The structure of the temporal variability of temperature records has been investigated by means of different statistical methods including also fractal analysis. Data both from meteorological stations and averaged over wider networks show very similar behaviour; combine a long-run persistence (characterized by a fractal dimension of D = 1.2-1.3) and short-run properties indicating high year-by-year variability. Synthetic temperature records were created with the use of Mandelbrot's fast fractional Gaussian noise generating algorithm. These fractal sets show the same stochastic properties as real temperature records do, and have even a very similar appearance. The results suggest that the fractal reconstruction algorithm could be used to extrapolate the present tendencies to the future and to forecast future fluctuations.
ASJC Scopus subject areas
- Atmospheric Science