Quantifying greenhouse-gas emissions from atmospheric measurements: A critical reality check for climate legislation

Abstract

Emissions reduction legislation relies upon 'bottom-up' accounting of industrial and biogenic greenhouse-gas (GHG) emissions at their sources. Yet, even for relatively well-constrained industrial GHGs, global emissions based on 'top-down' methods that use atmospheric measurements often agree poorly with the reported bottom-up emissions. For emissions reduction legislation to be effective, it is essential that these discrepancies be resolved. Because emissions are regulated nationally or regionally, not globally, top-down estimates must also be determined at these scales. High-frequency atmospheric GHG measurements at well-chosen station locations record 'pollution events' above the background values that result from regional emissions. By combining such measurements with inverse methods and atmospheric transport and chemistry models, it is possible to map and quantify regional emissions. Even with the sparse current network of measurement stations and current inverse-modelling techniques, it is possible to rival the accuracies of regional 'bottom-up' emission estimates for some GHGs. But meeting the verification goals of emissions reduction legislation will require major increases in the density and types of atmospheric observations, as well as expanded inverse-modelling capabilities. The cost of this effort would be minor when compared with current investments in carbon-equivalent trading, and would reduce the volatility of that market and increase investment in emissions reduction.

Full text

April 2011 , published 18
, doi: 10.1098/rsta.2011.0006369 2011 Phil. Trans. R. Soc. A
Ray F. Weiss and Ronald G. Prinn
check for climate legislation
atmospheric measurements: a critical reality
Quantifying greenhouse-gas emissions from
References
related-urls
http://rsta.royalsocietypublishing.org/content/369/1943/1925.full.html#
Article cited in:
l.html#ref-list-1
http://rsta.royalsocietypublishing.org/content/369/1943/1925.ful
This article cites 55 articles, 1 of which can be accessed free
Subject collections
(27 articles)geochemistry  (128 articles)climatology  (58 articles)atmospheric science 
collections
Articles on similar topics can be found in the following
Email alerting service herein the box at the top right-hand corner of the article or click
Receive free email alerts when new articles cite this article - sign up
http://rsta.royalsocietypublishing.org/subscriptions
go to: Phil. Trans. R. Soc. ATo subscribe to
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

Phil. Trans. R. Soc. A (2011) 369, 1925–1942
doi:10.1098/rsta.2011.0006
Quantifying greenhouse-gas emissions from
atmospheric measurements: a critical reality
check for climate legislation
BYRAY F. WEISS1,*AND RONALD G. PRINN2
1Scripps Institution of Oceanography, University of California, San Diego,
La Jolla, CA 92093-0244, USA
2Center for Global Change Science, Massachusetts Institute of Technology,
Building 54-1312, Cambridge, MA 02139-4307, USA
Emissions reduction legislation relies upon ‘bottom-up’ accounting of industrial and
biogenic greenhouse-gas (GHG) emissions at their sources. Yet, even for relatively well-
constrained industrial GHGs, global emissions based on ‘top-down’ methods that use
atmospheric measurements often agree poorly with the reported bottom-up emissions.
For emissions reduction legislation to be effective, it is essential that these discrepancies
be resolved. Because emissions are regulated nationally or regionally, not globally, top-
down estimates must also be determined at these scales. High-frequency atmospheric
GHG measurements at well-chosen station locations record ‘pollution events’ above the
background values that result from regional emissions. By combining such measurements
with inverse methods and atmospheric transport and chemistry models, it is possible
to map and quantify regional emissions. Even with the sparse current network of
measurement stations and current inverse-modelling techniques, it is possible to rival
the accuracies of regional ‘bottom-up’ emission estimates for some GHGs. But meeting
the verification goals of emissions reduction legislation will require major increases
in the density and types of atmospheric observations, as well as expanded inverse-
modelling capabilities. The cost of this effort would be minor when compared with current
investments in carbon-equivalent trading, and would reduce the volatility of that market
and increase investment in emissions reduction.
Keywords: global warming; greenhouse-gas emissions; climate legislation
1. Introduction
Entering the second decade of the twenty-first century, legislation to stem
the adverse effects of anthropogenic climate change by requiring reductions
in anthropogenic carbon dioxide (CO2) and non-CO2long-lived greenhouse-
gas (GHG) emissions is becoming increasingly widespread. In addition to the
multi-national Kyoto Protocol, established under the United Nations Framework
*Author for correspondence (rfweiss@ucsd.edu).
One contribution of 17 to a Discussion Meeting Issue ‘Greenhouse gases in the Earth system:
setting the agenda to 2030’.
This journal is ©2011 The Royal Society1925
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

1926 R. F. Weiss and R. G. Prinn
Convention on Climate Change (UNFCCC), there are many national, bilateral,
multi-lateral and regional initiatives to limit GHG emissions. Emissions of non-
CO2GHGs, many of which have global warming potentials (GWPs) that are
tens to thousands of times greater than CO2per unit mass, represent roughly
35 per cent of current GHG emissions on a carbon-equivalent basis [1], and
play a significant role in the current emissions trading market. Among the long-
lived non-CO2GHGs are a number of high-GWP anthropogenic gases that are
not regulated by the Kyoto Protocol, but which still contribute significantly
to anthropogenic radiative forcing, especially the chlorofluorocarbons (CFCs)
and other stratospheric ozone-depleting substances that are regulated by the
Montreal Protocol.
Current emissions reduction legislation is based on accounting methods that
are prescribed under the UNFCCC for calculating inventories of emissions of
industrial and biogenic GHGs at their sources, so-called ‘bottom-up’ emissions
reporting. Detailed guidelines for emissions reporting have been developed under
the auspices of the Intergovernmental Panel on Climate Change (IPCC; [2] and
subsidiary volumes). These prescribed procedures are based on activity metrics
such as economic and land-use databases, emission factors relating these activities
to GHG emissions, and time delays between GHG production and release. There is
capacity for including uncertainties in these estimates, but they are often reported
as ‘unknown’. This is a complex task involving estimates for a very wide range
of GHG emission sources; each reported individually and then aggregated. The
resulting national GHG emission inventories for UNFCCC Annex I developed
countries are reported with many digits of resolution, but usually without
uncertainties [3]. However, estimated emissions of GHGs of primarily industrial
origin and with limited types of sources are generally held to have greater accuracy
than emissions of GHGs from primarily biogenic sources that are more difficult
to quantify.
It is important to note that the Kyoto Protocol and other similar legislation
requires carbon-equivalent emissions reductions relative to a base period that are
often specified with resolutions of 5 per cent or less, with required ‘certification’
according to the IPCC or other formal bottom-up reporting procedures. But do
these procedures yield actual emissions? This is a critical question because the
anthropogenic affect on climate is ultimately driven by actual emissions of GHGs
to the atmosphere, not by reported ones.
2. Emission estimation approach
For emissions control legislation to be effective, and considering that enforcement
is likely to be practical only by bottom-up methods, it is essential that significant
discrepancies between bottom-up emissions estimates and ‘top-down’ emissions
estimates based on atmospheric measurements be resolved. But because emissions
control legislation is national or regional in nature, not global, it is also essential
that top-down emission estimates be determined at these same geographic scales.
Atmospheric GHG measurements and inverse modelling, when proceeding in
tandem, allow observations to be used to answer important scientific, as well
as regional, emission questions. Modelling can also help to define measurement
strategies (target species, site location, measurement precision and frequency)
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

