CHARACTERIZATION OF CLIMATE VARIABILITY AND INTERHEMISPHERIC CLIMATE LINKAGES ON MILLENNIAL TIME SCALES
We find that a first-order autoregressive (AR(1)) stochastic climate model with a physical timescale of t=600 +/- 300 years is a self-consistent explanation for the Antarctic record during the glacial period. While the character of the rapid warming events in Greenland precludes any autoregressive process from being a complete explanation of the data, AR(1) with t=400 +/- 200 years is a better characterization than stochastic resonance as it has been formulated in the literature. Adding a simple threshold rule to AR(1) can account for the asymmetries (though not the 1500-year spectral peak, which remains enigmatic).
Our analyses support recent results showing that 10 to 25% of the variance in the Byrd record at sub-Milankovitch timescales may be explainable by the time-integral of the GISP2 record. However, most of this shared variance occurs during longer episodes often associated with Heinrich events, rather than the shorter Dangaard-Oeschger events. Even for the larger North Atlantic events, regional (non-global) variability dominates in both records. One would therefore not expect that individual warming or cooling events can be directly correlated between these records. One would also not necessarily expect strong correlation with tropical climate records, even if the tropics were the driver of climate variability on these timescales. Finally, we find that the characteristic timescales for these records are significantly shorter during the Holocene, suggesting that the processes determining the pacing of millennial-scale variability is fundamentally different during glacial vs. interglacial regimes.