Radio-echo sounding measurements and ice-core synchronization at Dome C, Antarctica

We present a compilation of radio-echo sounding (RES) measurements of five radar systems (AWI, BAS, CReSIS, INGV and UTIG) around the EPICA Dome C (EDC) drill site, East Antarctica. The aim of our study is to investigate the differences of the various systems in their resolution of internal reflection horizons (IRHs) and bedrock topography, penetration depth, and quality of imaging the basal layer. We address the questions of the compatibility of existing radar data for common interpretation, and the suitability of the individual systems for Oldest Ice reconnaissance surveys. We find that the most distinct 5 IRHs and IRH patterns can be identified and transferred between most data sets. Considerable differences between the RES systems exist in range resolution and depiction of the basal layer. Considering both aspects, which we judge as crucial factors in the search for old ice, the CReSIS and the UTIG systems are the most valuable ones. In addition to the RES data set comparison we calculate a synthetic radar trace from EDC density and conductivity profiles. We identify ten common IRHs in the measured RES data and the synthetic trace. The reflection-causing conductivity sections are determined by sensitivity studies with the 10 synthetic trace. In this way, we accomplish an accurate two-way travel time to depth conversion for the reflectors, without having to use a precise velocity-depth function that would accumulate depth uncertainties with increasing depth. The identified IRHs are assigned with the AICC2012 time scale age. Due to the isochronous character of these conductivity-caused IRHs, they are a means to extend the Dome C age structure by tracing the IRHs along the RES profiles.


Introduction
To predict the future evolution of ice sheets, the knowledge of their past response to climate changes is inevitable.Ice cores are the perfect archives to study the climate of the past.In contrast to other climate archives, they contain actual paleo-atmosphere in the form of air bubbles that are trapped in the ice.With this advantage, they are the only means to answer some important questions in climate studies with respect to greenhouse gases, e.g., why did the glacial-interglacial cycles change from 40 ka to 100 ka at the mid-Pleistocene transition (MPT) and what drove the 40 ka cycles (Raymo et al., 2006)?For this reason the International Partnerships in Ice Core Sciences (IPICS) included the retrieval of the "Oldest-Ice" ice core as one of their The Cryosphere Discuss., doi:10.5194/tc-2016Discuss., doi:10.5194/tc- -147, 2016 Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License.with the vacuum wave speed c 0 and the real part of the complex relative dielectric permittivity where ε 0 is the ordinary dielectric permittivity of vacuum and ω the circular frequency.
As the accuracy of the DEP measurement performed at EDC does not allow for an inversion of the complex-valued permittivity of a two-phase mixture, as described by Eisen et al. (2006), we use the real-valued dielectric mixture equation by Looyenga (1965) to calculate ε (z) from ρ(z): with the measured density ρ(z) and the pure-ice values for density and permittivity ρ ice = 917 kg m −3 and ε ice = 3.17.The determination of the latter value is described in Sect.2.3.Neglecting the complex character of the relative permittivity in the two-phase mixture leads to errors in the permittivity, especially in the firn (Wilhelms, 2005).However, for the purpose of reproducing reflections, not the absolute value but the changes of conductivity are important.Though the positions of the reflections in the two-way travel time (TWT) domain are shifted for incorrect ordinary permittivities (the real part of the complex relative permittivity), we avoid these errors by calibrating the synthetic with the measured radar trace (see Sect. 2.3).
The permittivity record is smoothed with a 0.2m running mean to prevent the masking of the conductivity-induced reflections and the too quick reduction of the propagating energy in the synthetic radar trace by a multitude of permittivity-induced reflections in the firn section.More extensive reasoning and effects of this proceeding can be found in Eisen et al. (2006).
Permittivity and conductivity, processed as described above, are input parameters for the 1D-FD model "emice" (Eisen et al., 2004) that calculates synthetic radar traces by solving Maxwell's equations.The depth increment of the model domain is 0.02 m.The maximum depth is 3165.2m and an absorbing boundary is implemented.The time increment is 0.02 ns, which fulfills the Courant Criterion.
Following Eisen et al. (2006), we use a source wavelet of two and a half 150MHz cycles that is relatively short for determining reflector depth in high resolution, but long enough to reproduce some interference effects.
The envelope of the calculated trace corresponding to the reflected energy is obtained by conducting a Hilbert magnitude transformation.Finally, the trace is smoothed with a Gaussian running mean of 150 ns.The result of this step is the synthetic trace that we use for comparison with the measured RES data, as described in Sect. 5.