Greenhouse-gas emissions verification 1927
and priorities. Using the so-called minimum variance Bayesian inverse methods,
optimal estimates of fluxes by process and/or region in a chemical transport
model (CTM) can be obtained by minimizing the errors (variances) in the
estimated emissions given the errors in the observations. These inverse studies
can use measurements from a number of independent observational networks,
after conversion to a common calibration scale using results of inter-laboratory
measurement comparisons.
(a)Statistical methods
There are a number of statistical approaches to flux (source and sink)
estimations [4–6]. As an illustration here, we discuss specifically the discrete
Kalman filter (DKF). A desirable feature of the DKF is its capability for
objective estimation of the errors in the estimated fluxes and for inclusion of
both observational and certain CTM errors in the measurement error treatment.
A detailed introduction to the DKF in vector/matrix form for application to
estimating sources and sinks for atmospheric trace gases can be found in Prinn
[6]. To illustrate some of the key aspects of the use of the DKF for emissions
estimation, we can look at the highly simplified but informative one-dimensional
case of estimating sequential emissions from a single region using a sequence
of observations at a single site where vectors and matrices are now replaced
by scalars.
In this case, the filter is effectively minimizing the square of the error (pk=s2
xk )
in the estimated emissions (xk) at discrete time k. We define yo
kas the ‘observation’
(mole fraction) at time k,rkas the square of the error in the observation at time
k(rk=(so
yk )2), xf
kas the ‘forecast’ value for xk(value before using observation k),
xa
kas the ‘analysis’ value for xk(corrected value after using observation k),
hk=dyk/dxk(sensitivity of model mole fraction to model emissions) and yk=hkxf
k
as the model estimate for observation k. If we also define pf
kand pa
kas the
forecast and analysis values, respectively, for pk(values before and after using
observation k) and mas a scalar that multiplies the analysis values from a prior
time step to serve as a forecast for the current time step, then the recursive filter
equations to estimate emissions and their errors are simply
xf
k=mk−1xa
k−1, (2.1)
xa
k=xf
k+kk(yo
k−hkxf
k), (2.2)
kk=pf
khk
h2
kpf
k+rk
=1
hk+rk/(hkpf
k)=Kalman gain (scalar) at time k, (2.3)
pf
k=m2
k−1pa
k−1(2.4)
and pa
k=(1 −kkhk)pf
k=1−1
1+[rk/(h2
kpf
k)]pf
k. (2.5)
Note that because pa
k≤pf
k, then the square of the error in the estimated emissions
after use of the observation is less than its forecast value. In addition, for
the desired large pkreduction (or large reductions in emission errors from
their forecasts), we want rkh2
kpf
k. For the desired large emission correction
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

1928 R. F. Weiss and R. G. Prinn
atmospheric greenhouse-gas measurements
analysed observed winds
surface-emissions models (uncertain parameters)
atmospheric chemistry models
predictions of concentrations,
emission parameters,
sensitivities to emission parameters
corrected
emission
parameters
optimal estimation method
for emission parameters
global three-
dimensional
atmospheric
transport
models
alternative structures
Figure 1. Schematic illustrating the general inverse method for estimating emission fluxes or
parameters. The Kalman filter is an example of an optimal estimation method. The concentrations,
emission parameters and sensitivities to emission parameters refer to y,xand dy/dxin the simple
example discussed in the text. The Model of Atmospheric Transport and Chemistry (MATCH)
model, containing an atmospheric chemistry sub-model where appropriate, and driven by analysed
observed winds (National Centers for Environmental Prediction (NCEP), the European Centre
for Medium Range Weather Forecasts (ECMWF)) is an example of a suitable three-dimensional
model. Surface-emission models range from substantial codes that simulate the surface source and
sink processes down to simple specifications of time- and space-varying emissions based on in situ
flux measurements and information about anthropogenic sources. Alternative structures refer to
different choices for the emission models.
(xa
k−xf
k), for a given large difference (yo
k−hkxf
k) between observation and model,
we want the Kalman gain kk→1/hk(its maximum value), which occurs when
rkh2
kpf
k. That is, for both desirable outcomes, we want (so
yk )2(dyk/dxk)2(sf
xk )2
or equivalently so
yk <(dyk/dxk)sf
xk (i.e. the error in the observation yo
kneeds to be
less than the error in the model value yk=hkxf
kfor the observation). A schematic
of the general approach is given in figure 1.
It is important to ensure that model and observation imperfections are
accounted for properly (e.g. [4,6]). The processes and parameters to be estimated
are chosen so that they have effects on the emission patterns in space and
time that are sufficiently distinct from each other to ensure uniqueness and
stability (e.g. [7,8]). The estimations are also formulated so that small fractional
changes in concentrations do not yield unacceptably large fractional changes in
deduced emissions. Observational uncertainties that are associated with absolute
calibration, instrument precision and inadequate sampling in space and time are
incorporated into the observation error whenever possible. It is also important
to recognize when observation errors are correlated, thus possibly violating a
condition of optimal filters like the DKF or complicating the definition of the
observational error treatment. Various methods exist to address this (see [4,6]
for reviews). Weak nonlinearities in the chemical models (e.g. the lowering of the
hydroxyl radical (OH) when methane (CH4) emissions increase) are routinely
handled by recalculating the time-dependent sensitivities (hkin the above
simplified equations) after each run through all the data, and repeating the inverse
method to ensure convergence. Structural errors and random and systematic
transport errors (i.e. errors in hkabove) are handled through utilization of
multiple model versions and Monte Carlo methods (e.g. [9]) and/or by increasing
the measurement error to include the error owing to the model (e.g. [6,10]).
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

Greenhouse-gas emissions verification 1929
observed CH
4
(ppb)
modelled CH
4
(ppb)
modelled CH
4
(ppb)
observed CH
4
(ppb)
1640
1680
1720
1640
1680
1720
1760
1760
1640
1680
1720
1760
(a)
(b)
observations 1640
1680
1720
1760
Feb Jan Mar AprNov Dec May
1998/1999
model
model
model
mirror plot
mirror plot
observations
observations
observations
Samoa
Samoa
La Niña winds
NAO (+) winds
NAO (–) winds
El Niño winds
model
Feb Jan Mar AprNov Dec May
1997/1998
150°–150°180°
150° –150°180°
–30°
0°
1996
1997
1998
1999
2000
2001
4
SOI
NAO index
–4 0
–30°
0°
observed CH
4
(ppb) observed CH
4
(ppb)
modelled CH
4
(ppb) modelled CH
4
(ppb)
Jan Feb Mar Apr May
1995
1500
1700
1900
2100 1500
1700
1900
2100
1500
1700
1900
2100 1500
1700
1900
2100
60°
60°
–30°0°
–30°0°
4–4 0
1992
1993
1995
1994
1996
1997
1998
1999
Jan Feb Mar Apr May
1996
mirror plot
mirror plot
Figure 2. MATCH simulates the very significant effects of temporal and inter-annual variability of
circulation patterns on atmospheric methane (CH4) mole fractions. Advanced Global Atmospheric
Gases Experiment (AGAGE) CH4observations (black) versus MATCH simulations (red) are shown
at (a) Samoa and (b) Mace Head, Ireland [7,8]. The El Niño Southern Oscillation Index (SOI) and
the North Atlantic Oscillation (NAO) index are shown adjacent, respectively. The comparison
in (a) is for the same months during the 1998 El Niño (top) and 1999 La Niña (bottom) with
the January–May average surface wind fields for the 2 years, and the comparison in (b)isfor
the same months during the 1995 positive NAO (top) and 1996 negative NAO (bottom) with the
January–May average surface winds also shown. Note that observations and model estimates are
plotted on scales of opposite direction so that an exact mirror image of the two datasets implies
perfect agreement. The CH4measurement station location is shown by a cross on each surface
wind field map.
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

