Oxygen isotopic evidence from the deep oceans, as well as geomorphological and radiometric evidence from Iceland suggest that northern hemispheric ice sheets formed much later than the Antarctic ice sheets. Northern hemisphere ice sheets may have started to form in the late Miocene, about 6 million years ago, but they grew to reach sea-level all around the Arctic Ocean only by about 3.1 million years ago. A major increase in ice-volume occurred about 2.75 to 2.45 million years ago. For the last 2.45 million years large ice sheets on the Northern hemisphere waxed and waned cyclically, with a larger amplitude from 0.9 Ma on, at the specific Milankovich orbital frequencies. There has been much interest in the process of development of northern hemisphere ice sheets during the last 2.5 million years: this period is so recent that many organisms alive at the beginning of the formation of these ice sheets, or their close relatives, are still living, so that we know their environmental preferences. Deposits formed during this most recent period are at the surface in many regions, and thus can be studied in land exposures, and from the ocean floor in short (<10m long) piston cores, which can be taken fairly easily from non-specialized research vessels.

Land studies had demonstrated that there had been more than one 'ice age', a period during which large ice masses covered the northwest European and North American continents. The number of such periods, glaciations, was thought to be four or five (on the European mainland: Donau, Günz, Mindel, Riss, Würm). From studies of ocean cores, however, it became readily apparent that many more fluctuations in extent of the ice sheets had occurred: about 22 fluctuations were recognized over the last 800,000 years, up to 65 over the last 2 million yr. The extent to which ice-sheets reached southward varied from period to period, so that not each "ice age" was as extreme; many "partial ice ages" occurred These fluctuations were documented first by studies of calcareous nannofossils and planktonic foraminifera in the northern Atlantic and Pacific Oceans.

In the present oceans these planktonic microfloras/faunas occur in roughly latitudinal assemblages (modified by surface current systems). We can thus analyze the foraminiferal assemblage as we find them in the top of cores (i.e., the foraminifera as we find them in the sediment, reflecting present-day circumstances). We can then compare the faunal assemblages as we find them in the core tops with the temperature of the surface waters above the core sites. We can also compare the faunas in core tops with older faunas from samples lower in the cores. In the past the boundaries between the faunal assemblages have been at different latitudes from the present position, and the faunal boundaries, and thus presumably the boundaries between surface water masses of varying temperatures, have been moving north and south continuously. More quantitative work was needed to make estimates of the magnitudes of these environmental changes.

The next step was to translate the information from faunal assemblages (so many percent of species A, so many percent of species B, and so on, through about 29 species) into an estimate of the temperature of the surface waters for present day faunas for which water temperatures can be measured. We do this by submitting the paleontological census data (relative abundance data, % of each species) to a multivariate analysis (i.e., a statistical analysis using many variables at the same time). We then derive new parameters which consist of some combination of the old parameters (species percentages), and then relate the multivariate, derived parameters to observed oceanographic data (mainly sea surface temperature, SST). We derive a set of equations called transfer functions in which faunal percentages can be expressed as sea-surface temperature. We can derive both summer and winter temperatures, because different species reflect coldest and warmest months temperatures. A transfer function is thus a set of equations relating oceanographic parameters to faunal data (more in general: in a transfer function parameters are expressed as parameters with other dimensions). We can, for instance, use information on diatom populations in lake to calculate the acidity of lake waters, and then use fossil diatom data to reconstruct the history of pH of lakes.

Using transfer functions, we can thus derive estimates for surface water temperatures in the past, and we can derive a set of temperatures for winter and summer. These records show very strong fluctuations over time, with the magnitude of the fluctuations varying by latitude: the largest fluctuations were found at the middle-latitudes (40 to 60 ^oN).

Shown here is a temperature reconstruction for a core taken at about 50^oN in the North Atlantic, using data on the abundance of 20 species of planktonic foraminifera. Note the rapid end of the ice age just after 15,000 years ago, and the large changes in ocean surface temperatures at this mid latitude location.

The fluctuations in faunal content were mirrored in global fluctuations in the oxygen isotopic values values of benthic and planktonic foraminifera. The oxygen isotopic records of planktonic foraminifera show an overall global difference between glacial and non-glacial periods of about 1.9^o/_oo (heavier during cold periods). Early interpretations were that this record was mainly the result of temperature fluctuations of about 6-7 ^oC even in the tropics (about 1.4^o/_oo temperature, 0.5 ^o/_oo ice volume effect). Comparison of benthic and planktonic records, however, demonstrated that oxygen isotope fluctuations in both are of similar magnitude. At locations where the temperature of bottom waters is now about 0.5 ^oC, however, the deep water temperature cannot have been 6 to 7 ^oC lower during a glacial (that would mean frozen deep oceans). Thus there must be a large ice volume effect in both planktonic and benthic records. If the temperature change in the bottom waters was just a few degrees, than that in the surface waters at low latitudes must have been similar. This conclusion was supported by transfer function analysis of planktonic foraminiferal faunas from low latitudes, where the temperature of surface waters apparently fluctuated by only a few degrees C at most (1.5 + 1.5^oC).

Remember that these periodicities are:

• 1. Variation in ellipticity of the Earth's orbit; periodicity about 100,000 years.
• 2. Variation in obliquity, i.e., the tilt of the Earth's , periodicity 41,000.
• 3. Precession of the equinoxes, periodicity about19,000 and 23,000 years.

