24The stable carbon isotope ratio of atmospheric CO 2 (! 13 C atm ) is a key parameter to decipher 25 past carbon cycle changes. Here we present ! 13 C atm data for the last 24,000 years derived 26 from three Antarctic ice cores. We conclude that a pronounced 0.3‰ decrease in ! 13 C atm 27 during the early deglaciation can be best explained by upwelling of old, carbon-enriched 28 waters in the Southern Ocean. Later in the deglaciation, regrowth of the terrestrial 29 biosphere, changes in sea surface temperature, and ocean circulation governed the ! 13 C atm 30 evolution. During the Last Glacial Maximum, ! 13 C atm and CO 2 were essentially constant, 31suggesting that the carbon cycle was in dynamic equilibrium and that the net transfer of 32 carbon to the deep ocean had occurred before then. showing pronounced differences in atmospheric CO 2 rates of change in the course of the 47 last glacial/interglacial transition (3). Many processes have been involved in attempts to 48 explain these CO 2 variations, but it has become evident that none of these mechanisms 49 alone can account for the 90 ppmv increase in atmospheric CO 2 . A combination of 50 processes must have been operating (4, 5), with their exact timing being crucial. However, 51 a unique solution to the deglacial carbon cycle changes has not been yet found. 52 53

Reconstructions of atmospheric CO 2 concentrations based on Antarctic ice cores 1,2 reveal significant changes during the Holocene epoch, but the processes responsible for these changes in CO 2 concentrations have not been unambiguously identified. Distinct characteristics in the carbon isotope signatures of the major carbon reservoirs (ocean, biosphere, sediments and atmosphere) constrain variations in the CO 2 fluxes between those reservoirs. Here we present a highly resolved atmospheric d 13C record for the past 11,000 years from measurements on atmospheric CO 2 trapped in an Antarctic ice core. From mass-balance inverse model calculations 3,4 performed with a simplified carbon cycle model, we show that the decrease in atmospheric CO 2 of about 5 parts per million by volume (p.p.m.v.). The increase in d 13C of about 0.25% during the early Holocene is most probably the result of a combination of carbon uptake of about 290 gigatonnes of carbon by the land biosphere and carbon release from the ocean in response to carbonate compensation of the terrestrial uptake during the termination of the last ice age. The 20 p.p.m.v. increase of atmospheric CO 2 and the small decrease in d 13C of about 0.05% during the later Holocene can mostly be explained by contributions from carbonate compensation of earlier land-biosphere uptake and coral reef formation, with only a minor contribution from a small decrease of the land-biosphere carbon inventory.The Holocene is the current interglacial period, starting about 11,000 years before present (11 kyr BP, where present is defined as AD 1950) following the Transition (here defined as 18-11 kyr BP) from the last glacial maximum. Variations in the atmospheric concentration of CO 2 during the Holocene were significant but small compared to glacial-interglacial changes of typically 100 p.p.m.v. (refs 5, 6). Yet a decrease of about 5 p.p.m.v. from 11-7.5 kyr BP could be observed, followed by an increase of about 20 p.p.m.v. to the pre-industrial level of about 280 p.p.m.v. (refs 1, 2, 7). Different explanations for these variations were discussed 7,8 , such as changes in the carbon inventories of vegetation, soils and peatlands 9 , in anthropogenic land use 10,11 , in sea surface temperature (SST) 7,12 , coral reef growth 13,14 or carbonate compensation 15 . The latter is a multi-millennial equilibration process of the atmosphere-ocean-sediment system and the weathering cycle. Moreover, model simulations of atmospheric CO 2 and d 13C during the Holocene have not provided an unambiguous quantitative explanation 7,8,16 . The major stumbling block has been the scarcity of reconstructions of d We focus on the evolution of the carbon isotopes on a timescale of a few thousand years. Therefore, we calculated a spline and its 1s uncertainty bands with a cut-off period of 5 kyr (Fig. 2). In a Monte Carlo simulation, standard deviations smaller than 0.07% were

The equilibration method is the present-day standard method for measuring delta18O in water samples. The mass-to-charge ratio of 45 is measured at the same time but generally not used for further analysis. We show that an improved equilibration method can be used for precise determination of delta17O in addition to that of delta18O, and therefore can estimate 17O excess values to a precision of better than 0.1 per thousand. To control the masking effect of the 14 times more abundant 13C on mass 45, we propose to use a chemical buffer in the water samples to keep the pH value and therefore the fractionation during the equilibration process of the 13C constant. With this improved method, the precision for the delta18O value could also be slightly improved from 0.05 to 0.03 per thousand. Furthermore, we discuss the influences of the amount of water, the temperature, the CO2 gas pressures, and changes in the pH during the measuring procedure on oxygen and carbon isotopes. We noticed that measured delta45 values are a good control for delta18O measurements. This study tries to fathom the possibilities and limitations of the equilibration method for measuring 17O excess values of water samples.