Atmospheric Oxygen and Associated Tracers from
Flask Sampling and Continuous Measurements: Tools
for Studying the Global Carbon Cycle
Inauguraldissertation
der Philosophisch-naturwissenschaftlichen Fakultät
der Universität Bern
vorgelegt von
Patrick Sturm
von Niederstocken (BE)
Leiter der Arbeit:
PD Dr. M. Leuenberger
Prof. Dr. T. F. Stocker
Abteilung für Klima- und Umweltphysik
Physikalisches Institut der Universität Bern
Atmospheric Oxygen and Associated Tracers from
Flask Sampling and Continuous Measurements: Tools
for Studying the Global Carbon Cycle
Inauguraldissertation
der Philosophisch-naturwissenschaftlichen Fakultät
der Universität Bern
vorgelegt von
Patrick Sturm
von Niederstocken (BE)
Leiter der Arbeit:
PD Dr. M. Leuenberger
Prof. Dr. T. F. Stocker
Abteilung für Klima- und Umweltphysik
Physikalisches Institut der Universität Bern
Von der Philosophisch-naturwissenschaftlichen Fakultät angenommen.
Bern, den 27. Januar 2005
Der Dekan:
Prof. Dr. P. Messerli
Contents
Thesis Summary
7
1 Introduction
1.1
1.2
1.3
The Global Carbon Cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Atmospheric Oxygen as a Tracer for Carbon Cycle Processes . . . . . . . . . . 12
Outline of Chapters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2 Measurement Methods and Tests
2.1
2.2
2.3
2.4
2.5
2.6
9
Introduction . . . . . . . . . . . . . . . . . . . . . . . . .
Equipment . . . . . . . . . . . . . . . . . . . . . . . . .
2.2.1 Mass Spectrometer System . . . . . . . . . . . .
2.2.2 Pressure Regulators and Reference Cylinders . .
Mass Spectrometer Performance and Artifacts . . . . . .
2.3.1 Instrument Precision . . . . . . . . . . . . . . . .
2.3.2 Isobaric Interferences . . . . . . . . . . . . . . . .
2.3.3 Pressure Imbalance Eect . . . . . . . . . . . . .
2.3.4 Sample/Standard Asymmetry . . . . . . . . . . .
2.3.5 Repeated Measurement Drift . . . . . . . . . . .
CO2 Mixing Ratio Measurements by Mass Spectrometry
2.4.1 N2 O Background . . . . . . . . . . . . . . . . . .
2.4.2 Cross Contamination . . . . . . . . . . . . . . . .
Air Standard Measurements . . . . . . . . . . . . . . . .
2.5.1 Working and Reference Standards . . . . . . . .
2.5.2 CO2 Primary Standards . . . . . . . . . . . . . .
Summary . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
19
19
19
19
21
22
22
23
23
25
25
26
26
27
29
29
33
34
3 Permeation of Atmospheric Gases Through Polymer O-Rings Used in Flasks
for Air Sampling
37
3.1
3.2
3.3
Introduction . . . . . . . . . . . . . . . .
Permeation Process . . . . . . . . . . . .
3.2.1 Theory . . . . . . . . . . . . . . .
3.2.2 Measurement of Time-Dependent
Flask Storage Drift . . . . . . . . . . . .
3.3.1 Permeation Inuence on O2 /N2 .
3.3.2 Permeation of Water Vapor . . .
3.3.3 Permeation of Ar and CO2 . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
Permeation Rate
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
38
38
38
40
41
41
43
46
4
CONTENTS
3.4
3.5
Double O-ring Valves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Conclusions and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4 Atmospheric O2 , CO2 and δ 13 C Measurements from the Remote Sites Jungfraujoch, Switzerland, and Puy de Dôme, France
55
4.1
4.2
4.3
4.4
Introduction . . . . . . . . . . . . . . . . . . . . . . . .
Sampling Sites and Methods . . . . . . . . . . . . . . .
4.2.1 Sampling Locations . . . . . . . . . . . . . . . .
4.2.2 Flask Sampling . . . . . . . . . . . . . . . . . .
4.2.3 O2 /N2 Analysis . . . . . . . . . . . . . . . . . .
4.2.4 CO2 Analysis . . . . . . . . . . . . . . . . . . .
4.2.5 δ 13 C Analysis . . . . . . . . . . . . . . . . . . .
4.2.6 Data Selection and Drift Corrections . . . . . .
Results and Discussions . . . . . . . . . . . . . . . . .
4.3.1 Seasonal Variability of Atmospheric O2 /N2 and
4.3.2 δ 13 C of Atmospheric CO2 . . . . . . . . . . . .
Summary and Conclusions . . . . . . . . . . . . . . . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
. . .
CO2
. . .
. . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
56
56
56
57
57
58
58
59
60
60
62
64
5 Atmospheric O2 , CO2 and δ 13 C Observations from Aircraft Sampling over
Grin Forest, Perthshire, UK
67
5.1
5.2
5.3
5.4
Introduction . . . . . . . . . . . . . . .
Sampling Sites and Methods . . . . . .
5.2.1 Flight Location . . . . . . . . .
5.2.2 Aircraft and Sampling Protocol
5.2.3 Flask Sampling . . . . . . . . .
5.2.4 Flask Analysis . . . . . . . . .
Results and Discussions . . . . . . . .
Summary and Conclusions . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
68
68
68
69
69
71
73
79
6 Continuous Observations of CO2 , 222 Rn, O2 /N2 , Ar/N2 and Stable Isotopes
of CO2 , O2 , N2 and Ar at Bern, Switzerland
83
6.1
6.2
6.3
6.4
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Sampling and Analysis Techniques . . . . . . . . . . . . . . . . . .
6.2.1 Sampling Site . . . . . . . . . . . . . . . . . . . . . . . . . .
6.2.2 CO2 Mixing Ratio . . . . . . . . . . . . . . . . . . . . . . .
6.2.3 δ 13 C and δ 18 O of CO2 . . . . . . . . . . . . . . . . . . . . .
6.2.4 Elemental and Isotopic Ratios of Air . . . . . . . . . . . . .
6.2.5 222 Rn activity . . . . . . . . . . . . . . . . . . . . . . . . . .
Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . .
6.3.1 Temperature Dependent Fractionation at the Air Intake . .
6.3.2 The CO2 Record . . . . . . . . . . . . . . . . . . . . . . . .
6.3.3 δ 13 C of CO2 , δ 18 O of CO2 and O2 /N2 Measurements . . . .
6.3.4 δ 29 N2 , δ 34 O2 and δ 36 Ar of Air . . . . . . . . . . . . . . . . .
6.3.5 222 Rn Tracer Method to Estimate Regional CO2 Emissions
Summary and Outlook . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
83
84
84
85
85
85
86
86
86
88
90
96
97
98
CONTENTS
5
7 Development of Continuous O2 and CO2 Analyzer Systems
7.1
7.2
7.3
7.4
7.5
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Paramagnetic and Electrochemical Oxygen Sensors . . . . . . . . .
O2 and CO2 Analyzer System Design at Jungfraujoch, Switzerland
The CARIBIC O2 Analyzer . . . . . . . . . . . . . . . . . . . . . .
7.4.1 The CARIBIC Project . . . . . . . . . . . . . . . . . . . . .
7.4.2 Analyzer System Design . . . . . . . . . . . . . . . . . . . .
Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
103
103
104
105
107
107
107
109
A About Straight Line Regression Models when Both Variables are Subject
to Error
111
A.1 Functional Regression Models . . . . . .
A.1.1 Ordinary Least Squares . . . . .
A.1.2 Measurement Error Model . . . .
A.1.3 Orthogonal Distance Regression .
A.1.4 Geometric Mean Regression . . .
A.2 Structural Regression Models . . . . . .
A.3 Bivariate Correlated Errors and Intrinsic
A.4 Keeling Plots . . . . . . . . . . . . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
Scatter
. . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
111
112
113
114
115
115
116
116
Publications
120
Acknowledgements
121
Curriculum Vitae
122
Thesis Summary
Atmospheric oxygen measurements, together with carbon dioxide, are a useful tool to deduce
and disentangle carbon uxes due to surface exchange and atmospheric transport processes.
This thesis presents measurements of atmospheric oxygen and associated tracers and discusses
experimental artifacts that have to be considered for high precision O2 measurements. Different analysis techniques for O2 measurements have been established and tested. Within the
European project AEROCARB ask samples of atmospheric air from the high-altitude site
Jungfraujoch, Switzerland, the mountain site Puy de Dôme, France, as well as from aircraftbased vertical proling at Grin Forest, UK, were analyzed for O2 /N2 , CO2 and δ 13 C by
isotope ratio mass spectrometry.
The four-year records of Jungfraujoch and the three-year records of Puy de Dôme show
distinct seasonal cycles and superimposed long-term trends in the measured parameters. At
Jungfraujoch the seasonal variations are about two times smaller than at Puy de Dôme. The
seasonal O2 :CO2 correlation gives at both sites slopes of about 2 mol O2 /mol CO2 . Stable
carbon isotope ratios of source CO2 showed depleted values in wintertime and isotopically
enriched values in summer.
At the Grin Forest site, the peak-to-peak amplitude of the seasonal cycle of O2 /N2
decreases from 171 per meg at 800 m to 113 per meg at 3100 m. Furthermore the seasonal cycle
is shifted from low to high altitudes with a lag of about 1 month. The same features are
observed for CO2 with a decrease in the peak-to-peak amplitude of the seasonal cycle from
17.6 ppm at 800 m to 11.4 ppm at 3100 m. The vertical proles show decreasing O2 /N2 ratios
in summer and increasing O2 /N2 ratios in wintertime with increasing sampling height, due to
surface exchange of oxygen with the land biosphere. The O2 :CO2 exchange ratio for selected
vertical proles varies between −1.5 and −2.4 mol O2 /mol CO2 .
Technical aspects of the mass spectrometric method and gas handling procedures played
a key role in achieving the required precision for O2 measurements. In particular, glass ask
storage drift due to permeation of atmospheric gases through polymer seals proved to be a
crucial issue for δO2 /N2 ask sampling programs.
Continuous observations of CO2 , 222 Rn, O2 /N2 , and stable isotopes of CO2 at Bern,
Switzerland, shed light on diurnal and seasonal patterns of the carbon cycle in an urban atmosphere. There is considerable variance in nighttime δ 13 C and δ 18 O of source CO2 throughout the year, however, with generally lower values in winter compared to summertime. The
O2 :CO2 oxidation ratio during the nighttime build-up of CO2 varies between −0.96 and −1.54.
Concurrent Ar/N2 measurements showed again the importance of artifacts like thermal fractionation for precise measurements of O2 /N2 . Using the correlation from short term uctuations of CO2 and 222 Rn we estimated a mean CO2 ux density between February 2004 and
April 2004 in the region of Bern of 95 ± 39 tC km−2 month−1 .
Additionally, analyzer systems for continuous O2 and CO2 measurements were developed.
8
THESIS SUMMARY
In the framework of the European CarboEurope-IP project O2 and CO2 will be measured
continuously at Jungfraujoch. Atmospheric O2 is measured by paramagnetic as well as fuel cell
techniques. A second O2 and CO2 analyzer system dedicated for making measurements during
long distance ights in a passenger aircraft was developed in the context of the European
CARIBIC project. Our instrument is based on electrochemical cells for O2 analysis and a
NDIR analyzer for CO2 measurements.
In summary, this thesis contributes to the observational database for atmospheric O2
and CO2 measurements and reveals important technical issues for accurate measurements of
atmospheric composition.
Chapter 1
Introduction
There is increasing evidence that human activities may have dramatic eects on Earth's climate and on marine and terrestrial ecosystems. Direct observations of atmospheric and oceanic
temperature as well as numerous indirect observations of sea ice extent, snow cover and glacier
retreats all point to a warming world. The scientic consensus is growing that this warming is
primarily due to man-made emissions of greenhouse gases [IPCC , 2001; Tett et al., 2002; Stott
et al., 2004]. Atmospheric greenhouse gases are transparent to short-wave solar radiation,
but absorb long-wave (infrared) radiation, emitted by the Earth's surface, thus trapping heat
within the atmosphere. The main greenhouse gases include natural abundant species such
as water vapor (H2 O), carbon dioxide (CO2 ), methane (CH4 ), tropospheric ozone (O3 ) and
nitrous oxide (N2 O) and predominantly man-made compounds such as chlorouorocarbons
(CFCs). The largest contribution (about 60 %) to the anthropogenic increase in radiative
forcing since pre-industrial times is from CO2 . Since the beginning of the industrialization
in the 18th century the atmospheric CO2 concentration has risen from about 280 ppm (ppm
= parts per million) to 376 ppm in the year 2003 (Figure 1.1) [Neftel et al., 1985; Etheridge
et al., 1996; Keeling and Whorf , 2004]. This increase in CO2 concentration over the last 250
years is mainly due to combustion of fossil fuels and land use changes, such as deforestation.
Measurements of air entrapped in ice cores from Antarctica show that the present CO2 concentration has not been exceeded during the past 620,000 years [Petit et al. 1999; U. Siegenthaler,
personal communication], which highlights the dimension of the recent anthropogenic perturbation. Greenhouse gas concentrations are projected to further increase over the next century
and the climate and ecosystems of our planet will face unprecedented and perhaps irreversible
changes. It is therefore urgent to improve our understanding of the mechanisms and feedbacks
in the climate system in order to condently forecast future climate change and assess options
for mitigation. Quantifying the eect of anthropogenic emissions of greenhouse gases on the
climate requires precise knowledge of the global biogeochemical cycles.
1.1 The Global Carbon Cycle
The atmospheric CO2 concentration is regulated by the biogeochemical carbon cycle that
involves the exchange of carbon with the terrestrial biosphere, the ocean and the marine
biota, as well as interactions with sediments and the lithosphere (Figure 1.2). The CO2 in the
atmosphere constitutes only a small fraction of the total carbon stored in the other reservoirs.
Presently, about 790 PgC (1 PgC = 1015 grams of carbon) are in the atmosphere, about
10
1. INTRODUCTION
Atmospheric CO2 (ppm)
380
380
360
360
340
340
320
320
1960
300
1970
1980
1990
2000
280
260
1000
1200
1400
1600
Year
1800
2000
Figure 1.1: The evolution of atmospheric CO2 during the last millennium. The grey symbols denote mea-
surements of air bubbles trapped in ice cores recovered from several sites in Antarctica [Neftel et al., 1985;
Etheridge et al., 1996; Indermühle et al., 1999; Monnin et al., 2004]. The black symbols show data of direct
atmospheric measurements from Mauna Loa, Hawaii, since 1958 [Keeling and Whorf , 2004]. Monthly mean
values shown in the inset reveal seasonal variations due to the terrestrial biosphere.
2300 PgC are stored in the terrestrial biosphere (including soils), and about 38,000 PgC are in
the ocean, most of it as dissolved inorganic carbon. Sedimentation and rock weathering are
processes that take place on timescales of several thousand years and may thus be neglected on
timescales of up to a few hundred years. Currently, we are emitting about 6.5 PgC yr−1 to the
atmosphere from fossil fuel burning and cement production [Marland et al., 2003], and about
2 PgC yr−1 from land use change [Houghton , 2003], leading to a steady increase in atmospheric
CO2 concentration. Atmospheric CO2 has been monitored at many sites world-wide for up to
46 years [e.g. GLOBALVIEW-CO2 , 2004]. These measurements have shown that during the
past few decades about half of the annual input from fossil fuel emissions have remained in
the atmosphere [Prentice et al., 2001]. The rest has been absorbed by the ocean and the land
ecosystems.
The ocean is the largest of the fast exchanging reservoirs, which is due to the chemical
reactivity of CO2 with water. In contrast to most of the other air constituents (e.g. O2 , N2
and Ar), gaseous CO2 is not only physically dissolved in water, but reacts with water to form
2−
bicarbonate (HCO−
3 ), carbonate (CO3 ) and carbonic acid (H2 CO3 ). The uptake capacity
for CO2 of the ocean is governed by biological processes and the inorganic chemistry of CO2 ,
which is dependent on seawater temperature, salinity, CaCO3 formation and dissolution (in
shells and corals), and global and regional ocean circulation patterns. It is estimated that
the oceans have been a net sink of atmospheric CO2 in the 1990s of about 2 (±0.6) PgC yr−1
[Prentice et al., 2001; Plattner et al., 2002; Keeling and Garcia , 2002; Bopp et al., 2002].
The terrestrial biosphere takes up carbon from the atmosphere through photosynthesis
(gross primary production). Part of this assimilated carbon is respired by the plants (autotrophic respiration). The rest is allocated within the plant to make up its roots, wood and
leaves (net primary production). On the other hand, carbon is lost to the atmosphere by
1.1. THE GLOBAL CARBON CYCLE
11
Figure 1.2: Reservoirs and uxes in the global carbon cycle estimated for the 1980s. Black arrows show the
pre-industrial steady-state uxes (in petagrams of carbon per year, 1 Pg = 1015 g) and red arrows denote the
net uxes of anthropogenic carbon. The red numbers in the reservoirs indicate the changes resulting from
human activities since pre-industrial times (Figure from Sarmiento and Gruber [2002]).
the consumption of organic matter by soil organisms (heterotrophic respiration). The net
accumulation of carbon by an ecosystem (net ecosystem production) is the balance of net primary production and heterotrophic respiration. The net biome production, nally, is the net
carbon ux from the atmosphere to the terrestrial biosphere when other losses such as forest
clearance and re are accounted for. The magnitude and also the mechanisms and spacial
distribution of the terrestrial carbon balance are still debated [Gurney et al., 2002; Rödenbeck
et al., 2003; Janssens et al., 2003]. It is estimated that the terrestrial biota has been a net
sink of carbon in the 1990s at a rate of about 1 (±0.8) PgC yr−1 [Joos et al., 1999; Keeling
and Garcia , 2002; Plattner et al., 2002; Bopp et al., 2002], probably mainly due to changes in
land use and management. CO2 fertilization and deposition of xed nitrogen to the soils may
also contribute to the current terrestrial sink [Caspersen et al., 2000; Körner , 2000]. However, the land biosphere is a highly variable reservoir. The terrestrial carbon cycle is strongly
linked to inter-annual climate variability, and may thus be inuenced by phenomena such as
the North Atlantic Oscillation (NAO) and El Niño/Southern Oscillation (ENSO) [Joos et al.,
1999; Rödenbeck et al., 2003].
