Reduced chaotic noise to improve the accuracy of estimates of monthly flow (case study: Nahandchai, Aharchai and Lighvanchai Rivers)

The nonlinear and complex relationship between the components of a system in hydrological processes and the dynamic behavior between them makes it necessary to use intelligent models for modeling. Usually, in a variety of studies, to increase the accuracy of modeling results, newer models with more computational capabilities are used. In addition to the computational abilities of the models, the use of correct input information is also important to them, and it is necessary to achieve the appropriate accuracy in a variety of modeling methods. Because the error is usually in the hydrologic data, the purpose of this study is to investigate the effect of eliminating possible errors in hydrologic systems in increasing the accuracy of models. In this paper, the monthly flows of Nahandchai, Lighvanchai, and Aharchai rivers have been investigated in two cases the original series and noise reduced time series. Then, for both cases, the runoff predictions with the ANN model have been done. Chaos theory used for the separation of input noise data. The results base on the evaluation criteria shows that this method, providing a possible higher accuracy in data without noise. The amount of Nashatcliff coefficient increased (43.2%, 27.9%, and 5.9%) respectively at the stations of Nahandchai, Aharchai, and Lighvanchai, and RMSE decreased to 65.2%, 65.5% and 7.7% at these stations