1930 R. F. Weiss and R. G. Prinn
(b)Models
A basic requirement for inverse modelling is an accurate and realistic multi-
dimensional atmospheric CTM. Even apparently small transport errors can lead
to significant errors in estimated sources or sinks [11,12]. Three-dimensional
models are essential to resolve pollution events and solve for regional sources and
sinks. In addition, three-dimensional models must possess realistic atmospheric
circulations. These models may compute quantities at fixed grid points (Eulerian)
or compute them following the trajectories of the air parcels (Lagrangian).
As an example of a three-dimensional Eulerian model, a large number of
inverse studies have used the Model of Atmospheric Transport and Chemistry
(MATCH) [10,13,14]. This is an offline global three-dimensional transport
model that uses meteorological fields derived from forecast-centre analyses.
MATCH has been successfully driven by meteorological data from the National
Centers for Environmental Prediction (NCEP), the European Centre for
Medium Range Weather Forecasts (ECMWF) and the Goddard Space Flight
Center/National Aeronautics and Space Administration (GSFC/NASA) Data
Assimilation Office (DAO) analysis [14]. Sub-grid mixing processes, which include
dry convective mixing, moist convective mixing and large-scale precipitation
processes, are computed in the model. MATCH can be used at a horizontal
resolution as fine as T62 (1.8◦×1.8◦), with either 42 or 28 levels in the vertical.
MATCH inversions have been used for many GHGs, including chlorofluorocarbon-
11 (CF3Cl) [10], CO2[15–18], CH4[7,8], nitrous oxide (N2O) [19] and carbon
tetrachloride (CCl4)[20]. The ability of MATCH to accurately simulate the
effects of transport on long-lived trace gases is well illustrated by the methane
simulations in figure 2.
Another common modelling approach is based on back-trajectories computed
from meteorological data using a Lagrangian model (e.g. [12]). By dividing
the trace gas observations into ‘background air events’ and ‘pollution events’
and computing air mass back-trajectories, and thus air mass transit times
over predefined emission regions, the time/space average emissions from these
predefined regions can be determined. The method obviously requires accurate
definition of both back-trajectories and eddy diffusive fluxes. The Hybrid Single-
Particle Lagrangian Integrated Trajectory (HYSPLIT) model [21,22]ofthe
National Oceanic and Atmospheric Administration (NOAA) has been used for
this purpose [23]. The UK Met Office Lagrangian particle model (Numerical
Atmospheric dispersion Modelling Environment; NAME) has also been applied
extensively to determine European source and sink strengths for a variety of
species (e.g. [24–29]). Similarly, Australian regional emissions of various species
have been determined using inverse studies and regional transport models [30,31].
The FLEXPART Lagrangian particle dispersion model has also been used to
determine emissions of halocarbon GHGs regionally and globally (e.g. [32]).
While two-dimensional models are not suitable for regional emission
estimation, they have their uses (e.g. solving for global or hemispheric emissions
or for model error analysis). Three-dimensional models, being computationally
expensive, do not always lend themselves well to doing very long time integrations,
and multiple runs are required to address uncertainties (e.g. thousands of runs
for Monte Carlo treatments of model transport, rate constant and absolute
calibration errors). A two-dimensional model is better suited to a full uncertainty
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

Greenhouse-gas emissions verification 1931
analysis because its transport is ‘tunable’ to simulate observed latitudinal
gradients (e.g. [9,33]). For example, measurements of N2O have been used
in conjunction with three-dimensional atmospheric CTMs transport models
to attribute source and sink strengths [19,34], and use of a complementary
two-dimensional model to assess the effects of model error using a Monte
Carlo approach indicates that the errors emanating from the three-dimensional
inversions alone need to be augmented significantly to account for model
errors [19].
3. Global and regional emissions
There are large uncertainties associated with emissions of the biogenic
components of some of the most important anthropogenic GHGs such as CO2,
CH4and N2O—emissions associated with land-use changes, agriculture and waste
processing. For these gases, there is also a need to separate anthropogenic
emissions from natural emissions. As a result, the accuracies of bottom-up
emissions inventories are most easily assessed for global emissions of purely
industrial long-lived anthropogenic GHGs that are emitted from defined sources
and are easily quantified from atmospheric measurements. Furthermore, emissions
of industrial non-CO2GHGs, with their very high GWPs, are particularly
important to quantify because they represent a disproportionately large share
of the global carbon-equivalent trading market, even though their contributions
to global warming are much less than that of CO2.
Atmospheric abundances of a wide range of GHGs, including all of the
non-CO2GHGs currently regulated by the Kyoto and Montreal Protocols,
are routinely measured at a few select locations around the world by several
independent research programmes including our Advanced Global Atmospheric
Gases Experiment (AGAGE) programme [35,36]. These measurements, made
in real time at remote stations, in flask samples and in archived air samples,
yield accurate trends for the background atmospheric composition that extend
back three decades for most of these gases. They can also be extended well
before their industrial production when combined with measurements of air
trapped in polar firn or ice cores (e.g. [37]). Using such data and the estimation
approaches discussed above, it is possible to calculate top-down global emission
rates from the measured trends, especially for long-lived gases with simple
chemistry, and then to compare these values with the reported bottom-up
emission rates.
The current knowledge of methane emissions exemplifies the issues surrounding
gases with large biogenic, as well as anthropogenic, sources. After nearly a decade
of little net change, the mole fractions of this gas began to rise in both hemispheres
in 2006–2008 [38]. The inverse modelling by Rigby et al.[38] implied that the
increase in growth rate was due either to increasing tropical and high-latitude
emissions or to a smaller high-latitude emissions increase along with a few percent
tropical OH decrease (or to some combination of the two). A later study of
spatial gradients and Arctic isotopic signals by Dlugokencky et al.[39] suggested
that the rise was primarily due to increased wetland emissions in both the high
latitudes and tropics, with little influence from OH variations. Bousquet et al.[40]
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

