Carbon monoxide (CO) is a regulated pollutant and one of the key components
determining the oxidizing capacity of the atmosphere. Obtaining a reliable
record of atmospheric CO mixing ratios ([CO]) since preindustrial times is necessary
to evaluate climate–chemistry models under conditions different from today and to
constrain past CO sources. We present high-resolution measurements of CO
mixing ratios from ice cores drilled at five different sites on the Greenland
ice sheet that experience a range of snow accumulation rates, mean surface
temperatures, and different chemical compositions. An optical-feedback
cavity-enhanced absorption spectrometer (OF-CEAS) was coupled with continuous
melter systems and operated during four analytical campaigns conducted between
2013 and 2019. Overall, continuous flow analysis (CFA) of CO was carried out
on over 700
Carbon monoxide (CO) is a reactive trace gas that plays a crucial role in the
interactions between climate and atmospheric chemistry. CO strongly affects
the global oxidative capacity of the atmosphere by acting on the budgets of
both the hydroxyl radical (OH) and ozone (
CO is emitted by various surface processes and produced by the atmospheric
oxidation of different gaseous precursors. Atmospheric oxidation of
The understanding of the past CO trends is crucial for monitoring how
anthropogenic activities have impacted CO emissions, atmospheric chemistry, and atmospheric
composition. Over the last decades, analyses of past CO trends have been made
possible by a growing number of direct and indirect observations. Monitoring
of atmospheric CO, initiated by NOAA (National Oceanic and Atmospheric
Administration) in 1988, has revealed a global decrease in tropospheric CO of
Ground-based and satellite-derived CO data are only available, however, for
the last 3 decades. Thus, ancient air preserved in glacial ice and firn is
a unique archive for reconstructing the past atmospheric CO record prior to
the 1990s. The firn is the upper layer of an ice sheet where snow is slowly
transformed into ice. A large amount of air can be sampled from the
interconnected open pores. Mean ages of atmospheric gas increase with firn
depth. CO has only been measured in firn air at a few Northern Hemisphere
sites, including Summit, the North Greenland Ice Core Project (NGRIP), and the
North Greenland Eemian Ice Drilling (NEEM) site
Analysis of air trapped in bubbles in solid ice below the firn layer is
required to extend such reconstruction further back in time, beyond
1950 CE. The pioneering measurements conducted on the Eurocore ice archive
Over the last decade, continuous flow analysis (CFA) of ice core
In this study, we expand on this exploratory investigation by reporting
continuous CO data measured on a set of five additional Greenland ice
cores. By combining the analysis of these new ice core records and by separating
the high-frequency CO signals driven by in situ production from baseline
concentrations, we reveal (i)
an upper-bound estimate of past atmospheric CO abundance and (ii) atmospheric
[CO] trends in the Arctic and for the last three centuries. Climate–chemistry models and/or Earth system models can produce
simulated atmospheric [CO] at ground level in central Greenland, from the
preindustrial era to the present day. Such models are presently intercompared
within the Aerosol Chemistry Model Intercomparison Project (AerChemMIP;
Five ice cores extracted from Greenland (Table
Locations, site characteristics, and other relevant information for ice cores featured in this study.