Our understanding of the terrestrial carbon sink and its controlling mechanisms and vulnerability to changes in climate and land management are still very limited. For accurate
predictions of atmospheric CO2 loadings in future, expected to result from continued fossil
fuel combustion, an important question to be answered is what fraction of the current CO2
enters the oceans, and what fraction is taken up by the land biosphere. This is of importance
because the carbon storage in these two reservoirs is dierent. Increased biomass or soil organic matter appears to be susceptible to human intervention and climate change. In contrast,
if CO2 is taken up by the oceans, most of the carbon is not likely to reenter the atmosphere
soon, because of the slow mixing time of the oceans. Thus, on timescales of 50 to 100 years,
oceanic uptake may play a more signicant role in mitigating anthropogenic CO2 emissions.
12
1. INTRODUCTION
1.2 Atmospheric Oxygen as a Tracer for Carbon Cycle Processes
Direct measurements of carbon inventory changes in the ocean and the terrestrial biosphere
are dicult. The net uxes between the atmosphere, ocean and biosphere are only a few
percent of the gross exchange uxes and changes in the carbon inventory are small compared
to the total inventory of the vegetation and soils and of the ocean (Figure 1.2). Furthermore,
large inherent spatial and temporal inhomogeneities in the biosphere and the ocean complicate
estimates of the strength of carbon sinks and sources. Therefore, we have to rely primarily on
indirect methods to determine the partitioning of anthropogenic CO2 between the ocean and
the terrestrial biosphere.
One method is based on measurements of the ratios of two stable carbon isotopes, 13 C and
12 C in CO . During photosynthesis, terrestrial plants assimilate preferentially the lighter 12 C,
2
thereby enriching the 13 C of the CO2 left behind in the atmosphere. In contrast, exchanges
of CO2 between the atmosphere and the oceans leave the 13 C/12 C ratio nearly unchanged.
Thus, changes in the 13 C/12 C ratio of atmospheric CO2 give a measure of terrestrial carbon
storage [Keeling et al., 1989; Ciais et al., 1995; Battle et al., 2000]. Fossil fuel CO2 is depleted
in 13 C relative to 12 C, leading to a long-term decrease of the 13 C/12 C ratio of atmospheric
CO2 .
The second powerful approach to constrain the partitioning was developed by Keeling and
Shertz [1992] and is based on measurements of atmospheric oxygen (O2 ). O2 and CO2 are
exchanged in relatively xed stoichiometric ratios during the burning of fossil fuels as well as
during photosynthesis and respiration by plant, animals, and bacteria. By contrast, dissolution
of CO2 in the ocean has no eect on atmospheric O2 . Therefore, only uptake or release of CO2
by the biosphere will leave an imprint on atmospheric O2 . By knowing the fossil fuel emissions
and the exact values of the stoichiometric ratios, one can separate on timescales of a few years
the total CO2 uptake into land and ocean components, as shown graphically in Figure 1.3.
Strictly speaking, this method can only distinguish between net non-biological ocean uptake
and net biospheric uptake, which includes both the terrestrial and marine biosphere. However,
since biological O2 uptake in the ocean is not expected to have changed signicantly during
recent decades because of nutrient limitation in most parts of the ocean, this inferred biospheric
uptake is attributed to the land [Prentice et al., 2001]. Because of the temperature dependent
solubility, increases in ocean temperatures [Levitus et al., 2000] induce small outgassing uxes
of O2 that have to be taken into account. Additionally, impacts on atmospheric O2 caused
by changes in the ventilation of deeper, oxygen depleted waters occur on inter-annual and
possibly also on longer timescales [Keeling et al., 1993; Bender et al., 1996].
Measuring variations in atmospheric O2 concentration is challenging because of the high
O2 concentration in the atmosphere of 20.95 %. For example, a plant that removes by photosynthesis one CO2 molecule per million air molecules and releases an equivalent number of O2
molecules to the atmosphere will leave a 1/376 or 0.27 % signal in the background CO2 mixing ratio, but only a 1/209500 or 0.00048 % signal in the O2 concentration. This makes high
demands on the quality of O2 data to be useful in constraining the global carbon budget. The
required accuracy of O2 data exceeds the requirements for other concentration measurements,
and technical aspects like gas handling and measurement artifacts are particularly important
for O2 analysis.
Several independent analytical techniques for high precision measurements of atmospheric
1.3. OUTLINE OF CHAPTERS
O2 concentration,
difference from standard (ppm)
-20
13
1990
-25
-30
-35
fossil fuel burning
-40
-45
-50
2000
-55
outgassing
-60
-65
atmospheric increase
land
uptake
ocean uptake
-70
350
355
360
365
370
375
CO2 concentration (ppm)
380
385
Figure 1.3: Partitioning of fossil fuel CO2 uptake using O2 measurements. Solid circles are annual averages
of the observed O2 (vertical axis) and CO2 concentrations (horizontal axis) [Keeling et al., 1996; Manning ,
2001; Battle et al., 2000]. The arrow labelled fossil fuel burning indicates the change in atmospheric O2
and CO2 concentration that would have occurred if all CO2 emitted remained in the atmosphere. The arrow
labelled outgassing denotes O2 changes from oceanic outgassing, primarily due to changes in the marine
biogeochemical cycle. Carbon uptake by land and ocean is constrained by the known O2 :CO2 stoichiometric
ratios of the these processes, dening the slopes of the respective arrows (Figure adapted from Plattner et al.
[2002]).
O2 have been developed. The rst technique successful in achieving the required precision was
based on interferometry and measured changes in relative refractivity of air [Keeling , 1988;
Keeling et al., 1998]. Bender et al. [1994] have measured the O2 /N2 ratio with similar precision on an isotope ratio mass spectrometer. More recently, high precision techniques have also
been developed using paramagnetic [Manning et al., 1999; Manning , 2001], vacuum ultraviolet adsorption [Stephens , 1999], gas chromatographic [Tohjima , 2000], and electrochemical
methods [Stephens et al., 2001].
1.3 Outline of Chapters
This thesis describes rst results of high precision O2 measurements from three sites in Europe
and developments and improvements of measurement techniques for O2 and other species.
The primary focus was on establishing atmospheric O2 measurements at dierent sites in the
framework of the European project AEROCARB. The key objective of AEROCARB was to
demonstrate the feasibility of estimating and monitoring the net European carbon balance on
monthly to decadal timescales based on a synergy of atmospheric measurements, mesoscale
atmospheric transport models, high resolution surface emission data, and diagnostic models
of land ecosystems carbon exchange.
14
1. INTRODUCTION
We could use the infrastructure at the high-altitude site Jungfraujoch, Switzerland, as
well as at the mountain site Puy de Dôme, France, to take weekly or bi-weekly ask air
samples. Additionally, with a custom made sampling unit dedicated for aircraft sampling
we regularly obtained samples of vertical proles above Grin Forest in Scotland, UK. All
ask samples were analyzed primarily for O2 and CO2 concentrations in our laboratory using
mass spectrometry. Further useful constraints of carbon cycle processes were provided by
analysis of the stable carbon isotopes of CO2 . Some diagnostic information was obtained by
the concurrent analysis of dierent molecular and isotopic ratios including Ar/N2 and isotopes
of N2 and O2 .
A description of our mass spectrometric method for measuring O2 /N2 and CO2 is provided
in Chapter 2. Various tests as well as air standard measurements with the mass spectrometer
are presented and specic technical issues related to analytical or gas handling aspects are
discussed. Along those lines is the work presented in Chapter 3. Permeation of atmospheric
gases through elastomeric O-ring seals can have important eects on the integrity of atmospheric air samples collected in asks. The inuence of permeation is discussed for O2 /N2 and
related tracers. Results of sample storage tests for various ask and valve types and dierent
storage conditions are presented, and compared with theoretical calculations.
Chapter 4 presents the rst time series of atmospheric O2 /N2 measurements at Jungfraujoch and Puy de Dôme. The ask samples have also been analyzed for CO2 and δ 13 C. The
four-year record of Jungfraujoch and the three-year record of Puy de Dôme revealed distinct
seasonal cycles and superimposed long-term trends in the measured parameters. Regular
vertical aircraft sampling was performed in the lower troposphere above Grin Forest, Scotland. The seasonal cycles of O2 /N2 and CO2 and their vertical gradients are investigated in
Chapter 5.
Multiple species including CO2 , O2 , their stable isotopes and 222 Rn, were measured periodically in a continuous way during a one year period at Bern (Chapter 6). Diurnal as well as
seasonal patterns were observed. Using 222 Rn measurements the CO2 emission rate in the region of Bern was estimated. The use of multiple species allowed again to detect measurement
artifacts associated with fractionation processes at the air intake.
As a contribution to the European projects CarboEurope-IP and CARIBIC, continuous O2
analyzers based on paramagnetic and electrochemical techniques have been developed. The
analytical design and gas handling schemes are presented in Chapter 7.
Finally, Appendix A briey reviews dierent straight line regression techniques in use when
both variables are subject to error.
REFERENCES
15
References
Battle, M., M. L. Bender, P. P. Tans, J. W. C. White, J. T. Ellis, T. Conway, and R. J. Francey
(2000), Global Carbon Sinks and Their Variability Inferred from Atmospheric O2 and δ 13 C, Science,
287, 24672470.
Bender, M., T. Ellis, P. Tans, R. Francey, and D. Lowe (1996), Variability in the O2 /N2 ratio of
Southern Hemisphere air, 1991-1994: Implications for the carbon cycle, Global Biogeochem. Cycles,
10 (1), 921.
Bender, M. L., P. P. Tans, T. J. Ellis, J. Orchardo, and K. Habfast (1994), A high precision isotope
ratio mass spectrometry method for measuring the O2 /N2 ratio of air, Geochim. Cosmochim. Acta,
58 (21), 47514758.
Bopp, L., C. Le Quéré, M. Heimann, and A. C. Manning (2002), Climate-induced oceanic oxygen
uxes: Implications for the contemporary carbon budget, Global Biogeochem. Cycles, 16 (2), doi:
10.1029/2001GB001445.
Caspersen, J. P., S. W. Pacala, J. C. Jenkins, G. C. Hurtt, P. R. Moorcroft, and R. A. Birdsey (2000),
Contributions of Land-Use History to Carbon Accumulation in U.S. Forests, Science, 290, 11481151.
Ciais, P., P. P. Tans, M. Trolier, J. W. C. White, and R. J. Francey (1995), A Large Northern
Hemispheric Terrestrial CO2 Sink Indicated by the 13 C/12 C Ratio of Atmospheric CO2 , Science, 269,
10981102.
Etheridge, D. M., L. P. Steele, R. L. Langenfelds, R. J. Francey, J.-M. Barnola, and V. I. Morgan
(1996), Natural and anthropogenic changes in atmospheric CO2 over the last 1000 years from air in
Antarctic ice and rn, J. Geophys. Res., 101, 41154128.
GLOBALVIEW-CO2 (2004), Cooperative Atmospheric Data Integration Project - Carbon Dioxide,
CD-ROM, NOAA CMDL, Boulder, Colorado [Also available on Internet via anonymous FTP to
ftp.cmdl.noaa.gov, Path: ccg/co2/GLOBALVIEW].
Gurney, K. R., R. M. Law, S. A. Denning, P. J. Rayner, D. Baker, P. Bousquet, L. Bruhwiler, Y.-H.
Chen, P. Ciais, S. Fan, I. Y. Fung, M. Gloor, M. Heimann, K. Higuchi, J. John, T. Maki, S. Maksyutov,
K. Masarie, P. Peylin, M. Prather, B. C. Pak, J. Randerson, J. Sarmiento, S. Taguchi, T. Takahashi,
and C.-W. Yuen (2002), Towards robust regional estimates of CO2 sources and sinks using atmospheric
transport models, Nature, 415, 626630.
Houghton, R. A. (2003), Revised estimates of the annual net ux of carbon to the atmosphere from
changes in land use and land management 1850 − 2000, Tellus, 55B, 378390.
Indermühle, A., T. F. Stocker, F. Joos, H. Fischer, H. J. Smith, M. Wahlen, B. Deck, D. Mastroianni,
J. Tschumi, T. Blunier, R. Meyer, and B. Stauer (1999), Holocene carbon-cycle dynamics based on
CO2 trapped in ice at Taylor Dome, Antarctica, Nature, 398, 121126.
IPCC (2001), Climate Change 2001: The Scientic Basis. Contribution of Working Group I to the
Third Assessment Report of the Intergovernmental Panel on Climate Change, edited by J. T. Houghton,
Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and C. A. Johnson,
Cambridge University Press, Cambridge, UK.
Janssens, I. A., A. Freibauer, P. Ciais, P. Smith, G.-J. Nabuurs, G. Folberth, B. Schlamadinger,
R. W. A. Hutjes, R. Ceulemans, E.-D. Schulze, R. Valentini, and A. J. Dolman (2003), Europe's
Terrestrial Biosphere Absorbs 7 to 12 % of European Anthropogenic CO2 Emissions, Science, 300,
15381542.
16
1. INTRODUCTION
Joos, F., R. Meyer, M. Bruno, and M. Leuenberger (1999), The variability in the carbon sinks as
reconstructed for the last 1000 years, Geophys. Res. Lett., 26 (10), 14371440.
Keeling, C. D., and T. Whorf (2004), Atmospheric CO2 records from sites in the SIO air sampling
network, in Trends: A Compendium of Data on Global Change, Carbon Dioxide Information Analysis
Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, Tenn., U.S.A.
Keeling, C. D., R. B. Bacastow, A. F. Carter, S. C. Piper, T. P. Whorf, M. Heimann, W. G. Mook,
and H. Roelozen (1989), A three-dimensional model of atmospheric CO2 transport based on observed
winds: 1. Analysis of observational data, in Aspects of Climate Variability in the Pacic and the
Western Americas, Geophys. Monogr. Ser., vol. 55, edited by D. H. Peterson, pp. 165236, AGU,
Washington D.C.
Keeling, R., and S. Shertz (1992), Seasonal and interannual variations in atmospheric oxygen and
implications for the global carbon cycle, Nature, 358, 723727.
Keeling, R., R. Najjar, M. Bender, and P. Tans (1993), What atmospheric oxygen measurements can
tell us about the global carbon cycle, Global Biogeochem. Cycles, 7 (1), 3767.
Keeling, R., S. Piper, and M. Heimann (1996), Global and hemispheric CO2 sinks deduced from
changes in atmospheric O2 concentration, Nature, 381, 218221.
Keeling, R., A. Manning, E. McEvoy, and S. Shertz (1998), Methods for measuring changes in atmospheric O2 concentration and their application in Southern Hemisphere air, J. Geophys. Res., 103 (D3),
33813397.
Keeling, R. F. (1988), Development of an interferometric oxygen analyzer for precise measurement of
the atmospheric O2 mole fraction, Ph.D. thesis, Harvard University, Cambridge, Mass., U.S.A.
Keeling, R. F., and H. E. Garcia (2002), The change in oceanic O2 inventory associated with recent
global warming, Proc. Nat. Acad. Sci. USA, 99 (12), 78487853.
Körner, C. (2000), Biosphere Responses to CO2 Enrichment, Ecological Applications, 10 (6), 1590
1619.
Levitus, S., J. I. Antonov, T. P. Boyer, and C. Stephens (2000), Warming of the World Ocean, Science,
287, 22252229.
Manning, A. C. (2001), Temporal variability of atmospheric oxygen from both continuous measurements and a ask sampling network: Tools for studying the global carbon cycle, Ph.D. thesis, University of California, San Diego, California, U.S.A.
Manning, A. C., R. F. Keeling, and J. P. Severinghaus (1999), Precise atmospheric oxygen measurements with a paramagnetic oxygen analyzer, Global Biogeochem. Cycles, 13 (4), 11071115.
Marland, G., T. Boden, and R. J. Andres (2003), Global, Regional, and National Fossil Fuel CO2
Emissions, in Trends: A Compendium of Data on Global Change, Carbon Dioxide Information Analysis
Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, Tenn., U.S.A.
Monnin, E., E. J. Steig, U. Siegenthaler, K. Kawamura, J. Schwander, B. Stauer, T. F. Stocker,
D. L. Morse, J.-M. Barnola, B. Bellier, D. Raynaud, and H. Fischer (2004), Evidence for substantial
accumulation rate variability in Antarctica during the Holocene, through synchronization of CO2 in
the Taylor Dome, Dome C and DML ice cores, Earth and Planetary Science Letters, 224, 4554.
Neftel, A., E. Moor, H. Oeschger, and B. Stauer (1985), Evidence from polar ice cores for the increase
in atmospheric CO2 in the past two centuries, Nature, 315, 4547.