1932 R. F. Weiss and R. G. Prinn
earlier suggested that the stable period for CH4preceding 2006 was caused by
a decreasing wetland source countering an increasing fossil-fuel-related source,
with the sink of CH4owing to OH potentially playing a role in the observed
atmospheric variability. For comparison, Chen & Prinn [8] attributed the CH4
variations over 1996–2001 when compared with the literature values for the years
before that to decreased fossil-fuel-related emissions and increased rice-paddy
emissions, with the 1998 positive anomaly owing to increased global wetland and
wildfire emissions. While the suggestion of very large methane emissions from
vascular plants has been challenged [1], Chen & Prinn [8] have noted that their
computed increased rice emissions (about 25 Tg methane yr−1) could also be due
to neighbouring non-rice wetland emissions.
The estimation of emissions for the essentially purely anthropogenic
(industrial) greenhouse gases is more straightforward than for those gases
whose budgets involve both biogenic and anthropogenic components. However,
challenges still remain for accurately inferring regional emission estimates for
these industrial gases. Examples for four high-GWP anthropogenic GHGs are
presented here.
Carbon tetrafluoride (CF4) is the longest-lived GHG regulated by the Kyoto
Protocol, with an atmospheric lifetime of 50 000 years and a GWP of 7390 on a
100 yr time horizon [1]. It is emitted principally as a by-product of aluminium
production, with lesser but significant emissions from the electronics industry.
The trend of CF4concentration in the atmosphere is being measured in real time
by AGAGE, and with the aid of stored samples, its trend in both hemispheres has
been extended back to the 1970s [41]. There is also a small natural source of CF4
from the continental lithosphere [42], which, because of its very long atmospheric
lifetime, accounts for about half of the approximately 78 ppt (parts per trillion,
dry air mole fraction) in the current atmosphere. But this flux is negligible when
compared with the anthropogenic flux that drives the current trend. Modelling
the AGAGE atmospheric CF4trends using a 12box inverse model [41] yields a
top-down global anthropogenic CF4emission flux that peaked in 1980 at about
18 Gg yr−1and tapered to about 16 Ggyr−1in 1990 and about 11 Gg yr−1in 2006,
with a modelled uncertainty of about 5 per cent. This modelled top-down global
emission history is shown in figure 3, together with various reported bottom-up
CF4emission estimates.
Comparison with bottom-up estimates reported to the UNFCCC by the
Annex I industrialized countries [3] shows that in 1990, when reporting began,
reported emissions accounted for only about 60 per cent of the measured
atmospheric increase, and that by 2006, this fraction had decreased to about 40
per cent. In other words, while the rate of CF4accumulation in the atmosphere
decreased by about 30 per cent during this period, the Annex I reported
CF4emissions decreased by about 50 per cent. Although aluminium-producing
countries like China, India and Brazil are not included in Annex I reporting,
and production by these countries certainly increased significantly over this time
period, it is difficult to reconcile that by 2006, non-Annex I countries accounted
for about 60 per cent of global CF4emissions. Reinforcing this concern are the
bottom-up results of the International Aluminium Institute (IAI) report series
(e.g. [43]), which include worldwide CF4emissions from Annex I as well as non-
Annex I aluminium production, but do not include emissions from the electronics
industry. When the IAI estimates are added to the much smaller electronics
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

Greenhouse-gas emissions verification 1933
1975
0
5
CF4 emissions (Gg yr–1)
10
15
20
1980
UNFCCC
Annex I countries
global
aluminium (IAI)
global electronics (EDGAR v. 4)
global aluminium
(IAI) plus global
electronics
(EDGAR v. 4)
global atmospheric
measurements (AGAGE)
1985 1900 1995
y
ear
2000 2005 2010
Figure 3. Top-down global emissions of carbon tetrafluoride (CF4) modelled from global
atmospheric measurements in the AGAGE programme [41], with shaded combined modelling and
measurement uncertainties (±1 standard deviation), compared with the bottom-up global emissions
estimates for the aluminium industry (e.g. [43]), for the electronics industry [44] and for these two
sources combined. Also plotted are bottom-up emissions reported to the UNFCCC for the Annex I
developed countries [3]. Adapted from Mühle et al.[41].
19751970
0
SF6 emissions (Gg yr–1)
2
6
4
8
1980
UNFCCC
Annex I countries
EDGAR v. 4
global
EDGAR v. 4
Annex I countries
global atmospheric
measurements
(AGAGE)
1985 1900
year
1995 2000 2005 2010
Figure 4. Top-down global emissions of sulphur hexafluoride (SF6) modelled from AGAGE
programme global atmospheric measurements [48], with shaded combined modelling and
measurement uncertainties (±1 standard deviation) compared with bottom-up global emissions
reported to the UNFCCC for the Annex I developed countries [3] after correction of pre-1995
reported Japanese emissions [46]. Also plotted are the EDGAR global-estimated emissions [44]
that take into account bottom-up estimates as well as atmospheric measurements, and EDGAR
estimates for SF6emissions only from Annex I countries that are roughly double those reported to
the UNFCCC. Adapted from Rigby et al.[48].
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

1934 R. F. Weiss and R. G. Prinn
industry global CF4emissions estimates compiled by v. 4.0 of the Emission
Database for Global Atmospheric Research (EDGAR) [44], the results account
for only about 75 per cent of the AGAGE measured increases in 1990, and by
2006, this fraction decreases to about 45 per cent. In other words, the combined
IAI and EDGAR electronics estimates are in reasonable agreement with emissions
reported to the UNFCCC, even though the former values include non-Annex I
countries and the latter ones do not. For the most recent part of the record, neither
of these sets of reported CF4emissions account for even half of the measured
increase in the global atmosphere.
Sulphur hexafluoride (SF6) is the most potent GHG regulated by the Kyoto
Protocol, with a GWP on a 100 yr time horizon of 22 800 and an atmospheric
lifetime of 3200 years [1]. It is used principally as a dielectric to prevent arcing
in high-voltage equipment, and also in industrial applications where a heavy
and chemically inert gas is required. SF6is also emitted naturally from the
continental lithosphere, but in such small quantities relative to its atmospheric
lifetime that its natural pre-industrial background concentration was less than
0.006 ppt [45]. Its present atmospheric abundance of about 6.8 ppt is therefore
effectively entirely anthropogenic. Recent trends in global atmospheric SF6
distributions have been measured and modelled extensively and independently
by several research programmes, including the University of Heidelberg [46,47],
AGAGE [48] and the NOAA Earth System Research Laboratory [49,50]. The SF6
measurements of each of these three programmes are generally in good agreement
and lead to essentially the same conclusion, namely that global SF6emissions are
greatly underestimated by bottom-up emissions reported to the UNFCCC by
Annex I countries.
The modelled SF6global emissions history based on AGAGE global
atmospheric measurements [48] is compared with various bottom-up emissions
estimates in figure 4. After correction of pre-1995 reported Japanese
emissions [46], all the datasets show that throughout the mid-1990s, only about
40 per cent of total emissions were reported by Annex I countries, and that by
2006, this fraction had reduced to about 20–25%. If the Annex I countries have
indeed reported correctly, then the non-Annex I countries emitted about 1.5 times
as much SF6as the Annex I countries in the mid-1990s, and by 2006, they were
emitting four or five times as much. A more likely explanation that is supported
by the v. 4.0 EDGAR [44] interpretation of this mismatch is that Annex I SF6
emissions are severely under-reported and that they actually represented about
80 per cent of the total in the mid-1990s and about 60 per cent of the total in
2006. In other words, the Annex I countries collectively have likely under-reported
SF6emissions by more than a factor of 2.
With respect to the use of the EDGAR [44] database for comparisons of
top-down and bottom-up emission fluxes, it is important to note that because
EDGAR’s objective is to provide the best emission estimates, both bottom-
up and top-down methods are used to arrive at its recommended values of
such parameters as emissions factors and emissions delays. Discrepancies such
as the ones discussed above, in which measured global atmospheric trends have
disagreed significantly with EDGAR bottom-up assessments, have resulted in
major revisions of EDGAR emission values in subsequent release versions. For
CF4and SF6, the current (v. 4.0) EDGAR global total and Annex I emission
estimates are neither bottom-up nor top-down, but rather are in effect a synthesis
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

