Modelling trends and volatility in atmospheric carbon dioxide concentrations

Atmospheric carbon dioxide concentration (ACDC) is a crucial variable for many environmental simulation models, and is regarded as an important factor for predicting temperature and climate changes. However, the conditional variance of ACDC levels has not previously been examined. This paper analyses the trends and volatility in ACDC levels using monthly data from January 1965 to December 2002. The data are a subset of the well known Mauna Loa atmosphere carbon dioxide record obtained through the Carbon Dioxide Information Analysis Center. The conditional variance of ACDC levels is modelled using the generalised autoregressive conditional heteroscedasticity (GARCH) model and its asymmetric variations, namely the GJR and EGARCH models. These models are shown to be able to capture the dynamics in the conditional variance in ACDC levels and to improve the out-of-sample forecast accuracy of ACDC.