REFERENCES
17
Petit, J. R., J. Jouzel, D. Raynaud, N. I. Barkov, J.-M. Barnola, I. Basile, M. Bender, J. Chappellaz,
M. Davis, G. Delaygue, M. Delmotte, V. M. Kotlyakov, M. Legrand, V. Y. Lipenkov, C. Lorius,
L. Pépin, C. Ritz, E. Saltzman, and M. Stievenard (1999), Climate and atmospheric history of the
past 420,000 years from the Vostok ice core, Antarctica, Nature, 399, 429436.
Plattner, G.-K., F. Joos, and T. F. Stocker (2002), Revision of the global carbon budget due to
changing air-sea oxygen uxes, Global Biogeochem. Cycles, 16 (4), doi:10.1029/2001GB001746.
Prentice, I. C., G. D. Farquhar, M. J. R. Fasham, M. L. Goulden, M. Heimann, V. J. Jaramillo,
H. S. Kheshgi, C. Le Quéré, R. J. Scholes, and D. W. R. Wallace (2001), The Carbon Cycle and
Atmospheric Carbon Dioxide, in Climate Change 2001: The Scientic Basis. Contribution of Working
Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, edited
by J. T. Houghton, Y. Ding, D. J. Griggs, M. Noguer, P. J. van der Linden, X. Dai, K. Maskell, and
C. A. Johnson, pp. 183237, Cambridge University Press, Cambridge, UK.
Rödenbeck, C., S. Houweling, M. Gloor, and M. Heimann (2003), CO2 ux history 1982−2001 inferred
from atmospheric data using a global inversion of atmospheric transport, Atmos. Chem. Phys., 3,
19191964.
Sarmiento, J. L., and N. Gruber (2002), Sinks for anthropogenic carbon, Physics Today, 55 (8), 3036.
Stephens, B., P. Bakwin, P. Tans, and R. Teclaw (2001), Measurements of atmospheric O2 variations
at the WLEF tall-tower site, in Sixth International Carbon Dioxide Conference, Extended Abstracts,
vol. I, pp. 7880, Tohoku Univ., Sendai, Japan.
Stephens, B. B. (1999), Field-based Atmospheric Oxygen Measurements and the Ocean Carbon Cycle,
Ph.D. thesis, University of California, San Diego, California, U.S.A.
Stott, P. A., D. A. Stone, and M. R. Allen (2004), Human contribution to the European heatwave of
2003, Nature, 432, 610614.
Tett, S. F. B., G. S. Jones, P. A. Stott, D. C. Hill, M. R. Mitchell, J. F. B. Allen, W. J. Ingram,
T. C. Johns, C. E. Johnson, A. Jones, D. L. Roberts, D. M. H. Sexton, and M. J. Woodage (2002),
Estimation of natural and anthropogenic contributions to twentieth century temperature change, J.
Geophys. Res., 107 (D16), 4306, doi:10.1029/2000JD000028.
Tohjima, Y. (2000), Method for measuring changes in the atmospheric O2 /N2 ratio by a gas chromatograph equipped with thermal conductivity detector, J. Geophys. Res., 105 (D11), 14,57514,584.
Chapter 2
Measurement Methods and Tests
2.1 Introduction
The high demands on data quality that are made for using atmospheric O2 concentration measurements in global carbon cycle studies involve that gas handling and measurement related
issues are of great importance. During the last decade a lot was learnt about such technical
and analytical problems specic to high precision measurements of atmospheric O2 [Langenfelds , 2002; Bender et al., 1994; Keeling et al., 1998, 2004]. In the course of this thesis the
ability to perform high precision O2 measurements in our laboratory was further developed
and improved. For ask sample analysis we use an isotope ratio mass spectrometer, basically
following the method described by Bender et al. [1994].
This chapter provides a description of our mass spectrometric method for measuring O2 and
CO2 and discusses specic technical issues related to existing or potential artifacts and corrections that were applied to the data.
2.2 Equipment
2.2.1 Mass Spectrometer System
All ask air samples presented here have been analyzed by an isotope ratio mass spectrometer
(IRMS). The sample gas is ionized by electron impact ionization. Resulting ions are focused
and accelerated under high voltage. A magnetic eld deects the ions by an angle depending
on their mass to charge ratio m/z . The ion beams of interest are then detected by several
Faraday cups.
We use a Finnigan MAT DELTAplus XL instrument (Thermo Electron, Bremen, Germany).
It has been upgraded in March 2002 to a DELTAplus XP unit. The new electronics and software
allowed to fully use the extended ion optics in the m/z range of 28 to 44. The molecular
ratios of the most abundant air constituents O2 /N2 ((m/z 32)/(m/z 28)), Ar/N2 (40/28) and
CO2 /N2 (44/28) as well as the isotopic ratios 29 N2 /28 N2 , 33 O2 /32 O2 , 34 O2 /32 O2 and 36 Ar/40 Ar
can be monitored simultaneously.
All IRMS measurements are reported as relative deviations in the ratio from a reference gas,
such that for two components A and B:
µ
¶ (
1000 %
(A/B)sample − (A/B)reference
δ(A/B) =
·
(2.1)
(A/B)reference
106
per meg
20
2. MEASUREMENT METHODS AND TESTS
where (A/B)sample is the ratio of the sample gas and (A/B)reference is the ratio of an arbitrary
reference gas. In the δ notation isotopic ratios are often expressed in units of per mil (%),
whereas elemental ratios commonly are reported in per meg units (1 per meg = 0.001%).
Isotopic ratios are usually abbreviated, i. e. δ 29 N2 , δ 34 O2 (or δ 18 O of O2 ), δ 33 O2 (or δ 17 O of
O2 ) and δ 36 Ar correspond to δ(29 N2 /28 N2 ), δ(34 O2 /32 O2 ), δ(33 O2 /32 O2 ) and δ(36 Ar/40 Ar),
respectively.
The reason why oxygen concentrations are commonly reported as δO2 /N2 values and not
for example as mole fractions is twofold. First, the O2 mole fraction is sensitive to variations
in trace gases. For example, an increase in the CO2 mole fraction of 1 µmole mole−1 decreases
the O2 mole fraction by 0.21 µmole mole−1 without any O2 being added to or removed from
the air sample. Normalization to N2 eliminates such diluting eects associated with changes
in humidity, CO2 concentration or other trace gases. Secondly, it is presently not possible to
determine the absolute O2 mole fraction to the required precision (about 1 in 106 ), whereas
the relative deviations from a reference gas can be measured at this level. Furthermore, the
δO2 /N2 value is the directly observed quantity of the mass spectrometric method.
The atmospheric N2 mole fraction is only aected by a small, thermally driven, seasonal
air-sea ux [Keeling et al., 1993, 2004]. Depending on the application this N2 variation is often
neglected and the O2 /N2 ratio can then directly be used to investigate the biogeochemical
inuences on atmospheric O2 . Because the mole fraction of O2 in dry atmospheric air is
about 20.95% [Machta and Hughes , 1970], it follows that for small changes in mole fraction,
a change of 1 part per million (ppm) O2 is equivalent to 1/0.2095 = 4.8 per meg in δO2 /N2 .
Since an international scale for reporting oxygen concentrations is missing so far, all δO2 /N2
results are given on our local PIUB scale, as dened in Section 2.5.
To achieve the precision for O2 /N2 measurements required in global carbon cycle studies
of about 1 per meg [World Meteorological Organisation , 2005], the conventional dual inlet system of the IRMS with metal bellows was modied. We have developed a new inlet system,
which is based on an open split design as known from gas chromatography/mass spectrometry
(GC/MS) systems. Sample and reference gases are admitted to the mass spectrometer via
fused silica capillaries. A glass open split housed in a temperature and pressure controlled
box serves as a change-over device for the sample and reference gas. A detailed description of
our gas inlet system is given by Leuenberger et al. [2000a] and Sturm [2001]. Figure 2.1 shows
a schematic of the inlet system for O2 /N2 analysis with the multi-port valves for the ask
samples and the reference gas cylinders. All connections from the asks and the cylinders
are made up of fused silica capillaries to reduce surface adsorption eects. The temperature
in the change-over box is controlled to 30.00 ± 0.05 ◦ C and the pressure is about 200 hPa,
which allows to analyze ask samples with a sample pressure below atmospheric pressure. An
exhaust capillary which connects the change-over box with the vacuum system of the IRMS
compensates the inow of working and sample gas in order to keep the pressure constant in
the change-over box. The ask samples can be attached to the inlet system in horizontal position in a closed rack, which is placed on top of the mass spectrometer. The rack is insulated
by foamed material to prevent inuences from short-term temperature variations and small
fans mounted inside ensure a homogeneous temperature. Tests have revealed that even small
temperature gradients of a few tenths of a degree result in measurable dierences between the
individual ports. The rack can hold up to 16 asks at the same time.
2.2. EQUIPMENT
21
working
standard
DELTAplusXP
ion source
Y-shaped
open split
change-over box
temperature and pressure
controlled (30°C, 200hPa)
4-port switching valve
10-port valve
16-port valve
flask samples
reference gas
cylinders
pump
insulated box
Figure 2.1: Schematic of the inlet system for O2 /N2 analysis on the DELTAplus XP instrument.
2.2.2 Pressure Regulators and Reference Cylinders
A crucial component for high precision O2 /N2 measurements are the reference gases and their
stability. Our working and reference gases are contained in 50 L metal high pressure gas cylinders. They are lled to about 200 bar with dry atmospheric air and purchased from Carbagas
(Gümligen, Switzerland). From a set of cylinders lled at dierent days we select those which
are suitable as reference gases according to their CO2 and O2 /N2 concentration. Initial tests
showed that only high quality regulators intended for high purity applications are suitable for
O2 /N2 analysis. For example, the HBS 315 regulators from Air Liquide (Carbagas, Gümligen,
Switzerland), that we often use for other applications, not only showed large fractionation
due to temperature and pressure variations in the laboratory but also revealed problems with
the humidity of the sample air. This eect occurs mainly because of our small ow rates
used for O2 /N2 measurements. The water vapor was either sticking persistently on the walls
or permeating in through the polymeric seals of the regulator. Therefore, the high pressure
cylinders are now tted with dedicated single stage, chrome-plated brass DLRS Nr.5 regulators with PCTFE seats manufactured by Air Liquide, or high purity two-stage stainless
steel regulators (64-3400) with PCTFE seats manufactured by Tescom Corp. (Elk River, MN,
USA). A suit of 8 primary standards contained in 48 L aluminium cylinders and purchased
from the Carbon Cycle Gases Group of the Climate Monitoring and Diagnostics Laboratory
/ National Oceanic and Atmospheric Administration (CCGG/CMDL/NOAA), Boulder, CO,
USA, are tted with ultra-high purity single-stage Tescom 74-2400 regulators. The 74-2400
regulators show a change in outlet pressure per change in inlet pressure of −7 mbar/bar, the
22
2. MEASUREMENT METHODS AND TESTS
δO2/N2 (per meg)
-2280
-2290
-2300
-2310
-2320
0
1
2
3
Exhaust flux (mL min-1)
4
5
Figure 2.2: Test of the δO2 /N2 ratio of air drawn from a 10 L glass ask, where additional exhaust capillaries
with dierent ow rates were mounted.
two-stage Tescom 64-3400 pressure reducing regulators have a decaying inlet characteristic of
0.08 mbar/bar for inlet pressures > 20 bar and −7 mbar/bar for inlet pressures <20 bar.
The reference cylinders for O2 /N2 measurements are placed horizontally in order to minimize
potential concentration gradients inside the tanks due to thermal diusion or gravitational
settling of air [Keeling et al., 1998]. The pressure regulators are insulated with foamed material.
For analysis we use a gas ux entering the inlet system of about 0.25 mL min−1 . Because
the high pressure regulators are not well adapted for such low ow rates, we additionally
employ for the working and reference standards an exhaust capillary, where gas with a ow
rate of about 3 mL min−1 permanently escapes. This still guarantees a lifetime of about 5
years for the high pressure cylinders, but improves the stability of the standard gases drawn
from the cylinders. The exhaust and sample capillaries are mounted with a 2-hole ferrule
on the pressure regulator. To ensure that dividing the ow into an exhaust and a sample
line does not introduce fractionation, tests with air from a 10 L glass ask were conducted.
The δO2 /N2 ratio was monitored for exhaust ow rates of 04 mL min−1 . A fractionation in
δO2 /N2 relating to the exhaust ow is expected to be constant over time, and as shown in
Figure 2.2, it is, if present at all, smaller than ±5 per meg.
2.3 Mass Spectrometer Performance and Artifacts
2.3.1
Instrument Precision
The measurement precision of the DELTAplus XP mass spectrometer is listed in Table 2.1. It
is dened here as the standard deviation of eight consecutive standard/sample cycles. The
idle time, i.e. the time period over which the signals are discarded to allow for stabilization
after each standard/sample switch, and the integration time are generally set to 30 s and 8 s,
respectively. These times proved to be a viable compromise between a high measurement
precision for all elemental and isotopic ratios (including δCO2 /N2 ) and a short analysis time.
The measured standard deviations (middle column in Table 2.1) are the mean of 336 reference
gas measurements between August 2003 and September 2004 with a m/z 28 intensity of about
4000 mV. The statistical limitation of the precision caused by the Poisson statistics of the
number of ions produced and detected in the mass spectrometer is calculated for an integration
2.3. MASS SPECTROMETER PERFORMANCE AND ARTIFACTS
σmeasured (per meg)
δO2 /N2
δAr/N2
δAr/O2
δCO2 /N2
δ 29 N2
δ 34 O2
δ 33 O2
δ 36 Ar
4.5 ± 1.9
12.2 ± 5.7
11.9 ± 5.0
60.6 ± 58.6
16.9 ± 5.4
41.3 ± 14.7
113.7 ± 37.4
193.2 ± 64.8
23
σstatistical (per meg)
4
13
13
60
20
54
124
217
Table 2.1: Mass spectrometer precision based on the standard deviation of repeated standard/sample cycles
with idle and integration times of 30 s and 8 s, respectively, and a m/z 28 signal intensity of about 4000 mV.
time of 8 s and a m/z 28 intensity of 4000 mV and listed in the last column of Table 2.1. The
comparison of the measured with the expected statistical precision shows that the instrument
precision can be explained by statistical limitations of ion detection.
2.3.2 Isobaric Interferences
Isobaric interferences with m/z 28 and 32 can potentially introduce errors in the measured
value of δO2 /N2 . This mainly originates from fragments of CO2 , which interfere at m/z 28
(12 C16 O) and 32 (16 O16 O) with the primary N2 and O2 signals. The production of NO in the
ion source [Leuenberger et al., 2000b] leads also to an interference at m/z 32 from 14 N18 O.
However, this interference is very small and of identical magnitude on sample and reference
side of the mass spectrometer. Measurements from Langenfelds [2002] on a DELTAplus XL and
from Bender et al. [1994] on a MAT251 mass spectrometer gave a sensitivity of δO2 /N2 to
isobaric interference from fragmentation of CO2 in the ion source of −0.1 per meg/ppm CO2 .
This value is used to correct all δO2 /N2 data relative to a CO2 mixing ration of 380 ppm.
For atmospheric samples from remote areas, where the CO2 variations are typically within
±10 ppm, this correction is in the order of 1 per meg. For sampling sites with larger CO2
variations (e.g. see Chapter 6) the correction may become signicant. For example, the slope
of O2 :CO2 correlations increases (is less negative) by about 0.02 mol O2 /mol CO2 with this
correction. Interference eects on other elemental and on isotopic ratios are expected to be
negligible for atmospheric air samples.
2.3.3 Pressure Imbalance Eect
The ion source pressure and ion beam intensity of the sample and standard gas streams
entering the ion source have to be balanced as exactly as possible, because the measured
elemental ratios are dependent on the ion source pressure. Thus, a pressure imbalance leads
to a systematic bias in the measured ratios. Our inlet system, which is based on an open split
design, circumvents the need of adjusting the ion beam intensity on sample and standard side,
because the ow rate of gases into the mass spectrometer is mainly determined by the pressure
in the change-over box. However, second order variations are caused by dierent sample
pressures. For example, a higher pressure of a ask sample results in a larger ux into the
change-over box, which in turn causes a small increase of the dynamic pressure in the Y-split.
24
2. MEASUREMENT METHODS AND TESTS
δO2/N2 (per meg)
30
20
10
0
-10
δAr/N2 (per meg)
30
20
10
0
-10
δCO2 /N2 (per meg)
400
200
0
-200
-400
-0.8
-0.4
0
0.4
Imbalance (VSA -VST) / VST (%)
0.8
Figure 2.3: Sensitivity of δO2 /N2 (top), δAr/N2 (middle), and δCO2 /N2 (bottom) to pressure imbalances
between the sample and standard side of the inlet system as measured in September 2003. Note that there is
an asymmetry between sample and standard side (δ -value 6= 0) in absence of any pressure imbalance.