Greenhouse-gas emissions verification 1935
of both, which improves upon the purely bottom-up approach by adjusting
emission factors. In the case of SF6, these relationships are discussed by Maiss &
Brenninkmeijer [51] and Rigby et al.[48]
Nitrogen trifluoride (NF3) is not regulated by the Kyoto Protocol, and its
emissions from Annex I countries are therefore not required to be reported to
the UNFCCC, but its long lifetime of about 550 years and high GWP of about
16 800 on a 100 yr time horizon [52], and its use as a replacement in electronics
manufacturing for perfluorocarbons that are currently regulated, make it a strong
candidate for future emissions regulation. The first measurements of the trend of
NF3in the global atmosphere were made by the AGAGE programme [53]. They
showed that its present global abundance is about 0.5 ppt and that its emissions
in 2006 were roughly four times greater than the only published bottom-up global
NF3emission estimate for that year. When recently unpublished industry-wide
estimates of global NF3usage provided by Air Products Corporation, a major
producer, are compared with the modelled global atmospheric NF3emission
results, they suggest that about 9 per cent of current global NF3usage is emitted
to the atmosphere, 4.5 times greater than the 2 per cent emission factor that has
been used widely in bottom-up estimates [54].
Although each of the above examples finds significantly greater global emissions
than are estimated by bottom-up accounting methods, this is not uniformly the
case. Sulphuryl fluoride (SO2F2) is a fumigant that is used increasingly to kill
termites and other pests, often as a replacement for methyl bromide, which is
restricted by the Montreal Protocol because of its role in stratospheric ozone
depletion. SO2F2is not regulated by the Kyoto Protocol, and its emissions
from Annex I countries are therefore not reported to the UNFCCC. The
first measurements of SO2F2in the global atmosphere [55], coupled with
laboratory studies of its optical properties and reactivity [56], showed that its
GWP is about 4800 on a 100 yr time horizon and its atmospheric lifetime
is about 36 years. The observed global trend of SO2F2, presently at about
1.6 ppt and rising at about 5 per cent per year, is best explained by the
emission to the atmosphere of only about two-thirds of its estimated global
usage [55]. Either the usage data are overestimated, or unidentified mechanisms
destroy about one-third of the gas before it is emitted to the atmosphere so
that the flux to the atmosphere is reduced without changing the modelled
atmospheric lifetime.
The picture that emerges from these four examples of global top-down and
bottom-up comparisons for these industrial gases is that the discrepancies can
be quite large, and that more often than not, the measured accumulations
of industrially produced GHGs in the atmosphere are substantially greater
than can be explained by the emissions that have been reported. Certainly,
the discrepancies are large enough to call into serious question the reliability
of the emission factors that are used in bottom-up emissions accounting, the
many significant digits with which these emissions are typically reported, and
the viability of GHG emissions reduction legislation that depends solely on
bottom-up reporting procedures. Also called into question are the viability of
emissions trading that depends upon bottom-up emissions accounting, and the
feasibility of implementing legislation that requires reductions in emissions that
are small when compared with the ability of prescribed bottom-up methods to
resolve them.
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

1936 R. F. Weiss and R. G. Prinn
4. Future prospects for improvements
The uncertainties of current regional emission estimates either by top-down
or bottom-up approaches are commonly greater than 10–20%—sometimes very
much greater—and thus are grossly inadequate for verifying claims of emission
reductions by nations (e.g. mandated reductions under the Kyoto Protocol are
generally only a few to 10%).
Looking to the future, it is clear that the spatial density of precise high-
frequency atmospheric trace-gas measurements, whether using in situ or remotely
sensed methods, needs to be increased by an order of magnitude or more.
Equally important, the knowledge (theory, observation) embraced in models of
industrial or ecosystem fluxes should be incorporated into the model system
to enable estimation of uncertain parameters in these flux models as opposed
to simply the fluxes themselves. In essence, this approach combines the best
features of the bottom-up and top-down methods in flux estimation. Specifically,
the use of an adjoint of MATCH or other similar model, coupled to models
of the surface fluxes, would enable the estimation of uncertain parameters
or controls in the flux model with a more powerful statistical approach. An
adjoint of a model code is a complementary code that relates anomalies in
model outputs (e.g. mole fractions) to changes in model inputs (e.g. emissions)
or model parameters. Using the mathematical algorithms (but not the exact
goal) of control theory (e.g. [57]), one can formulate the state estimation
problem using a cost function J(pkin the simplified discussion earlier) that is
augmented with a demand for model consistency using the so-called Lagrange
multipliers. Any variable that can be affected by changes in any of the
control variables is called active, while variables that remain unaffected are
called passive. Thus, the computations consist of an active part in which the
coupling of the surface fluxes or atmospheric destruction rates is considered
and that would be improved as part of the estimation problem to minimize
J, and a passive part in which all the other elements of the system are
considered that do not change throughout the optimization (e.g. the MATCH
atmospheric circulation).
Through variation of the controls and initial conditions of the system, a
solution of the state vector (xkin the simplified discussion earlier) is sought that
minimizes J. The general structure of Jconsists of four sums measuring: (i) the
departure of the initial state from a first guess, (ii) the difference between the
observations and the model projections of them, (iii) the deviation of the controls
from a prior, and (iv) the demand that the state vector satisfies the various
model equations through the introduction of Lagrange multipliers. Besides its
key role in the J-minimization process, and thus in the state vector and model
parameter estimations, the three-dimensional model and flux model adjoints
could also be used to analyse the origins of observed mole-fraction anomalies
in terms of specific flux model parameters and initial conditions. This linking of
effects to causes enables observation-based corrections to the industry and natural
emission models.
Improvements to the driving circulations for CTMs are also needed.
The widely used meteorological circulation re-analyses (NCEP, ECMWF,
etc.) show significant differences, particularly in regions with very sparse
meteorological observations to correct the underlying weather forecasting models.
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

Greenhouse-gas emissions verification 1937
The modelling of convection and other sub-grid scale phenomena in both
chemical transport and weather forecasting models is another area needing
significant improvement.
Besides inclusion of all available surface and aircraft measurements of trace-gas
mole fractions, future inversions should also include column abundances obtained
from satellite and ground-based remote sensing. The above control system
approach also enables use of direct measurements of industrial and ecosystem
fluxes in the estimation algorithms. Besides accurate calibration comparisons
among laboratories, a key issue in combining multiple types of measurements
is an accurate estimation of the uncertainties in each type as they will be the
basis for the weighting contained in the inverses of the various measurement
uncertainty matrices involved in the inversions.
Finally, for GHGs that have natural, anthropogenic, industrial and biogenic
emissions, such as CO2,CH
4and N2O, measurements of atmospheric abundances
alone are inadequate to differentiate precisely among these different sources.
High-frequency in situ measurements of not just the total mole fractions of
these gases, but also their isotopic compositions are a new frontier in global
monitoring and hold the promise of revolutionizing understanding of the natural
cycles of these gases and verifying claims of emission reductions. At present,
stable isotopic measurements for CO2and CH4are carried out routinely only
by collecting air samples weekly to monthly at network stations for analysis
in a central laboratory by conventional gas-source magnetic-sector isotope
ratio mass spectrometry, but this sampling frequency is far too limited to
be used to accurately constrain estimates of sources and sinks by process
and by region. Also, deployment of these instruments at remote stations is
inhibited by their high costs, maintenance needs and power requirements,
which hinder reliable automation. Measurements of N2O isotopic composition
in the troposphere are even scarcer. Bulk nitrogen-15 (15N) data are available
(e.g. [58]), and some intra-molecular 15N measurements have also been made
(e.g. [59]).
However, high-frequency in situ isotope measurements are now becoming
feasible using optical (laser) detection. Recent improvements in mid-infrared
quantum cascade lasers (QCL) enable continuous wave (CW) operation near
room temperature (RT) with higher power, narrower line widths and higher
spectral mode purity than previously possible. The application of CWRT-QCLs
has greatly extended detection limits for atmospheric trace-gas measurements
without cryogenic cooling of the laser. CWRT-QCLs have been applied to
detection of the isotopes of CO2,CH
4and N2O (e.g. [60,61]). In addition to
the QCL work, there has been recent progress with isotopic monitoring using
other laser sources (e.g. [62,63]). For CH4and N2O, automated cryogenic pre-
concentration will probably be necessary to measure their isotopic compositions
with the precisions needed to differentiate their various surface fluxes (biogenic,
anthropogenic) and photochemical sinks.
The prospects for remote sensing from satellite, aircraft and surface platforms
of some of the above isotopomers and isotopologues are less clear, but
should also be pursued. These isotopically resolved trace-gas measurements of
ambient air would need to be accompanied by accurate field and laboratory
measurements of the isotopic signatures of relevant industrial, biological and
chemical processes.
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