The relative permittivity of ice
Equation (1) defines the dependence of the electromagnetic wave speed in ice on the relative permittivity of ice.For a smaller permittivity the wave speed is higher and a distinct reflection does thus appear earlier.This time difference increases with the absolute depth of the reflector.We determine ε ice for the Dome C region by comparing the synthetic traces of model runs with different ε ice in reference to the AWI RES data (see Sect. 3.1).For too small or too large values of ε ice , the time lags between RES and synthetic trace for identified reflections increase with depth.The best result is obtained with ε ice = 3.17.For this value, the identified reflections occur at the same TWT for both traces.This value is close to the commonly used ε ice = 3. 15 (e.g., The Cryosphere Discuss., doi:10.5194/tc-2016-147, 2016 Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License.-Morales et al., 2014) and to the pure isotropic ice value of ε ice = 3.16 found in laboratory experiments (Bohleber et al., 2012).

Rodriguez
3 Radio-echo sounding data The profiles closest to the drill site were selected from the RES data in the Dome C area.The distance between the drill site and the furthest profile is less than 2 km.The positions of the RES profiles relative to Dome C are shown in Fig. 1.As the influence of different radar systems on the recorded radargram is to be examined, we use the data of five different institutes: the Alfred Wegener Institute (AWI), Bremerhaven, Germany, the British Antarctic Survey (BAS), Cambridge, UK, the Center for Remote Sensing of Ice Sheets (CReSIS), Lawrence, USA, the Istituto Nazionale di Geofisica e Vulcanologia, Rome, Italy (INGV) and the University of Texas Institute for Geophysics (UTIG), Austin, USA.The characteristics of the different systems and the data processing are described next.The system characteristics are summarized in Table 1.

AWI
The airborne radar system of AWI is a burst system with a carrier frequency of 150 MHz that was operated in toggle mode with 60 ns and 600 ns bursts (Nixdorf et al., 1999).Measurements were conducted in austral season 2007/08 with the DC-3T aircraft Polar 5. We use the data with 60 ns bursts and stack it 10-fold.The stacked data have a trace distance of about 75 m and a vertical sampling interval of 13.33 ns.The profile passes the Dome C drill site in 280 m distance.

BAS
The BAS profile was recorded in season 2005/06 with an airborne radar system on a Twin Otter in 450 m distance to the drill site.The source is a 4µs chirp wavelet with a center frequency of 150 MHz and a bandwidth of 10 MHz.The vertical sampling interval is 45.45 ns.The data are unprocessed except for a horizontal smoothing with a 49 sample moving-average filter and 10-fold stacking.The trace distance after stacking is 45 m.

CReSIS
CReSIS had one campaign in the Dome C area in season 2013/14, using a 15-channel radar system on an Orion P3 aircraft.
The source wavelets are a 1, 3, and 10µs chirps, each running linearly through the frequency range of 180 MHz to 210 MHz (Gogineni, 2012).We use the L1B-data CSARP_standard file, processed with pulse compression, focused SAR processing, and array processing with multilooking (CReSIS, 2016).The final product has a vertical sampling interval of 33 ns and a trace distance of 30 m.The profile passes the drill site at 1745 m distance.

INGV
The passes the drill site in 65 m distance.A spiking deconvolution, low pass filtering and gain adjustment is conducted.We stack the data 10-fold.

