Glacial Puzzles

See allHide authors and affiliations

Science  04 Sep 1998:
Vol. 281, Issue 5382, pp. 1467-1468
DOI: 10.1126/science.281.5382.1467

One of the most perplexing and enduring puzzles in paleoclimatology has been the cause of the 100,000-year rhythm of the major glacial-interglacial cycles during the past 1 million years. The smaller warmings and coolings superimposed on this pattern are more-or-less linear responses to variations in the distribution of solar heating caused by changes in Earth's orbital position, known as “Milankovitch” variations, after an early investigator (1), but no such simple orbital mechanism has been found that can explain the 100,000-year cycle. Although Earth's orbital eccentricity, and hence average distance from the sun, varies with a 100,000-year cycle, the resulting changes in solar heating are believed to be too small to be climatically significant.

How can such a strong climate response arise from such a seemingly weak forcing? And why did the 100,000-year cycle only appear about 800,000 years ago? From about 3 to 1 million years ago, smaller ice sheets varied at an almost metronomic 41,000-year rhythm, the period of changes in orbital tilt. Hypotheses seeking to explain this 20-year-old paradox have generally fallen into one of three camps: mechanisms that posit that Earth's ice-atmosphere-ocean climate system maintains an internal oscillation near 100,000 years that can get phase-locked to external orbital forcing (2), mechanisms that invoke highly nonlinear responses of this system to weak forcing by eccentricity (3), and mechanisms that instead invoke temporal variations in the inclination of Earth's orbit relative to the solar system (4), another orbital parameter that varies with a periodicity of about 100,000 years. Each of these explanations has difficulty accounting for some aspect of the climate record, and hence none have achieved broad acceptance. Recently, a fourth type of hypothesis has been proposed (5), one that elegantly avoids some of the shortfalls of earlier models and draws attention to simple relations apparent in the most recent and accurate ice volume and insolation records [see, for instance, (6)].

Inspired by the observation that models of ocean thermohaline circulation have multiple steady states, Paillard (5) investigated a climate system with three steady states and a set of predefined rules for moving between them. The three climatic regimes are interglacial, mild glacial, and full glacial; transitions between regimes occur whenever a specific threshold is passed. Very simply, if incoming solar radiation (insolation) at high summer latitudes falls below a specific threshold, an interglacial to mild glacial transition occurs. If insolation then stays below a critical threshold for an extended period of time, a mild glacial to full glacial transition will occur. The model returns to interglacial conditions only from a full glacial state by exceeding a defined insolation threshold; it cannot go directly from a full glacial to a mild glacial or from a mild glacial to an interglacial state. This simple linear differential model, which is forced by insolation and which allows ice volume to change continuously, shows an impressive match to the geologic record of ice volume change (see figure). In particular, by showing a strong interglacial response at marine isotope stage 11, a time of weak eccentricity forcing, this model does not fall victim to the classic “stage 11 shortfall” problem of the oscillator and nonlinear response models.

How realistic are the assumptions of this multiple-state model when compared with our best geologic records of the ice ages? We know that one of the defining characteristics of the “100,000-year world” is terminations: rapid and abrupt shifts from extreme glacial to extreme interglacial conditions that are not observed earlier in the Pleistocene. These terminations, which conclude in an interglacial-glacial couplet, require for their genesis the buildup of unusually large, presumably unstable ice sheets. The inherent instability of such ice sheets can then explain the extraordinary rapidity of the deglaciation. This “extra” ice appears to grow only during extended periods of low summer insolation at high northern latitudes (see figure). When insolation next increases to high levels (that is, above a certain threshold), terminations are triggered much as described by Paillard's model.

Glacial cycles.

Summer solar radiation in the Northern Hemisphere [after (7) (top)], the oxygen isotopic record for ice volume (8) (middle), and Paillard's (5) modeled ice volume over the last 1 million years (bottom). The model reproduces the oxygen isotope records. Most of the full glacial episodes correspond to extended times of low summer insolation (shaded peaks, top plot). Marine isotope stage designations are shown at the bottom.

From insolation and ice volume records (see figure), it is also apparent that interglacial-glacial cycles are “quantum” in the sense that they are either four or five precessional cycles long depending on the particular interactions of obliquity and eccentricity with precession. This characteristic is captured by Paillard's model. What the model does not appear to capture is the lack of large ice sheet growth and, subsequently, a termination, during marine isotope stages 7 to 8 and 13 to 14 (see figure). The data suggest that a mild glacial to mild interglacial transition can occur, a sequence of events precluded by Paillard's initial model.

This brings one back to the big question, the one asked by our students: What causes the 100,000-year cycle? I would answer that it is a pseudoperiodic cycle varying in length from about 80,000 to 120,000 years. It is caused by the periodic buildup of large ice sheets during times of unusually low summer insolation at high latitudes that occur roughly every 100,000 years as dictated by the interaction of precession with eccentricity and, to a lesser degree, obliquity. A cycle ends abruptly, with a termination, when insolation increases above a threshold value that causes the ice sheet to become unstable and melt rapidly.

Why then did the 100,000-year cycle begin, apparently abruptly, about 800,000 years ago? Along with Paillard, I suspect that a gradual secular decrease in the strength of Earth's greenhouse is to blame (shown schematically by the fiducial line on the insolation curve in the figure). If atmospheric CO2 concentrations were higher in the past, then, regardless of orbital configuration, it may never have been cold enough, for long enough, for massive ice sheets to grow on North America and Scandinavia.

Furthermore, given that precession so completely dominates the insolation regime at high latitudes in summer (see figure), a far more perplexing mystery may be how one explains the first 2 million years of the major Northern Hemisphere ice ages when global ice volume and deep-sea temperature change varied predominantly at the 41,000-year period of orbital obliquity. If the canonical view that summer radiation at high latitudes exerts the ultimate control on ice sheet mass balance were correct, then far more precessional variance would be observed in high-latitude climate records of the late Pliocene and early Pleistocene. No existing climate model has yet been able to successfully reproduce the almost sinusoidal growth and decay of the ice sheets at 41,000 years observed during the first two-thirds of the Northern Hemisphere ice ages.

Hence, even though the “41,000-year world” is often held up as the textbook example of orbital control of climate, the mechanisms by which that control is exerted are far from obvious. The 41,000-year world may prove even more mysterious than the 100,000-year world, which has baffled scientists for decades. Still, Paillard has made an important and insightful contribution to our understanding of the pacing mechanisms of climate change over the past 700,000 years and has pushed the study of the 100,000-year cycle in an important new direction. Our understanding of the mechanisms of climate change, the “rules” that govern transitions between states, is still far from perfect. However, they involve ocean-atmosphere-ice processes that can change atmospheric temperatures, CO2, methane, ocean circulation, and sea surface temperatures, on time scales as short as centuries, even as they are paced by orbital changes that occur over many millennia. Only with the accumulation of more data and thoughtful modeling like Paillard's can we hope to solve the puzzle of paleoclimate.


View Abstract

Stay Connected to Science

Navigate This Article