1938 R. F. Weiss and R. G. Prinn
Finally, measurements of radiocarbon (14 C) are particularly valuable in
distinguishing fossil sources of CO2and CH4, but unfortunately the techniques
described above do not yet hold promise for measuring 14C in these gases in situ
with the required high frequency, sensitivity and precision. For the foreseeable
future, the highest temporal resolution 14C measurements will probably continue
to be made by accelerator mass spectrometry (e.g. [64]), which does not lend
itself to in situ operation.
5. Conclusions
The examples cited here show that the discrepancies between reported bottom-up
GHG emissions and measured top-down accumulations of these emissions in the
atmosphere can be substantial. There are many possible explanations for these
large discrepancies. Statistical uncertainties in emission factors used in bottom-up
protocols are always possible, but such errors ought to be mostly random, and
thus do not explain the tendency for the actual emissions to exceed the reported
ones, more often than not. When emissions from industrial processes are measured
at their sources to establish emission factors, the equipment may be adjusted to
minimize emissions, so that the measured values may be lower than they are under
typical day-to-day operating conditions, and this would lead to under-reporting.
Furthermore, the possible existence of unaccounted or unidentified sources, such
as fugitive emissions during industrial production or transportation, would also
lead to under-reporting. In addition, the negative impact of GHG emissions on
climate, and the financial value of emissions reductions in carbon-equivalent
trading markets, both create incentives to under-report actual emissions, whether
consciously or subconsciously.
Because effective emissions control legislation ultimately must depend upon
enforcement by reliable bottom-up methods, it is essential that these discrepancies
be resolved. But since the legislation is national or regional in scale, not global,
top-down emission estimates must be determined at these same scales. In addition
to recording background GHG trends driven by global and hemispheric emissions,
high-frequency atmospheric measurements at well-chosen ground-based station
locations also record GHGs that have been elevated above their background
values because of regional emissions. By analysing these and other atmospheric
measurements with three-dimensional atmospheric transport and mixing models
using inverse numerical methods, it is possible to map and quantify regional
emissions over time. To reconcile discrepancies between top-down assessments
and those obtained by bottom-up protocols, we propose that optimal estimation
and control-theory methods be used to identify and correct the most likely causes
of the observed discrepancies.
Even with the relatively sparse existing network of ground-based measurement
stations, using current modelling techniques, it is possible to rival the accuracies
of bottom-up emissions estimates for some GHGs in some regions. Awareness by
policy-makers of the large discrepancies which have been found between reported
bottom-up emissions and emissions determined from atmospheric measurements
has, so far, been limited, but there is an example for European CH4emissions
in which German CH4emissions for 2001 reported to the UNFCCC have been
revised upward substantially, and thus were brought into far better agreement
with the modelled atmospheric observations [65].
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

Greenhouse-gas emissions verification 1939
Meeting the goals of effective top-down GHG emissions verification and
refining bottom-up protocols to bring about convergence is a research problem
that will require substantial increases in the density and scope of atmospheric
measurements, as well as improvements in inverse modelling capabilities, but
many of the basic components exist already and need only to be increased in scale.
Such an initiative would be large when compared with most current atmospheric
research programmes, but could easily be supported by an annual investment of
less than 1 per cent of the $144 billion US$ currently invested in global carbon-
equivalent trading markets [66], with the added benefit of reducing the volatility
of these markets and thereby increasing investment in emissions reductions.
We thank the convenors of the Discussion Meeting ‘Greenhouse gases in the Earth system: setting
the Agenda to 2030’ for the opportunity to present our views on this timely issue. We also thank
our colleagues in the AGAGE community for their many contributions to the work discussed here,
and NASA’s Upper Atmosphere Research Program for its continuing support of AGAGE.
References
1 Forster, P. et al. 2007 Changes in atmospheric constituents and in radiative forcing. In Climate
change 2007: the physical science basis. Contribution of Working Group I to the Fourth
Assessment Report of the Intergovernmental Panel on Climate Change (eds S. Solomon, D. Qin,
M. Manning, Z. Chen, M. Marquis, K. B. Averyt, M. Tignor & H. L. Miller). Cambridge, UK:
Cambridge University Press.
2 IPCC. 2006 IPCC Guidelines for National Greenhouse Gas Inventories. Prepared by the
National Greenhouse Gas Inventories Programme. In 2006 Institute for Global Environmental
Strategies (IGES) (eds H. S. Eggleston, L. Buendia, K. Miwa, T. Ngara & K. Tanabe), Japan.
See http://www.ipcc-nggip.iges.or.jp/public/2006gl/index.html.
3 UNFCCC. 2010 National inventory submissions. See http://unfccc.int/national_reports/
annex_i_ghg_inventories/national_inventories_submissions/items/5270.php.
4 Enting, I. G. 2002 Inverse problems in atmospheric constituent transport. Cambridge, UK:
Cambridge University Press.
5 Kasibhatla, P., Heimann, M., Rayner, P., Mahowald, N., Prinn, R. G. & Hartley, D. E. (eds)
2000 Inverse methods in global biogeochemical cycles. Geophysical Monograph Series, vol. 114.
Washington, DC: American Geophysics Union.
6 Prinn, R. G. 2000 Measurement equation for trace chemicals in fluids and solution of its
inverse. In Inverse methods in global biogeochemical cycles (eds P. Kasibhatla et al.), pp. 3–18.
Geophysical Monograph Series, vol. 114. Washington, DC: American Geophysics Union.
7 Chen, Y.-H. & Prinn, R. G. 2005 Atmospheric modeling of high- and low-frequency methane
observations: importance of interannually varying transport. J. Geophys. Res. 110, D10303.
(doi:10.1029/2004JD005542)
8 Chen, Y.-H. & Prinn, R. G. 2006 Estimation of atmospheric methane emissions between
1996–2001 using a 3D global chemical transport model. J. Geophys. Res. 111, D10307.
(doi:10.1029/2005JD006058)
9 Prinn, R. G. et al. 2005 Evidence for variability of atmospheric hydroxyl radicals over the past
quarter century. Geophys. Res. Lett. 32, L07809. (doi:10.1029/2004GL022228)
10 Mahowald, N. M., Prinn, R. G. & Rasch, P. J. 1997 Deducing CCl3F emissions using an inverse
method and chemical transport models with assimilated winds. J. Geophys. Res. 102, 28 153–
28 168. (doi:10.1029/97JD02086)
11 Hartley, D. E. & Prinn, R. G. 1993 On the feasibility of determining surface emissions of trace
gases using an inverse method in a three-dimensional chemical transport model. J. Geophys.
Res. 98, 5183–5198. (doi:10.1029/92JD02594)
12 Mulquiney, J. E., Taylor, J. A., Jakeman, A. J., Norton, J. P. & Prinn, R. G. 1998 A new
inverse method for trace-gas flux estimation. II. Application to tropospheric CFCl3fluxes. J.
Geophys. Res. 103, 1429–1442. (doi:10.1029/97JD01811)
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

