Page 2: Climate model evaluation and improvement
Extension of the PMIP model evaluation framework to ocean, and coupled ocean-atmosphere models will be central to the success of CLIVAR. Unfortunately, our understanding of paleocean dynamics is lagging behind, because past boundary conditions are much less well defined than for the atmosphere. In order to reconstruct quasi-steady state oceanic conditions at certain times in the past, the surface buoyancy field has to be known in a detailed way. Recent modeling studies (two- and three-dimensional) have shown that steady states are very sensitive to the north-south buoyancy contrast in the world ocean (Stocker et al., 1992; Fichefet et al., 1994; Manabe and Stouffer, 1995). Additional constraints, such as tracer distributions ( 14 C and 13 C), must be incorporated. Within PAGES, those efforts that aim at a reconstruction of past sea surface conditions (e.g., maps of SST, SSS) have to be enhanced, while information from the ocean interior, such as the mix of water masses, is indispensable for checking the steady states of various ocean models.
  • Simulation of natural variability

As noted earlier, we currently lack an established and accepted theory of natural decade- to century-scale climate variability. Models do not replace theory-they are based on it. From recent limited model results (Table 2), we have learned that several modes of natural variability can be generated in a few geographical areas of the climate system, but the processes involved are not well understood. More clues are needed from the paleoclimatic record, as well as from efforts to simulate aspects of observed past change. Simplified models have to be developed and employed where individual processes can be analyzed. At the same time, efforts need to be made to identify a few well-defined patterns (spatial and frequency domain) that all models should strive to simulate. The definition of such patterns can only come from the analysis of paleoclimatic data.

PAGES/CLIVAR interaction centered on seasonal-to-interannual ocean variability is limited somewhat to low latitudes where a good collection of coral records is becoming available. At higher latitudes, annually dated paleoceanographic data are more limited, but extensive networks of

Reducing Uncertainty.
Numerical models are the backbone of modern climate prediction, yet results from the large number of models now in use suggest a wide range of possible future climatic changes. PAGES-CLIVAR interaction will focus efforts to use the extensive record of past climatic changes to evaluate and improve the performance of predictive models.