Monthly Streamflow Prediction of Yesilirmak Basin by Using Chaotic Approach

Abstract

Streamflow prediction relies highly on reliable hydrological data. However, hydrological data are often inadequate because of ungauged orpoorly gauged basins. Obtaining reliable streamflow time series in developing countries such as Turkey is required when planning and managing water resources. Although the streamflow time series usually show seasonality, the high instability of affecting factors leads to the chaotic and relatively random behavior. This behavior makes it difficult to predict and modelling streamflow time series. In this study, it was intended to predict short-term monthly streamflow by using chaotic approach. Thus, the phase space was reconstructed by using monthly streamflow data
for different stations located in Yesilirmak Basin in Turkey which is used as a case study. The observation period was taken to be 1969-2011 for all stations. Phase space has two parameters that are time delay (τ) and embedding dimension (m). Firstly, Mutual Information Function (MIF) was obtained in order to find an optimum time delay. Then, False Nearest Neighbor (FNN) algorithm was applied to define embedding dimension. After obtaining phase space parameters, monthly streamflow data were predicted successfully by using local prediction method. Reliable streamflow time series obtained with the use of the chaotic approach will provide an important contribution to water resources planning and management.