The eects of pressure imbalances between sample and standard side on the measured isotopic
and elemental ratios were quantied with a dedicated experiment conducted in September
2003. Air was drawn from a high pressure cylinder with two fused silica capillaries, one
attached to the sample side and the other to the standard side of the inlet system. On the
sample side the ow rate of gas was varied using dierent lengths of capillary, resulting in small
pressure imbalances. Figure 2.3 shows the anomalies of δO2 /N2 , δAr/N2 and δCO2 /N2 as a
function of the relative dierence in sample and standard voltage. The measured sensitivities
for the elemental ratios δO2 /N2 , δAr/N2 , and δCO2 /N2 are listed in Table 2.2. No signicant
eect could be observed for the isotopic ratios δ 29 N2 , δ 33 O2 , δ 34 O2 , and δ 36 Ar. All measured
elemental ratios are corrected for this eect of pressure imbalance. Generally, variations in the
sample pressure of atmospheric ask samples are in the order of ±0.2 %, leading to corrections
of < 5 per meg for δO2 /N2 and δAr/N2 and a correction of < 150 per meg (equivalent to about
< 0.1 ppm) for δCO2 /N2 . These corrections may be dependent on the mass spectrometer
settings, e.g. the ion source settings, and may thus be varying with time. Although any errors
due to this eect are likely to be small for atmospheric air samples, the magnitude of the
pressure imbalance eect should be regularly monitored, especially when mass spectrometer
2.3. MASS SPECTROMETER PERFORMANCE AND ARTIFACTS
δO2 /N2
δAr/N2
δAr/O2
δCO2 /N2
Pressure imbalance eect
(per meg/%)
Sample/standard asymmetry
(per meg)
19 ±2
−14 ±3
−33 ±3
−679 ±16
7 ±2
6 ±6
0 ±7
−21 ±29
25
Table 2.2: Measured sensitivity of the elemental ratios to pressure imbalances between sample and standard
side (left) and absolute oset between sample and standard side (right).
settings are changed.
2.3.4 Sample/Standard Asymmetry
Another feature apparent in Figure 2.3 and also arising from other tests is an asymmetry
between sample and standard side in absence of any pressure imbalance. If the same gas is
admitted on the sample as well as on the standard side, the δ -value of the elemental ratios is
not exactly zero (Table 2.2). A possible explanation for this oset is fractionation in the Y-split
due to small geometrical asymmetries of the Y-split or the change-over capillaries. Constant
corrections of 7 and 6 per meg are applied to the δO2 /N2 and δAr/N2 data, respectively,
whereas no correction was applied to δAr/O2 and δCO2 /N2 . Again, for isotopic ratios no
sample/standard asymmetry could be observed.
2.3.5 Repeated Measurement Drift
A potentially large source of noise in ask measurements is the process of transferring air
from the ask to the inlet system. The standard procedure for preparing ask samples before
analysis involves dierent steps, where mass-dependent fractionation and/or surface eects
can occur. After the asks have been mounted on an inlet port in the rack, the lines to the
asks are evacuated to a pressure of ≤ 2 × 10−3 mbar (as measured immediately in front of
the rotary vane pump). Subsequently the sample gas is expanded into the connecting part
of the ask valve, the adapter tting (1/2 inch Ultra-Torr) and the connecting capillary up
to the multi-port valve, that selects between the dierent inlet ports. This expansion volume
is about 10 mL. For the analysis of the 0.5 L asks a glass piston is usually inserted into the
connecting part of the valve, which reduces the expansion volume to about 5 mL.
Repeated analysis of the same ask samples showed a constant drift after each measurement. Table 2.3 lists the weighted mean of the drift rate of nine 0.5 L asks measured between
8 and 42 times each.
This drift may be attributed to selective surface adsorption and/or diusive fractionation
during the sample gas expansion. The observed drifts are negligible for isotopic ratios. For
δCO2 /N2 the drift is masked by large and varying background eects (see Section 2.4). Since
most ask samples are measured only twice, the drift for δO2 /N2 and δAr/N2 is comparable
to measurement precision. The δO2 /N2 and δAr/N2 data from 0.5 L ask samples are corrected for the repeated measurement drift, resulting in a correction of the mean of duplicate
measurements of less than 1.5 per meg. Tests with 1 L ask samples showed that this eect is
smaller for these asks than for 0.5 L asks. This is expected due to the larger ask volume
26
2. MEASUREMENT METHODS AND TESTS
Drift rate
(per meg measurement−1 )
δO2 /N2
δAr/N2
δAr/O2
δ 29 N2
δ 34 O2
δ 33 O2
δ 36 Ar
−2.9 ± 0.2
2.1 ± 0.2
6.4 ± 0.3
0.7 ± 0.1
1.9 ± 0.3
1.1 ± 0.6
−0.4 ± 0.1
Table 2.3: Repeated measurement drift for 0.5 L ask samples. The values represent the weighted mean from
9 asks measured between 8 and 42 times each.
leading to a stronger dilution of any modifying eect. No drift correction was applied to 1 L
ask sample measurements.
2.4
CO2 Mixing Ratio Measurements by Mass Spectrometry
CO2 is usually measured by gas chromatography (GC) or non-dispersive infrared absorption
(NDIR) techniques. These techniques are compared to the mass spectrometric method less
costly in instrumentation and operation and have a better measurement precision. However,
since we get with the mass spectrometer the δCO2 /N2 simultaneously with the other elemental
and isotopic ratios, we have developed procedures to use this information for inferring the CO2
concentration of the air samples.
2.4.1
N2 O Background
Precise measurements of CO2 mixing ratio by mass spectrometry are complicated by dierent
factors. The major diculty arises from the production of N2 O from exited N2 and O2
molecules or fragments of those in the ion source [Leuenberger et al., 2000b; Sturm , 2001].
N2 O interferes with CO2 , since both have a mass to charge ratio of 44. The N2 O production
results in a m/z 44 signal, which is of the same order of magnitude as the m/z 44 signal
originating from the CO2 of an air sample. To quantify this large m/z 44 background we
regularly measured the δ(m/z44)/(m/z28) ratio of CO2 -free air (Figure 2.4). The m/z 44
background shows large variations with time. It was highest after vacuum breakdowns of the
ion source and then gradually decreased over several months. It is also dependent on the ion
beam acceleration potential and focussing settings. This indicates that the N2 O production is
dominated by surface exchange eects in the ion source. However, on timescales of days it was
found to be constant. Hence, it was determined at the end of each measurement day using
CO2 -free air. Because CO2 concentrations are commonly reported as mixing ratios and not in
δ -notation, the measured δCO2 /N2 values have to be converted into mixing ratios. Firstly, this
requires that the CO2 mixing ratio of the working standard is known. Furthermore another
reference gas with known CO2 mixing ratio has to be analyzed to correct for the inuence of
the N2 O background. Under the two assumption that the N2 O background is the same for
the sample, the working standard and the reference gas and that changes in CO2 mixing ratio
27
-300
832
-400
535
-500
356
-600
238
-700
153
-800
89
-900
40
1-Sep-01 1-Mar-02
1-Sep-02 1-Mar-03
1-Sep-03 1-Mar-04
m/z 44 background (ppm)
δ (m/z 44)/(m/z 28) (‰)
2.4. CO2 MEASUREMENTS BY MASS SPECTROMETRY
1-Sep-04
Figure 2.4: Measured δ(m/z44)/(m/z28) (in %) of CO2 -free air. If no background component was present,
the δ(m/z44)/(m/z28) would yield for CO2 -free air −1000 %. The right axis shows the equivalent concentration in ppm of this background component. Times of vacuum-breakdown are indicated by dashed lines.
are proportional to changes in CO2 /N2 ratio, the δCO2 /N2 value can be converted to a CO2
mixing ratio according to
[CO2 ]sa =
(δCO2 /N2 )sa
· ([CO2 ]ref − [CO2 ]w.st. ) + [CO2 ]w.st.
(δCO2 /N2 )ref
(2.2)
where [CO2 ] denotes the CO2 mixing ratio in ppm.
2.4.2 Cross Contamination
Compared with other air components, CO2 preferentially adsorbs on surfaces, leading to
signicant background signals even after long evacuation times and the possibility of biases
due to cross contamination, i.e. the contamination of the sample gas with traces of working
standard gas (and vice versa) in the ion source. Cross contamination has been identied
as a prominent cause of inter-laboratory dierences for stable carbon and hydrogen isotope
measurements [Meijer et al., 2000; Verkouteren et al., 2003], but has not been investigated up
to now for elemental ratios measured by IRMS. Obviously, a mixing of sample and standard
gases can be attenuated by longer idle times. For some sticky gases like CO2 , however,
an idle time of 30 s or more may not be long enough to get completely rid of the rest of
the previous gas in the ion source. To asses the importance of cross contamination for CO2
measurements, the formalism of Meijer et al. [2000] was used, where δtrue and δm are the true
and measured delta values, respectively, and η is the cross contamination coecient
δtrue =
δm
.
1 − 2η − ηδm
(2.3)
28
2. MEASUREMENT METHODS AND TESTS
CO2 cross contamination η
0.008
0.006
0.004
0.002
0
1-Nov-02
1-Feb-03 1-May-03
1-Aug-03
1-Nov-03
Figure 2.5: Measurements of η for δCO2 /N2 and an idle time of 30 s
η is the fraction of the sample CO2 that is admixed to the working standard gas in the ion
source during the measurement (and vice versa). It can be determined using measurements
of the CO2 /N2 raw ratio. For this purpose a sample gas with a highly enriched or depleted
δCO2 /N2 value in relation to the working standard is measured and compared with measurements of the CO2 /N2 ratio if the working gas itself is used as a sample. Figure 2.5 shows cross
contamination measurements for δCO2 /N2 deduced from CO2 -free air and with an idle time
of 30 s. η is in the range of 0.0010.006. The scatter of results mainly reects the imprecision
of the determination of η due to the instability of the CO2 /N2 raw ratios, although mixing
eects are also likely to vary depending on ion source conditions [Verkouteren et al., 2003].
Figure 2.6 illustrates the potential inuence of cross contamination on the measured CO2 mixing ratio for dierent values of η . Here, the working standard and reference gas for background
correction are assumed to have CO2 mixing ratios and δ -values of [CO2 ]w.st. = 356.7 ppm,
[CO2 ]ref = 0 ppm, and (δCO2 /N2 )ref = −600 %. This calculation indicates that the bias
caused by cross contamination is somewhere between 0.0 and 0.4 ppm, which is comparable to
our measurement precision. Nevertheless, it shows that the CO2 mixing ratio of the working
standard should preferentially be chosen in the upper range of the CO2 variability of atmospheric air samples to minimize the inuence of mixing eects. So far we did not correct our
CO2 data for cross contamination, but a regular measurement of and correction for it could
further improve the accuracy of the CO2 measurements by mass spectrometry.
Cross contamination also potentially inuences δO2 /N2 . However, from dierent tests
performed to estimate η of O2 /N2 no clear conclusions could be drawn so far. Short-term
variations in the O2 /N2 raw ratio, which are always present at timescales of a few minutes (and
which are overcome by measuring sample/standard cycles and calculating δ -values), impair
the determination of η . In any case, cross contamination eects for O2 /N2 are expected to
be even smaller than for CO2 . A value of η = 0.005, for example, would lead to a O2 /N2
scale contraction of about 1 %. A possible source of error due to cross contamination can
therefore not be ruled out. A recently initiated international intercomparison program for
O2 /N2 measurements will help to address such questions.
CO2 cross contamination effect (ppm)
2.5. AIR STANDARD MEASUREMENTS
29
η = 0.0060
0.4
η = 0.0035
0.3
η = 0.0010
0.2
0.1
0
-0.1
0
-0.2
100
200
300
Sample CO2 concentration (ppm)
400
500
-0.3
-0.4
Figure 2.6: Calculated cross contamination inuence on CO2 measurements for dierent values of η and a
CO2 mixing ratio of the working standard of 356.7 ppm.
2.5 Air Standard Measurements
2.5.1 Working and Reference Standards
Until 11 September 2003 all air samples have been measured against air from the cylinder
LK560944. We then replaced our working standard LK560944 with the cylinder LK560962,
because the O2 /N2 and Ar/N2 ratios of LK560962 were much closer to atmospheric values than
the air contained in LK560944. LK560944 is now further used as reference standard. All data
presented here are referenced against LK560962 and measurements before 11 September 2003
were converted to the LK560962 scale using the values listed in Table 2.4. Reference standards
were measured on each day that O2 /N2 analyses were conducted. Air standards were analyzed
immediately before and after a sample measurement sequence consisting of up to 16 asks.
The δO2 /N2 , δAr/N2 , δ 29 N2 , δ 34 O2 , δ 33 O2 and δ 36 Ar records of three reference standards
(LK560913, LK560944 and LK557104) relative to LK560962 are plotted in Figures 2.7 and 2.8.
Several features can be seen in the δO2 /N2 plot. First, LK560913 shows the largest variability,
especially during the rst year of analysis. Since the other two standards (LK560944 and
LK557104) are more stable during this period, the variability must originate from LK560913
rather than from the working standard. The reason for this behavior is unclear. The gas
handling and analysis procedures are the same for all three reference gases. A second important
feature are long-term variation over timescales of months to years. Since the drift is not
unidirectional, it excludes a steady enrichment or depletion of O2 /N2 in cylinder air due to
pressure-dependent fractionation as an explanation. All high pressure cylinders, including
δO2 /N2
δAr/N2
δAr/O2
δCO2 /N2
δ 29 N2
δ 34 O2
δ 33 O2
δ 36 Ar
−1171 ± 4
−3465 ± 7
−2298 ± 11
15861 ± 728
−16 ± 8
8 ± 20
−12 ± 48
114 ± 90
Table 2.4: Assigned dierence (in per meg) between LK560944 and LK560962.
30
2. MEASUREMENT METHODS AND TESTS
Figure 2.7: δO2 /N2 , δAr/N2 , and δ 29 N2 measurements of reference standards relative to LK560962.
2.5. AIR STANDARD MEASUREMENTS
Figure 2.8: δ 34 O2 , δ 33 O2 and δ 36 Ar measurements of reference standards relative to LK560962.
31
32
2. MEASUREMENT METHODS AND TESTS
Figure 2.9: CO2 measurements of three O2 /N2 reference standards.
the working standard, are placed horizontally in a rack under a table, tted with the same
regulators and used with the same ow rates. Thus, they are actually expected to be aected
by the same processes. A possible explanation might be a sensitivity of the measurements
to the ion source condition, which is changing due to ion-induced metal sputtering and redeposition within the ion source. However, in that case the inuence should be smallest for
gases like LK560913, which have the same or very similar concentrations compared to the
working standard.
The δAr/N2 is generally well correlated with δO2 /N2 , indicating that fractionation rather
than surface adsorption processes are the dominant cause of the variability. The δ 29 N2 is
stable within measurement precision, with the exception of small long-term variations of about
25 per meg. These drifts are presumably related to changing ion source conditions, since δ 29 N2
is sensitive to the amount of moisture in the ion source. Protonation of 28 N2 due to the presence
of water produces 14 N14 NH+ , which interferes with 29 N2 . A positive correlation of δ 29 N2 with
the water signal, as inferred from ion beam intensity at m/z 18, could be measured on ask
samples with elevated moisture content. There is always a substantial background of water
in the ion source, which reects the strong tendency of moisture to absorb onto surfaces and
slowly desorb under vacuum. Even though all sample and standard gases admitted to the
mass spectrometer were dried during sampling, there is a constant supply of moisture to the
ion source from remaining traces of water in these gases. The background signal of water also
depends on the vacuum history of the ion source.
The isotopic ratios of O2 (δ 34 O2 and δ 33 O2 ) are stable within measurement precision,
whereas δ 36 Ar (lowermost panel in Figure 2.8) shows signicant long-term variations also
related to ion source conditions. Major shifts in δ 36 Ar occurred after revision of the ion
source.
Figure 2.9 shows the CO2 records of the three reference standards in use for δO2 /N2 measurements. The variability in CO2 mixing ratio decreased in the course of these measurements
as we gained more experience in CO2 analysis.
2.5. AIR STANDARD MEASUREMENTS
33
Measured - assinged CO2 (ppm)
0.4
0
y = 0.006 x -2.5
R2 = 0.73
-0.4
-0.8
-1.2
-1.6
-2
160
200
240
280
320
360
CO2 mixing ratio (ppm)
400
440
Figure 2.10: Dierence between the measured CO2 mixing ratio and the assigned value for eight high pressure
gas cylinders calibrated by CCGG/CMDL/NOAA. The CO2 mixing ratio of these primary standards are in
the range of 192363 ppm. The observed trend can be removed by assuming a 2.15 ppm contamination of the
reference gas used for background correction.
2.5.2
CO2 Primary Standards
The eight high pressure gas cylinders purchased from and calibrated by CCGG/CMDL/
NOAA constitute primary standards for CO2 , CO, N2 O, CH4 and δ 18 O and δ 13 C of CO2 .
They were analyzed for CO2 by mass spectrometry twice during the last two years. The CO2
mixing ratio of these primary standards are in the range of 192363 ppm. Figure 2.10 shows the
dierence between the measured CO2 mixing ratio and the assigned value for all eight cylinders. There is a concentration dependent oset of about 0 to −1.5 ppm. This can be explained
by a contamination of about 2.02.3 ppm of the reference air used for background correction.
This CO2 -free reference air is decanted into 0.5 L asks from a high pressure cylinder, where
the cylinder CO2 mixing ratio has been independently determined to be < 0.4 ppm. Although
the asks are evacuated before lling, CO2 might still be adsorbed on the interior surface of the
ask and on the O-rings of the valves, from where it then slowly desorbs. Tests with GC-MS
as well as with air where the CO2 was cryogenically trapped indicated that the CO2 mixing
ratio could indeed be about 1.11.7 ppm, which explains at least part of the nonlinearity. A
possible cross contamination correction would even increase the observed discrepancy. There
is also a storage related eect. CO2 mixing ratio is expected to increase in these asks due
to permeation by about 0.4 ppm per month, and for some measurements asks with storage
times of up to 4 months have been used. Removing the trend results in a mean dierence
of −0.3 ± 0.21 ppm. At least part of the observed variations might also be real deviations
from the assigned value. Unfortunately, we do not have any calibrated CO2 standards with a
mixing ratio above 363 ppm. However, if the observed trend is extrapolated to higher mixing
ratios it follows that for atmospheric air samples with a CO2 mixing ratio of 380 ± 10 ppm the
deviations are < 0.3 ppm.
