rary, however; since polar regions cooled dramatically in Late Pliocene time, polar outbreaks have produced very cold winter temperatures in the southeastern United States.
In summary, simulation experiments using GCMs to test the climatic effects of factors such as plateau uplift and isthmus connections have yielded numerous regional trends that agree with observed climatic changes, as well as some that do not. Basically, however, these experiments support the major conclusion from earlier simulations (Barron, 1985) that changes in geography alone are inadequate to explain the large amplitude of late-Cenozoic cooling. An additional factor is needed, and the most likely explanation is a long-term decrease in atmospheric CO2 levels. Two factors are important in determining long-term CO2 levels: (1) rates of input from volcanoes, and (2) rates of removal by chemical weathering of silicate rocks on land.
One hypothesis links long-term CO2 changes mainly to rates of seafloor spreading at midocean ridge crests (Berner et al., 1983). Globally averaged spreading rates are proposed to control rates of CO2 emission from island-arc volcanos, where ocean crust and sediments are consumed and destroyed in ocean trenches. However, over the past 30 m.y., global mean rates of seafloor spreading have been roughly constant and so cannot explain the pronounced climatic cooling.
A second hypothesis invokes increased chemical weathering due to uplift of plateaus and mountains as the mechanism for drawing down atmospheric CO2 levels (Raymo et al., 1988; Raymo and Ruddiman, 1992; see also Chamberlin, 1906). As noted above, plateau uplift produces a monsoonal increase in rainfall. Uplift also exposes fresh silicate rock on faulted slopes to attack by weathering, which consumes more atmospheric CO2 than is subsequently released to the ocean during secretion of carbonate in plankton. Whereas the Berner et al. (1983) model assumed that chemical weathering is a function of both temperature and global sea-level, the uplift hypothesis takes into account the fact that orography has a powerful control on weathering rates (Edmond, 1987). Steep slopes also flush away the products of chemical erosion, keeping weathering rates high.
An additional factor is the partitioning of carbon and alkalinity between the deep and the intermediate portions of the ocean. Because the ocean is the largest reservoir of carbon on Earth, this partitioning is potentially a critical factor in regulating atmospheric CO2. At present, however, the partitioning even in the late Pleistocene (last glacial) ocean is poorly understood.
Whatever the ultimate cause, falling CO2 levels during the late Cenozoic must eventually have passed through a series of critical climatic thresholds. At some point, sea ice began to expand across larger areas of the Arctic Ocean; GCM experiments show that its thickness and extent must have oscillated considerably in response to orbitally driven changes in insolation (Kutzbach and Gallimore, 1988). Modeling studies also indicate that one critical effect of more extensive sea ice would have been to increase the intensity of winter outbreaks of polar air masses across east-central North America and the rate of extraction of heat from the western North Atlantic (Raymo et al., 1990).
Variations in the Earth's orbit have affected climate throughout the history of the Earth. Changes in the Earth's inclination (or tilt) relative to the plane of the ecliptic alter seasonal insolation at high latitudes at a period of 41,000 yr. Precession of the equinoxes around the elliptical orbit influence seasonal insolation at all latitudes, mainly at periods of 23,000 and 19,000 yr. Changes in the eccentricity of the orbit at periods near 100,000 and 410,000 yr do not directly result in variations in seasonal insolation at these rhythms, but instead modulate the amplitude of the precessional signal.
Most of the clear evidence for orbital control of climate has come from continuously deposited, well-dated marine sediments of Pleistocene age (Hays et al., 1976). Although orbital variations have no doubt influenced climate through all of geologic time, once ice sheets appeared in the Northern Hemisphere, they provided a particularly effective means of amplifying insolation forcing from within the climate system. For the first 1.5 m.y. or more of the Northern Hemisphere ice age, these ice sheets varied mainly at periods of 41,000 and 23,000 yr, in response to direct orbital variations in summer insolation (Raymo et al., 1989). Ice-volume changes lag several thousand years behind the direct insolation forcing at 41,000- and 23,000-yr periods, because of the inherently slow time constants of ice response.
For the larger ice sheets of the past 0.9 Ma, the dominant rhythm of ice-sheet change has been centered near 100,000 yr, despite the lack of direct insolation forcing at this period (Hays et al., 1976; Shackleton and Opdyke, 1976, 1977). This rhythm of change appears to reflect either a highly nonlinear response to insolation forcing (linked to the modulation of the precession signal by eccentricity) or else some kind of natural resonance that has recently developed within the climate system and is paced by insolation.