The fifth ice core in this study is a new core, retrieved from central
Greenland during June 2015: the PLACE ice core. The PLACE core was drilled
1
The five sites investigated experience accumulation rates ranging from 8 to
41
Since the first continuous, high-resolution, CO measurements were performed
along the NEEM-S1 ice core in 2011
The ice cores listed in Table
Our samples were all analyzed at the Desert Research Institute (DRI), Reno
(NV, USA) during two analytical campaigns (Tunu13, D4, NEEM-SC, and NGRIP in
2013; PLACE in 2015). The NEEM-2011-S1 core was analyzed at DRI in 2011, as
described by Faïn et al. (2014). The PLACE core was also analyzed at the
Institute of Environmental Geosciences (IGE, Grenoble, France) in
2017. Specific descriptions of the DRI and IGE setups have already been provided by
A unique spectrometer, using optical-feedback cavity-enhanced absorption
spectrometry
The degassing membranes currently operated in gas-CFA systems extract bubbles
from the sample flow but do not recover dissolved gases from the water phase
efficiently. Carbon monoxide (or methane) have higher solubility than
The CO procedural blank was evaluated for each analytical campaign. Two
different evaluation approaches were applied, which are described in Sect. S1.5. A relatively high CO blank of
Internal precision and stability of gas-CFA measurements can be evaluated by
Allan variance tests
External precision of the continuous CO measurements (i.e., including all
sources of errors or bias) was investigated by melting replicate ice sticks on
different days during each analytical campaign. Specifically, we defined the
external precision as the pooled standard deviation calculated on the
differences of CO concentrations from main and replicate analyzed ice sticks,
averaging continuous CO data over intervals of a few centimeters in length (Sect. S1.7). This approach estimates an external precision of 5.7
CFA-based gas records must be corrected for under-recovery of gases dissolved
in the water stream to obtain absolute values on the WMO-CO X2014A calibration
scale. The magnitude of gas dissolution is monitored using the calibration
loop (Sect.
In this study, we used the calibration loop operated with standard gases
(Table S3) for absolute calibration of the CO dataset. Such an approach has
been successfully applied for
To further evaluate the robustness of the CFA procedure and this absolute
calibration approach, five discrete samples of the PLACE core were analyzed
with a discrete CO setup (Sect. 1.8.4, Fig. S7). The
agreement between the discrete and CFA-based CO dataset (both IGE and DRI) was
excellent, with differences in the CO mixing ratio ranging from 0.4
The mixing of gases and meltwater during the sample transfer from the melt
head to the laser spectrometer induces a CFA experimental smoothing of the
signal. The extent of the CFA-based damping was determined for each analytical
campaign by performing step tests, i.e., switches between two synthetic
mixtures of degassed DI water and synthetic air standards of different
CO and ice chemistry data collected at DRI were mapped onto depth scales using
high-resolution (0.1–0.5
The occasional entry of ambient air into the analytical system as breaks in the
core were encountered and caused contamination. The OF-CEAS spectrometer
simultaneously measures CO and
Uncertainty in the CO mixing ratios is established as 1
Finally, we used high-resolution water isotope datasets to match the DRI and IGE
PLACE depth scales. Water isotopic ratios were measured simultaneously with
gas concentrations in both laboratories using laser spectroscopy
To investigate if elevated or highly variable CO levels observed
in the Greenland ice core could originate from the melting process (i.e., “in
extractu” CO production from trace organics in the ice), CO concentrations
were determined in tests with 20 discrete ice and firn samples (sized between
272 and 1089
The CO concentration in the gas available in the vessel headspace is
determined by gas chromatography combined with a mercuric oxide reduction
detector (RGD; Peak Performer 1, Peak Laboratories, USA). This is carried out at three different
stages: before melting, during melting, and after melting. CO concentrations
are calibrated using four synthetic air gas standards with nominal CO
concentrations ranging from 50 to 500
Three large-diameter (27
During the two DRI analytical campaigns, melted ice core samples were analyzed
continuously by inductively coupled plasma mass spectrometry (ICP-MS) and CFA
for chemical species. These analytical methods have been reported previously
The new CO records available from the PLACE, NGRIP, NEEM, D4, and Tunu13
ice cores are reported in Fig.
The first five panels (from top to bottom) show the Greenland continuous CO records available from the PLACE (DRI), NGRIP, NEEM, D4, and Tunu13 ice cores, respectively, and are plotted for the period spanning from 1650 to 1960 CE. The bottom panel shows the CO baseline levels from each ice record, defined as the 5th percentile of data every 4 years over a 15-year moving window, with the shaded regions on each baseline representing the uncertainty.
