Perfluorocarbons (PFCs) are very potent and long-lived greenhouse gases in
the atmosphere, released predominantly during aluminium production and
semiconductor manufacture. They have been targeted for emission controls
under the United Nations Framework Convention on Climate Change. Here we
present the first continuous records of the atmospheric abundance of
Perfluorocarbons (PFCs) are very potent greenhouse gases (about
7000–11 000 times more powerful than
Sinks of these PFCs are dominated by unintentional thermal destruction during
high-temperature combustion at ground level, giving atmospheric lifetimes for
Atmospheric measurements of greenhouse gases are the only reliable way to
verify estimates of global emissions to ensure that we can predict the
effect of emissions on radiative forcing and to guide mitigation options.
Air extracted from firn (the layer of unconsolidated snow overlying an ice
sheet) or bubbles in polar ice provides a reliable way to reconstruct
atmospheric composition prior to direct atmospheric measurements.
Here we present measurements of
The firn and ice core measurements used in this work come from ice or firn
air collected at the following sites:
DE08 and DE08-2 are located 16 km east of the summit of Law Dome
( DSSW20K is 20 km west of the deep DSS (Dome Summit South) drill site near the summit of Law Dome
in East Antarctica EDML (EPICA Dronning Maud Land) is the EPICA drill site near Kohnen Station ( NEEM 2008 firn air was extracted from a borehole near the NEEM (North Greenland Eemian Ice Drilling Project) deep
ice core drilling site ( South Pole 2001, Antarctica (
In addition to the firn and ice core measurements, we use archived and in
situ measurements from
Measurements were made using two different measurement systems and primary
calibration scales. The measurements in
To characterise the age of the air in the firn and ice samples, we use a
numerical model of the processes that occur in firn and ice (mainly diffusion
of air in the firn layer, advection of snow downwards as new snow falls at
the surface and gradual trapping of the air into bubbles). These processes
mean that air contained in firn or ice corresponds to atmospheric air over a
range of times rather than a single age. We use the CSIRO firn model
The depth profile of diffusivity in the firn and other diffusivity-related
parameters in the firn models need to be calibrated for each site that we
model. To do this we tune the models to fit firn measurements of trace gases
for which we know the past atmospheric history. Calibration of the CSIRO firn
model for DE08 and DE08-2 (which are modelled as identical sites), as well as
DSSW20K, NEEM 2008 and South Pole 2001, is described in
The diffusion coefficients we use for the PFCs relative to
As the firn model is linear, and the physical processes in firn are taken as
constant in time, we can characterise the firn models using Green's functions
(also known as age distributions, age spectra or pulse response functions)
that relate the mole fraction of a trace gas at the measurement depths to
atmospheric mole fraction of that gas over a range of times
In order to incorporate the effect of uncertainty in the firn models into our
inversion calculations, we use an ensemble of Green's functions for each
site, constructed as follows. When we calibrate the CSIRO firn model, in
addition to finding the diffusivity profile that gives the best fit to
calibration observations, we also create some alternative diffusivity
profiles that approximately represent the 95 % confidence interval of the
firn model parameters, as described by
We begin with an inversion of the air archive and in situ PFC measurements at
the monthly timescale and semi-hemispheric spatial scale, to infer emissions
of
We then use an inversion similar to
The firn and archive data do not have adequate information content to
constrain semi-hemispheric emissions, so InvE2 infers annual global emissions
with a fixed north–south distribution (for this we use the estimated
north–south distribution from InvE1 for 1990). The AGAGE 12-box model is used
to relate annual high-latitude mole fraction in each hemisphere to annual
global emissions with the fixed north–south distribution, creating Green's
functions that are denoted
We may expect to see a shift in the north–south distribution of emissions over time in recent decades, when global emissions have gone from being predominantly due to aluminium production to now include semiconductors and as developing nations such as China have increased their fraction of global emissions. InvE1 is capable of estimating such a shift in emissions, although with the caveat that derived emissions at the semi-hemispheric level are known to be sensitive to uncertainties in the model transport parameters. Use of a constant north–south distribution of emissions in InvE2 is the best choice prior to the 1980s when the emissions distribution was probably more stable than in recent decades, and the firn and ice core measurements would not contain adequate information to resolve distribution changes anyway. However, use of the constant emissions distribution does degrade the fit to observations in recent decades. We can use the emissions distribution already estimated by InvE1 to improve InvE2. We do this by subtracting the (modelled) contribution to all mole fraction measurements of the monthly semi-hemispheric emissions after 1980 inferred from InvE1, before inverting for additional emissions with the constant north–south gradient. These additional emissions estimated by InvE2 will mostly be emissions before 1980, although they could include small adjustments (positive or negative) to the emissions after 1980, but the adjustments will have the constant (1990) spatial distribution. In this way, we are combining the strengths of the higher-resolution InvE1 inversion for the monthly in situ measurements with the InvE2 inversion for the ice, firn and early archive measurements, to give our best estimate for emissions.
