To better constrain the response of the Earth's climate system to continuing emissions, a better understanding of past climate change is essential. A key advance would be to understand the transition in the climate response to changes in orbital forcing during the 'mid-Pleistocene transition' (900 to 1200 ka ago). Marine records indicate that during this time the periodicity of the glacial cycles changed from 40 ka to our current 100 ka e.g.. The driver of this change is not well understood and in particular the role of the atmospheric CO2 and other greenhouse gases is of great interest. Therefore, a key goal in the ice-core community is to retrieve a continuous climate record of this transition as only ice cores contain the unique quantitative information about past climate forcing and atmospheric responses . An Antarctic ice core extending 1.5 Ma back in time (termed the “Oldest Ice” core) would provide not only a local Antarctic climate history but also global greenhouse gas concentrations . This would be key to unravel the linkages between the carbon cycle, ice sheets, atmosphere, and ocean behaviour. However, so far continuous ice-core records that may provide essential evidence about past mechanisms of climate change more than 1 Ma ago, have not been retrieved.

Under the umbrella of the International Partnerships in Ice Core Sciences (IPICS), the European “Beyond EPICA – Oldest Ice” (BE-OI) consortium and its international partners aim to retrieve an ice core up to 1.5 Ma old. As part of the pre-site survey plan, several extensive airborne operations have already been carried out. Campaigns have been conducted in the Dome C region, revealing several sites of interest and pinpointing areas for targeted exploration , and extending the known ages of the EPICA ice core to surrounding sites . Results from these surveys have provided additional constraints for modelling efforts to improve site predictions or to estimate geothermal heat fluxes . At Dome Fuji, two ice cores have been retrieved extending 330 and 720 ka back in time, respectively . The second deep ice coring project reached a depth of 3028 m where the ice is an estimated 720 ka and close to the melting point . However, in the wider region around the ice core site (Fig. ), the subglacial topography has so far been undersampled. The latest compilation of Antarctic ice thicknesses Bedmap2, relies mainly on Soviet airborne data from the 1970s with a large navigational uncertainty, and on ground-based Japanese surveys with a limited spatial coverage. Here, we present an updated ice thickness dataset for the Dome Fuji region including new data from two radar surveys carried out by the Alfred Wegener Institute, Helmholtz Centre for Polar and Marine Research (AWI). In this region, there is a likely presence of old ice according to earlier evaluations . We use the new dataset to update this prediction and issue recommendations for future field campaigns in the region.

We apply the one-dimensional thermokinematic model described in to the OIR data. The model calculates the minimum required geothermal heat flux that will cause the bed to reach the pressure melting point. It is based on the simplified model of , wherein the one-dimensional temperature equation is solved while neglecting horizontal advection caused by ice-flow and further assuming steady-state conditions. κ∂2T∂z2-w∂T∂z=0, where w is the vertical velocity, and κ=K/(ρc), where K is the thermal conductivity, ρ is the density, and c the heat capacity of ice. This approach neglects several processes that may influence the resulting geothermal heat flux, including changes over time in ice thickness or flow regime and horizontal advection. This is in particular problematic for areas that have experienced higher velocities than what is currently observed. For our study domain, ice thicknesses are thought to have varied between 50 and 250 m and are thus unlikely to have experienced substantial changes in flow regime.

In the original study by , the model was applied to the entire Antarctic Ice Sheet on a 5 km resolution grid. The input parameters were horizontal ice velocities (assumed to be equal to balance velocities in our region of interest, see Fig. S7 in the Supplement), surface mass balance , and surface temperature . The balance velocities were obtained from the balance fluxes, assuming that the mass of ice flowing out of any area exactly equals the sum of the inflow and the mass accumulated over the area e.g.. The vertically averaged horizontal velocities thus depend on ice thickness and the prescribed mass balance. The ice is assumed to be internally deforming, according to Glen's flow law with exponent n=3 cf.. Note that the balance velocities are based on the old ice thickness data. The geothermal heat fluxes are from three different studies ( and ). In our study, all above-mentioned parameters with the exception of ice thicknesses are identical to the fields used in but regridded to 1 km and 500 m resolutions using nearest neighbour interpolation. Thus, the difference in ice thickness between Bedmap2 and the OIR data is the main cause for the difference in results. As all the input parameters are smoothly varying on these spatial scales, we also expect a smooth minimum geothermal heat flux map especially as the geothermal heat flux datasets have a sparse spatial resolution of around 100 km. This implies that while the large-scale pattern is robust, detailed interpretation of features that are on a scale smaller than a few kilometres is not realistic.

