Modelling Long-Term Variability in Daily Air Temperature Time Series for Southern Hemisphere Stations

Customarily, climate studies of long-range temperature variability have been carried out using annual or monthly averages. The approach mixes the details of short- and long-range variability that are different for air temperature series. This work shows that a useful method for eliminating short-range variability on long-range variability is to apply a sufficiently long (about 2 months) time step to the daily series. An autoregressive integrated moving average model is fitted to daily maximum and minimum temperature anomalies from the mean seasonal cycle, using data from a number of Australian and New Zealand weather stations. The fitted model can be considered as a sum of random walk plus white noise. This enables us to obtain a quantitative long-term description of the temperature variability.