Noise Reduction Effect on Chaotic Analysis of Nazluchay River Flow

Considering the dynamic and nonlinear nature of river flow, it is expected that the river flow time series is obtained from a deterministic chaotic system. Since that the time series obtained from the natural phenomena are generally contaminated by noise, the presence of noise limits the chaotic analysis and consequently makes limitations in the prediction of time series. For this reason, in this study the chaotic analysis, including the evaluation of the presence of chaos using correlation dimension and simulating the river flow using Local Approximation Method, was investigated on daily series of Nazluchay River during the 1990 to 2012 period. Afterwards, in order to evaluate the noise effect on the process of analysis, noise reduction of time series was carried out by a nonlinear method based on phase space reconstruction. The results showed 6.07% decrease in correlation dimension and an increase in model accuracy for the noise reduced time series with respect to the original series (1.09% increase in R2 and 48% decrease in RMSE). Finally, by the selected simulation model, prediction of the river flow was done using the original and noise reduced time series for the 2012-2013 period. The model results predicted with the noise reduced series were found to be more accurate than those with the original series.