The analysis of annual series of temperature of surface air (TSA) shows that three components can be distinguished in them: a long-term trend, a set of harmonious components, and anomalies of the stochastic process. We offer for allocation of quasiperiodic fluctuations to use wavelet-transformation of the initial series. In this case, distributions of factors of transformation allow allocating fluctuations of various scales both close to harmonic, and having character of non-stationary oscillatory process. Further, extrapolation forward is spent on factors of wavelet-transformation of the allocated scales in view of their time dynamics, and the oscillatory component of the series is restored by inverse wavelet-transformation. The offered approach is shown by the examples of some annual series of TSA to stations Syktyvkar and Tomsk, with time period more than 100 years.
wavelet-transform; prognosis; ground air temperature; harmonic decomposition