| description abstract | A 50-year integration of a simple two-layer atmospheric model coupled to a prognostic-depth ocean mixed layer is used for a preliminary exploration of the potential for global climate prediction on seasonal to interannual time scales. Despite a number of quantitative deficiencies, the model simulation permits the investigation of a wider range of climate predictors and predictands than is usually possible from observations: ocean mixed-layer temperature, depth, heat content, and surface heat fluxes are tested as statistical predictors of atmospheric thickness, static stability, and zonal/meridional winds. The field significance of correlations of the atmospheric predictands lagging ocean predictors in different latitude sectors is assessed, and the predictive power and consistency of the ocean variables are determined as a function of time lag, latitude, atmospheric predictand, and season. It is found that the ocean variables demonstrate a modest potential for predicting atmospheric climate at lags up to three seasons, but no more than about 40% of the local variance of an atmospheric field is explained by any ocean predictor. The most powerful predictors are situated in the tropics, while Northern Hemisphere subtropical and midlatitude ocean variables are substantially better predictors than their Southern Hemisphere counterparts. Ocean mixed-layer temperature is the strongest predictor, while the thickness is the most predictable atmospheric field. Spring and autumn ocean variables are most effective, and winter variables least effective, in predicting subsequent atmospheric seasonal states, but the atmosphere is more predictable in summer and winter than in the transitional seasons. The implications of these results for global climate prediction are discussed, and some possible future research priorities are proposed. | |
