(IPCC, 1990); for a sudden doubling, decades to a century may elapse before a substantial fraction of this response will occur. The study of internal variability, and of its interaction with the external forcing, will require improving each type of model, and enhancing the intercomparisons between model results. A more intensive use of the hierarchical modeling concept can be guided by the ideas of nonlinear dynamics, which provide a common thread in the analysis of complex behavior for time-dependent models with arbitrary spatial detail.
Given the shortness of instrumental records, model validation on the decade-to-century time scale requires the use of sophisticated methods for their analysis and interpretation. A number of such methods have been applied recently to climate records, with promising results, and are being extended from the purely temporal to the combined space-time domain. These methods need to be refined further and applied systematically to model results, as well as to the existing records. Similar features detected in a consistent manner, by applying the same statistical methods, in models and data will increase our confidence in the former, our understanding of the latter, and hence our ability to predict climate change on the time scales of interest.