We identified the CO baseline levels from each ice record by calculating the
5th percentile of data every 4 years over a 15-year moving window. A
15-year moving window corresponds to a time interval shorter than the full width
at half maximum of the CO age distribution at these sites (Table 1). We chose
to use the 5th percentile CO baselines because they can be calculated identically
for all datasets and are not based on the subjective operator decisions
involved in identifying and removing peaks to derive a baseline level. The
temporal trends in records calculated using the 10th percentiles are similar. The 5th
percentiles were not computed when more than 50
Figure
The true extent of high CO variability may be masked, however, by CFA
analytical smoothing. Figure
High-resolution TOC,
High-frequency variability within CO records can be quantitatively
characterized using the MAD (median absolute deviation). The MAD values calculated for all CO
records collected at DRI every 4 years over a 15-year moving window
(similarly to baseline signals, see Sect. 3.1.1) are reported in Fig. S18. For the 1700–1950 CE time period, we observe that the NEEM,
NGRIP, Tunu13, and PLACE records exhibit stable and similar MAD values, with
mean MAD values over the entire 1700–1950 CE period ranging from
9.9 to 14.2
The observation of high-frequency CO variability along the NEEM-2011-S1 core
led
We observe that abrupt CO spikes of similar magnitude occur in ice cores
extracted by dry drillings (e.g., PLACE, D4, Tunu13) and when a drilling fluid
was used (e.g., NEEM, NGRIP) (Fig.
By comparing the storage time of different ice cores (storage without light
exposure) prior to CO analysis with the same CFA setup (NEEM-2011-S1 and D4),
In 2013, at DRI, a 4
As described above, CO analyses of 20 discrete ice and firn core samples (including the PLACE core) were carried out at the University of Rochester, with the goal of investigating the possibility of the in extractu production of CO from trace organics in the ice during melting of ice core samples (Sects. S2.3 and S1.11, Table S4).
All of the different ice and firn sample types, including samples of different
mass and a gas-free ice sample, showed a consistent excess CO growth occurring
during the melting step, with excess CO increasing by 6.4
Average excess in the CO mixing ratios produced during the melting of discrete samples
measured by gas chromatography combined with a mercuric oxide reduction
detector. The CO values shown have been corrected for a small
(
If the CO production mechanism involves organic compounds present in the ice
lattice, it would be expected that variations in the concentrations of
organics from samples collected at different sites/from different depth
levels as well as variations in the melted ice mass would result in
significant differences in excess CO. Some of the PLACE ice samples were
specifically selected because they exhibited low and stable [CO] in the
continuous CFA record, whereas others were selected because they exhibited elevated and
spiky [CO] behavior, suggesting the possibility of instantaneous CO production
during melting (Fig.
We now explore if the past atmospheric concentrations of CO can be extracted from
the low-frequency variability in the CO ice records' baselines
(Fig.
In our earlier study on the NEEM-S1-2011 core (Fain et al., 2014), we reported
that 68
Typical CO, TOC, and
Typical CO, TOC, and
High-resolution TOC,
The significant analytical smoothing for the Tunu13 ice core suggests that
annual winter minima in CO (associated with winter minima in TOC) may be
unresolvable, which would lead to an overestimation of the baseline even when
using the 5th percentile values. In contrast, repeated analysis of the higher-snow-accumulation PLACE core on two different gas-CFA setups (DRI and IGE,
Table S2) shows that, despite the DRI setup (2015 configuration) exerting a
greater smoothing effect than the IGE one (Sect. S1.9, Figs. S10 and S11), the CO minima values are well-resolved in both cases
(Fig.
The high-resolution TOC measurements collected along the PLACE core exhibit a
stable TOC background value of 13.5 ppbC (baseline defined as the 5th
percentile), with 17 spikes above 40 ppbC (70
Running averages (10
In situ CO production within Greenland ice archives is very often co-located
in ice layers where TOC levels are high (e.g.,
Figs.
