A large perturbation in atmospheric CO 2 and O 2 or bioproductivity will result in a drastic pulse of 17 O change in atmospheric O 2 , as seen in the Marinoan Oxygen-17 Depletion (MOSD) event in the immediate aftermath of a global deglaciation 635 Mya. The exact nature of the perturbation, however, is debated. Here we constructed a coupled, four-box, and quick-response biosphereatmosphere model to examine both the steady state and dynamics of the MOSD event. Our model shows that the ultra-high CO 2 concentrations proposed by the "snowball' Earth hypothesis produce a typical MOSD duration of less than 10 6 y and a magnitude of 17 O depletion reaching approximately −35‰. Both numbers are in remarkable agreement with geological constraints from South China and Svalbard. Moderate CO 2 and low O 2 concentration (e.g., 3,200 parts per million by volume and 0.01 bar, respectively) could produce distinct sulfate 17 O depletion only if postglacial marine bioproductivity was impossibly low. Our dynamic model also suggests that a snowball in which the ocean is isolated from the atmosphere by a continuous ice cover may be distinguished from one in which cracks in the ice permit ocean-atmosphere exchange only if partial pressure of atmospheric O 2 is larger than 0.02 bar during the snowball period and records of weathering-derived sulfate are available for the very first few tens of thousands of years after the onset of the meltdown. In any case, a snowball Earth is a precondition for the observed MOSD event. (3,(6)(7)(8)(9). Available data show that for the past 750 My, tropospheric O 2 has had a small magnitude of 17 O depletion, except for one unusual episode: the immediate aftermath of the Marinoan glacial meltdown at 635 Mya, when the Δ 17 O O2 probably reached values as negative as ∼−40‰ (6), compared with the −0.34‰ in today's atmospheric O 2 (10). An ultra-high pCO 2 condition was proposed to explain this Marinoan Oxygen-17 Depletion (MOSD) event (3, 6), which is consistent with the "snowball" Earth hypothesis that argues for a completely icecovered globe lasting several million years (11,12). However, it is not known if there are other potential atmosphere-biosphere scenarios that might produce distinctively negative Δ 17 O O2 , such as a condition in which pCO 2 is moderate but pO 2 and/or biosphere O 2 flux are low, as proposed recently in an effort to reconcile organic and inorganic carbon isotope data from the postglacial cap carbonates (13). It therefore is imperative to examine the Δ 17 O O2 variability among geologically reasonable scenarios. Previous modeling studies on Δ 17 O O2 vary the ratio of pO 2 /pCO 2 , but with a fixed O 2 residence time, or focus on steady state without considering dynamic evolution of pO 2 , pCO 2 , or O 2 production fluxes (3, 13) or consider only minor Δ 17 O O2 difference between recent glacial and interglacial periods (14). In addition, a critical piece of information, the evolution of pO 2 before and after the meltdown of the most severe global glaciation in Earth history, w...