The δO2 /N2 and δAr/N2 of these CO2 primary standard have not been precisely measured
up to know. Large drifts in δO2 /N2 were partly present during the CO2 measurements. More
34
2. MEASUREMENT METHODS AND TESTS
cylinder number
δO2 /N2
(per meg)
δAr/N2
(per meg)
δAr/O2
(per meg)
CO2 measured
(ppm)
CO2 assigned
(ppm)
CA04556
CA04183
CA04200
CA04565
CA04568
CA04561
CA04574
CA03901
-3400
60
60
-2200
150
-1650
130
200
-4500
-1950
-2000
-3100
-2000
-2750
-2080
-1950
-1100
-2000
-2050
-900
-2200
-1100
-2200
-2150
191.18
191.39
218.19
222.18
258.06
294.25
295.49
362.89
192.66
192.70
219.05
223.34
258.88
295.02
296.53
363.01
Error estimate
±30
±50
±50
±0.2
±0.1
Table 2.5: Approximate δO2 /N2 , δAr/N2 , and δAr/O2 values of the CCGG/CMDL/NOAA standards. The
two last columns contain the measured and assigned CO2 mixing ratios.
attention has to be paid for δO2 /N2 analysis to prevent fractionation associated with drawing
air from the cylinder through the regulator into the inlet system. Therefore, Table 2.5 only
lists approximate values of δO2 /N2 , δAr/N2 , and δAr/O2 . The mean of the measured and
the assigned CO2 mixing ratios are also listed in Table 2.5.
2.6 Summary
The dierent tests and corrections presented in this chapter emphasize the challenges faced
in making highly precise O2 /N2 measurements. Several aspects relating to the mass spectrometric method and gas handling procedures for O2 /N2 and CO2 analysis are discussed. One
large source of noise still present in our measurements presumably originates from temperature variations in the laboratory. Diurnal temperature drifts of up to 3 ◦ C are common and
they can contribute to all sorts of variability relating for example to changes in instrument
sensitivity or fractionation of the reference and sample gases. Variability associated with admitting the gases to the mass spectrometer could probably be reduced by actively controlling
the temperature of the box containing the sample asks. Pressure or temperature induced
fractionation of O2 relative to N2 are among the most important modifying processes, but
the conditions under which it occurs and the physical processes causing it are still not fully
understood. The problem of possible fractionation during sampling is not addressed here, but
may also contribute to noise in atmospheric time series. Another important modifying process
specic to ask sampling programs, namely permeation of gases through polymeric seals of
the ask's valves, is discussed in Chapter 3.
An urgent need for using O2 /N2 measurements from dierent laboratories in global carbon
cycle studies is to establish inter-laboratory calibration strategies. Up to now, no absolute
O2 /N2 reference scale exists, which mainly arises from technical diculties in preparing, storing and analyzing O2 /N2 standards to the required precision. Each laboratory involved in
O2 /N2 measurement programs relies on its own calibrations scale. A recently initiated international intercomparison program strives to merge the existing O2 /N2 scales from the dierent
laboratories and will allow to compare the data and the long-term stability of the calibration
scales.
REFERENCES
35
References
Bender, M. L., P. P. Tans, T. J. Ellis, J. Orchardo, and K. Habfast (1994), A high precision isotope
ratio mass spectrometry method for measuring the O2 /N2 ratio of air, Geochim. Cosmochim. Acta,
58 (21), 47514758.
Keeling, R., R. Najjar, M. Bender, and P. Tans (1993), What atmospheric oxygen measurements can
tell us about the global carbon cycle, Global Biogeochem. Cycles, 7 (1), 3767.
Keeling, R. F., B. B. Stephens, R. G. Najjar, S. C. Doney, D. Archer, and M. Heimann (1998), Seasonal
variations in the atmospheric O2 /N2 ratio in relation to the kinetics of air-sea gas exchange, Global
Biogeochem. Cycles, 12 (1), 141163.
Keeling, R. F., T. Blaine, B. Paplawsky, L. Katz, C. Atwood, and T. Brockwell (2004), Measurement
of changes in atmospheric Ar/N2 ratio using a rapid-switching, single-capillary mass spectrometer
system, Tellus, 56B (4), 322338.
Langenfelds, R. L. (2002), Studies of the global carbon cycle using atmospheric oxygen and associated
tracers, Ph.D. thesis, Univ. of Tasmania, Hobart, Tasmania, Australia.
Leuenberger, M., P. Nyfeler, H. Moret, P. Sturm, and C. Huber (2000a), A new gas inlet system for
an isotope ratio mass spectrometer improves reproducibility, Rapid Commun. Mass Spectrom., 14,
15431551.
Leuenberger, M., P. Nyfeler, H. Moret, P. Sturm, A. Indermühle, and J. Schwander (2000b), CO2
concentration measurements on air samples by mass spectrometry, Rapid Commun. Mass Spectrom.,
14, 15521557.
Machta, L., and E. Hughes (1970), Atmospheric Oxygen in 1967 to 1970, Science, 168, 15821584.
Meijer, H. A. J., R. E. M. Neubert, and G. H. Visser (2000), Cross contamination in dual inlet isotope
ratio mass spectrometers, International Journal of Mass Spectrometry, 198, 4561.
Sturm, P. (2001), Entwicklung eines neuen Einlasssystems für die massenspektrometrische Messung
des O2 /N2 Verhältnisses, Master's thesis, Physics Institute, University of Bern, Bern, Switzerland.
Verkouteren, R. M., S. Assonov, D. B. Klinedinst, and W. A. Brand (2003), Isotopic metrology of
carbon dioxide. II. Eects of ion source materials, conductance, emission, and accelerating voltage on
dual-inlet cross contamination, Rapid Commun. Mass Spectrom., 17, 777782, doi:10.1002/rcm.906.
World Meteorological Organisation (2005), Recommendations of the 12th WMO/IAEA Meeting of
Experts on Carbon Dioxide Concentration and Related Tracer Measurement Techniques, Toronto,
Canada, 1518 Sep. 2003, in press.
Chapter 3
Permeation of Atmospheric Gases
Through Polymer O-Rings Used in
Flasks for Air Sampling
P. Sturm1 , M. Leuenberger1 , C. Sirignano2 , R.E.M. Neubert2 , H.A.J. Meijer2 , R. Langenfelds3 , 4 , W.A. Brand5 , and Y. Tohjima6
Published in Journal of Geophysical Research, Vol. 109, D04309,
doi:10.1029/2003JD004073, 2004.
Abstract
Permeation of various gases through elastomeric O-ring seals can have important eects on
the integrity of atmospheric air samples collected in asks and measured some time later.
Depending on the materials and geometry of asks and valves, and on partial pressure differences between sample and surrounding air, the concentrations of dierent components of
air can be signicantly altered during storage. The inuence of permeation is discussed for
O2 /N2 , Ar/N2 , CO2 , δ 13 C in CO2 and water vapor. Results of sample storage tests for various ask and valve types and dierent storage conditions are presented, and compared with
theoretical calculations. Eects of permeation can be reduced by maintaining short storage
times and small partial pressure dierences and by using a new valve design that buers exchange of gases with surrounding air, or by using less permeable materials (such as Kel-F)
as sealing material. General awareness of possible permeation eects helps to achieve more
reliable measurements of atmospheric composition with ask sampling techniques.
1
Climate and Environmental Physics, Physics Institute, University of Bern, Bern, Switzerland.
2
Centrum voor IsotopenOnderzoek, University of Groningen, Groningen, Netherlands.
3
Commonwealth Scientic and Industrial Research Organisation Atmospheric Research, Aspendale, Australia
4
Institute of Antarctic and Southern Ocean Studies, University of Tasmania, Hobart, Tasmania, Australia.
5
Max Planck Institute for Biogeochemistry, Jena, Germany.
6
Atmospheric Environment Division, National Institute for Environmental Studies, Tsukuba, Japan.
38
3. PERMEATION OF GASES THROUGH POLYMER O-RINGS
3.1 Introduction
Flask sampling of atmospheric air is an important tool in climate research to investigate
atmospheric composition changes on regional to global spatial scales and on timescales from
hours to decades. Especially for remote sites and for airborne campaigns, often the only way
to obtain measurements is to collect air in glass asks in situ and to analyze the samples
later in a laboratory equipped with measuring instruments. This increases the risk of altering
the composition of the air either during the sampling procedure or during the storage prior to
measurement. To minimize such adverse eects one must choose appropriate lling procedures
and ask materials.
In the case of atmospheric O2 concentrations these gas handling aspects prove to be particularly crucial [Keeling et al., 1998]. Measurements of the O2 concentration can provide
important constraints on the global carbon cycle [Keeling et al., 1993]; however, they are
challenging because the variations occur only at the parts per million (ppm) level relative to
the large 21 % O2 content. Any small disturbance of sample air can thus result in a similar or
bigger signal compared to the real atmospheric variations.
To reduce adsorption and desorption of gas molecules on surfaces, the asks most commonly used for O2 measurements are made of glass. There exist dierent valve designs; however, all asks currently in use for this task are sealed by a polymer material, which is pressed
against the glass surface of the valve. All polymers are subject to permeation of gases, which
therefore limits the eectiveness of the seal if partial pressure dierences between sample and
surrounding air exist.
Processes other than permeation may also inuence the stability of the sample composition.
For instance, selective outgassing or physical adsorption of gases on O-rings as well as oxidation
of grease used to lubricate the O-rings can cause changes of the air composition. Keeling et al.
[1998] reported a depletion in O2 /N2 ratio and CO2 concentration in 5 L glass asks tted
with Viton O-rings and lled to 1000 hPa pressure with 20 extra O-rings inserted in the asks.
Flasks with no extra O-rings, however, showed no signicant drifts over 313 days, in accordance
with ndings of Bender et al. [1996].
In this paper we report on the permeation process of dierent gases and various O-ring materials used for glass valves. Because of the sensitivity of high precision O2 /N2 measurements
to small interferences we focus particularly on O2 /N2 . Additionally, water vapor, Ar/N2 , CO2
and δ 13 C in CO2 are discussed. Section 3.2 summarizes theoretical considerations on the permeation process and shows the time dependence of the permeation rate for helium through a
Viton O-ring. Section 3.3 deals with changes of sample air composition with increasing storage
time and compares measured to calculated permeation rates. Tests with various O-ring materials and dierent storage conditions are presented. In section 3.4 storage drift measurements
and permeation calculations of asks sealed with new double O-ring valves are discussed.
Finally, section 3.5 summarizes our major ndings.
3.2 Permeation Process
3.2.1
Theory
Permeation is commonly viewed as a three-step process. First, gas dissolves into the solid's
surface, then it diuses through the solid; and nally, desorbs from it. The time dependence of
the permeation rate is determined by the diusion process, because the solution and desorption
3.2. PERMEATION PROCESS
39
occur on much shorter timescales than the diusion through the solid [Perkins , 1973]. The
concentration c of the diusing species in the solid's surface is given by Henry's law
(3.1)
c = Sp,
where S is the solubility coecient and p the gas pressure. Diusion of the dissolved gas into
the interior is governed by Fick's second law
∂c
= D∇2 c,
(3.2)
∂t
where D is the concentration-independent, isotropic diusion coecient. If we consider a
planar membrane with a large area-to-thickness ratio, equation (3.2) can be reduced to a
one-dimensional form
µ 2 ¶
∂c
∂ c
=D
.
(3.3)
∂t
∂x2
Assuming that the membrane is initially free of gas, and the concentrations of gas at the two
surfaces are xed at c1 and c2 , respectively, the initial and boundary conditions are
c(x, t = 0) = 0,
(3.4)
c(x = 0, t) = c1 ,
(3.5)
c(x = d, t) = c2 ,
(3.6)
where d denotes the membrane thickness. The solution of equation (3.3) consistent with (3.4),
(3.5) and (3.6) is [Barrer , 1941]
µ
¶
∞
x 2 X c2 cos nπ − c1
nπx
Dn2 π 2 t
c(x, t) = c1 + (c2 − c1 ) +
sin
exp −
.
(3.7)
d π
n
d
d2
n=1
The ux of diusing species is given by Fick's rst law
(3.8)
J = −D∇c.
The gas ux through the low-pressure surface of the wall can then be calculated from equations
(3.7) and (3.8)
µ
¶
∞
D(c1 − c2 ) 2D X
Dn2 π 2 t
n
J(x = d, t) =
+
((−1) c1 − c2 ) exp −
.
(3.9)
d
d
d2
n=1
For large values of t, one obtains the steady state ux
J(x = d, t → ∞) =
D(c1 − c2 )
DS(p1 − p2 )
=
.
d
d
(3.10)
The product
(3.11)
K = DS
is the permeation coecient. It has, like the diusion and solubility coecient, a strong
(approximately exponential-like) dependence on temperature. Finally, the steady state permeation ux F of a gas exposed with a partial pressure dierence (p1 − p2 ) to a membrane of
thickness d and surface area A is
F = AJ(x = d, t → ∞) = K
A(p1 − p2 )
.
d
(3.12)
40
3. PERMEATION OF GASES THROUGH POLYMER O-RINGS
A
cylindrical shaft
B
tapered shaft
C
buffer volume
Figure 3.1: Schematic diagram of dierent ask valves; (a) valve with cylindrical shaft (Louwers Hapert,
Netherlands), (b) valve with tapered shaft (Glass Expansion, Melbourne, Australia), and (c) double O-ring
valve with buer volume.
3.2.2
Measurement of Time-Dependent Permeation Rate
Flasks used at the Physics Institute, University of Bern are each equipped with two highvacuum valves (Louwers Hapert, Netherlands). The valves are opened and closed by moving
a cylindrical plug in and out of a 9 mm ID valve tube (see Figure 3.1a). An O-ring sitting in
a notch on the plug seals the ask volume from surrounding air. The O-ring material is Viton
r A).
(FKM 70-Copolymer-compound 51414-GenuineViton °
Although the property of a large area-to-thickness ratio is not true for an O-ring, the
one-dimensional consideration of the permeation process is still justied for this valve type
because the sealing O-ring is mounted on and enclosed by impermeable cylindrical glass, which
acts as a boundary in two dimensions. According to the size of the valves and O-rings, the
eective thickness and exposed surface area of an O-ring can be estimated. For our asks
the O-ring thickness and surface area are ∼1.7 mm and ∼20 mm2 , respectively, per valve.
Unless otherwise specied, all calculations presented in this paper are based on these O-ring
dimensions.
To make sure that the changes in sample composition we observe in these asks are primarily due to permeation and not a result of leaks, we conducted an experiment using these
asks and a helium leak detector. The previously evacuated asks had been lled to 950 hPa
with helium, and the ux through the Viton O-ring valve was subsequently measured by
the leak detector. The temperature was ∼20 ◦ C. Figure 3.2 shows the permeation ux as a
function of time. Instead of a sharp signal increase as expected from a real leak, the leak
3.3. FLASK STORAGE DRIFT
41
Permeation flux [10-13 m3s-1 STP]
1.8
1.6
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
0
10
20
30
Time [min]
40
50
60
Figure 3.2: Helium permeation through a Viton O-ring. An initially evacuated ask was lled with helium
to 950 hPa. The helium ux was then measured as a function of time (circles). The O-ring has a thickness
of ∼1.7 mm and an exposed surface area of ∼20 mm2 . A least squares t of equation (3.9) to the measured
data (line) yields permeation and solubility coecients of KHe = 15.1 · 10−15 m2 s−1 hPa−1 and SHe = 2.3 ·
10−5 hPa−1 , respectively.
detector's helium signal slowly increased because of permeation. Steady state permeation was
achieved in ∼40 min. The solid line in Figure 3.2 is a least squares t of equation (3.9) to the
measured data. The tted parameters are the permeation and solubility coecients K and
S , respectively. We obtain a solubility coecient of SHe = 2.3 · 10−5 hPa−1 and a permeation
coecient of KHe = 15.1 · 10−15 m2 s−1 hPa−1 , which is in agreement with the values listed in
Table 3.1.
Given the O-ring geometry and permeation coecients for dierent gases, we can estimate
the inuence of permeation on sample air composition for dierent sample pressures and ask
volumes. However, the uncertainty in the O-ring dimensions and the large range of some permeation coecients shown in Table 3.1 restricts the precision of such an estimate. The scatter
of results in Table 3.1 is probably mainly due to modications in polymer manufacturing
over the last decades and the fact that the polymers may vary between dierent manufacturers. The composition of a polymer can cause properties to dier by orders of magnitude. It
is therefore essential to compare theoretical calculations with measured drifts in sample air
composition as a function of storage time.
3.3 Flask Storage Drift
3.3.1 Permeation Inuence on O2 /N2
We express changes in elemental and isotopic ratios as relative deviations from a reference gas.
In the case of O2 /N2 , variations are reported in units of per meg (where 1 per meg = 0.001 %)
according to δO2 /N2 = [(O2 /N2 )sample /(O2 /N2 )ref erence − 1] · 106 .
