We measured the methane mixing ratios of enclosed air in five ice core sections drilled on the East Antarctic Plateau. Our work aims to study two effects that alter the recorded gas concentrations in ice cores: layered gas trapping artifacts and firn smoothing. Layered gas trapping artifacts are due to the heterogeneous nature of polar firn, where some strata might close early and trap abnormally old gases that appear as spurious values during measurements. The smoothing is due to the combined effects of diffusive mixing in the firn and the progressive closure of bubbles at the bottom of the firn. Consequently, the gases trapped in a given ice layer span a distribution of ages. This means that the gas concentration in an ice layer is the average value over a certain period of time, which removes the fast variability from the record. Here, we focus on the study of East Antarctic Plateau ice cores, as these low-accumulation ice cores are particularly affected by both layering and smoothing. We use high-resolution methane data to test a simple trapping model reproducing the layered gas trapping artifacts for different accumulation conditions typical of the East Antarctic Plateau. We also use the high-resolution methane measurements to estimate the gas age distributions of the enclosed air in the five newly measured ice core sections. It appears that for accumulations below 2
The East Antarctic Plateau is characterized by low temperatures and low accumulation rates. This creates the conditions for the presence of very old ice near the domes in this region
However, the gas records in low-accumulation ice cores cannot be interpreted as perfect records of the atmospheric history. Indeed, due to the process of gas trapping in the ice, two distinct effects create discrepancies between the actual history of the atmosphere and its imprint in the ice. The first one is due to the heterogeneous structure of the firn when transforming into airtight ice with bubbles. The overall structure of the firn column is characterized by a progressive increase in density with depth, associated with the constriction of the interstitial pore network
In order to properly evaluate the composition of past atmospheres based on the gas records in polar ice cores, it is thus necessary to characterize the specificities of those two effects. Gas age distributions can be calculated for the purpose of estimating firn smoothing. In the case of modern ice cores this may be accomplished by using gas trapping models parameterized by firn air and pore closure data
Layered gas trapping, on the other hand, originates from firn heterogeneities and is a stochastic process. Moreover, current gas trapping models do not fully represent the centimeter-scale variability in the firn. The proper modeling of layered gas trapping is thus limited by the lack of knowledge of glacial firn heterogeneities and the difficulty to model gas trapping in a layered medium.
For this work, we analyzed methane concentrations in five ice core sections from the East Antarctic Plateau. These sections cover accumulation and temperature conditions representative of East Antarctica for both the glacial and interglacial periods. The new data are used to test the layered gas trapping model proposed by
The newly measured ice core sections originate from the East Antarctic sites of Vostok, Dome C, and Lock-In. The three sites are displayed on the map in Fig.
Summary of the different ice core sections studied in the paper with their associated atmospheric references.
The five ice core sections were analyzed for methane concentrations using a continuous flow analysis (CFA) system, including a laser spectrometer based on optical-feedback cavity-enhanced absorption spectroscopy
The obtained records present numerous gaps, ranging from a few centimeters to several meters. Several reasons explain the presence of such gaps. First, the space between consecutive melting sticks lets modern air enter the CFA system, resulting in a contamination and abnormally high methane concentrations. Moreover, the presence of cracks and fractures in the ice might also let modern air enter the measured ice stick itself, also resulting in abnormally high concentrations. The moments of potential air intrusions were recorded during the measurement campaigns, and the data were screened to remove the resulting contaminations, creating gaps in the record
In order to correct the measured mixing ratios for the preferential dissolution of methane in water, a correction factor is applied to the data to raise them to absolute values
The discrete methane measurements from the high-accumulation West Antarctic Ice Sheet (WAIS) Divide (WD) ice core published by
Note that in the case where our CFA data and the independent absolute measurements do not originate from the same ice core, the match between the datasets is performed on sections with low methane variability to reduce the influence of firn smoothing. The signals after correction for the dissolution of methane are displayed in light blue in the upper panels of each part of Figs.
Illustration of a corrected CFA record (in black) by matching an already-calibrated record
In Sect.
