Estimation of missing streamflow data using principles of chaos theory

In this paper, missing consecutive streamflows are estimated, using the principles of chaos theory, in two steps. First, the existence of chaotic behavior in the daily flows of the river is investigated. The time delay embedding method of reconstructing the phase space of a time series is utilized to identify the characteristics of the nonlinear deterministic dynamics. Second, the analysis of chaos is used to configure two models employed to estimate the missing data, artificial neural networks (ANNs) and K-nearest neighbor (K-nn). The results indicate the utility of using the analysis of chaos for configuring the models. ANN model is configured using the identified correlation dimension (measure of chaos), and (K-nn) technique is applied within a subspace of the reconstructed attractor. ANNs show some superiority over K-nn in estimating the missing data of the English River, which is used as a case study.