UTIG
The radar profile of the UTIG was collected in season 2008/09 from a DC-3T aircraft in 150 m distance to the drill site.The system uses a 1µs chirp wavelet running linearly through the frequencies from 52.5 MHz to 67.5 MHz.The recorded data were filtered with a 10MHz band notch filter and a convolution.An automatic gain control is conducted and the data are stacked horizontally coherently ten times, log detected and incoherently stacked five times so the final trace rate is 4 Hz.This gives a trace distance of about 22 m for this product (Young et al., 2011;Cavitte et al., 2016).The vertical sampling interval is 20 ns.
4 Assembling the data sets Different system characteristics and processing result in different appearance of the RES data.Our aim is to compare the RES and synthetic radar data in terms of identifying distinct reflectors that can be found in all data sets and that can confidently be matched in between the different data sets.Our basis for determining the origin of the IRHs in the RES data is by relating them to the conductivity record.For this reason, we neglect the first few microseconds, i.e., the upper few hundred meters, where most reflections are due to density changes.
Like Karlsson et al. (2016), we found that this matching can best be achieved manually.But additionally we use a combination of two different ways of imaging: 1. Single traces as reflected energy versus TWT (A-scope) to compare the reflection peaks' shapes and positions.For every RES profile, the trace (of stacked data) closest to the EDC drill site is selected and plotted as a single trace.This is described in Sect.5.1 and shown in Fig. 2. To make the different data comparable, we first shift them in time so that the surface reflections are at TWT zero.Here, we use the maximum of the surface reflection peaks for the systems with chirp wavelet (BAS, CReSIS and UTIG), but its steepest slope for the pulse systems.This is motivated by the systems differing in signal generation and digitization.The depth of a reflector is depicted by the maximum of its reflection for the chirp systems, but by the rise of its reflection for pulse systems.Note that this is also valid for internal reflections.The position of the surface-reflection pick for all traces is marked by the vertical black line in the left panel of Fig. 2. The exponential trend is removed from every trace.The peak amplitudes of the reflections decline in a different manner for the different data, depending on the source wavelet of the radar system and the processing.For the data with strongly declining peak amplitudes we plot the logarithm of the amplitude for the deepest part (right panel) to make potential reflections in the basal layer more visible.
2. Radar profiles of several kilometer length around the closest trace to the drill site, plotted as TWT vs. trace number with amplitude values in gray scale (Z-scope), as shown in Fig. 3 and described in Sect.5.2.This way of imaging is especially suitable to compare specific sequences of reflections and to check whether the reflections matched in the A-scope image are spatially coherent and representative over larger regions, e.g. for extrapolation.Again, the surface reflections are shifted to 6 The Cryosphere Discuss., doi:10.5194/tc-2016-147,2016 Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License.
TWT zero and we plot the logarithm of the amplitudes for all RES data.For the deepest approximately third (bottom panel) we use differently processed, i.e. 2d-focused, CReSIS and UTIG data for an improved visibility of the deep internal structure.

Results
In this section we compare the different RES data and the synthetic radar data in order to match some reflections or reflection patterns distributed over the depth range.In Sect.5.3 we determine the depth origin of the identified horizons by means of sensitivity studies with the conductivity record.

Single traces, A-scope
Figure 2 shows the traces closest to EDC of all RES profiles, and the synthetic trace.The positions of the measured traces are marked by crosses in Fig. 1.The left panel of Fig. 2 shows the reflections of the air-ice interface, which marks TWT zero for each trace.The air-ice interface is determined by the location of the maximum, or of the first break of the peak for the chirp, and the pulse systems, respectively.
The second part of Fig. 2 shows the traces for the majority of the ice column.In this section we find a number of distinct reflections that can be identified in some or all of the traces.We highlight ten of them (H1-H10), shaded in gray, for which we are confident to have them matched correctly and use them for further discussion.Those events are also used for the sensitivity studies with the conductivity record.The conductivity profile that is used for calculating the synthetic trace is plotted in the bottom panel of Fig. 2. The parts of the profile plotted in red color are the conductivity signals that have to be removed for the identified reflections to disappear from the synthetic trace.In that way, we are able to assign the reflections with their depth and age, as described in Sect.5.3.What is striking, when comparing the different RES traces, is the comparatively low resolution of the INGV and BAS data.Multitudes of peaks from the other data are not separately resolved by the INGV and BAS systems, but combined to wider peaks.However, it is not always obvious, which peaks join together into one peak.Nevertheless, there are still reflections that are visibly similar to those in the other traces.E.g. the INGV and BAS peaks at about 15 µs (H5) can be matched with the one in the AWI data or the INGV peak just before 14 µs (H4) and the double peak at 17 µs (H6) with the UTIG data.In between these reflections, though, the appearance differs considerably from the other data.
There are small time shifts for the identified reflections in the different RES data.In the CReSIS trace, for example, the peaks are usually about 50-100 ns earlier than in the other traces.The time differences for the peaks in between the other data are much smaller.
The third panel shows the bed reflections of the measured RES data and, just before the bed reflection, the section of the basal layer.In this last section, a reflection can be found at about 30 µs that fits well in the CReSIS, UTIG high gain, INGV, BAS and synthetic data (H10).
The TWTs for the bed reflections fit well for AWI and UTIG data, with the UTIG reflections having a longer slope than the