For the modern Lock-In and Dome C sections, we used the discrete measurements from the WD ice core, also used for correcting the methane preferential dissolution
Yet, we observed that the NEEM record cannot be directly used as an atmospheric input for the DO21 period. As explained in Sect. S1 of the Supplement, the first feature of the DO21 (around the depth of 1260 m in Fig.
Measured and modeled modern
Same as Fig.
It has been observed in the high-accumulation firn of DE08 that not all layers close at the same depth
In accordance with
The usage of high-resolution measurements allow us to easily distinguish layering artifacts as abrupt spikes exceeding the analytical noise in the record. However, with a lower-resolution technique, such as discrete measurements, it is possible to inadvertently measure a layering artifact without realizing that it is not representative of its surrounding concentrations, which would result in an anomalous point in the record. As a specific example,
Important late-closure artifacts are unusual in methane gas records. The new measurements confirm this observation. Late-closure artifacts that should appear as positive methane anomalies at the onset of the DO events are almost absent in our measurements. This can be explained by the fact that the surrounding firn layers are already sealed and prevent gas transport to greater depths, which in turn prevents the late-closure layers from enclosing young air
In orange: discrete methane measurements of the DO8 event in the EPICA Dome C record
The EDC96 methane data suggest a lack of layering artifacts affecting the DO6 event (around the depth of 700 m in Fig.
To evaluate the potential impact of ions on layered gas trapping artifacts in the EDC96 methane record, we compared the high-resolution methane data with total calcium data measured in the ice phase
The ability to model layered gas trapping helps quantify and predict its impact on gas records. For this purpose, we revised the simple layered gas trapping model proposed by
The model requires a densification rate for the firn–ice transition zone. Observations of density profiles in Dome C and Vostok proposed by
The depth ranges of the expected modeled artifacts are displayed below the newly measured signals in Figs.
Same as Fig.
Since layered gas trapping artifacts are not due to chronologically ordered atmospheric variations, a proper interpretation of gas records requires them to be removed from the datasets. As pointed out by
As the layering artifacts can be visually distinguished, it would be possible to remove them manually from the data. However, such a procedure would be cumbersome due to the large number of CFA data points. Hence, to clean the three new methane signals presenting layering artifacts (EDC96, EDC99 and Vostok) we use a recursive cleaning procedure similar to the algorithm described by
This procedure was successfully applied to the Holocene section of EDC99 that only exhibits a few clear artifacts near 323 m. The resulting cleaned signal is displayed in the upper panel of Fig.
Direct application of this algorithm to the DO6-9 EDC96 and DO21 Vostok records leads to an ineffective removal of layering artifacts in periods of fast methane rise. After a detailed investigation, it appears that the main issue the algorithm is facing is the determination of the signal free of artifacts by the running median. In some sections of the ice cores, the signal displays high numbers of artifacts with methane anomalies of 50 ppbv or more, as seen for instance in the onset of the DO8 in Fig.
The five ice core sections measured in this paper all originate from the East Antarctic Plateau. The low temperatures and aridity of this region result in a slow densification of the firn and a slow bubble closure. Therefore, the gas enclosed in a given ice layer tends to have a broad age distribution. This leads to an significant smoothing of atmospheric fast variability observed in East Antarctic ice cores
Since smoothing in an ice core record is a direct consequence of the gas age distribution (GAD), the knowledge of GADs for various temperature and accumulation conditions is necessary to predict the impact of smoothing on gas signals. We thus apply the GAD extraction method proposed by
This method, designed to estimate the GAD of low-accumulation records, is based on the comparison with a weakly smoothed record derived from a high-accumulation ice core used as an input atmospheric reference (see Sect. 3.2). The idea of the method is to find a GAD that is able to smooth the atmospheric reference to the level of the low-accumulation record. We thus searched for the GAD that minimizes the root-mean-square deviation between the CFA measurements and the smoothed version of the atmospheric reference. In order to have a well-defined problem in a mathematical sense, the GAD of the low-accumulation ice core is assumed to be a log-normal function. Such a log-normal distribution is fully defined by two independent parameters. Finding the best GAD to smooth the atmospheric reference is then reduced to the recovery of a pair of optimal parameters. Nonetheless, log-normal distributions exhibit a large range of shapes that can adequately represent age distributions
The gas age distributions obtained for the five new ice core sections are displayed as solid lines in Fig.