1940 R. F. Weiss and R. G. Prinn
13 Lawrence, M. G., Crutzen, P. J., Rasch, P. J., Eaton, B. E. & Mahowald, N. M. 1999 A model
for studies of tropospheric photochemistry: description, global distributions, and evaluation.
J. Geophys. Res. 104, 26 245–26 277. (doi:10.1029/1999JD900425)
14 Rasch, P. J., Mahowald, N. M. & Eaton, B. E. 1997 Representations of transport, convection,
and the hydrologic cycle in chemical transport models: implications for the modeling of short-
lived and soluble species. J. Geophys. Res. 102, 28 127–28 138. (doi:10.1029/97JD02087)
15 Dargaville, R., McGuire, A. D. & Rayner, P. 2002 Estimates of large-scale fluxes in high
latitudes from terrestrial biosphere models and an inversion of atmospheric CO2measurements.
Clim. Change 55, 273–285. (doi:10.1023/A:1020295321582)
16 Gurney, K. R. et al. 2002 Towards robust regional estimates of CO2sources and sinks using
atmospheric transport models. Nature 415, 626–630. (doi:10.1038/415626a)
17 Gurney, K. R. et al. 2003 Transcom 3 CO2inversion intercomparison. I. Annual mean
control results and sensitivity to transport and prior flux information. Tellus B 55, 555–579.
(doi:10.1034/j.1600-0889.2003.00049.x)
18 Law, R. M., Chen, Y.-H., Gurney, K. R. & TransCom 3 modellers. 2003 TransCom3 CO2
inversion intercomparisons. Part 2. Sensitivity of annual mean results to data choices. Tellus B
55, 580–595. (doi:10.1034/j.1600-0889.2003.00053.x)
19 Huang, J. et al. 2008 Estimation of regional emissions of nitrous oxide from 1997 to 2005 using
multinetwork measurements, a chemical transport model, and an inverse method. J. Geophys.
Res. 113, D17313. (doi:10.1029/2007JD009381)
20 Xiao, X. et al. 2010 Atmospheric three-dimensional inverse modeling of regional industrial
emissions and global oceanic uptake of carbon tetrachloride. Atmos. Chem. Phys. 10, 10 421–
10 434. (doi:10.5194/acp-10-10421-2010)
21 Draxler, R. R. & Hess, G. D. 1997 Description of the HYSPLIT_4 modeling system. NOAA
Technical Memorandum ERL ARL-224, NOAA Air Resources Laboratory, Silver Spring, MD.
22 Draxler, R. R. & Hess, G. D. 1998 An overview of the HYSPLIT_4 modelling system for
trajectories, dispersion and deposition. Australian Met. Mag. 47, 295–308.
23 Kleiman, G. & Prinn, R. G. 2000 Measurement and deduction of emissions of trichloroethene,
tetrachloroethene and trichloromethane (chloroform) in the northeastern United States and
southeastern Canada. J. Geophys. Res. 105, 28 875–28 893. (doi:10.1029/2000JD900513)
24 Biraud, S., Ciais, P., Ramonet, M., Simmonds, P., Kazan, V., Monfray, P., O’Doherty, S., Spain,
T. G. & Jennings, S. G. 2000 European greenhouse gas emissions estimated from continuous
atmospheric measurements and radon 222 at Mace, Head Ireland. J. Geophys. Res. 105, 1351–
1366. (doi:10.1029/1999JD900821)
25 Derwent, R., Simmonds, P. G., O’Doherty, S., Ciais, P. & Ryall, D. 1998 European source
strengths and northern hemisphere baseline concentrations of radiatively active trace gases at
Mace Head, Ireland. Atmos. Environ. 32, 3703–3715. (doi:10.1016/S1352-2310(98)00093-4)
26 Derwent, R. G., Simmonds, P. G., O’Doherty, S. & Ryall, D. B. 1998 The impact of
the Montreal Protocol on halocarbon concentrations in Northern hemisphere baseline and
European air masses at Mace Head, Ireland from 1987–1996. Atmos. Environ. 32, 3689–3702.
(doi:10.1016/S1352-2310(98)00092-2)
27 Manning, A. J., Ryall, D. B., Derwent, R. G., Simmonds, P. G. & O’Doherty, S. 2003 Estimating
European emissions of ozone-depleting and greenhouse gases using observations and a modeling
back-attribution technique. J. Geophys. Res. 108, 4405. (doi:10.1029/2002JD002312)
28 Ryall, D. B., Derwent, R. G., Manning, A. J., Simmonds, P. G. & O’Doherty, S. 2001 Estimating
source regions of European emissions of trace gases from observations at Mace Head. Atmos.
Environ. 35, 2507–2523. (doi:10.1016/S1352-2310(00)00433-7)
29 Ryall, D. B., Maryon, R. H., Derwent, R. G. & Simmonds, P. G. 1998 Modelling long-
range transport of CFCs to Mace Head Ireland. Q. J. R. Meteor. Soc. 124, 417–446.
(doi:10.1002/qj.49712454604)
30 Cox, M. L., Sturrock, G. A., Fraser, P. J., Siems, S. T., Krummel, P. B. & O’Doherty, S.
2003 Regional sources of methyl chloride, chloroform and dichloromethane identified from
AGAGE observations at Cape Grim, Tasmania, 1998–2000. J. Atmos. Chem. 45, 79–99.
(doi:10.1023/A:1024022320985)
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

