On Modelling Temperature Volatility Dynamics

Meteorological variables are not constant but shows some regular variations. This is also true for temperature during day and night time. Asymmetric variations also observed due to seasonality or other causes. In particular, for a long term observed temperature data, the correlation between conditional volatility and unexpected temperature behaviour is negative in winter and positive in summer. This asymmetric volatility dynamics of temperature can be modelled as GARCH process with conditional mean and conditional standard deviation. In this article we use the weekly Dutch temperature data to model the asymmetric volatility. Both in-sample results and out-of-sample performance indicate strong potential for the use of non-linear GARCH model to the temperature data.