Gas age distributions in various ice cores. In order to be more easily compared, the age distributions are all drawn relative to their median gas age. This is why parts of the distributions have negative ages in the graph.
Parameters defining the log-normal gas age distributions derived by comparison with weakly smoothed records. The location and scale parameters correspond to the
The degree of smoothing in a gas record is strongly linked to the accumulation rate under which the gases are trapped, with low-accumulation sites exhibiting a stronger degree of smoothing
Our results suggest that during the glacial period, East Antarctica ice cores are affected by the same level of smoothing that can be represented with the same gas age distribution. We produced two new glacial period gas age distributions (DO6-9 EDC96 and DO21 Vostok) that are to be added to the previously published GAD obtained for the Vostok site during the DO17 event using the same GAD extraction method
However, the uncertainty analysis in Sect. S3 reveals that the age distribution of the EDC96 record is poorly constrained. This indicates that the smoothing of the DO6-9 period is weakly sensitive to the choice of GAD and therefore that a large range of GADs results in adequate smoothing for the EDC96 record. On the other hand, the DO21 period is very sensitive to the choice of GAD, and much fewer age distributions are able to reproduce the smoothing of the DO21 Vostok record. Our understanding is that because of its shape and fast atmospheric rate of change
Our results also indicate that the glacial and interglacial smoothings of East Antarctic ice cores are relatively similar. Indeed, comparing the typical GAD of the glacial period with the ones obtained for the Holocene and modern periods indicates that the latter lead to a slightly smaller degree of smoothing. This is illustrated with the smoothing of the 8.2 ka event displayed in Fig.
Finally, the expected smoothing using the firn pumping GAD estimation for modern Vostok is also displayed in Fig.
Compared to the GADs of the Holocene and modern periods, the GADs of the glacial period exhibit a stronger degree of skewness (see Fig.
East Antarctica is a region of particular interest for the drilling of deep ice cores. Indeed, thanks to low accumulation rates, it likely contains the oldest stratigraphically undisturbed ice on earth. There is currently a search for very old ice, with ages potentially dating back 1.5 million years
In this section, we estimate the potential differences between atmospheric signals and their measurements in a theoretical 1.5-million-year-old ice core, for methane and carbon dioxide. Here, we consider that the gas trapping occurs under an accumulation rate of 2
As a first case, we study the alterations of a methane record measured using a CFA system analogous to the one currently used at IGE. For the sake of simplicity, we assume that the atmospheric history recorded in this theoretical 1.5-million-year-old ice core is similar to the DO events 15 to 17 of the last glacial period. Hence, we simply use the WD CFA methane measurements as the atmospheric reference
Impacts of layered trapping and smoothing on a hypothetical 1.5-million-year-old methane record. The signal without smoothing or artifacts is displayed in dashed orange. The recorded signal with layering is shown in light blue, and the recorded signal with smoothing only is shown in black. The synthetic CFA measurements taking into account the artifacts are displayed in red, while the synthetic measurements without artifacts are displayed in green. The synthetic depth scale is expressed relative to an arbitrary point of the record. For clarity the depth scale is expressed in centimeters.
The internal smoothing of the CFA system also tends to deteriorate information recorded in the ice core by attenuating fast variability. It is therefore important to determine to what extent CFA smoothing adds to the already-present firn smoothing. To study this point, we compare the frequency responses of the CFA system and firn smoothing. These frequency responses represent the attenuation experienced by a sine signal, as a function of the sine period. The smoothing function of the CFA system (its impulse response) was determined by
So far in this paper only methane records have been considered. However, smoothing and layering artifacts also potentially affect other gaseous records, such as carbon dioxide records. Since
The results are displayed in Fig.
We also compared the rate of change in carbon dioxide in the atmospheric reference and in the measured record. For some of the fast variations in atmospheric
Impacts of layered trapping and smoothing on a synthetic 1.5-million-year-old carbon dioxide record.