Radargrams, Z-scope
Figure 3 shows Z-scopes of the five RES profiles.Unlike in Fig. 2, here the TWT serves as the y-axis.In the leftmost panel the synthetic trace is shown.In the second panel this trace is adjacently plotted 200 times with the amplitude in gray scale.
White noise is superimposed on each synthetic trace to get the appearance of a measured radargram.The other panels show the measured radar data of the five radar systems, processed as described in Sect.3. We use about 5 km profile length for all of the images with the exception of the INGV profile, which is only 2.7 km long.
Like with the single traces, again some reflectors can be matched nicely.Here, especially sequences of IRHs are striking.For example the three closely spaced reflectors at about 6.5 µs TWT in the synthetic radargram (red arrow in Fig. 3) that can also be found in the measured data, although with a slightly different appearance.In the AWI data, the first reflector is the most pronounced one, and in the UTIG data they are rather blurred into one broad reflector.Another nice example is the strong reflector just below 8 µs, followed by the wider sequence below 9 µs (H1 and H2, yellow double arrow) that catch one's eye in all of the RES data.Those can also be matched with events in the synthetic data, whereas in the latter, there exist more strong reflectors in between.Hereby, the one at 9 µs is especially distinct and has a counterpart only in the CReSIS data.The section from about 13 to 16 µs TWT (blue lines), with densely spaced, relatively strong reflectors, is also similar in all of the RES data.It is starting with a double-reflection, corresponding to H3 in Fig. 2, that can also be found in the synthetic radargram.
In the middle section of Fig. 3

Depths of the RES reflectors
We determine the depths of the IRHs, identified in Fig. 2 and Fig. 3, by sensitivity studies with the measured conductivity record (bottom panel of Fig. 2).This is possible, because the conductivity profile, measured on the ice core, is closely resembled by the synthetic radar trace and strong peaks can be matched confidently.The resemblance is not as obvious when comparing the conductivity profile and synthetic trace as a whole, but it clearly is, when looking at subsections.An exception is the very uppermost part ( ˜400 m), where the reflectivity is also influenced by density variations.The identified peaks are removed from the conductivity profile and the model is run with the changed profile as input.In the calculated synthetic radargram the according reflections are missing.The bottom trace of Fig. 2 shows the synthetic trace calculated with original (red) and changed (black) conductivity profile, and the bottom panel shows the corresponding conductivity profiles.If a reflection in the RES data can confidently be matched with one of the synthetic radargram, the depth of the reflection can immediately be transferred from the conductivity profile.It has to be taken into account, however, that the determined depth is the horizon's depth at the ice-core location, and may differ from its depth at the RES trace.Even so, the horizon can be assigned with an age.Due to the conductivity-induced IRHs being isochronous, their age is the same at the position of the RES profile, even if the depth is somewhat different.If matched correctly, the uncertainty of the reflector's age depends only on the width and number of the reflection-causing conductivity peaks and the accuracy of the age scale.The advantage over converting TWT to depth using ice-core densities and a velocity function is that the depth uncertainties do not accumulate with depth, but are independent of the absolute depth.
We remove sections from the conductivity record so that the gray-shaded reflections from Fig. 2 disappear from the synthetic trace.In that way we find that e.g., reflection H1 is caused by the conductivity peaks in the depth range of 700.54 m-702.64m, or that the sequence H2 is caused by a multitude of conductivity peaks, spanning about 22 m depth.All identified reflections, the depth ranges of their inducing conductivity sections, and their age ranges with standard deviation according to the AICC2012 timescale (Veres et al., 2013;Bazin et al., 2013) are listed in Table 2. Age uncertainties for the IRHs due to reflector width or shifts in TWT are discussed in Sect.6.2.2.