Atmospheric ask samples taken in asks at the high altitude station Jungfraujoch (3580 m
above sea level (asl), 46◦ 33'N, 7◦ 59'E) at ambient pressure and stored in the laboratory in Bern
(560 m asl) show a strong dependence of the measured O2 /N2 ratio on the storage time. This
42
3. PERMEATION OF GASES THROUGH POLYMER O-RINGS
δO 2 /N 2 [p e r m e g ] (v s. B e rn re fe re n ce g a s)
2000
1900
1800
1700
1600
1500
1400
0
50
100
150
200
250
S tora ge tim e [d a ys]
300
350
400
Figure 3.3: Measurements of δO2 /N2 as a function of storage time. The samples are from Jungfraujoch and
are stored in 0.5 L asks, lled at ambient pressure at 3580 m above sea level and at a mean temperature of
−7 ◦ C. Mean sample pressure at room temperature is 720 hPa, and mean ambient pressure is 950 hPa. Ordinary
least squares regression yields a slope of 407 ± 11 per meg yr−1 .
correlation is attributed to permeation of gases through the Viton O-rings of the asks' valves.
In Figure 3.3, δO2 /N2 measurements from Jungfraujoch are plotted as a function of storage
time. The 0.5 L asks were stored at room temperature and at a mean ambient pressure
of 950 hPa over a period of up to 381 days. The mean sample pressure during the storage
time was ∼720 hPa. Ordinary least squares regression yields a slope of 407 ± 11 per meg yr−1 .
Hence, even after a storage time of ∼100 days, the eect of permeation clearly exceeded any
real atmospheric signal. For example, the seasonal cycle amplitude at this site is expected
to be in the order of 100 per meg [Keeling et al., 1993]. With permeation coecients KO2 =
1.1 · 10−15 m2 s−1 hPa−1 and KN2 = 0.3 · 10−15 m2 s−1 hPa−1 for O2 and N2 respectively, one
would expect a slope of 379 per meg yr−1 , very close to the measured value.
On the basis of these results, laboratory experiments were conducted to further test the
stability of sample air composition in these asks. Four 0.5 L asks were lled to 655 hPa with
dry air from a high pressure cylinder and stored at ambient pressure (950 hPa). The samples
were analyzed four times during the storage period of 299 days, in which sample pressure
decreased to ∼560 hPa. Figure 3.4 shows δO2 /N2 deviations from the rst measurement of
each ask sample. These measurements agree well with predicted drift rates, indicating that
the O-ring dimensions and permeation coecients used to calculate predicted δO2 /N2 drift
rates seem to be justied.
Storage tests were also conducted by the Commonwealth Scientic and Industrial Research
Organisation (CSIRO) in their 0.5 L glass asks tted with dierent O-ring materials (Viton,
peruoroalkoxy (PFA) and polytetrauoroethylene (PTFE)) and at a range of sample pressures from 1000 to 2100 hPa. The results demonstrate the dependence of permeation rates
on O-ring material and partial pressure dierences [Langenfelds , 2002]. Each CSIRO ask
is tted with two valves (Glass Expansion, Melbourne, Australia), which dier from those
manufactured by Louwers Hapert in that the valve shafts are tapered and seals are formed by
3.3. FLASK STORAGE DRIFT
43
700
600
∆δO 2 /N 2 [p er m e g ]
500
400
300
200
100
0
0
50
100
150
S to ra g e Tim e [d a ys]
200
250
300
Figure 3.4: Changes in δO2 /N2 due to storage in 0.5 L asks. Measurements of the asks UBE-107 (circles),
UBE-108 (squares), UBE-113 (diamonds) and UBE-114 (plus sign), lled to 655 hPa and stored at 950 hPa,
agree well with the calculated increase (solid line). Note that δO2 /N2 does not increase linearly because of the
pressure decrease of the sample air after each decanting.
compression of the O-rings (Figure 3.1b). Figure 3.5 shows changes in δO2 /N2 over storage
times of up to 15 months in ask air samples lled to 1500 hPa. The higher sample pressure
compared with a mean ambient pressure of 1000 hPa led to δO2 /N2 being depleted in these
samples. All data points represent individual ask samples analyzed on a single occasion.
Lowest drift rates of −61 per meg yr−1 were observed with Viton O-rings.
To account for the geometric dierences between Glass Expansion (GE) and Louwers
Hapert (LH) valves, we measured helium permeation rates with a helium leak detector for
both valve types tted with Viton O-rings. This experiment showed that GE valves have a
lower helium permeation rate than LH valves. Depending on the valve and on how tightly
the GE valve is closed, the steady state permeation ux was a factor of 410 smaller for GE
valves than for LH valves. This can be explained by the smaller area of the O-ring that is
exposed to the gas in the closed GE valves. Because of the conical tip of the piston the O-ring
is compressed into the groove cut in this tip, and the gap between the glass piston and the
mating conical seat is smaller than for LH valves. Assuming a 7 times smaller area (3 mm2
instead of 20 mm2 , d = 1.7 mm), one would expect a δO2 /N2 drift rate of −59 per meg yr−1 ,
which is consistent with the observations.
Figure 3.6 shows observed rates of drift as a function of sample pressure. Drift rates are
near zero at sample pressures of 1000 hPa and increase approximately linearly with sample
pressure, providing compelling evidence that permeation is indeed the dominant modifying
process.
3.3.2 Permeation of Water Vapor
Variations of the amount of moisture in the sample air can induce dierent artifacts, like oxygen
isotope exchange during ask storage between CO2 and water condensed on the ask walls
44
3. PERMEATION OF GASES THROUGH POLYMER O-RINGS
100
0
∆δO 2 /N 2 [p e r m e g ]
-1 0 0
-2 0 0
-3 0 0
-4 0 0
-5 0 0
-6 0 0
-7 0 0
0
50
100
150
200
250
300
S to ra g e Tim e [d a ys]
350
400
450
500
Figure 3.5: Changes in δO2 /N2 due to storage in Commonwealth Scientic and Industrial Research Organisation's 0.5 L glass asks tted with Viton (circles, with linear regression represented by the line), peruoroalkoxy
(plus signs), and polytetrauoroethylene (squares) O-rings.
δO 2 /N 2 d rift ra te [p e r m e g yr -1 ]
0
-2 0 0
-4 0 0
-6 0 0
-8 0 0
-1 0 0 0
-1 2 0 0
-1 4 0 0
800
1000
1200
1400
1600
S a m p le p re ssu re [h P a ]
1800
2000
2200
Figure 3.6: Rate of drift in δO2 /N2 as a function of pressure and O-ring material (symbols same as for
Figure 3.5), as determined by linear regression to data in Figure 3.5 (corresponding to the sample pressure
of 1500 hPa) and data for other sample pressures (not shown). Error bars denote the standard error of the
regression for data sets with more than two points.
[Gemery et al., 1996]. Depending on the analysis technique, water vapor can also adversely
aect the measurement itself. In particular, for precise O2 /N2 measurements it is a prerequisite
that the analyzed sample air is dry. Therefore ask air samples are usually dried at the
time of sampling either by a chemical drying agent (e.g. anhydrous magnesium perchlorate,
Mg(ClO4 )2 ) or by a cryogenic cold trap. However, because of the large permeability coecients
of most polymers for water vapor (Table 3.1), initially dry samples stored in the laboratory
are intensely subject to water vapor permeation. Figure 3.7 shows the increasing amount of
moisture as a function of storage time for air samples in 0.5 L and 2.5 L asks stored in the
3.3. FLASK STORAGE DRIFT
45
Table 3.1: Permeation Coecients K for Various Polymer Materials and Gasesa
Material
Viton
PTFEc
NBRd
Neoprene
Polyurethane
Silicone
PCTFEe
N2
O2
0.050.3
1.01.1
1.1
0.3
0.17
0.45
0.233
0.14
4.7
2.5
1.442.4
2.3
0.22.0
0.1771.89
0.81.2
0.012
0.19
0.41.1
75210
128
0.0040.3
1.7
0.04
8.2
3.377.5
7.5
0.76.0
0.726.15
34
3.0
1.4
1.13.6
0.8
76460
195450
286
0.020.7
Ar
1.65
CO2
H2 O
He
Temp., ◦ C
Sourceb
5.86.0
40
216
916
2030
25
20
25
2330
2030
25
23
25
20
2030
2025
2030
25
20
2030
2025
2030
room
20
2030
25
1
2
3
4
5
1
2
6
5
7
1
5
1
5
7
1
5
1
5
3
1
6
3.5
2.4
0.12
4.4
1.2
450
27
23.3
7.51
5.748
5.63
1920
13.919.2
1030
10.5
4602300
10282280
0.041
760
1400
260-9500
8000
0.22
4.9
15.0
12.8
570
30.1523
530
5.26
7.48
1011
0.67.5
7.2
3.6
238263
156
a
The units are 10−15 m2 s−1 hPa−1 .
b
Sources are 1, Peacock [1980]; 2, Ma et al. [1995]; 3, Beckmann [1991]; 4, Laurenson and Dennis [1985]; 5,
Parker Hannin Corporation [2001]; 6, O'Hanlon [1989]; and 7, Holland et al. [1974].
c
PTFE is polytetrauoroethylene.
d
NBR is acrylonitrile butadiene copolymer.
e
PCTFE is polychlorotriuoroethylene.
laboratory. Water vapor concentration was inferred by mass spectrometry from ion beam
intensity at m/z 18, measured as a voltage V18 . After subtraction of the background signal,
which is the m/z 18 intensity of a dry standard gas, V18 was converted into a pressure using the
m/z intensity of N2 (V28 ) and a correction factor, which accounts for the dierent ionization
eciencies of water vapor and N2 .
The 0.5 L ask air samples show an increase in moisture content of 3.87 ± 0.17 hPa yr−1 ;
the 2.5 L samples are enriched by 0.79 ± 0.22 hPa yr−1 . The 2.5 L asks are a factor of 4.9 ± 1.4
less inuenced by water vapor corresponding to the 5 times larger volume of these asks. Both
ask types had a sample pressure of 1000 hPa.
Since for initially dry samples the water vapor pressure dierence between sample and surrounding air is independent of sample pressure, the water vapor permeation ux is controlled
by the humidity of the air at the asks' storage location. With an average ambient vapor
pressure of 14 hPa, which corresponds to ∼60% relative humidity at 20 ◦ C, one obtains from
the observed drift rates a permeation coecient for water vapor through our Viton O-rings
(d = 1.7 mm and A = 20 mm2 ) of about KH2 O = 220 · 10−15 m2 s−1 hPa−1 , which is consistent
with the values listed in Table 3.1.
46
3. PERMEATION OF GASES THROUGH POLYMER O-RINGS
6
W a te r va p o r [h P a ]
5
4
3
2
1
0
0
50
100
150
200
250
S to ra g e Tim e [d a ys]
300
350
400
450
Figure 3.7: Water vapor as a function of storage time for 0.5 L (circles) and 2.5 L asks (plus signs). The
steeper slope for the 0.5 L asks (factor of 4.9 ± 1.4) reects the dierence in ask volume.
Thus permeation of water vapor may have adverse eects on trace gas measurements if
sample air is not dried at the time of analysis. The increasing water vapor concentration leads
to dilution of trace gas concentrations in sample air and might also lead to other storage- or
analysis-related measurement artifacts for some species, including δO2 /N2 .
3.3.3
Permeation of Ar and CO2
Changes in δAr/N2 and δCO2 /N2 for the same samples as in Figure 3.4 are shown as a function
of storage time in Figure 3.8. The δAr/N2 values of these four samples show an increase of
399 ± 32 per meg after 299 days. Due to larger mass-dependent fractionation processes at
the time of analysis and the poorer overall precision of δAr/N2 measurements, the storage
drift is not as well resolved as for δO2 /N2 . The δAr/N2 drift rate is slightly smaller than
that for δO2 /N2 . This is consistent with measurements from Langenfelds [2002]. However,
an estimate based on a permeation coecient of KAr = 1.65 · 10−15 m2 s−1 hPa−1 (Table 3.1)
gives an expected drift rate of 1070 per meg in 299 days for these samples. The predicted and
observed drift rates dier by a factor of 2.7, suggesting that the value of KAr in Table 3.1 is
too large.
The initial CO2 concentrations of these samples were between 401.6 and 406.9 ppm. During the storage period of 299 days a δCO2 /N2 change of ∼6 % was observed (Figure 3.8,
bottom). This result can be explained with a CO2 permeation coecient of KCO2 = 6 ·
10−15 m2 s−1 hPa−1 (Table 3.1) and a mean CO2 concentration of ambient laboratory air of
450 ppm, i.e., ∼ 20 % above a background atmospheric CO2 level of 370 ppm (solid line, Figure 3.8).
Unlike for O2 , N2 and Ar, where concentration changes in ambient air are negligible in
respect of permeation, the mole fraction of CO2 in laboratory air can change substantially.
Changes in sample CO2 concentration are therefore highly dependent on the storage conditions. The CO2 partial pressure dierence and the permeation ux are likely to vary with
storage location and also with time. As the ask samples may be stored in a CO2 -enriched en-
3.3. FLASK STORAGE DRIFT
47
∆δA r/N 2 [p e r m e g ]
1000
800
600
400
200
0
8
7
∆δC O 2 /N 2 [‰ ]
6
5
4
3
2
1
0
0
50
100
150
S to ra g e Tim e [d a ys]
200
250
300
Figure 3.8: Changes in δAr/N2 and δCO2 /N2 plotted as a function of the storage time and as a change
relative to the rst measurement for the same samples as in Figure 3.4: UBE-107 (circles), UBE-108 (squares),
UBE-113 (diamonds) and UBE-114 (plus signs). (top) The δAr/N2 values show an increase of 399 ± 32 per meg
after 299 days. Based on KAr = 1.65 · 10−15 m2 s−1 hPa−1 one would expect an increase of 1071 per meg (line).
(bottom) The CO2 /N2 ratios (in %) are enriched by 5.8±0.3 % after 299 days. The initial CO2 concentrations
were between 401.6 and 406.9 ppm. The line is an estimate based on KCO2 = 6 · 10−15 m2 s−1 hPa−1 and an
assumed mean CO2 concentration of 450 ppm of laboratory air.
vironment or a well-ventilated room or as the presence of people in the laboratory may cause
diurnal CO2 concentration variations, it is dicult to accurately predict CO2 permeation
uxes. Under some circumstances, there may be little or no change in the CO2 concentration
of sample air, for example, where a background air sample at a pressure of 1200 hPa is stored
in an environment with CO2 enriched by 20% above background level.
In terms of a mixing ratio the permeation of other gases may also become important. The
increasing amount of CO2 in the samples of Figure 3.8 is diluted by the simultaneous inux
of water vapor. If the sample air is not dried again at the time of analysis and assuming
an ambient water vapor pressure of 14 hPa and KH2 O = 220 · 10−15 m2 s−1 hPa−1 , one would
obtain a CO2 concentration drift after 299 days of −0.2 ppm for the conditions as outlined
in Figure 3.8. If the sample air is dried prior to measurement, the δCO2 /N2 increase of 6 %
corresponds to a CO2 concentration change of +2.4 ppm. Thus CO2 measurements from ask
samples are not only dependent on ask volume and sample pressure but are also highly
sensitive to ambient CO2 mixing ratios and analysis techniques. Changes in CO2 of the
magnitude reported here are potentially much larger than targets for CO2 inter-comparability
48
3. PERMEATION OF GASES THROUGH POLYMER O-RINGS
among laboratories of ±0.1 ppm in the Northern Hemisphere or ±0.05 ppm in the Southern
Hemisphere recommended by the World Meteorological Organisation [2003].
Eects of permeation will also inuence the isotopic composition of CO2 in ask samples.
It is likely that isotopic fractionation is associated with the permeation process itself; that
is, the lighter isotopomeres are favored over the heavier ones, comparable, for example, to
diusion of CO2 through air. Moreover, the inux of water vapor might lead to 18 O isotopic
exchange between water and CO2 .
Permeation can aect the isotopic composition, even in the absence of permeation fractionation eects. Changes in δ 13 C occur if asks are stored in an environment enriched by biogenic
CO2 , such as in a laboratory. Assuming an average enrichment of 80 ppm with δ 13 C = −25 %
(versus Vienna Peedee belemnite) compared to −8 % for background air, the drift rate for
0.5 L asks with 600 hPa sample pressure would be about −0.05 % yr−1 . Thus changes in
isotopic composition become potentially signicant if samples are stored for long periods in a
CO2 -rich environment.
In many laboratories it is the practice to store pure CO2 samples for a limited period of
time in test tubes sealed by an O-ring valve. On some occasions, for example, when shipping
such tubes to other laboratories, storage times can be up to several weeks, and permeation
eects for these samples with their large CO2 gradient can become large. These eects were
already observed in the 1980s [Keeling et al., 1989] but were not studied in detail at that time.
New investigations, now with the aid of the permeation model presented here, are under way
at the Centrum voor IsotopenOnderzoek (CIO), Groningen, Netherlands. They will produce,
among other things, δ 13 C and δ 18 O permeation fractionation factors.