Greenhouse-gas emissions verification 1941
31 Wang, Y.-P. & Bentley, S. 2002 Development of a spatially explicit inventory of methane
emissions from Australia and its verification using atmospheric concentration data. Atmos.
Environ. 36, 4965–4975. (doi:10.1016/S1352-2310(02)00589-7)
32 Stohl, A. et al. 2009 An analytical inversion method for determining regional and global
emissions of greenhouse gases: sensitivity studies and application to halocarbons. Atmos. Chem.
Phys. 9, 1597–1620. (doi:10.5194/acp-9-1597-2009)
33 Cunnold, D. M., Fraser, P. J., Weiss, R. F., Prinn, R. G., Simmonds, P. G., Miller, B. R.,
Alyea, F. N. & Crawford, A. J. 1994 Global trends and annual releases of CCl3F and CCl2F2
estimated from ALE/GAGE and other measurements from July 1978 to June 1991. J. Geophys.
Res. 99, 1107–1126. (doi:10.1029/93JD02715)
34 Hirsch, A. I., Michalak, A. M., Bruhwiler, L. M., Peters, W., Dlugokencky, E. J. & Tans, P. P.
2006 Inverse modeling estimates of the global nitrous oxide surface flux from 1998–2001. Glob.
Biogeochem. Cycles 20, GB1008. (doi:10.1029/2004GB002443)
35 Miller, B. R., Weiss, R. F., Salameh, P. K., Tanhua, T., Greally, B. R., Mühle, J. & Simmonds,
P. G. 2008 Medusa: a sample preconcentration and GC/MS detector system for in situ
measurements of atmospheric trace halocarbons, hydrocarbons and sulfur compounds. Anal.
Chem. 80, 1536–1545. (doi:10.1021/ac702084k)
36 Prinn, R. G. et al. 2000 A history of chemically and radiatively important gases in air deduced
from ALE/GAGE/AGAGE. J. Geophys. Res. 105, 17 751–17 792. (doi:10.1029/2000JD900141)
37 Butler, J. H., Battle, M., Bender, M. L., Montzka, S. A., Clarke, A. D., Saltzman, E. S., Sucher,
C. M., Severinghaus, J. P. & Elkins, J. W. 1999 A record of atmospheric halocarbons during
the twentieth century from polar firn air. Nature 399, 749–755. (doi:10.1038/21586)
38 Rigby, M. et al. 2008 Renewed growth of atmospheric methane. Geophys. Res. Lett. 35, L22805.
(doi:10.1029/2008GL036037)
39 Dlugokencky, E. J. et al. 2009 Observational constraints on recent increases in the atmospheric
CH4burden. Geophys. Res. Lett. 36, L18803. (doi:10.1029/2009GL039780)
40 Bousquet, P. et al. 2006 Contribution of anthropogenic and natural sources to atmospheric
methane variability. Nature 443, 439–443. (doi:10.1038/nature05132)
41 Mühle, J. et al. 2010 Perfluorocarbons in the global atmosphere: tetrafluoromethane,
hexafluoroethane, and octafluoropropane. Atmos. Chem. Phys. 10, 5145–5164. (doi:10.5194/
acp-10-5145-2010)
42 Deeds, D. A., Mühle, J. & Weiss, R. F. 2008 Tetrafluoromethane in the deep North Pacific
Ocean. Geophys. Res. Lett. 35, L14606. (doi:10.1029/2008GL034355)
43 IAI. 2009 The International Aluminium Institute Report on the aluminium industry’s global
perfluorocarbon gas emissions reduction programme. Results of the 2008 Anode Effect Survey.
International Aluminium Institute, London, UK.
44 EDGAR. 2009 Emission database for global atmospheric research, release v. 4.0. European
Commission, Joint Research Centre/Netherlands Environmental Assessment Agency. See
http://edgar.jrc.ec.europa.eu.
45 Vollmer, M. K. & Weiss, R. F. 2002 Simultaneous determination of sulfur hexafluoride and
three chlorofluorocarbons in water and air. Mar. Chem. 78, 137–148. (doi:10.1016/S0304-
4203(02)00015-4)
46 Levin, I. et al. 2010 The global SF6source inferred from long-term high precision atmospheric
measurements and its comparison with emission inventories. Atmos. Chem. Phys. 10, 2655–
2662. (doi:10.5194/acp-10-2655-2010)
47 Maiss, M. & Levin, I. 1994 Global increase of SF6observed in the atmosphere. Geophys. Res.
Lett. 21, 569–572. (doi:10.1029/94GL00179)
48 Rigby, M. et al. 2010 History of atmospheric SF6from 1973 to 2008. Atmos. Chem. Phys.
10, 10 305–10 320. (doi:10.5194/acp-10-10305-2010)
49 Geller, L. S., Elkins, J. W., Lobert, J. M., Clarke, A. D., Hurst, D. F., Butler, J. H. & Myers,
R. C. 1997 Tropospheric SF6: observed latitudinal distribution and trends, derived emissions
and interhemispheric exchange time. Geophys. Res. Lett. 24, 675–678. (doi:10.1029/97GL00523)
50 Peters, W. et al. 2004 Toward regional-scale modeling using the two-way nested global model
TM5: characterization of transport using SF6.J. Geophys. Res. 109, D19314. (doi:10.1029/
2004JD005020)
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from

1942 R. F. Weiss and R. G. Prinn
51 Maiss, M. & Brenninkmeijer, C. M. 1998 Atmospheric SF6: trends, sources, and prospects.
Environ. Sci. Technol. 32, 3077–3086. (doi:10.1021/es9802807)
52 Prather, M. J. & Hsu, J. 2008 NF3, the greenhouse gas missing from Kyoto. Geophys. Res.
Lett. 35, L12810. (doi:10.1029/2008GL034542)
53 Weiss, R. F., Mühle, J., Salameh, P. K. & Harth, C. M. 2008 Nitrogen trifluoride in the global
atmosphere. Geophys. Res. Lett. 35, L20821. (doi:10.1029/2008GL035913)
54 Lee, J.-Y., Lee, J. B., Moon, D. M., Souk, J. H., Lee, S. Y. & Kim, J. S. 2007 Evaluation method
on destruction and removal efficiency of perfluorocarbons from semiconductor and display
manufacturing. Bull. Korean Chem. Soc. 28, 1383–1388. (doi:10.5012/bkcs.2007.28.8.1383)
55 Mühle, J. et al. 2009 Sulfuryl fluoride in the global atmosphere. J. Geophys. Res. 114, D05306.
(doi:10.1029/2008JD011162)
56 Papadimitriou, V. C., Portmann, R. W., Fahey, D. W., Mühle, J., Weiss, R. F. & Burkholder,
J. B. 2008 Experimental and theoretical study of the atmospheric chemistry and global warming
potential of SO2F2.J. Phys. Chem. A. 112, 12657–12666. (doi:10.1021/jp806368u)
57 Wunsch, C. 1996 The ocean circulation inverse problem. New York, NY: Cambridge University
Press.
58 Rahn, T. & Wahlen, M. 1997 Stable isotope enrichment in stratospheric nitrous oxide. Science
278, 1776–1778. (doi:10.1126/science.278.5344.1776)
59 Kaiser, J., Röckmann, T. & Brenninkmeijer, C. A. M. 2003 Complete and accurate mass
spectrometric isotope analysis of tropospheric nitrous oxide. J. Geophys. Res. 108, 4476.
(doi:10.1029/2003JD003613)
60 Nelson, D., McManus, J., Herndon, S., Zahniser, M., Tuzson, B. & Emmenegger, L. 2008
New method for isotopic ratio measurements of atmospheric carbon dioxide using a 4.3 mm
pulsed quantum cascade laser. Appl. Phys., B Lasers Opt. 90, 301–309. (doi:10.1007/
s00340-007-2894-1)
61 Nelson, D. D., McManus, J. B., Urbanski, S., Herndon, S. & Zahniser, M. S. 2004 High
precision measurements of atmospheric nitrous oxide and methane using thermoelectrically
cooled mid-infrared quantum cascade lasers and detectors. Spectrochim. Acta A 60, 3325–3335.
(doi:10.1016/j.saa.2004.01.033)
62 Castrillo, A., Casa, G., Kerstel, E. & Gianfrani, L. 2005 Diode laser absorption spectrometry
for 13CO2/12 CO2isotope ratio analysis: investigation on precision and accuracy levels. Appl.
Phys., B Lasers Opt. 81, 863–869. (doi:10.1007/s00340-005-1949-4)
63 Waechter, H. & Sigrist, M. 2007 Mid-infrared laser spectroscopic determination of isotope
ratios of N2O at trace levels using wavelength modulation and balanced path length detection.
Appl. Phys., B Lasers Opt. 87, 539–546. (doi:10.1007/s00340-007-2576-z)
64 Graven, H. D., Guilderson, T. P. & Keeling, R. F. 2007 Methods for high-precision 14C AMS
measurement on atmospheric CO2at LLNL. Radiocarbon 49, 349–356.
65 Bergamaschi, P., Krol, M., Dentener, F., Vermeulen, A., Meinhardt, F., Graul, R., Ramonet,
M., Peters, W. & Dlugokencky, E. J. 2005 Inverse modelling of national and European
CH4emissions using the atmospheric zoom model TM5. Atmos. Chem. Phys. 5, 2431–2460.
(doi:10.5194/acp-5-2431-2005)
66 Kossoy, A. & Ambrosi, P. 2010 State and Trends of the Carbon Market 2010. Washington, DC:
The World Bank.
Phil. Trans. R. Soc. A (2011)
on March 12, 2014rsta.royalsocietypublishing.orgDownloaded from