Page 3: Climate model evaluation and improvement
time series are available for land areas. Mid- to high-latitude paleoceanographic prospects are more encouraging for studies of decade- to century-scale variability given the increasing availability of high accumulation-rate sediment cores. Emphasis must be put on the Labrador Sea, since recent observations showed a great variability in the period 1970-1980, notably in relation to the Great Salinity Anomaly that propagated around the margins of this sea (Lazier, 1980, 1988; Wallace and Lazier, 1988). Best use can be made of a network of high-resolution data with SST and SSS as the target variables. On the other hand, PAGES may suggest additional areas where models should exhibit defined patterns of interannual- to century-scale variability, and so serve as an independent check for these models.

The problem of sparse data may also be attacked by defining a series of paleoclimate indices. This approach has proven to be a fruitful and efficient step in characterizing and describing ENSO (Philander, 1990). It should be employed also for paleoclimatic studies as a useful guide for more quantitative studies. Such indices must be defined in a way characteristic of decadal to		 century variability (e.g., SST, SSS differences across ocean basins, phase-lagged SST, and SSS). PAGES can provide the records and interpretation from which these indices could be constructed.
  • Simulation of abrupt change

Given the significance of abrupt paleoclimatic events for understand-ing the true range of climate vari-ability and predictability, the PAGES community must strive to ensure that data describing past abrupt events is reliable. Data are essential to indicate whether an abrupt change may have been caused or triggered by a perturbation of the climate system (e.g., the collapse of an ice sheet), or whether the abrupt event was the manifestation of natural but rapid changes. High-resolution records should then be used in combination to indicate leads and lags of different climatic variables. For a verification of climate models, paleoclimatic data analysis must not focus on a single region, such as the North Atlantic, but should document amplitudes, timescales, and phase relations of abrupt change on the entire globe. From such an analysis, spatial patterns can be extracted that will be essential in assessing climate model performance.