ABSTRACT: The development of high quality age models for paleoclimate records is critical for constraining leads, lags, and rates of change within the climate system, which helps to elucidate the mechanisms of past climate changes. For example, improvements to age models for paleoclimate records from marine sediment cores were central to proving Milankovitch theory.

In this dissertation, I compiled a quality-controlled database of almost 300 previously published benthic foraminiferal oxygen isotope (d18O) records from globally distributed sites. The only way to directly date marine sediments is by radiocarbon dating, but this technique can only be used back to about 40,000 years ago and is subject to uncertainties about past surface reservoir age changes. Chapter 2 presents an averaged high-latitude North Atlantic reservoir age reconstruction based on aligning 33 deep North Atlantic benthic d18O records and comparing high- and low-latitude planktonic foraminiferal radiocarbon dates. This reconstruction shows increased reservoir ages during the beginning of the last deglaciation and the Younger Dryas, which suggests that previous chronologies based on the assumption of constant reservoir age need to be revised by up to about 1,000 years. North Atlantic reservoir ages are also a proxy for Atlantic meridional overturning circulation strength, so deglacial reservoir age excursions provide insight into the timing and role of ocean circulation changes during the deglaciation.

Regional benthic d18O stacks and chronologies fill an important niche between global stacks and studies comparing a few well-dated records. Chapters 3 presents seven different regional benthic d18O stacks with independent planktonic 14C-based age models for 0-40,000 years ago. The stacks show regional differences in the timing of benthic d18O change of up to 4,000 years during the last deglaciation. In Chapters 4 and 5, the Atlantic, Pacific, and Indian benthicd18O stacks are extended beyond the limits of radiocarbon to 145,000 years ago and combined into a single volume-weighted global stack. The regional stacks aligned to proxy data with well-constrained absolute age estimates provide improved chronostratigraphic constraints for the last glacial cycle. The volume-weighted global stack demonstrates errors of up to 4,000 years in the currently accepted global benthic d18O stack.