Because primary aluminium production is known much more precisely than
emission factor for the PFCs, we also formulate the inversion to directly
estimate PFC emission factors (in kilograms per metric tonne, kg t
For the InvEF inversion, we first subtract from the observations the effect
of the InvE1 emissions after a selected date (
There are a number of contributions to the uncertainties in inferred
emissions and atmospheric abundance
The contribution to the error that is most difficult to deal with is due to
the fact that we are solving an inverse problem that is ill-conditioned, so
that the solution is not unique. In our case, the ill-conditioning is partly
due to lack of data (mole fraction in the firn is not known at all depths)
but also importantly it is a consequence of the smoothing (and therefore lost
information) by the firn and bubble trapping processes. As noted by
Regularisation (e.g. minimising the length of the solution;
In the InvE2 and InvEF inversions, we use regularisation similar to
Uncertainties in estimated emissions are calculated by repeating the
inversion many times with perturbations to model inputs, including (a) firn
model Green's functions, (b) observations perturbed according to their
uncertainty, as in bootstrapping
Apart from the regularisation term involving the year-to-year changes in
emissions (or emission factor), the cost function consists of the squared
model–data mismatch weighted by the observation uncertainties. We do not
include the prior estimate in the cost function. Previous studies
Prior emissions for the InvE2 inversion after 1980 (or 1983 for
Inputs and results for the
Inputs and results for the
Figures
The third panel shows the estimated history of PFC mole fraction in the atmosphere for the high-latitude northern (dashed) and southern (solid) latitudes calculated with the InvE2 inferred emissions. The annual values of atmospheric PFC mole fraction that are used in the inversion are shown by the black circles for the Southern Hemisphere (based on the Cape Grim air archive and in situ data) and the grey circles for the Northern Hemisphere (based on northern hemispheric tanks and Mace Head in situ data).
The fourth panel shows the emissions estimated by the three inversions. The
red, inner black and blue lines show our preferred solution for each
inversion. The pink shading, outer black lines and blue shading show the
estimated 95 % confidence intervals. The InvEF inversion estimates emission
factors, and we combine the inferred emission factor with aluminium
production to calculate the corresponding emissions that are shown here. For
InvE2 and InvEF, the confidence intervals come from the full ensemble of
Green's functions plus other components of the uncertainty as described in
Sect.
The fifth panel shows emission factors inferred by the InvEF inversion, with
95 % confidence intervals. We also show estimates of recent
Our inversions simultaneously match almost all firn and ice core measurements
very well, showing consistency between the different sites and between the
firn, ice core and atmospheric observations. Our inferred emissions between
1975 and 2008 are very much like those in
The emissions from the InvE2 inversion are very similar to those from the
InvE1 inversion. The only notable differences are that the InvE2 inversions
for
The estimated emissions from InvEF are generally quite similar to those from
InvE2, but there are significant differences in the temporal variability. It
is important to note that in InvE2, regularisation is applied to emissions
(minimising the year-to-year variability in emissions along with the
model–data mismatch, as described in Sect.
The estimated emission factors are quite high in the first few decades of the
20th century, but this is multiplied by very small aluminium production,
leading to small emissions. The uncertainty in emission factor in this period
is large and is dominated by the assumed background mole fraction levels.
Our estimated
Figure
Figure
The pre-anthropogenic background level assumed in our inversions influences
the emissions estimated for the early part of the calculation, and we include
uncertainty in the background level in our uncertainty calculation, using
conservative ranges of 33.66–34.82 ppt for
The results shown by the lines in Figs.
Mole fraction depth profiles at the bottom of the firn and in the
DE08-2 ice for
Figure
A particular strength of this work comes from the fact that the inferred histories of emissions give a good match to overlapping atmospheric, firn and ice core measurements from eight different firn and ice core sites with very different climate and snow accumulation characteristics, collected at different times, with measurements made at two different laboratories and interpreted using two different firn models. This gives us increased confidence that the firn and ice core data provide a consistent and reliable picture of 19th and 20th century greenhouse gas changes.
Our oldest samples are from DE08 and contain air with
The DE08 ice core measurements of
The natural
Inversions InvE2 and InvEF for
Emission factors for PFC emissions from aluminium production have decreased
markedly since the early 20th century, from around 2.1–4.4 kg t
The InvE1 inversion gives estimates of monthly semi-hemispheric emissions.
