Causes, evolution, and dynamics of ice ages in the last 3 million years

Abstract

This thesis concerns the response of global climate to variations in Earth’s orbital configuration and atmospheric CO2 levels since major Northern Hemisphere ice sheets appeared three million years ago. A first chapter focuses on determinants of climate variability in the late Pliocene and early Pleistocene, 3-1 Ma. Climate in this period has been described as varying with a steady rhythm of 41,000-year (41 ky) cycles that follow variations in obliquity, the angle of earth’s axial tilt. A technique for detecting orbital variability in climate proxies is introduced that circumvents shortcomings in commonly-used methods such as spectral analysis and bandpass filtering. Use of this technique to analyze early-Pleistocene proxies of global ice- volume and deep-ocean temperature reveals significant variability following climatic precession, the combined rotation of earth’s spin axis and of the orbit itself that causes 18-24 ky variations in the intensity and duration of seasons. Whereas exclusive obliquity-pacing would suggest complicated mechanisms controlling ice- sheet variability that cancel the precession component, combined obliquity and precession pacing implies that early Pleistocene climate more straightforwardly followed variations in Northern Hemisphere summer insolation intensity in accord with Milanković (1941).

The dominant feature of climate in the late Pleistocene, ∼1-0 Ma, is the presence of ∼100 ky glaciation cycles that cause sea-level changes at times exceeding 100 meters. An unresolved question is whether those cycles are caused by orbital variations or represent an internal mode of climate variability that merely becomes synchronized to the orbit. In Chapter 2, I extend analysis of ∼100 ky cycles to the entire epoch of recent Northern Hemisphere glaciation, 2.8-0 Ma, and identify four dis- tinct intervals in which they occur. Intervals of ∼100 ky cycles are shown to consistently arise under conditions of large-amplitude orbital forcing, strongly suggesting an external origin for the glacial cycles.

A lack of continuous atmospheric CO2 records prior to 0.8 Ma hinders inferences as to the greenhouse-gas forcing of early Pleistocene climate. Reconstructions of past CO2 levels from foraminiferal Boron isotope ratios indicate that late- Pleistocene sea level responded more sensitively to CO2 radiative forcing than did early-Pleistocene sea level, engendering proposals that the controls on ice-sheet response to radiative forcing fundamentally changed in the middle Pleistocene (Chalk et al., 2017). In Chapter 3, a hierarchical Bayesian methodology is introduced for inferring the statistical relationship between orbital variations, sea level, and atmospheric CO2 by way of an energy balance model that is paired with a simple ice sheet. The model reproduces the observed temporal shift in sea-level sensitivity on the basis of ice-albedo feedbacks and the nonlinear relationship between ice-sheet volume and length, without a change time in parameters over time. This result suggests a stable, albeit nonlinear, relationship between sea level and atmospheric CO2 over at least the past 2 Ma. Large uncertainties in individual sea-level reconstructions propagate into model predictions of early-Pleistocene CO2. In Chapter 4 the Bayesian model from Chapter 3 is used to make an ensemble prediction of early-Pleistocene CO2 using five different sea-level reconstructions. The ensemble estimate proposes that CO2 levels between 2 and 0.8 Ma averaged 241 ppm, and remained between 206 ppm and 275 ppm throughout 95% of that interval, compared with values of 225 ppm, 192 ppm, and 274 ppm after 0.8 Ma. When accounting for sparse sampling and the temperature dependence of ice accumulation rates, the range of model early-Pleistocene CO2 is within 2 ppm of that of measurements in discontinuous ice segments from the Allan Hills Blue Ice Area in East Antarctica, strongly supporting the ensemble prediction. One hypothesis for the cause of the lower late-Pleistocene CO2 during glacial periods involves sea-level modulation of CO2 emissions from volcanoes on land and at mid-ocean ridges. Spectral peaks at periods of orbital variations have been identified in bathymetry profiles of late-Pleistocene crust near mid-ocean ridges, supporting the hypothesis that sea-level variations can influence the production rate of carbon-bearing magma. In Chapter 5, a series of numerical experiments are undertaken to explore the conditions under which Pleistocene sea-level variations could influence bathymetry by inducing faults at mid-ocean-ridges. ∼100 ky sea- level variations are predicted to control fault spacing at ridges with intermediate spreading half-rates (2.5-4 cm/year), and 41 ky sea-level variations are expected to control spacing at faster-spreading ridges. These predictions readily accord with observed periodicities in a global collection of mid-ocean-ridge bathymetry profiles.