Discussion
In the sections below we discuss the results of our comparison of the different data sets.Section 6.1 addresses the comparison of the measured RES data among each other and gives possible reasons for differences between the data.In Sect.6.2 we look at the connection of the RES data with the synthetic radar trace, and thus the ice-core data, including an uncertainty assessment

Comparability of RES data
As pointed out in Sect.5.1 and 5.2, the RES systems depict the major characteristics of the conductivity profile in a very similar way and there are common features notable in all of the RES data (e.g., the strong IRH at 8 µs).Especially the AWI, BAS, CReSIS and UTIG radargrams show the same patterns of reflectors, like many densely spaced reflectors (e.g., starting from The Cryosphere Discuss., doi:10.5194/tc-2016-147,2016 Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License.
13 µs), or of lacking reflectors (22 µs, 32 µs) at the same depths.The distinct reflectors (that are usually chosen for tracing) can be identified and traced (at least for the lengths of the investigated profile sections) in those RES data.However, there are conspicuous differences, when comparing the data with respect to resolution and penetration depth.The differences and reasons are discussed in the sections below.

Resolution
There are dissimilarities in the RES data in markedness and vertical expansion of the reflectors.Those can partly be explained by the different range resolutions of the various radar systems, due to different source wavelets, receiver bandwidths, sampling rates and post-processing.The sampling intervals vertically range from 13.33 ns (AWI) to 45.45 ns (BAS) (see Sect. 3).
This gives approximately one sample every meter, and 3.8 m, respectively.The vertical resolution due to source wavelet and received-antenna system bandwidth ranges from about 3 m to 17 m, as listed in Table 1.To some extent, the differences in the RES data can therefore be attributed to the varying range resolution of the different systems.The systems with lower range resolution are not able to capture multiple closely-spaced conductivity changes, and these closely-spaced layer variations are only represented by a single reflector (of potentially complex shape).Regarding the resolution of IRHs at intermediate depths, we attest the AWI, CReSIS and UTIG systems the best quality, which is expected per the smaller range resolution identified in Table 1.The CReSIS system shows the most detailed structure, while the AWI system has the least penetration depth of the three systems.Obviously in the INGV and BAS radargrams, but also in the AWI and UTIG data, there are examples (at 10 µs and 20 µs) where a series of reflectors in CReSIS data are depicted as only one wider reflection.The detailed structure in the CReSIS data is advantageous for synchronizing IRHs at a specific location in high resolution, like we do in this study.
However, Cavitte et al. (2016), who use several radar data sets to connect the EDC and Vostok drill sites, point out that the high vertical resolution might make it difficult to trace the IRHs over wide distances as the IRHs can thin out more easily than for systems with lower vertical resolution and therefore more robust IRH returns.As for the compatibility of the data sets, we assume that there are no major issues in combining AWI, CReSIS and UTIG data at one location where the profiles are close to each other, due to their very similar reflector patterns.Yet, the additional resolution of the CReSIS data could add ambiguity to combined data interpretations that include tracing of IRHs (Cavitte et al., 2016).A similar problem arises if the lower resolved BAS and INGV data are included, as it might become difficult to decide which reflector to continue when going from lower to higher resolution data.
The horizontal resolution of the bed reflection is best in the CReSIS data.Whereas in the other data the bed reflection is somewhat blurred, we get a well-defined bedrock topography from the CReSIS data.However, these quality differences could BAS system, compared to the other two systems, applies also for the basal layer.

Synchronization of RES and ice-core data
We find that distinct patterns of IRHs in the RES data can also be found in the synthetic trace.The pronounced reflectors that are identified in all RES data can also be matched with the synthetic data.In this way their depth origin at the drill site and thus their age can be determined, as described in Sect.5.3.The uncertainties on depth and age are discussed in Sect.6.2.2.
However, like for the measured RES data, there are also pronounced differences in appearance between the RES and synthetic data.Possible reasons are discussed in Sect.6.2.1.

