Abstract

This research analyses time series of air temperature and wind speed. Trends and periodical components are defined for both time

series with no significant statistical relationship between them. The spectral analysis of temperature series with no trends shows

the peak frequency fc = 1/12, corresponding to a seasonal periodical component. The graph of the spectral density function of the

wind speed time series does not manifest a seasonal periodic component. Forecasting of both time series is carried out using

exponential smoothing with a moving average and an ARIMA model. Prediction for 2012 is computed using a multiplicative

approach bases on Winters and ARIMA models.