The reconstruction of δ13CO2 using Antarctic ice cores promises a deeper understanding on the causes of past atmospheric CO2 changes. Previous measurements on the Taylor Dome ice core over the last 30,000 years (Smith et al., 1999) indicated marine processes to be dominating the significant δ13CO2 changes over the transition, whereas glacial δ13CO2 was only slightly depleted relative to the Holocene (Leuenberger et al., 1992; Smith et al., 1999). However, significant uncertainty and the low temporal resolution of the Taylor Dome δ13CO2 data prevented a more detailed interpretation.Recently, substantial improvements have been made in the analysis and the resolution of ice core δ13CO2 records (Elsig et al., 2009; Lourantou et al., 2010). With these and new measurements presented here, three independent δ13CO2 data sets over the last glacial/interglacial transition have now been derived from the two EPICA and the Talos Dome ice cores. Two of the methods use traditional dry extraction techniques with a reproducibility of 0.07-0.1. The third method uses a novel sublimation technique with a reproducibility of 0.05. Here we compare the data sets, their analytical setups and discuss their joint information as well as their differences. The three records provide a more detailed picture on the temporal evolution of δ13CO2 and confirm two pronounced isotope minima between 18-12,000 years BP in parallel to the two major phases of CO2 increase (Lourantou et al., 2010; Smith et al., 1999) as also reflected in marine sediments (Marchitto et al., 2007; Skinner et al., 2010). Accordingly, a release of old carbon from the deep ocean is most likely responsible for a large part of the long-term increase in atmospheric CO2 in this time interval. However, the fast CO2 jumps at a round 12,000 and 14,000 years BP may be partly of terrestrial origin (Elsig, 2009; Köhler et al., 2010b). The new sublimation data set provides also unambiguous δ13CO2 data for clathrate ice in the LGM. This shows a rather constant δ13CO2 level, which is only about 0.1 lower than the Holocene, despite significant changes in the terrestrial and marine carbon storage. Accordingly, during the LGM the changes in the different processes acting on the glacial carbon cycle largely compensate each other with respect to δ13CO2 as predicted by carbon cycle modeling (Köhler et al., 2010a).References:Elsig, J. (2009), PhD thesis, University of Bern.Elsig, J. et al. (2009), Nature 461, 507-510.Köhler, P. et al. (2010a), Paleoceanogr. 25, doi:10.1029/2008PA001703.Köhler, P. et al. (2010b), Climate of the Past Disc. 6, 1473-1501.Leuenberger et al. (1992), Nature 357, 488-490.Lourantou, A. et al. (2010), Global Biogeochem. Cycles 24, doi:10.1029/2009GB003545.Marchitto et al. (2007), Science 316, 1456-1459.Skinner, L. C. et al. (2010), Science 328, 1147-1151.Smith, H. J. et al. (1999), Nature 400, 248-250.