Long-range Forecasting of Monthly and Seasonal Air Temperatures Considering Periodical Nonstationarity

B. G. Sherstyukov

A statistical method is proposed for prediction of monthly mean, seasonal mean, and annual mean temperatures with 12-month projection. The method is based upon revealing latent rhythms of the climate system which force sequences of interannual disturbances in air temperature. In selection of the rhythms with prognostic information, it is taken into account that season-to-season transition violates the time series stationarity. The forecast verification shows advantage of the proposed method as compared with climatologic forecasts.