Although the estimates of the north–south distribution of emissions are
sensitive to uncertainties in model transport parameters, we can draw some
general conclusions about shifts in the distribution over time. Between 1980
and 2010, we see an increase of the proportion of global
Our study adds an extra 6 years of measurements compared to
Global emissions of
While the initial reduction of PFC emission factors last century was a
consequence of measures to reduce electricity consumption during aluminium
production, in recent decades there has been a concerted effort by both the
aluminium and semiconductor industries to reduce PFC emissions. However, the
rate of decrease of emissions appears to have slowed and possibly stopped in
recent years. Other than the 2009 dip,
We have reconstructed emissions and atmospheric abundance of
The 19th century levels of
At the recent end of the record, we see temporary reductions in
The firn, ice core and archive PFC measurements and the reconstructed histories of mole fraction, emissions and emission factor are available
in the Supplement. The in situ measurements are available on the CDIAC website
The present study combines firn and in situ PFC measurements made on the
Medusa system (reported on the SIO calibration scale) with firn and ice core
measurements from DE08, DE08-2 and EDML made only at the University of East
Anglia (reported on the UEA calibration scale). To combine these data we need
to ensure that they are on the same calibration scale, so we convert the UEA
measurements to the SIO calibration scale (SIO-05 for
Figure
The linear least-squares fit to all
The fit to all
We have only one set of measurements for
The measurements at similar depths were not used to determine the fit, just
to give more points for comparison. A linear fit to the NEEM 2008
Air was extracted from DE08 and DE08-2 ice core sections at CSIRO using a “cheese grater” dry extraction system
At UEA the PFCs from EDML firn air and air released from the ice cores were
analysed using a high-sensitivity gas chromatograph/trisector mass
spectrometer system (Waters/Micromass Autospec) according to the procedures
detailed in
Analysis of firn air on the Medusa system was described in
EDML is located near Kohnen Station, Dronning Maud
Land, Antarctica
The LGGE-GIPSA firn model was tuned for DSSW20K using measurements of
To obtain annual mole fraction data for the high northern and southern
latitudes, we fit smoothing splines to measurements from Cape Grim (air
archive and in situ) and Mace Head (in situ) and the suite of old tanks from
the Northern Hemisphere. The splines are sampled at 1-year intervals to give
mole fraction corresponding to the start of each year. For the period covered
by the air archives, where there are not reliable measurements in every year,
we only retain annual values when there are air archive measurements around
the same time. The splines have 50 % attenuation at periods of 1 year.
Figure
C. M. Trudinger, P. J. Fraser, D. M. Etheridge and W. T. Sturges designed the study; M. K. Vollmer, D. R. Worton, B. R. Miller, J. Laube and F. Mani performed PFC measurements on firn or ice core samples; P. J. Fraser, W. T. Sturges, J. Mühle, P. B. Krummel, L. P. Steele, C. M. Harth and S. O'Doherty contributed to the in situ records and general PFC measurement capability; D. M. Etheridge, T. Blunier, J. Schwander and M. Battle collected ice or firn air; C. M. Trudinger developed and ran the CSIRO firn model and InvE2 and InvEF inversions; M. Rigby developed and ran the InvE1 inversions; P. Martinerie and E. Witrant developed and ran the LGGE-GIPSA firn model; P. J. Rayner contributed ideas on modelling and uncertainties; C. M. Trudinger wrote the paper with input from other authors.
This work has been undertaken as part of the Australian Climate Change Science Program, funded jointly by the Department of the Environment, the Bureau of Meteorology and CSIRO. We acknowledge support from the Australian Antarctic Science Program. This work is a contribution to the European Project for Ice Coring in Antarctica (EPICA), a joint European Science Foundation/European Commission scientific programme, funded by the European Union (EPICA-MIS) and by national contributions from Belgium, Denmark, France, Germany, Italy, the Netherlands, Norway, Sweden, Switzerland and the United Kingdom. This work was also funded by the CEC programme (EUK2-CT2001-00116, CRYOSTAT). NEEM is directed and organised by the Center of Ice and Climate at the Niels Bohr Institute and US NSF, Office of Polar Programs, and 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), United Kingdom (NERC) and the USA (US NSF, Office of Polar Programs). We acknowledge the support of the CSIRO GASLAB team. The operation of the AGAGE instruments at Mace Head and Cape Grim is supported by the National Aeronautic and Space Administration (NASA) (grants NAG5-12669 and NNX07AE89G to MIT; grants NNX07AF09G and NNX07AE87G to SIO), the Department of Energy and Climate Change (DECC, UK) contract GA01081 to the University of Bristol, CSIRO and Bureau of Meteorology (Australia). William Sturges recognises the CSIRO Fröhlich Fellowship for supporting a visit to CSIRO Aspendale. Martin Vollmer acknowledges a CSIRO Office of the Chief Executive Distinguished Visiting Scientist grant to CSIRO Aspendale for firn air measurements. Matthew Rigby is supported by an advanced research fellowship (NE/I021365/1) from the UK Natural Environment Research Council (NERC). Francis Mani was supported by a Marie Curie Fellowships in Antarctic Air-Sea-Ice Science award, David Worton by a NERC Studentship, and Johannes Laube by a NERC Fellowship (NE/I021918/1). We thank Cecelia MacFarling-Meure for ice extraction, Jean-Marc Barnola, Andrew Smith, Tas van Ommen, Dominic Ferretti and Mark Curran for helping to collect the Law Dome firn and ice samples and Pep Canadell and Roger Francey for helpful comments on the manuscript.Edited by: A. Jones Reviewed by: two anonymous referees