Spatial variability
The dissimilarities in the appearance of reflectors described for the comparison of the different RES data are even stronger when comparing any of the measured RES data with the synthetic trace.This is consistent with the explanation by different measuring sites and spatial variability in accumulation and ice thickness.The synthetic trace is calculated from the ice-core data.The EDC99 core and the RES profiles are a few hundred to two thousand meters apart.Additionally, the core, with its 0.1m diameter, samples a much smaller area than a radar system that averages over a footprint on the 10-100m scale.There were discussions about the spatial significance of ice-core signals (e.g., Fisher et al., 1985;Richardson and Holmlund, 1999;Veen and Bolzan, 1999).Palais et al. (1982) and Münch et al. (2015) examine the representativity of cores, using snow accumulation variability and stable-water isotope variations, respectively.Frezzotti et al. (2005), too, get slightly different accumulation rates from ice core and radar measurements and explain them with the different sample area.However, they find that for the Dome C region the differences of core and radar measurements of 3% and the spatial variability of accumulation are relatively small compared to other East Antarctic regions, and the smallest of their Dome C-Terra Nova Bay traverse.Wolff et al. (2005) compare the conductivity record of the EDC99 core with the one of the first drilling attempt (EDC96) that is 10 m away.They find that only in 45% of the cases the largest conductivity peak in a 10m section is also the largest peak in a 10m section of the parallel core.This is typical of low accumulation sites, because significant parts of a single year's signal can be lost at a location by post-depositional processes like snow drift.Gautier et al. (2016) evaluate the variability of the volcanic signal at the EDC site, using five 100m firn cores, drilled 1 m apart from each other.They find that the probability of missing volcanic events is 30% when using only a single core.The Tambora event (1815 CE), for instance, is not detected by their algorithm in two out of five cores.Wolff et al. (2005) suggest to use methods that sample a larger volume of ice to smooth out The Cryosphere Discuss., doi:10.5194/tc-2016-147,2016 Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License. the spatial inhomogeneity.RES being such a method explains that the RES data are more similar among each other than RES versus synthetic data.This implies also that a very strong RES reflector does not necessarily have to be a large peak in the conductivity profile or synthetic trace.

