Abstract
Many modern climate experiments consist of multiple General Circulation Model (GCM) integrations. Many experiments also contain climate integrations in which a climate equilibrium is not obtained for a considerable period of time, or in which the mean state slowly evolves with time in response to slowly changing external conditions (such as the concentration of CO2 In this paper we describe some relatively simple statistical techniques which test for the presence of trend and take its effects into account when studying simulated inter-annual variability. The limitations of these methods, the assumptions which are implicit in their use, and techniques for developing
tests of hypothesis are all discussed.