Carbon isotopic records also show changes between glacial and interglacial time periods. These records show a depletion oby about 0.8 ^o/_oo in benthic values during glacials (thus lighter values of total dissolved inorganic carbon during glacials); the benthic values can be seen as representing the average of the deep global ocean. This was interpreted originally as resulting from less storage of organic material in the terrestrial biosphere during glacials. The land biomass was supposedly smaller, because large, now forested areas (before human deforestation) were barren tundras during glacials. The shallow water shelves were dry during glacial periods (lower sea level), thus organic-rich sediments on the shelves could be eroded and their light organic carbon added to the oceans. This idea, however, led to the prediction that there would have been more CO₂ in the atmosphere during glacials (about 150 ppm), because less carbon would have been locked in the biosphere and shallow water sediment.

This prediction was incorrect: measurements of CO₂ concentration in bubbles of air trapped in polar ice demonstrated that there was a lower concentration of CO₂ in the atmosphere during glacials (by about 90 ppm, leading to a glacial CO₂ concentration of about 200ppm). The contradiction could be partially solved by comparing benthic and planktonic carbon isotopic records. The planktonic records show much less change between glacials and interglacials than the benthic records, so that the difference in d¹³C values in surface and deep waters (called Dd¹³C) was larger (about 2.1) during glacials than during interglacials (about 1.5). This suggests that productivity in the oceans was higher during glacials. Why would this be?

At the start of a glacial, sea level drops as a result of storage of water in ice caps. Thus the organic matter on shelves is eroded into the oceans, thus leading to decreased values of d¹³C in Total Dissolved Carbon in the oceans. This erosion, however, adds not only carbon, but also nutrients from the oxidized organic material, leading to increased nutrient contents in ocean water. The increased amount of nutrients (fertilizer) may then lead to increased oceanic primary productivity.

The temperature differences from high to low latitudes in atmosphere and ocean increased during glacial periods, because there was strong cooling at high latitudes, little or no cooling at low latitudes. These increased gradients may have led to stronger, more vigorous currents and winds, leading to more vigorous upwelling, which could lead to even higher surface ocean productivity. It has also been proposed that the ratio of N and P to C in organic matter (the Redfield ratio) might have differed during glacial times, which simplified means, that organisms might have been able to produce more organic material with less nutrients (N and P).

The effect of decreasing productivity during interglacials is to decrease Dd¹³C (the difference in carbon isotopic values for planktonic and benthic foramifera, i.e., the difference in carbon isotopic values of dissolved carbonate in surface and bottom waters), and bring the benthic and planktonic records closer together, as observed. But one would expect the planktonic curve to show the largest excursion (because that's where the change in productivity occurs), while the benthic values stayed almost unchanged, because the deep ocean is so much larger than the surface ocean. But the benthic values fluctuated most, while the planktonic values showed slight fluctuations only. This suggest that there was not only a productivity effect, but also a reservoir effect, resulting from decreased areas of forests during glacials - leading to lighter isotopic values in total dissolved carbon during glacials. For the planktonic foraminifera the effects of productivity and reservoir almost cancel each other out: during glacials there is higher productivity, leading to heavier values of total dissolved carbon and thus of the planktonic tests, but at the same time light carbon is being added to the oceans from the terrestrial biosphere reservoir, leading to lighter values of the total dissolved carbon. Presently, about 0.4 ^o/_oo of the changes in benthic foraminiferal d¹³C values is attributed to changes in terrestrial biosphere, about 0.4 ^o/_oo to oceanic productivity changes.

The description of glacials-interglacials as resulting from cyclical changes of insolation as a result of orbital changes is thus essentially probably correct, but much too simplified.

• Insolation changes are the result of the specifics of the motion of the Earth around the Sun, which did not really change that much during millions of years . So why ice ages only during short intervals of Earth history?
• Insolation changes result in very small temperature differences, and can not be expected to result in the large climatic changes from glacial to interglacial without some kind of positive feed-back system to amplify these small changes.
• The increased albedo as a result of the presence of ice caps may be one of the feedback mechanisms. Another such method of inforcement is through the CO₂-effect: once a glacial starts, CO₂ is subtracted from the atmosphere because more carbon dioxide can dissolved in colder waters (soda bottle effect), leading to even lower temperatures, and so on.
• The insolation changes are out-of-phase between the northern and the southern hemisphere, so that we would expect cold periods in the northern hemisphere at the same time as warm periods in the southern hemisphere. As far as we know, however, the glacials and interglacials occurred synchronously in both hemispheres, and are "in synch" with the insolation of the northern hemipshere.
• Both ice sheets and deep-sea cores show that climate change commonly occurs rapidly, at much shorter time scales than the changes at Milankovich frequencies. These shorter time scale changes have been called millennial climate changes, because the rapid changes occur over decades, and are spaced by one to a few thousand years (millennia).

This suggests that the insolation changes are the trigger, and that subsequent feedback reinforces the climate signal.

This change in deep-water formation in the North Atlantic must also have influenced the world's climate, because the North Atlantic water (before sinking) cools strongly, and thus gives off heat to the north Atlantic environment. Lack of this deep water formation may thus have helped cool the northeastern Atlantic during glacials.

The processes of "switching on and off" glacials are thus very complex, and the complexity is increasing rapidly with more and more knowledge. A more detailed knowledge of the way in which the ocean-atmosphere system works, and of the role of the biosphere in shaping climate (by its effects on the carbon cycle) is needed before we can begin to understand even this latest, best-known part of Earth history.