The result from the calculation of the minimum geothermal heat flux is denominated Gminpred. It is compared to the three different geothermal heat flux datasets, as well as Gmeandata, the average of the three datasets and σG their standard deviation. The difference ΔG is then defined as ΔG=Gminpred-Gmeandata. Potential candidate sites for Oldest Ice are assumed to be areas where ΔG is larger than 5 mW m-2 and the standard deviation between the different geothermal heat flux datasets is low (σG<25 mW m-2). These two threshold values are chosen according to mean geothermal heat flux values for the whole interior Antarctic Ice Sheet. The values restrict the dispersion of geothermal heat flux values between data sets and reduce the likelihood of melting. The number of sites is further constrained by only considering areas where the ice thickness is above 2000 m and the horizontal ice-flow velocities are <2 m a-1. We refer readers to for an in-depth discussion about the choice of these parameter values.

The result of the model applied to the OIR data is shown in Fig.  in colours with black lines outlining the prediction of using Bedmap2. The colours of the figure show the value of ΔG. Larger values of ΔG indicate that based on current observations it is likely that the geothermal heat flux is so low that the temperature of the bed is below the pressure melting point. Note that the old prediction is based on both the simple model presented above and results from an ensemble of runs with a more advanced three-dimensional ice-flow model. The new prediction roughly outlines the one based on Bedmap2. The areas of high likelihood of Oldest Ice fall into the same three areas: one approximately 250 km to the west of Dome Fuji, one in a large sector immediately south of Dome Fuji and one approximately 200 km further to the south. Notably, the area to the west has decreased in size and areas that were considered likely to contain old ice are now no longer probable candidates for the chosen conditions. The two other areas have increased slightly in size. However, repeating the ensemble runs with the OIR dataset might modify this prediction but that is beyond the scope of this study.

Given that the thermokinematic model only depends on local parameters, the difference between our results and those presented in reflects the updated ice thickness information as all other parameters are identical. Thus, our results are sensitive to the uncertainty in ice thickness.

In order to make the best informed decision on the optimum drill site for retrieving Oldest Ice, we now return to the radar data for additional information. Based on a manual assessment of basal signal reflection strength and specularity of the signal cf.,, we have identified several areas that exhibit signs of the presence of subglacial water (blue dots, Fig. ). In the following we refer to these areas as “lakes” although in our preliminary analysis it is not possible to distinguish between a lake or, for example, water-filled sediments. An example of a radargram with two clear lakes is shown in Fig.  along with the corresponding layer continuity (cf. Fig. b) and the value of ΔG from the model (cf. Fig. ). Four additional radargrams showing examples of lakes and the layer continuity can be seen in the Supplement (Figs. S3–S6). The figures also give insight into how the model resolves small-scale features. In Fig. S4, two small lakes separated by a topographic high are visible in the radargram. In the kriged ice thickness data, this topographic high is smoothed out and to the model the lakes thus appear as one. Previous studies have also identified lakes in the region but our analysis adds several new locations (Fig. a).

A few of the lakes from the above-mentioned study do not appear in our dataset. This might be due to the uncertain geolocation from the old radar data or to the angle of interception of the new radar lines. The identification of these lakes is somewhat subjective and probably underestimates the amount of liquid water present at the bed. Assuming that the subglacial water follows the steepest gradient in the hydropotential , we estimate the potential drainage routes of the water. Fig. a shows the hydropotential (coloured contours), subglacial lakes (black dots), and their drainage routes (blue lines). The updated Oldest Ice prediction is outlined in black. Although several lakes are in the vicinity of sites that may contain Oldest Ice, the water is draining away from these regions, indicating that subglacial water is not easily introduced to the sites.

One key priority for drilling for an Oldest Ice core is not just the existence of old ice but also a sequentially stacked and undisturbed ice column. The internal stratigraphy of the ice, as imaged by the radar data, may provide a valuable constraint. Therefore, the radar data are analyzed for layer continuity using the method of . This automatic method gives an indication of the continuity of the layers, i.e. how coherent and easily traceable they might be. The analysis was conducted only on the OIR data collected during the 2016/17 field season, as the method relies on consistency in the radar system settings and processing chain in order to be comparable between radar lines. Figure b shows the continuity index smoothed by a moving window of 100 horizontal samples (∼ 3.5 km). Evidently, the layer stratigraphy southeast of Dome Fuji is markedly more continuous than layers north and west of the station. Indeed, the candidate sites for the Oldest Ice do not have lower layer continuity compared to some of the regions outside of the candidate sites. However, it should be emphasized that this does not preclude tracing of the layer. The continuity index merely indicates that there are potentially fewer well behaved layers, or that in order to resolve the layers a survey would be needed with higher vertical or horizontal resolution on the order of metres and kilometres, respectively, for example, by ground-based radar.

