The location, strength and depth penetration of the major sinking regions of the ocean at high latitude (see Plate 4) are known to have changed during glacial cycles, emphasizing the importance of sea-ice cover in insulating the ocean from the atmosphere, preventing deep convection and physical sinking from occurring (e.g., Sarnthein et al., 1994). The contrasting effect of freezing sea-water is that salty brine is rejected from the ice, yielding a small but very dense volume of water that can contribute to sinking events. During the twentieth century lesser yet still significant shifts of the deep circulation (e.g., Molinari et al., 1998) have been verified by tracers and direct current measurements.
Abrupt changes in climate can occur with spatial patterns that in some way reflect the natural dynamics of atmosphere and ocean. These “modes” of circulation are seen in the seasonal, interannual and decadal variability of the system, and have great potential as an aid to understanding just how abrupt changes can occur. At work in establishing the modes are “teleconnections” both vertically, and across the globe. Various waves, particularly Rossby (or “planetary”) waves and Kelvin waves, and unstable waves on the time-averaged circulation, are involved, as is the direct transport of climate anomalies by the circulation.
Natural variability of climate is now occurring in the context of global warming, so the discussion of abrupt climate change during the period of instrumental records must acknowledge the presence of anthropogenic and natural change, and the possibility of strong interaction between them.
Instrumental records show that the climate is characterized by patterns or modes of variability, such as the polar annular modes and ENSO of the equatorial Pacific, as described below. The spatial patterns can provide regional intensification of climate change in quite small geographic areas. The strong couplings and feedbacks among at least the atmosphere, oceans and sea ice, and probably other elements of the climate system, allow a pattern to persist for periods of years to many decades. The different regional modes also interact with one another. For instance, Amazonian rainfall responds to a mode of tropical Atlantic variability, which itself might be responding to ENSO or the Arctic Oscillation.
The behavior of highly idealized models of the climate system suggests that climate change can be manifested as a shift in the fraction of the time that climate resides in the contrasting phases (for example, warm/cold or