Characterizing the fate of TOC in Greenland ice is a first requirement for a
better understanding of in situ CO production
processes. Figure
However, the relationships between TOC and ammonium suggest that other
processes can impact TOC after deposition at Tunu13. Forest fire debris
reaching central Greenland snow during spring or summer mainly consists of
ammonium formate, and organic carbon is mostly made of formate at Summit
Investigating the CO and TOC PLACE records did not reveal any evidence that
the CO baseline could be influenced by in situ production at that site. In
contrast, Tunu13 has a lower snow accumulation rate than other sites
(Table
While the mean CO mixing ratio is significantly correlated with the mean TOC concentrations in ice at Tunu13, this is not the case for the baseline CO level which does not exhibit significant correlations with mean or baseline TOC concentrations. However, we cannot rule out the fact that a redistribution of organic carbon along depth driven by OC post-deposition process (shifting the ammonium–formate equilibrium with the gas phase) specifically impacts the Tunu13 CO record by providing some additional organic substrates in winter layers.
CO mixing ratio collected (i) along the PLACE ice core, showing the continuous IGE CFA record with the 5th percentile baseline (gray line) and the discrete dataset (black dot); and (ii) along the Eurocore archive, showing the historical Eurocore discrete CO data (blue dot; Haan et al., 1998) and the CFA-based dataset measured in 2019 at IGE (red).
Over the 1700–1900 CE period, Eurocore discrete CO data are about
10
To further investigate this apparent offset between our new Greenland CO
records and the Eurocore CO record of
To summarize, we have not been able to reconcile the validity of the
Extracting atmospheric information from a single CO record retrieved from a
Greenland ice archive affected by in situ production features is challenging
(e.g., Faïn et al., 2014). However, we show here that the PLACE, NGRIP,
NEEM, and D4 records, reconstructed from ice archives originating from
different Greenland locations and drilled over a time period spanning more
than a decade, exhibit common patterns. We conservatively exclude Tunu13 from
paleoatmospheric interpretations, knowing that the Tunu13 baseline CO record
could be positively bias by analytical smoothing or the redistribution of organic
carbon with depth (see Sect. 3.3). Figure
A multisite composite CO baseline spanning 1700–1957 CE was generated by
considering the D4, NEEM, NGRIP, and PLACE records and excluding Tunu13. For
consistency, we only consider CO records collected on the same gas-CFA setup
(DRI). The multisite composite, reported in Fig.
Past atmospheric CO mixing ratios for the Northern Hemisphere high latitudes spanning from 1700 to 2018 CE. The black line denotes the multisite average obtained by combining the baselines (5th percentiles) of continuous CO records collected at four Greenland sites (PLACE, NEEM, NGRIP, and D4), representing an upper-bound estimate of past atmospheric CO levels. Light blue denotes the firn air records obtained by combining samples from NEEM, NGRIP, and Summit. Blue represents the average firn air record
The fact that the CO baseline records from the four different Greenland ice
cores are so consistent gives us confidence that our multisite reconstruction
provides a reliable reconstruction of past atmospheric changes. An additional
argument for this conclusion comes from the good overlap between our multisite
reconstruction and CO data from Greenland firn air measurements
(Sect. 3.5.2). However, we cannot exclude the possibility that past
atmospheric CO levels could have been lower than those shown in
Fig.
