) , 2011 .

[2]  Çagdas Hakan Aladag,et al.  Forecasting nonlinear time series with a hybrid methodology , 2009, Appl. Math. Lett..

[3]  Mohammad Ebrahim Banihabib,et al.  Comparative study of statistical and artificial neural network's methodologies for deriving global solar radiation from NOAA satellite images , 2013 .

[4]  Özgür Kişi,et al.  Daily River Flow Forecasting Using Artificial Neural Networks and Auto-Regressive Models , 2005 .

[5]  Lee-Ing Tong,et al.  Forecasting time series using a methodology based on autoregressive integrated moving average and genetic programming , 2011, Knowl. Based Syst..

[6]  Bo-Suk Yang,et al.  A hybrid of nonlinear autoregressive model with exogenous input and autoregressive moving average model for long-term machine state forecasting , 2010, Expert Syst. Appl..

[7]  M. E. Banihabib,et al.  NARX Recurrent Neural-Network Model for Long-Term Forecasting of Daily flow , 2014 .

[8]  Li Wang,et al.  An ARIMA‐ANN Hybrid Model for Time Series Forecasting , 2013 .

[9]  Yonghong Tan,et al.  Neural-network-based d-step-ahead predictors for nonlinear systems with time delay , 1999 .

[10]  Wen Wang.  Stochasticity, nonlinearity and forecasting of streamflow processes , 2006 .

[11]  François Anctil,et al.  Comparing Sigmoid Transfer Functions for Neural Network Multistep Ahead Streamflow Forecasting , 2010 .

[12]  M. Valipour,et al.  Comparison of the ARMA, ARIMA, and the autoregressive artificial neural network models in forecasting the monthly inflow of Dez dam reservoir , 2013 .

[13]  Durdu Ömer Faruk.  A hybrid neural network and ARIMA model for water quality time series prediction , 2010, Eng. Appl. Artif. Intell..

[14]  Kung-Sik Chan,et al.  Time Series Regression Models , 2008 .

[15]  Mohammad Ebrahim Banihabib,et al.  A dynamic artificial neural network for assessment of land-use change impact on warning lead-time of flood , 2015 .

[16]  Mohammad Ebrahim Banihabib,et al.  Performance of conceptual and black-box models in flood warning systems , 2016 .