This is a summary of the main CO2 results obtained from the Vostok core which have been presented in two papers recently published (Barnola and others 1987; Genthon and others 1987).

Previous results of ice-core analysis have already provided valuable information on atmospheric CO2 variations associated with anthropogenic activities (Neftel and others 1985, Raynaud and Barnola 1985[a], Pearman and others 1986) and with climatic variations back to about 40 ka ago (Delmas and others 1980, Neftel and others 1982, Raynaud and Barnola 1985[b]). The Antarctic Vostok ice core provides a unique opportunity for extending the ice record of atmospheric CO2 variations over the last glacial–interglacial cycle back to the end of the penultimate ice age, about 160 ka ago.

CO2 measurements were made at 66 different depth levels on the air enclosed in the 2083 m long core taken at Vostok Station. The air was extracted by crushing the ice, under vacuum, in a cold-room, and analysed by gas chromatography (Barnola and others 1983). The selected sampling corresponds to a time resolution between two neighbouring levels which range approximately from 2000 to 4500 years. The ages quoted in this abstract are based on the Vostok ice chronology given by Lorius and others (1985) and take into account the fact that the air is trapped in the firn well after snow deposition (between about 2500 and 4300 years after precipitation in the case of Vostok). The CO2 variations observed are compared directly with the changes in Antarctic temperature as depicted by the stable-isotope record of the Vostok ice (Jouzel and others 1988, this volume).

Furthermore, a CO2-orbital forcing-climate interaction is suggested by spectral analysis of the CO2 and temperature profiles, which both show a concentration of variance around orbital frequencies. The temperature profile is clearly dominated by a 40 ka period (which can be related to the obliquity frequency) (Jouzel and others 1988, this volume), whereas the CO2 record exhibits a well-defined 21 ka peak (which can be related to the precession frequencies) and only a weak and doubtful 40 ka peak. To check the relative influence of CO2 and orbital forcings on the temperature at Vostok, we modelled the temperature signal deduced from the stable-isotope record of the ice as a response to CO2, Northern Hemisphere ice volume and local insolation forcings. The results indicate that more than 90% of the temperature variance can be explained by these three kinds of forcing and that the contribution of the CO2 radiative effect associated with an amplification factor (which should reflect the long-term feed-back mechanisms) lies between 27 and 85% of the explained variance. This approach stresses the important role that CO2 may generally have played in determining the Earth’s climate during the late Pleistocene.

The most obvious feature of the Vostok CO2 record lies in its high correlation (r2 = 0.79) with the climatic record. The results obtained show high CO2 concentrations during warm periods (mean CO2 values of 263 ppm volume for the Holocene and 272 ppm volume for the last interglacial period) and low concentrations (between about 240 and 190 ppm volume) over glacial periods. Within the last glaciation, the CO2 concentrations are higher during the first part (mean CO2 value of 230 ppm volume between about 110–65 ka B.P.) than during the second part (203 ppm volume between 65–15 ka B.P.); the second part also indicates that climatic conditions were colder.

Our results point to some limitation on the possible mechanisms driving the atmospheric CO2 variations and, in particular, the influence of some oceanic areas or of changes in sea-level (see, for example, Broecker and Peng 1986). The weak 41 ka cycle (this cycle seems to be a characteristic of the spectra of the proxy data for high latitudes) in our CO2 record suggests that high latitudes may not have a major influence on CO2 variations. Furthermore, the phase relationship between CO2 and the temperature variations indicates that at the beginning of the two deglaciations around 145ka B.P. and 15ka B.P., taking into account the time resolution of our profile, the CO2 increases roughly in phase with the Vostok temperature. As surface-temperature changes around Antarctica are expected to lead to changes in sea-level (see, for instance, CLIMAP Project Members 1984), our results suggest that the CO2 increase cannot lag the increase in sea-level and thus that this parameter cannot initiate the CO2 variation recorded at the beginning of those two deglaciations. Nevertheless, this does not rule out influence of variations in sea-level on atmospheric CO2 for other periods of interest, in particular during the last interglacial–glacial transition, where the CO2 lags the Vostok temperature.