River Flow Forecasting using Wavelet Analysis

In the last two decades, researchers have been more interested in river flow forecasting by means of nonlinear models, Genetic Programming, Time-Series, Wavelet Analysis, etc.included. Wavelet transform by decomposition of signals into time and frequency, same as Fourier analysis has presented a new method for signal processing. Meyer discrete wavelet was used for prediction of average monthly Lighvan-Chai river flow, using 90% of data for testing. The results revealed that 10 levels was the most appropriate number of levels, the best monthly forecasting horizon was 12 months and the correlation coefficient between observed and anticipated ones was 0.92 at Lighvan station and 0.91 at Hervi station. Moreover, in time series, ARIMA ((1,0,1),(1,1,1))12 had the best results with correlation coefficient 0.87 at Lighvan station and 0.93 at Hervi station. According to data, time series has analyzed peak points better than the other one. Overall, with attention to correlation coefficients and attention to that complex series change into simple series with wavelet transform that series had analyzed easier, so wavelet transform is more proper than time series.