Impact of gas trapping processes and measurements on the
This work evaluated two gas trapping effects that affect the recorded atmospheric trace gas history in polar ice cores. The first one is the layered gas trapping that produces stratigraphic heterogeneities appearing as spurious values in the measured record
The characterization of layering artifacts in these new measurements is consistent with the physical mechanisms already proposed in the literature
In order to constrain the smoothing effect, we estimated the gas age distributions in each of the five ice core sections. For this purpose, we use the method proposed by
Finally, we applied our methodology to the theoretical case of a 1.5-million-year-old ice core drilled on the East Antarctic Plateau. Our results suggest that due to thinning, the layering artifacts will no longer be resolved during measurements. However, it appears that their potential influence on the end result measurements is rather low, with impacts below 10 ppbv and 0.5 ppmv for methane and carbon dioxide, respectively. In the case of methane variations during DO events, most of the record alterations originate from the firn smoothing. Yet, the limited spatial resolution of the CFA system induces a smoothing that is not entirely negligible compared to the one of the firn. For carbon dioxide, firn smoothing appears to significantly diminish the recorded rates of change in abrupt
The programs used for data processing and modeling were developed using Python 3.0 and available packages. They will be provided upon request to the corresponding authors.
The high-resolution Modern Dome C, Holocene Dome C, DO6-9 Dome C, and DO21 Vostok methane datasets are available on the World Data Center for Paleoclimatology:
The supplement related to this article is available online at:
This scientific project was designed by JC, XF, KF, and PM. JC and PM participated in the Lock-In drilling, and XF participated in the shallow Dome C drilling. AAE and VL made available and preprocessed the Vostok ice core section. The high-resolution methane measurements were carried out by XF and KF. The codes for data processing and modeling were developed by KF and PM. All authors contributed to the interpretation of the data. The paper was written by KF with the help of all coauthors.
The authors declare that they have no conflict of interest.
The opinions expressed and arguments employed herein do not necessarily reflect the official views of the European Union funding agency or other national funding bodies.
This work is a contribution to EPICA, a joint European Science Foundation–European Commission scientific program funded by the European Union and national contributions from Belgium, Denmark, France, Germany, Italy, the Netherlands, Norway, Sweden, Switzerland, and the United Kingdom. The drilling of the Vostok ice core and the logistics support were assured by the Russian Antarctic Expeditions. The Vostok ice core was made accessible in the framework of the Laboratoire International Associé (LIA) Vostok. The Lock-In drilling was supported by the IPEV (project no. 1153) and the European Commission, Seventh Framework Programme (ICE&LASERS (grant no. 291062)). We are grateful to the Lock-In field personnel David Colin, Phillipe Dordhain, and Phillipe Possenti for performing the drilling, as well as Patrice Godon for setting up the logistics. The shallow Dome C drilling was performed by Phillipe Possenti and supported by the IPEV project no. 902 and the French ANR program RPD COCLICO (grant no. ANR-10-RPDOC-002-01). We thank Robert Mulvaney and the British Antarctic Survey for the Fletcher Promontory ice core drilling and the permission to use the Fletcher methane data. We thank Grégoire Aufresne for his help processing the ice cores. We thank Grégory Teste for his help during the ice core processing and the CFA methane measurements. We thank Jochen Schmitt and Hubertus Fischer for their constructive comments. We are thankful for the two anonymous referees for reviewing this work, as well as for Hubertus Fischer for editing it. This is EPICA publication no. 313 and Beyond EPICA – Oldest Ice publication no. 10.
This publication was generated in the frame of Beyond EPICA. The project has been supported by the European Commission, Seventh Framework Programme (ICE&LASERS; grant no. 291062), and H2020 Research Infrastructures (BE-OI, grant no. 730258, and Beyond EPICA, grant no. 815384). It is supported by national partners and funding agencies in Belgium, Denmark, France, Germany, Italy, Norway, Sweden, Switzerland, the Netherlands, and the United Kingdom. Logistic support is mainly provided by ENEA and IPEV. This publication also benefited from financial support by the French CNRS INSU LEFE projects NEVE-CLIMAT and HEPIGANE, by the French ANR program RPD COCLICO (grant no. ANR-10-RPDOC-002-01), and by the Russian Science Foundation (grant no. 18-17-00110).
This paper was edited by Hubertus Fischer and reviewed by two anonymous referees.