The paper is devoted to application of the wavelet network technique to nonlinear models for timeseries analysis and the nonlinear approximation by the method is able to construct a nonlinear function that is used to process temperature data, resulting in good fitting accuracy.