Uncertainty assessment
Uncertainties incorporated in the synthetic data are the uncertainties from the ice-core measurements, inaccuracies in determining the permittivity and the neglect of temperature and anisotropy.The uncertainties on the density decrease with depth, due to higher absolute densities and Hörhold et al. (2011) give a value of 0.66% at 100 m depth.The errors in the permittivity, induced by neglecting the complex character in the air-ice mixture, are negligible, as we exclude the firn section and look only at depths greater than 600 m, where the air partition is minor.Uncertainties on wave velocity due to the dependence of the permittivity on temperature and anisotropy both stay below 1% (Gough, 1972;Matsuoka et al., 1997;Fujita et al., 2000).
The errors in the synthetic trace have an influence only on the TWT of the synthetic reflectors, and thus eventually on matching those with the RES reflectors, but not on the actual depth and age assignment.Sensitivity studies with the conductivity profile define the depth range in which a synthetic reflection peak has its origin.Thus the depth uncertainties of the IRHs are given by the depth uncertainties of the DEP measurements and the resolution of IRHs in the synthetic trace The latter is adjustable, as it is defined by the bandwidth of the used source wavelet and the smoothing filter, as shown by e.g., Cavitte et al. (2016).The depth uncertainties of the DEP measurement are reported as about 2 mm by Wolff et al. (1999).Added to that comes the depth uncertainty of the ice core itself, which is difficult to quantify.Due to breaks in an ice core, an inclined borehole, and core relaxation after drilling the logged ice-core length is always different from the true depth.Parrenin et al. (2012, Section 2.2.1) estimate the offset to reach up to several meters for a deep drilling.The direct transfer of reflector depths by the sensitivity studies is not completely true for the RES reflectors.Because of different bandwidths used in the RES data, there is an uncertainty associated in matching the peaks between the synthetic and RES data sets.All of the BAS and INGV reflections, for example, are wider than the ones in the synthetic trace and thus it is not clear which conductivity peaks are incorporated in one RES reflector.The age uncertainties due to reflector width increase with depth and can be inferred from Fig. 4, which shows the gradient of age with depth (blue curve), and with TWT (black), respectively.Added to that comes the age uncertainties from the AICC2012 age scale itself, given in Table 2 (Veres et al., 2013;Bazin et al., 2013).The curves of Fig. 4 also give the magnitude of error in age that is avoided by our method compared to using only a RES profile, in some distance to the core, and the age scale of the core.For example: The RES profile is 1 km away from the core and the IRH of interest has a slope of 10 m km -1 in the direction drill site-RES profile.When using only the RES profile, we would assign the reflector with an age off by 10 m on the ice core time scale.For a reflector at about 10 µs TWT that would correspond to an age shift of about 1 ka, for a reflector at 32 µs of about 8 ka.This shows the advantage of our approach of first matching the RES IRHs with the synthetic radargram and only then determining the depth and age of the reflectors at the ice-core location, as it eliminates one possible source of error.The extent of error reduction, however, does depend on IRH slopes and distances of the RES profiles to the ice-core site.
In our study we compare the data sets of five different RES systems, addressing the questions of their compatibility for combined usage and suitability for informing potential "Oldest Ice" site characterization.All RES profiles were recorded within a 2km radius around the EPICA Dome C drill site, where the current oldest ice sample was retrieved.We find that the data are broadly comparable at that location and that the most-pronounced reflectors can be found in the RES data.The main differences between the RES systems are constituted by their resolution of englacial structure and bedrock and their quality in imaging the basal layer.et al., 2013;Bazin et al., 2013).
AWI, and the UTIG high gain a longer one than the low gain bed reflection.In the BAS, INGV and CReSIS trace the bed The Cryosphere Discuss., doi:10.5194/tc-2016-147,2016 Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License.reflections occur a few hundred nanoseconds earlier.Possible reasons for the differences in the timing of the bed reflections and internal reflections are discussed in Sect.6.1.2.
the reflectors at 19.8 µs and 20.5 µs (H7) are the most striking ones in the synthetic data (light blue double arrow).The most alike counterpart of this sequence is to be found in the BAS data.But there is also a match in the other RES data.Notable is that the first of the two reflectors is more pronounced in the UTIG and INGV data, whereas it is the second in the CReSIS data.Striking differences between the RES systems exist in the quality of recording reflections from the basal layer, shown in the lowermost panels of Fig.3.In the AWI data, where the IRHs are nicely resolved in the upper two thirds, the visibility of reflectors ceases at 28 to 29 µs.The only distinct reflector after that is the bed reflection 10 µs later, leaving about 800 m of echo-free zone.The same is the case for the INGV data.In contrast, IRHs are clearly visible down to about 33.5 to 34.0 µs TWT in the BAS, CReSIS and UTIG data, ending with a relative strong continuous reflector with strong vertical variation (green arrows).In the BAS data there even are some signals, spatially coherent for a few kilometers, as deep as approximately 36 µs TWT.The same can be found in the CReSIS and UTIG profiles a few tens of kilometers away from EDC, which are not shown here.At the very bottom of the figure, the bed reflections of the RES data can be seen.These measured bed-reflection depths are not to be compared with the strong reflection at the bottom of the synthetic radargram.The latter marks solely the margin of the model, which is equal to the end of the DEP record.At least 100 m are missing to the actual bed, which makes a difference of more than 1 µs in TWT.The CryosphereDiscuss., doi:10.5194/tc-2016Discuss., doi:10.5194/tc--147, 2016     Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License.
easily be induced solely by the different processing techniques, or to some extent influenced by the different locations of the profiles.Thus a comparison at this stage is of little help for judging the actual systems' capabilities in depicting the bed topography.If a more quantitative comparison of this aspect is required, we propose a survey with all systems measuring a common profile with a length of five to ten ice thicknesses.The Cryosphere Discuss., doi:10.5194/tc-2016-147,2016 Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License.et al. (2009) under the former name echo-free zone.The lack of IRHs in this range not being an issue of the systems' power is also supported by the fact that there are regions with to some extent spatially coherent signals almost down to the bedrock, like in the BAS data and also in the UTIG and CReSIS profiles, outside the 5km range shown in Fig. 3.It is difficult to give a comparing judgement on the quality of the three RES systems in this deepest part.The profiles were not measured at the same location and the internal structure in the ice close to the bed differs strongly.But as we still see some small scale structures and coherent signals, we are confident that all three systems are able to image structures in the basal layer reliably.As the basal layer comprises the old ice, this is a crucial factor for Oldest Ice reconnaissance surveys.The lower range resolution of the

Figure 1 .
Figure 1.The deep drill sites in East Antarctica and close-up to Dome C with the RES profiles.The crosses mark each profile's closest trace to the drill site.The location of the EDC drill site is marked by the red hexagon.

