Person:

Wunsch, Carl

The consequences of the hypothesis that Milankovitch forcing affects the phase (e.g., termination times) of the 100 kyr glacial cycles via a mechanism known as “nonlinear phase locking” are examined. Phase locking provides a mechanism by which Milankovitch forcing can act as the “pacemaker” of the glacial cycles. Nonlinear phase locking can determine the timing of the major deglaciations, nearly independently of the specific mechanism or model that is responsible for these cycles as long as this mechanism is suitably nonlinear. A consequence of this is that the fit of a certain model output to the observed ice volume record cannot be used as an indication that the glacial mechanism in this model is necessarily correct. Phase locking to obliquity and possibly precession variations is distinct from mechanisms relying on a linear or nonlinear amplification of the eccentricity forcing. Nonlinear phase locking may determine the phase of the glacial cycles even in the presence of noise in the climate system and can be effective at setting glacial termination times even when the precession and obliquity bands account only for a small portion of the total power of an ice volume record. Nonlinear phase locking can also result in the observed “quantization” of the glacial period into multiples of the obliquity or precession periods.

The 100,000-year timescale in the glacial/interglacial cycles of the late Pleistocene epoch (the past ,700,000 years) is commonly attributed to control by variations in the Earth’s orbit. This hypothesis has inspired models that depend on the Earth’s obliquity (,40,000 yr; ,40 kyr), orbital eccentricity (,100 kyr) and precessional (,20 kyr) fluctuations, with the emphasis usually on eccentricity and precessional forcing. According to a contrasting hypothesis, the glacial cycles arise primarily because of random internal climate variability. Taking these two perspectives together, there are currently more than thirty different models of the seven late-Pleistocene glacial cycles. Here we present a statistical test of the orbital forcing hypothesis, focusing on the rapid deglaciation events known as terminations. According to our analysis, the null hypothesis that glacial terminations are independent of obliquity can be rejected at the 5% significance level, whereas the corresponding null hypotheses for eccentricity and precession cannot be rejected. The simplest inference consistent with the test results is that the ice sheets terminated every second or third obliquity cycle at times of high obliquity, similar to the original proposal by Milankovitch. We also present simple stochastic and deterministic models that describe the timing of the late-Pleistocene glacial terminations purely in terms of obliquity forcing.

Precession of the equinoxes has no effect on the mean annual insolation, but does modulate the amplitude of the seasonal cycle. In a linear climate system, there would be no energy near the 21,000 year precession period. It is only when a non-linear mechanism rectifies the seasonal modulation that precession-period variability appears. Such rectification can arise from physical processes within the climate system, for example a dependence of ice cover only on summer maximum insolation. The possibility exists, however, that the seasonality inherent in many climate proxies will produce precession-period variability in the records independent of any precession-period variability in the climate. One must distinguish this instrumental effect from true climate responses. Careful examination of regions without seasonal cycles, for example the abyssal ocean, and the use of proxies with different seasonal responses, might permit separation of physical from instrumental effects.

A new chronology of glaciation, spanning the last 780,000 years, is estimated from 21 marine sediment cores using depth as a proxy for time. To avoid biasing this ‘‘depth-derived’’ age estimate, the depth scale is first corrected for the effects of sediment compaction. To provide age uncertainty estimates, the spatial and temporal variability of marine sediment accumulation rates are estimated and modeled as an autocorrelated stochastic process. Depth-derived ages are estimated to be accurate to within ±9000 years, and within this uncertainty are consistent with the orbitally tuned age estimates. Nonetheless, the remaining differences between the depth and orbitally tuned chronologies produce important differences in the spectral domain. From the d18O record, using the depth-derived ages, we infer that there are weak nonlinearities involving the 100 kyr and obliquity frequency bands which generate interaction bands at sum and difference frequencies. If an orbitally tuned age model is instead applied, these interactions are suppressed, with the system appearing more nearly linear.