CHAOTIC INFLUENCES AND THE PROBLEM OF DETERMINISTIC SEASONAL PREDICTIONS

An analysis has been performed of three simulations made by forcing a global climatic model with observed sea-surface temperature and sea-ice distributions for the period 1979-1988. These simulations were started from different years of an extended control run, otherwise all conditions were identical. The analysis concentrates on the regional and local differences in the simulations caused by chaos attributable to the differing initial conditions interacting with the non-linearities of the model climatic system. Of particular interest is whether the presence of sea-surface temperature variations in individual years is able to overcome chaotic influences. This interest arises because of the increasing international appreciation that such temperature variations provide the basis for the development of multi-seasonal prediction schemes. The analysis shows that between about 20 degrees N and 20 degrees S a coherent climatic reponse is obtainable owing to the boundary value forcing associated with the sea-surface temperature field. At high latitudes the signal-to-noise ratio rapidly deteriorates. Even at low latitudes, in a sensitive variable such as rainfall, chaos creates substantial differences between the three simulations, except over the Pacific Ocean where a very coherent response is found. In general, over land areas, even those within recognized regions of ENSO (El Nino-Southern Oscillation) influence, the impact of chaos is much higher. Examination of Pacific North American teleconnection patterns for 1988 reveals substantial differences between the three simulations, and the associated impact on rainfall and surface temperatures over the North American continent. Agreement with observations ranges from rather good to poor, indicating the need for multiple runs for any given situation. Overall the results establish that sea-surface temperature variations can overwhelm chaos in appropriate circumstances and specific regions. Importantly, this indicates that given accurate predictions of sea-surface temperature anomalies a few seasons in advance, then model-based predictions will be possible over many regions of the world despite the omnipresence of chaos.