Based on the OIR ice thickness dataset, we identify two regions with the best potential for containing Oldest Ice. Our criteria include modelled ΔG values, ice-flow velocities, proximity to the Dome Fuji ice-core and local ice thickness. For the latter criterium, we make a qualitative assessment essentially assuming that shallower ice thicknesses result in lower time-resolution but we make no quantitative analysis here. The most promising region is also the most easily accessible region from the existing Dome Fuji station: the region immediately south and southeast of Dome Fuji (black box, Fig b). For this region, we favour two areas outlined with white and black rectangles in Fig. , that each contain two local maxima where ΔG is elevated. Although the entire region has a value of ΔG>5 mW m-2, we pinpoint the two areas in particular for two reasons: (1) the value of ΔG approaches and exceeds 12 mW m-2, thus the uncertainty in geothermal heat flux is less important for the robustness of the prediction. (2) The bed topography appears to consist of high plateaus rather than rugged mountain terrain. The second reason is important for layer continuity and to avoid disturbances to the lowermost part of the ice, that would inevitably impact the stratigraphy. The proximity to the ice core site also implies that extending the age–depth information from the ice core to a new drill site would be relatively straightforward. Especially the area “II” where the distance is small and the bed topography relatively smooth is a promising site.

The second region with a potential for Oldest Ice is the site west of Dome Fuji (blue dotted box, Fig. ). In the western part of this area, values of ΔG are also high, although at a first glance two properties make it less favourable: firstly, the horizontal surface velocities approach and exceed 2 m a-1 in this area, increasing the travel distance of the particles and thus making interpretation of an ice core more complicated; secondly, the ice is relatively thin in the region (typically less than 2.5 km) which may prove problematic for obtaining an adequate age resolution in an ice core. Furthermore, the distance (several hundreds of kilometres) to Dome Fuji makes it challenging to connect the internal layers to the ice-core chronology. Even so, layer tracing has been achieved across equal or longer distances in this part of Antarctica e.g., and a targeted campaign with radar systems optimized for layer clarity could address this issue.

To verify that the two regions could indeed provide a suitable drilling target for IPICS Oldest Ice objectives, we recommend first further investigations in these two regions. High-resolution radar measurements, ideally from the ground, are needed to identify the layer integrity especially in the basal region, i.e. the lowermost 20 % of the ice. Temperature measurements in boreholes in the upper ∼ 600 m of the ice column need to be merged with measurements of the vertical velocity by phase-sensitive radar systems e.g.. Using those datasets with ice-flow modelling and the age–depth distribution extrapolated from the Dome Fuji ice core would provide better estimates of the age near the bed as well as the respective annual layer thickness, which constrains the applicability of currently available ice core analytics . This would provide further constraints to localize areas suitable for rapid access drilling e.g. to be deployed in a second step, which would enable a preliminary analysis of climate proxies and thereby constrain the age of the ice at the sample site by comparing directly with marine climate records.

A new ice thickness dataset for the Dome Fuji region has been constructed from airborne radar data with the aim of improving predictions of sites that may contain ice that is older than 1.5 Ma. The new data resolve the topography in substantially higher detail than previously published data, revealing a landscape of valleys and highlands. A manual assessment of the data also identified several areas exhibiting signs of the presence of subglacial lakes, and our analysis indicates that any subglacial water drains away from potential Oldest Ice sites.

We use the new data to force a thermokinematic model and update the prediction for candidate sites. Based on the model results, we presented a new assessment of areas in the Dome Fuji region where the presence of ice older than 1.5 Ma is likely. We identified two regions where the available margin for geothermal heat flux uncertainties is large enough to sustain old ice over several glacial–interglacial cycles. One such region is south of the Dome Fuji station, and within this region especially two areas are of substantial interest. We recommend further targeted investigations to these areas to ascertain layer continuity and to establish approximate age–depth information.