An independent means to assess the validity of our atmospheric CO
reconstruction is to compare it with a reconstruction of Northern Hemisphere
high-latitude atmospheric CO mixing ratios from Greenland firn air, available
for the 1950–2008 CE period
PLACE and D4 high-resolution CO baseline records overlap with the firn
record. The firn air data suggest that the CO concentration spanned
138–148
The Atmospheric Chemistry and Climate Model Intercomparison Project
New continuous profiles of the CO mixing ratio have been measured along five
Greenland ice cores: the PLACE, NGRIP, D4, Tunu13, and NEEM-SC archives. We
also revisited the NEEM-2011-S1 dataset
We have presented a multisite average ice core reconstruction of past
atmospheric CO for the Northern Hemisphere high latitudes, covering the period
from 1700 to 1950 CE, providing an upper-bound estimate of past atmospheric
CO abundance. This paleoatmospheric signal was extracted from the low-frequency variability in the CO records' baselines of the PLACE, D4, NEEM, and
NGRIP archives, which all exhibit consistent trends despite different surface
accumulation snow rates and chemical compositions. From 1700 to 1875 CE, the
multisite average record is stable with a slight increase with time from 103
to 114
The high-resolution carbon monoxide datasets are accessible on PANGAEA (
The supplement related to this article is available online at:
This scientific project was designed by XF, JC, VVP, RR, EJB, and TB. The high-resolution carbon monoxide measurements were carried out by XF and RHR with support from KF, TB, JC, and EJB. The discrete CO analyses were carried out by PP and EC, and the chemical analyses were carried out by NC and JRM. XF, PP, and VVP participated in the PLACE drilling. RR participated in the Tunu13 drilling. The codes for data processing and modeling were developed by KF, JRM, and XF. All authors contributed to the interpretation of the data. The paper was written by XF with help from all co-authors.
The contact author has declared that neither they nor their co-authors have any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Grateful thanks go to Olivia Maselli, Larry Layman, Daniel Pasteris, Michael Sigl, and other members of the DRI team who assisted with the measurement campaigns. We also wish to thank Frederic Prié, Elise Fourre, and Amaelle Landais for their help with the continuous measurement of water isotopes along the PLACE core at IGE in 2017. We are grateful to Ray Langenfelds for the measurements of gas standards at the CSIRO in 2015. We thank Sophie Szopa and Kostas Tsigaridis for useful discussions. Polar Field Services and the 109th New York Air National Guard as well as the French Polar Institute (IPEV) provided logistical support for ice core drilling. We are grateful to the drillers and field teams. This work also benefited from data available in the public NOAA ESRL database. We thank the people involved in acquiring and analyzing these data as well as making them accessible. This paper was greatly improved by constructive reviews from Murat Aydin and Maria Elena Popa.
This research has been supported by the following programs: the French ANR projects RPD-COCLICO (grant no. 10-RPDOC-002-01 to Xavier Faïn) and NEEM (grant no. 07-VULN-09-001); the EU projects PEGASOS (FP7-IP grant no. ENV-2010/265148) and ICE&LASERS (FP7 ERC grant no. 291062 to Jérôme Chappellaz); the US NSF award nos. 1406236 (Vasilii V. Petrenko), 0221515 (Joseph R. McConnell), 0909541 (Joseph R. McConnell), 1204176 (Joseph R. McConnell), 1406219 (Joseph R. McConnell), and 0968391 (Edward J. Brook, Partnerships in International Research and Education project, PIRE); the Packard Fellowship for Science and Engineering (Vasilii V. Petrenko); the French national program LEFE/INSU (project GreenCO, Xavier Faïn); and, finally, the IPEV Arctic program (grant no. 1159). The NEEM project is directed by the Centre for Ice and Climate at the Niels Bohr Institute, Copenhagen, and the US NSF OPP. It is supported by funding agencies and institutions in Belgium (FNRS-CFB and FWO), Canada (NRCan/GSC), China (CAS), Denmark (FIST), France (IPEV, CNRS/INSU, CEA, and ANR), Germany (AWI), Iceland (RannIs), Japan (NIPR), Korea (KOPRI), the Netherlands (NWO/ALW), Sweden (VR), Switzerland (SNF), the United Kingdom (NERC), and the USA (US NSF, OPP). NGRIP is a multinational research program funded by participating institutions in Denmark, France, Germany, Japan, Sweden, Switzerland, Belgium, Iceland, and the USA.
This paper was edited by Alberto Reyes and reviewed by Murat Aydin and Maria Elena Popa.