3.4 Double O-ring Valves
Tests with dierent O-ring materials and the data in Table 3.1 show that Viton has comparably
low permeation coecients. So far, it is the preferred material for O-rings in glass asks used
for O2 /N2 measurement programs. A new approach to further reduce permeation inuences is
a new valve design, the so-called double O-ring valves. This design was, to our knowledge, rst
discussed and used by the National Institute for Environmental Studies, Tsukuba, Japan. In
contrast to conventional valves, where a single O-ring seals the sample air from the atmosphere,
the double O-ring valves incorporate a second O-ring, which creates an additional small buer
volume (Figure 3.1c). Initially, this buer volume also contains sample air, eliminating any net
permeation uxes between ask and buer volume. In rst order, permeation due to partial
pressure gradients between sample and ambient air only leads to changes of air composition
in the buer volume. Only these changes in the buer volume are eventually passed on to
the ask volume through permeation. The larger the buer volume is and the smaller the
ambient/buer permeation uxes are, the more eective the double O-ring valve is in reducing
modication of sample air composition. The construction of the double O-ring valve allows
a sequential opening, in order to be able to evacuate the buer volume prior to decanting
sample air into analytical inlet systems.
The permeation inuence with double O-ring valves can be calculated by solving two
coupled linear dierential equations for the partial pressures in the buer and ask volume,
respectively. In addition to calculations, dierent ask storage tests with double O-ring valves
were carried out. Figure 3.9 shows a storage drift test performed at the Max Planck Institute
for Biogeochemistry (MPI-BGC), Jena, Germany. Here 1.4 L asks equipped with two custom-
3.4. DOUBLE O-RING VALVES
49
20
∆δO 2 /N 2 [p e r m e g ]
0
-2 0
-4 0
-6 0
-8 0
-1 0 0
0
50
100
150
S to ra g e tim e [d a ys]
200
250
Figure 3.9: Changes in δO2 /N2 due to storage for nine 1.4 L asks equipped with two double O-ring valves
(symbols, dashed lines). The buer volume of each valve is 0.6 mL, and the sample air pressure is 2000 hPa.
The solid line represents the expected drift rate.
made double O-ring valves (Louwers Hapert, Netherlands) and lled with dry sample air to
2000 hPa were stored at ambient pressure and were measured several times during ∼200 days.
The buer volume of each valve is 0.6 mL. Six asks show a change in δO2 /N2 of −3 to
−31 per meg after 200 days storage in accordance with the calculated drift rate (solid line in
Figure 3.9). Three samples, however, are depleted in δO2 /N2 by −70 to −92 per meg, which
is close to the expected change for single O-ring asks of −116 per meg. Possible explanations
for this dierent behavior are dierent initial ask conditions, i.e. impurities in the asks, or
oxidation of grease used to lubricate the Viton O-rings.
A dierent experimental set-up was used at the CIO, Groningen, for a comparison test
between single and double O-ring valves. All asks had a volume of only 0.35 L and were lled
with dry air to ambient pressure (1000 hPa). During the storage time they were connected to
a vacuum line, which kept the pressure below 4 hPa in order to sustain a pressure gradient
of 1000 hPa across the O-rings. The asks were tted with two valves each (Louwers Hapert,
Netherlands), either with single or double O-rings. The double O-ring stopcocks were produced
at the workshop in Groningen. The buer volume of these valves is about 0.08 mL.
The experiment was carried out twice, once with 16 asks stored over a period of 98 days
and once with 8 asks stored over 66 days. Samples were measured after a range of storage
times and in pairs at a time, in each case including a single and a double O-ring ask. Each
sample was analyzed only once so that sample pressures were constant over the duration of
the storage test.
The results are plotted in Figure 3.10 as the deviation from the mean of the measurements
on the lling day. Solid symbols represent individual samples from double O-ring asks, and
open symbols represent samples from single O-ring asks. The single O-ring asks show a
change in δO2 /N2 of about −470 per meg after 98 days, which is in good agreement with the
predicted change for this test set-up (shaded line). The drift rate for the double O-ring asks
is clearly lower. However, the calculated drift rate leads to a depletion of 181 per meg (solid
line), whereas the measurements reveal a depletion of ∼280 per meg over 98 days storage,
50
3. PERMEATION OF GASES THROUGH POLYMER O-RINGS
0
∆δO 2 /N 2 [p e r m e g ]
-1 0 0
-2 0 0
-3 0 0
-4 0 0
-5 0 0
0
20
40
60
S to ra g e tim e [d a ys]
80
100
Figure 3.10: Comparison of δO2 /N2 drifts for 0.35 L asks equipped with either two single O-ring (open
symbols) or two double O-ring (solid symbols) valves. The buer volume of the double O-ring valves is
0.08 mL. Sample air pressure is 1000 hPa, and ambient storage pressure was kept below 4 hPa for both ask
types. The calculated changes for single and double O-ring asks are represented by the shaded and the solid
line, respectively.
δO 2 /N 2 im p ro ve m e n t fa cto r
80
60
40
20
0
0
60
120
180
240
S to ra g e tim e [d a ys]
300
360
Figure 3.11: Calculated ratio of single O-ring to double O-ring ask storage drifts (improvement factor)
for δO2 /N2 and double O-ring asks with a buer volume of 0.6 mL per valve as a function of time. This
improvement factor is nearly independent of ask volume and pressure gradients.
which corresponds to an improvement factor of only about 1.7 compared with the single Oring asks. Independent of the O-ring dimensions, the permeation coecients, and the size of
the buer volume, one would expect the double O-ring asks to perform at least a factor of 2
better with respect to permeation than the single O-ring asks because the overall thickness
of polymer for any gas to permeate is doubled with two O-rings. The reason why this is not
the case for the presented experiment is not yet known. Part of the explanation might be
the dierent absolute pressure gradients across the two O-rings, which causes dierent plastic
deformations of the two O-rings.
On the basis of theoretical considerations it is clear that the performance of double O-ring
3.5. CONCLUSIONS AND OUTLOOK
51
asks compared with single O-ring asks becomes less ecient the longer the storage period
is. Double O-ring asks with a buer volume of 0.6 mL per valve and O-ring dimensions of
d = 1.7 mm and A = 20 mm2 lead to a substantial improvement in δO2 /N2 stability only for
the rst few weeks of storage. The permeation inuence compared with single O-ring asks is
reduced after 1 month by a factor of 29 and after 1 year by a factor of only 3.5 (Figure 3.11).
The improvement factor is, to a good approximation, independent of ask volume and pressure
gradients.
Remarkably, a steady state is almost reached for CO2 and water vapor after only 1 week
because of the large CO2 and water vapor permeabilities. The buer volume is rapidly equilibrated with these gases, reducing the permeation inuence of sample air only by about a
factor of 2 compared with single O-ring valves. Even a 10 times larger buer volume would
result only in a moderate reduction of water vapor content of 6 % after a storage time of 1
year.
3.5 Conclusions and Outlook
Permeation of atmospheric gases through O-rings used to seal glass ask valves can substantially inuence sample air composition. Large storage drifts, which can impair atmospheric
δO2 /N2 studies, have been found for samples with signicant partial pressure dierences after
a storage period of up to 1 year. Likewise, changes of δAr/N2 , water vapor, and CO2 concentration occurred because of permeation uxes of these gases. Appropriate sampling and
storage strategies have to be chosen to ensure reliable ask measurements. Obviously, short
storage times improve the sample quality. Larger ask volumes can also help to attenuate
storage eects. However, for logistical reasons the ask volumes are however kept small (0.5
2.5 L). For O2 /N2 studies the asks must be lled preferentially to a pressure similar to the
ambient storage pressure in order to minimize permeation eects.
Flasks intended for CO2 measurements should be stored in an environment of known
CO2 concentration so that permeation eects can be well characterized and corrected for. In
practice, the most convenient way of storing asks in a CO2 -controlled environment is to store
them in a well-ventilated area where CO2 levels are always close to background.
The water vapor content increases for initially dried samples if they are stored at ambient
humidity. If water vapor aects the analysis, the sample air therefore has to be dried again
prior to the measurement after long storage periods.
The surface area and the thickness of the sealing material vary with dierent valve types.
Such geometric dierences can also result in signicant dierences in permeation rates. Where
samples must contain overpressure, e.g., for reasons concerning introduction of air samples to
analytical instruments, the double O-ring valves can considerably reduce permeation inuences, at least for storage times of up to a few weeks. Other methods for reducing permeation
eects are under investigation. One approach involves using seals constructed from Kel-F
(polychlorotriuoroethylene (PCTFE)). Compared with Viton, Kel-F has the advantage of
lower permeation coecients but the disadvantage of its greater hardness, making it more difcult to achieve reliable glass-polymer seals. Tests with Kel-F sealed glass asks are currently
being performed at the MPI-BGC, Jena, Germany.
52
3. PERMEATION OF GASES THROUGH POLYMER O-RINGS
Acknowledgments
This work was supported by the Swiss National Science Foundation and the EC Projects MILECLIM,
ALPCLIM and AEROCARB, an EC Project of the CARBOEUROPE cluster. We thank P. Nyfeler
for technical assistance, R. Francey and P. Steele for helpful discussions. Valuable comments were
provided by R. Keeling and an anonymous reviewer. We acknowledge the International Foundation
High Altitude Research Stations Jungfraujoch and Gornergrat (HFSJG), which made it possible to
take ask samples at Jungfraujoch.
Helium flux ( m3 s-1 STP)
Supplementary Figure
1x10-13
1x10-14
Viton, Louwers Hapert
PCTFE, QVF
Viton, Glass Expansion
1x10-15
0
10
20
30
40
Time (min)
50
60
70
80
Figure 3.12: Helium permeation through dierent valve types and sealing materials. Initially evacuated asks
were lled with helium to ∼ 950 hPa and the helium ux through the valves was measured by a helium leak
detector. The LH valves (Louwers Hapert, Netherlands) with Viton O-rings show the highest permeation rates
(crosses, equivalent to Figure 3.2). GE valves (Glass Expansion, Melbourne, Australia) with Viton O-rings
show a 410 times smaller permeation rate, depending on how tightly these valves are closed (circles). PCTFE
(Kel-F) sealed valves (QVF Labortechnik GmbH, Germany) have a helium permeation rate which is smaller
compared to LH valves but larger compared to GE valves (diamonds).
REFERENCES
53
References
Barrer, R. M. (1941), Diusion in and Through Solids, Cambridge Univ. Press, New York.
Beckmann, W. (1991), Gasdurchlässigkeit von Elastomeren, Kautsch. Gummi Kunstst., 44 (4), 323
329.
Bender, M., T. Ellis, P. Tans, R. Francey, and D. Lowe (1996), Variability in the O2 /N2 ratio of
Southern Hemisphere air, 1991-1994: Implications for the carbon cycle, Global Biogeochem. Cycles,
10 (1), 921.
Gemery, P. A., M. Trolier, and J. W. C. White (1996), Oxygen isotope exchange between carbon dioxide and water following atmospheric sampling using glass asks, J. Geophys. Res., 101 (D9), 14,415
14,420.
Holland, L., W. Steckelmacher, and J. Yarwood (1974), Vacuum manual, E. and F. N. Spon, London.
Keeling, C. D., R. B. Bacastow, A. F. Carter, S. C. Piper, T. P. Whorf, M. Heimann, W. G. Mook,
and H. Roelozen (1989), A three-dimensional model of atmospheric CO2 transport based on observed
winds: 1. Analysis of observational data, in Aspects of Climate Variability in the Pacic and the
Western Americas, Geophys. Monogr. Ser., vol. 55, edited by D. H. Peterson, pp. 165236, AGU,
Washington D.C.
Keeling, R., R. Najjar, M. Bender, and P. Tans (1993), What atmospheric oxygen measurements can
tell us about the global carbon cycle, Global Biogeochem. Cycles, 7 (1), 3767.
Keeling, R., A. Manning, E. McEvoy, and S. Shertz (1998), Methods for measuring changes in atmospheric O2 concentration and their application in Southern Hemisphere air, J. Geophys. Res., 103 (D3),
33813397.
Langenfelds, R. L. (2002), Studies of the global carbon cycle using atmospheric oxygen and associated
tracers, Ph.D. thesis, Univ. of Tasmania, Hobart, Tasmania, Australia.
Laurenson, L., and N. T. M. Dennis (1985), Permeability of common elastomers for gases over a range
of temperatures, J. Vac. Sci. Technol., A 3 (3), 17071710.
Ma, C., E. Shero, V. Nishith, S. L. Gilbert, and F. Shadman (1995), Permeation of Moisture and
Oxygen Through Polymeric O-Rings, J. IES, 38 (2), 4346.
O'Hanlon, J. F. (1989), A User's Guide to Vacuum Technology, 2nd ed., John Wiley, Hoboken, N. J.
Parker Hannin Corporation (2001), Parker O-Ring Handbook, Tech. Rep. ORD 5700, O-Ring Division, Lexington, KY.
Peacock, R. N. (1980), Practical selection of elastomer materials for vacuum seals, J. Vac. Sci. Technol.,
17 (1), 330336.
Perkins, W. G. (1973), Permeation and Outgassing of Vacuum Materials, J. Vac. Sci. Technol., 10 (4),
543556.
World Meteorological Organisation (2003), Report of the 11th WMO/IAEA Meeting of Experts on
Carbon Dioxide Concentration and Related Tracer Measurement Techniques, Tokyo, Japan, 2528
Sep. 2001, Tech. Rep. 148, World Meteorol. Organ. Global Atmos. Watch, Geneva, Switzerland.
Chapter 4
Atmospheric O2, CO2 and δ 13C
Measurements from the Remote Sites
Jungfraujoch, Switzerland, and Puy
de Dôme, France
P. Sturm1 , M. Leuenberger1 , and M. Schmidt2
Submitted to Journal of Geophysical Research, December 2004
Abstract
First atmospheric O2 , CO2 and δ 13 C ask measurements from the high altitude site Jungfraujoch, Switzerland (3580 m a.s.l., 46◦ 33'N, 7◦ 59'E), as well as O2 , CO2 and δ 13 C ask measurements from the mountain site Puy de Dôme, France (1480 m a.s.l., 45◦ 46'N, 2◦ 58'E) are
presented. The four-year records of Jungfraujoch and the three-year records of Puy de Dôme
show distinct seasonal cycles and superimposed long-term trends in the measured parameters. The mean peak-to-peak amplitudes of the respective seasonal cycles at Jungfraujoch are
79 per meg for O2 /N2 , 11 ppm for CO2 , and 0.45 % for δ 13 C, as derived from data tting. At
Puy de Dôme the seasonal variations are about two times larger than at Jungfraujoch. The
O2 and CO2 variations are in opposite phase at both sites. The spring time CO2 maximum
at Puy de Dôme appears in early march, whereas at Jungfraujoch it shows up 1 to 2 months
later. The O2 :CO2 correlation gives at both sites slopes of about 2 mol O2 /mol CO2 . Stable
carbon isotope ratios of source CO2 show depleted values in the wintertime and isotopically
enriched values in the summer.
1
Climate and Environmental Physics, Physics Institute, University of Bern, Sidlerstrasse 5, 3012 Bern,
Switzerland
2
Laboratiore des Science du Climat et de l'Environnement, UMR CEA-CNRS, CE Saclay, Orme des
Merisiers, 91191 Gif sur Yvette, France
56
4. O2 , CO2 AND δ 13 C FROM JUNGFRAUJOCH AND PUY DE DÔME
4.1 Introduction
The concentration of atmospheric carbon dioxide (CO2 ) is rising, primarily because of anthropogenic emissions from fossil fuel combustion and land use changes. This leads to a positive
radiative forcing of the climate and an expected global warming of surface temperatures.
However, less than about half of the CO2 emitted by burning of fossil fuel accumulates in
the atmosphere. The remaining amount of the CO2 is taken up by the oceans and the land
biosphere. Understanding the ocean-land partition and its temporal variation is essential for
reliable future projections of climate change.
Several dierent methods have been used to experimentally constrain this partition. One
approach involves measuring the rate of change of the O2 concentration in air. O2 and CO2
are inversely linked by photosynthesis, respiration, and combustion. However, because of their
dierent solubilities, atmospheric O2 is not directly aected by oceanic CO2 uptake. Thus,
precise measurements of atmospheric O2 , along with concurrent CO2 measurements, allow
to derive estimates of global oceanic and land biotic carbon sinks [Keeling and Shertz , 1992;
Keeling et al., 1993; Bender et al., 1994; Battle et al., 2000].
Another independent approach makes use of the isotopic composition of atmospheric CO2 ,
which contains information about the magnitude and distribution of sources and sinks of
carbon. Plants discriminate against 13 CO2 during photosynthesis by about -18 %, leaving
the atmosphere enriched in 13 CO2 , whereas only a few-per-mil fractionation occurs when CO2
is taken up or released by the surface ocean. Thus, δ 13 C of atmospheric CO2 can be used
to separate uxes of atmospheric CO2 between the terrestrial biosphere and the ocean [Ciais
et al., 1995; Battle et al., 2000].
A large number of globally distributed monitoring station are measuring atmospheric CO2
and other greenhouse gases, but to reduce uncertainties in net terrestrial carbon uxes a denser
observational network, including more continental sites [Gloor et al., 2000], and additional
tracers for source apportionment are needed. Atmospheric O2 measurements are still very
sparse, especially over the European continent. In the framework of the EU project Aerocarb
the high altitude site Jungfraujoch, Switzerland, was established as a ask sampling site for
O2 and CO2 measurements. Additionally, ask samples from the mountain site Puy de Dôme,
France, have been analyzed for O2 and CO2 concentration. We present results for O2 /N2 , CO2
and δ 13 C of a four-year measurement period at Jungfraujoch and a three-year record at Puy
de Dôme. Our ask sampling and analysis methods are described and preliminary conclusions
are discussed.
