The analysis of air bubbles trapped in polar ice permits the reconstruction of atmospheric components over various timescales. Past evolution of greenhouse gases, such as carbon dioxide (CO2), lies on the frontline of paleorecords understanding. Within this study, the glacial interglacial oscillations of CO2 will be examined for the last 160,000 years. This period encompasses two deglaciations.The simultaneous analysis of the stable carbon isotope composition (δ13CO2) allows to better constrain the global carbon cycle. Based on the different isotopic signatures of the ocean and the terrestrial biosphere (major reservoirs responsible for the CO2 oscillations on a glacial interglacial scale), δ13CO2 contributes in distinguishing the major sources of CO2 for the studied periods.The LGGE method of gas extraction from ice was used in combination with a new instrumental setup to investigate the CO2 mixing ratio and its stable carbon isotope composition in air from the two last deglaciations at the EPICA Dome Concordia site in Antarctica. Being challenged from the different ice properties corresponding to the two major periods (being in bubble form for the last and in clathrate form for the penultimate deglaciation), the resulting averaged 3-expansion 1-sigma uncertainty (0.98 and 1.87 ppmv for CO2, respectively), accompanied by an averaged 0.1 1-sigma for δ13CO2 for both periods were satisfying enough to exclude any artefact scenario in the experimental protocol. The resolution of our results (~250 and ~730 years, for last and penultimate deglaciation) allows us to divide Terminations (T) into sub-periods, based on the different slope CO2 experiences. For TI, the four sub-periods revealed climatic events for both hemispheres (e.g.: Heinrich I, Bölling/Alleröd, Antarctic Cold Reversal, Younger Dryas), as also shown from polar and oceanic proxies. For the case of TII, a similar dynamic pattern between CO2 and δ13CO2 is seen as for TI, but the synchronization of oceanic events in our atmospheric record is more delicate due to higher data uncertainties one encounters for such a time scale.Our results show a ~80 ppmv CO2 increase throughout TI, which is coherent with previously published studies. The δ13CO2 shows a deglacial ~0.6 decrease accompanying the CO2 rise, showing clear trends during the different sub-periods. TII shows similar trends as for TI but of a larger magnitude: we therefore observe a ~110 ppmv rise associated with a ~0.9 decrease. Several scenarii can explain the abrupt deglacial CO2 increase, but there is presently no consensus on the exact causes and their respective role. Still, it is presumed that the ocean reservoir contributes the most. As a first interpretation of the obtained TI coupled CO2 and δ13CO2 dataset, the use of two C cycle box models is applied, validating the initial dominant oceanic role. The use of polar and oceanic proxies for the atmosphere and the ocean, superposed with our atmospheric signal should provide some responses on the similarities and differences of both deglaciations. Similarities potentially concern forcing factors and the amplifying role of the climatic system towards the external forcing, while differences mainly concern the different relative timing and magnitudes.