Nonlinear time series prediction of O3 concentration in Istanbul

In this study, local prediction method is used to predict O3 concentration over Istanbul City at different stations. Observed single-variable time series data are used to reconstruct the attractor in the multidimensional space. Subsequently, the dynamic model to generate the attractor is estimated and the change of the trajectory is predicted by a polynomial approximation. In other words, the dynamics of system are described step by step locally in the phase space. Parameters needed to reconstruct the phase space are delay time and embedding dimension. The delay time is chosen as the lag time at which the autocorrelation function first reaches to zero. The embedding dimension is obtained from the plot of correlation coefficient vs. embedding dimension. This embedding dimension corresponds to minimum relative error between observed and predicted values. The relative error between model outputs and observations is within the practically acceptable limits pointing out that O3 concentration is governed by a deterministic chaotic system.