Figure 2 .
Figure 2. RES traces of the five radar systems and synthetic trace.The amplitudes are scaled individually and the exponential trend is removed.The surface reflection of each trace is shifted to time zero (left panel).Some distinct reflections that can be seen in some or all of the traces are gray-shaded.The bottom panel shows the conductivity profile.The conductivity peaks that are plotted in red cause the identified reflections H1-H10 in the synthetic trace.

Figure 3 .
Figure 3. Z-scope of synthetic and RES data sets of the five radar systems.The surface reflections are shifted to time zero as shown in Fig. 2 and the vertical red lines mark the positions of the traces of Fig. 2. The length of the RES profiles is indicated on the x-axes.For the depth axis we convert TWT to depth with a wave velocity of c = 168.5 m µs −1 and a firn correction of 10 m.For the bottom UTIG and CReSIS panels an extended focused is applied.The colored arrows and lines mark distinct reflector patterns, closer described in Sect. 5.

Figure 4 .
Figure 4. Gradient of the age from the AICC2012 age scale with depth (blue, left and bottom axes) and TWT (black, right and top axes) to infer age uncertainties due to reflector width and slope.

Table 1 .
The CReSIS data have the best horizontal resolution in the depth of the bed and are thus providing a well-defined subglacial bedrock topography.At this stage it is inconclusive, if this can be attributed to the CReSIS system, or rather to the different processing technique and profile location.If interested in well resolved IRHs at intermediate depths, the radar systems with the highest bandwidths, the AWI, CReSIS, and UTIG systems, are most suitable.However, we did not investigate the continuity of the IRHs beyond the 5 km profile lengths.Based on their close similarity in reflection patterns at the investigated location we assume that the AWI, CReSIS and UTIG data are smoothly combinable for common interpretation,although this yet has to be double-checked at cross-over points.The best quality in imaging the basal layer have the CReSIS, UTIG and BAS data, the latter, however, with comparably low resolution.For this reason we attest the CReSIS and UTIG systems the best suitability in our comparison for Oldest Ice reconnaissance surveys, as here the basal layer plays an important role.The AWI and INGV data in the current version are not as convenient for this purpose, as they fail to depict the internal structure in the deepest approximately third of the ice thickness at EDC.Nevertheless, the profiles could be used to close data gaps with respect to IRHs at intermediate depths and ice thickness.In addition to the profound comparison of the RES data, we synchronize the measured RES data with a synthetic radar trace for depth conversion.Input for the forward model for the synthetic trace are the EDC ice-core conductivity and density.We find that the RES data are more similar among each other than compared to the synthetic trace.This can be explained by the spatial variability of the strengths of single conductivity signals as sampled by ice cores and the smoothing effect of RES measurements due to their larger footprint and lower vertical resolution.Nevertheless, we are able to match 10 pronounced reflectors in the RES and synthetic data.The identified IRHs are conductivity-caused and thus isochronous.Hence the IRHs can be used to extend the age structure, provided by the Dome C ice core, to regions of interest for an Oldest Ice drill site.The CryosphereDiscuss., doi:10.5194/tc-2016Discuss., doi:10.5194/tc--147,2016Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License.Characteristics of the five RES systems.The fourth column gives either the bandwidth in case of the chirp systems, or the pulse length in case of the pulse and burst systems.The last column gives the best vertical resolution due to bandwidth (for distinguishing two reflectors), without taking into account any windowing of signals during processing.

Table 2 .
Identified layers of Fig.2, their approximate TWTs, depth ranges of their inducing conductivity sections and corresponding age with age uncertainties (average of published AICC2012 age uncertainties of the top and bottom depth) on the AICC2012 timescale (Veres TWT depth top depth bottom age top age bottom uncertainty The Cryosphere Discuss., doi:10.5194/tc-2016-147,2016 Manuscript under review for journal The Cryosphere Published: 14 June 2016 c Author(s) 2016.CC-BY 3.0 License.