4.2 Sampling Sites and Methods
4.2.1
Sampling Locations
The high alpine research station Jungfraujoch (3580 m a.s.l., 46◦ 33'N, 7◦ 59'E) is located on
a mountain crest on the northern edge of the Swiss Alps. The station, which is surrounded
by rocks and glaciers, is prevalently situated in the free troposphere, but can be inuenced
by convection of atmospheric boundary layer air during the afternoon in the warmer months
[Baltensberger et al., 1997; Lugauer et al., 1998]. The saddle position of the Jungfraujoch,
between the Jungfrau (4158 m) and Mönch (4089 m) mountains, channels the local horizontal
wind in a north-western and south-eastern direction. With north-westerly wind directions, air
from the Swiss plateau is advected to the Jungfraujoch, while during south-easterly directions,
4.2. SAMPLING SITES AND METHODS
57
the air comes from the inner alpine area. The annual mean temperature is about −8 ◦ C, and
the mean pressure is 653 hPa.
Air sampling was also conducted west of the Alps at Puy de Dôme station (1480 m a.s.l.,
45◦ 46'N, 2◦ 58'E), located at the summit of Puy de Dôme in the center of France. The nearest
larger town Clermont-Ferrand (254,000 inhabitants) is about 15 km away. The site lies in
winter mainly in the free troposphere, despite its relatively low elevation [Sellegri et al., 2003].
It is inuenced by both Atlantic air masses and continental air from northern Europe. The
station is surrounded by meadows and during winter the ground is largely snow covered.
4.2.2 Flask Sampling
Flask sampling at Jungfraujoch commenced in October 2000 on a biweekly basis. The sampling procedure had to be adapted several times up to July 2002 because the air analysis
revealed some deciencies for precise O2 /N2 measurements. Initially duplicate whole-air samples were collected into 0.5 L glass asks, equipped with two Viton O-ring high-vacuum valves
at both ends (Louwers Hapert, Netherlands). The asks and the sampling unit were placed
outside on the platform of the research station. The asks were ushed for 15 min at a ow
rate of about 12 L min−1 with a pump (KNF Neuberger, Switzerland, N022AN.18 with Neoprene diaphragm) located downstream of the asks. The resulting sample pressure was equal
to ambient pressure at Jungfraujoch (about 650 hPa). From July 2002 the sample air was pressurized to about 950 hPa at a ow rate of about 4 L min−1 . The overpressure was regulated by
a modied KNF relief valve. The air was sucked through about 15 m of PVC tubing into the
research station, where the sampling unit with the asks were located at room temperature.
The sampling was carried out by the custodians of the research station and took place
between 06:30 and 07:30 in the morning in order to guarantee that the samples represent clean
background air. Drying of the air was performed with magnesium perchlorate (Mg(ClO4 )2 ),
placed in a U-shaped glass tube and sealed at each end by glass wool plugs. A 2 µm-lter
avoided entrainment of material inside the sampling unit.
Duplicate asks were sampled at Puy de Dôme on a weekly basis from July 2001. However, only one ask was sent back to the Physics Institute, University of Bern (PIUB) for
O2 /N2 analysis. The second ask was analyzed by Laboratiore des Sciences du Climat et
de l'Environnement, CE Saclay (LSCE) for CO2 and other trace gases [Pépin et al., 2001].
Until November 2001 the air was dried using a stainless steel cartridge lled with Mg(ClO4 )2 ,
afterwards the drying was performed cryogenically to a dew point of about −60 ◦ C. 0.5 L asks
with two Viton O-ring valves were ushed at a ow rate of 5 L min−1 for 10 min at ambient
pressure without pressurizing the sample air. Mean sample pressure is about 900 hPa. The
time of sampling varied between 07:30 and 16:00.
Before shipment to the sampling site, all asks were conditioned by lling them with dry
air to the prospective sample pressure.
4.2.3 O2 /N2 Analysis
We use an isotope ratio mass spectrometer (IRMS) for determining the oxygen content in an
air sample, following the method described by Bender et al. [1994]. With the DELTAplus XP
IRMS (Thermo Electron, Bremen, Germany) we can simultaneously monitor the molecular
ratios of the most abundant air constituents N2 (mass to charge ratio 28), O2 (m/z 32), Ar
(m/z 40) and CO2 (m/z 44).
4. O2 , CO2 AND δ 13 C FROM JUNGFRAUJOCH AND PUY DE DÔME
58
Atmospheric O2 is reported as the relative dierence in the ratio of O2 /N2 measured
against an arbitrary reference gas, and expressed in per meg units,
µ
δ (O2 /N2 ) =
(O2 /N2 )sample − (O2 /N2 )reference
(O2 /N2 )reference
¶
· 106
(permeg).
(4.1)
Because the mole fraction of O2 in dry atmospheric air is about 20.95 % [Machta and Hughes ,
1970], it follows that for small changes in mole fraction, a change of 1 part per million (ppm)
O2 is equivalent to 4.8 per meg in δ O2 /N2 .
To achieve the precision required in global carbon cycle studies for O2 /N2 measurements
(< 5 per meg), the conventional dual inlet system of the IRMS with metal bellows was modied.
We have developed a new inlet system, which is based on an open split design as known
from gas chromatography/mass spectrometry (GC/MS) systems. Sample and reference gases
are admitted to the mass spectrometer via glass capillaries. A glass open split housed in a
temperature and pressure controlled box serves as a change-over device for the sample and
reference gas. A detailed description of our gas inlet system is given by Leuenberger et al.
[2000a] and Sturm [2001]. Samples are analyzed relative to a working gas, which in turn is
calibrated against reference gases from three high pressure cylinders. Flask intercomparison
was also performed between PIUB and the Centrum voor IsotopenOnderzoek, Groningen,
(CIO). All results are reported on the local PIUB scale, based on cylinder LK560962.
4.2.4
CO2 Analysis
The CO2 mixing ratio was inferred from δ CO2 /N2 (m/z 44/28 ratio) measured by mass
spectrometry, simultaneously with δ O2 /N2 . The production of N2 O from excited N2 and O2
molecules or fragments of those in the ion source of the IRMS yields a signal for m/z 44, which
is in the same magnitude as the m/z 44 signal originating from the CO2 of the air sample
[Leuenberger et al., 2000b]. This large m/z 44 background varies mainly depending on the
time that passed since the last vacuum breakdown of the ion source occurred. Hence, it was
determined at the end of each measurement day using zero CO2 air. With the known CO2
mixing ratio of the working standard of 356.45 ppm and the measured background correction
one can then convert the δ CO2 /N2 into a CO2 mixing ratio.
CO2 mixing ratios are reported in the WMO CO2 mole fraction scale. Primary standards
from NOAA/CCGG, Boulder, CO, USA, are used to calibrate the working and secondary
standards. The accuracy of the CO2 mixing ratio data is estimated to be about ±0.5 ppm.
4.2.5
δ 13 C Analysis
The δ 13 C of CO2 was determined with GC/MS and a syringe method, which uses only about
1 mL STP of air. This technique is adapted from δ 13 C analysis of very small air amounts
extracted from ice as it was developed in our laboratory [Leuenberger et al., 2003]. The
air sample is injected in a helium stream where a preconcentration system separates the
CO2 cryogenically from the air. The small CO2 amount is then released into a low helium
stream which forces the CO2 via an open split device to a DELTAplus XL mass spectrometer.
The precision is about ±0.1 %, compared to ±0.02 % using a conventional CO2 extraction.
However, our principle is fast and we can do several replicates. A N2 O correction of −0.23 %
is applied to the data. Carbon isotopic compositions are expressed on VPDB-CO2 scale.
4.2. SAMPLING SITES AND METHODS
59
4.2.6 Data Selection and Drift Corrections
Atmospheric ask samples taken at Jungfraujoch at ambient pressure (∼ 650 hPa) during the
rst year and stored in the laboratory in Bern (∼ 950 hPa) showed a strong dependence of the
measured O2 /N2 ratio on the storage time. This correlation was attributed to permeation of
gases through the Viton O-rings of the asks' valves [Sturm et al., 2004b]. The O2 /N2 storage
drift was estimated to be 1.19 ± 0.04 per meg/day and an average correction of 158 per meg
(ranging from 2 to 455 per meg), corresponding to an average storage time of 130 days, was
applied to the data up to July 2002. The eect of permeation and thus the applied correction
clearly exceeded any real atmospheric O2 /N2 variations (see Figure 4.1). This large and
simplied correction gives O2 /N2 data of somewhat minor value compared to the pressurized
samples, which have been taken afterwards to prevent this problem. The corrected data
of duplicate asks partly show large pair dierences. 8 data in this time period with pair
dierences of > 50 per meg were rejected from further analysis. Additionally 18 data were
rejected as outliers because of sampling or measurement error indicated by unusual values of
the measured parameters. After July 2002 measurements on duplicate asks revealed on 3
occasions a pair dierence > 15 per meg and these results were also rejected.
For CO2 an estimated storage drift of 0.0053 ppm/day was applied to the CO2 ask data
up to July 2002, which resulted in an average correction of 0.69 ppm (ranging from 0.01 to
2.02 ppm). CO2 measurements on duplicate asks were rejected if the pair dierence was
> 1 ppm (11 data up to July 2002 and 3 data from July 2002 to January 2003). Permeation
can also aect the isotopic composition. However, because it is dicult to accurately quantify
the permeation fractionation of δ 13 C, no storage correction is applied and the δ 13 C data up
to July 2002 are rejected from further analysis.
The mean pair dierence of the remaining samples after July 2002 is 3.9 per meg for O2 /N2 ,
0.18 ppm for CO2 and 0.07 % for δ 13 C. These pair reproducibilities are compatible with the
precision of the individual measurements.
The quality of the ask samples from Puy de Dôme was on some occasions aected by
insucient drying of the sample air. Therefore, 12 wet samples had to be rejected. Comparison
of the PIUB CO2 measurements with the LSCE measurements showed in 10 cases a dierence
of > 1 ppm. These samples together with 3 contaminated samples, which could be detected by
highly elevated CO2 mixing ratios, were considered as outliers. The mean dierence between
the remaining CO2 measurements of PIUB and LSCE is −0.15±0.50 ppm. The scatter of these
dierences is about a factor of two larger compared to ask samples from Grin Forest, UK
[Sturm et al., 2004a], which are analyzed in both laboratories in an analogous manner. This
discrepancy may be due to sampling problems encountered at Puy de Dôme or storage related
eects, which are expected to be larger for Puy de Dôme samples, because of their generally
longer storage times. Still, considering the measurement precision of the mass spectrometric
method, the accuracy of the CO2 data is regarded as satisfactory.
Storage drift corrections of 0.18 per meg/day and 0.0025 ppm/day were applied to all O2 /N2
and CO2 ask data, respectively, resulting in average corrections of 6.8 per meg (ranging from
0.2 to 47.3 per meg) for O2 /N2 and 0.11 ppm (ranging from 0.00 to 0.42 ppm) for CO2 .
The selected data were tted by a cubic smoothing spline for the long term trend and
three annual harmonics for seasonal variations [Nakazawa et al., 1997].
4. O2 , CO2 AND δ 13 C FROM JUNGFRAUJOCH AND PUY DE DÔME
60
800
700
Jungfraujoch
δO 2 /N2 (per meg)
600
500
400
300
300
200
390
390
370
-7.5
-8.0
-8.5
-9.0
-9.5
APO (per meg)
300
200
0
370
360
-8.0
-8.5
-9.0
-9.5
400
100
380
-7.5
δ13 C (‰)
δ13C (‰)
400
100
360
APO (per meg)
500
200
380
Puy de Dôme
600
CO 2 (ppm)
CO2 (ppm)
δO2/N2 (per meg)
700
(a)
2000
2001
2002
2003
2004
400
300
200
100
0
(b)
2000
2001
2002
2003
2004
Figure 4.1: (a) O2 /N2 ratio, CO2 mixing ratio, δ 13 C of CO2 , and atmospheric potential oxygen (APO) from
Jungfraujoch (3580 m a.s.l., 46◦ 33'N, 7◦ 59'E). Air samples were inuenced by storage drift (open circles) up to
July 2002. Grey closed circles show the corrected values. Solid lines are harmonic t curves through the data.
Dashed lines represent long term trends. δ O2 /N2 results are reported on the local PIUB scale. (b) Same as
(a) but for Puy de Dôme (1480 m a.s.l., 45◦ 46'N, 2◦ 58'E).
4.3 Results and Discussions
4.3.1
Seasonal Variability of Atmospheric O2 /N2 and CO2
Figure 4.1 shows the O2 /N2 ratios, CO2 mixing ratios, δ 13 C of CO2 , and atmospheric potential
oxygen (APO, dened below) together with the tted curves through the data for Jungfraujoch
from mid-2000 to mid-2004 and for Puy de Dôme from mid-2001 to mid-2004. Regular seasonal
cycles are observed at both sites and in all measured components. Fossil fuel combustion results
at both sites in the decreasing trend observed in O2 /N2 ratios, and the increasing trend in
CO2 (dashed lines in Figure 4.1).
The mean peak-to-peak amplitudes derived from the t at Jungfraujoch were 79 per meg
for δ O2 /N2 and 11.0 ppm for CO2 . The annual mean CO2 mixing ratio in 2003 was 375.0 ppm.
The average seasonal cycle of δ O2 /N2 reached a minimum in April and a maximum in September. The CO2 mixing ratio varies in opposite phase with the maxima at the end of April,
followed by a fast decrease towards the minimum at the beginning of September. These ampli-
4.3. RESULTS AND DISCUSSIONS
61
tudes are comparable to ask measurements from aircraft sampling above Grin Forest, UK,
where the seasonal cycle at 3100 m a.s.l. reveals an peak-to-peak amplitude of 113 per meg for
δ O2 /N2 and 11.4 ppm for CO2 [Sturm et al., 2004a].
At Puy de Dôme the mean peak-to-peak amplitudes were 190 per meg for δ O2 /N2 and
18.2 ppm for CO2 . The average seasonal cycle of O2 /N2 reached a minimum in early March
and a maximum in late August, again with the CO2 mixing ratio varying seasonally in opposite
phase with O2 /N2 . The annual mean CO2 mixing ratio derived from these data was 377.3 ppm
in 2002 and 379.5 ppm in 2003. The spring time CO2 maximum and δ O2 /N2 minimum seem
to occur 1 to 2 months earlier at Puy de Dôme than at Jungfraujoch.
Atmospheric potential oxygen is a useful tracer to distinguish between land and ocean
processes aecting the air masses at the sampling sites [Stephens et al., 1998]. If we neglect
minor inuences by CH4 and CO oxidation, APO can be dened as
APO = δO2 /N2 +
1.1
∆CO2
0.2095
(permeg),
(4.2)
where ∆CO2 is the dierence in the CO2 mixing ratio of the sample from an arbitrary reference
gas, in ppm, the factor 1/0.2095 converts CO2 from mixing ratio (ppm) to per meg units, and
−1.1 is the O2 :CO2 exchange ratio for land photosynthesis and respiration [Severinghaus ,
1995]. APO is invariant to land biotic exchanges and variations in APO can be caused only
by air-sea exchange of CO2 , O2 and N2 and by combustion of fossil fuels. An O2 :CO2 exchange
ratio of approximately −1.4 is generally assumed for fossil fuel combustion [Manning , 2001].
The curve ts through the APO records, consisting of two annual harmonics for the seasonal variation, reveal peak-to-peak amplitudes of 34 and 97 per meg at Jungfraujoch and Puy
de Dôme, respectively. Thus, at both sites about half of the seasonal cycle of δ O2 /N2 can be
attributed to air-sea exchange uxes.
A comparison of O2 /N2 ratios, CO2 mixing ratios, δ 13 C and APO at the two sites is shown
in Figure 4.2. The remoteness and the higher altitude of the Jungfraujoch compared to Puy
de Dôme results in smaller seasonal amplitudes. However, a part of the larger amplitude at
Puy de Dôme is also due to the fact that the time of ask sampling was not restricted to the
early morning hours. In summer, the CO2 mixing ratio can be signicantly lowered during
the day because of enhanced net ecosystem production. In winter months, the advection of
air from the planetary boundary layer with large amounts of fossil fuel CO2 can also lead to
diurnal variations. Air sampling during the day can therefore increase the apparent seasonal
amplitude, compared to air samples only taken in the early morning. A selection of the data
according to sampling time, wind direction and wind speed is not applicable, because of the
limited seasonal data coverage of the current records.
In Figure 4.3 the detrended O2 /N2 ratios are plotted against the detrended CO2 mixing
ratios. The slopes calculated by geometric mean regression and expressed in units of mol
O2 /mol CO2 are −2.1 ± 0.2 (R2 = 0.62) for Jungfraujoch and −2.2 ± 0.2 (R2 = 0.71) for
Puy de Dôme. These values are rather large compared to the land biota O2 :CO2 exchange
ratio of −1.1 and the globally averaged fossil fuel O2 :CO2 combustion ratio of −1.4. This
points again to a strong oceanic component contributing to the seasonal cycle of the δ O2 /N2
even at the continental sites Jungfraujoch and Puy de Dôme. Also, there seems to be at both
sites a seasonal dierence in the observed slopes. In autumn and winter when the δ O2 /N2 is
decreasing, the molar O2 :CO2 ratios are −2.4 ± 0.2 (R2 = 0.72) and −2.3 ± 0.2 (R2 = 0.69)
at Jungfraujoch and Puy de Dôme, respectively. When δ O2 /N2 is increasing during spring
and summer the O2 :CO2 correlation yields a respective slope of −1.1 ± 0.2 (R2 = 0.56) and
62
4. O2 , CO2 AND δ 13 C FROM JUNGFRAUJOCH AND PUY DE DÔME
CO2 (ppm)
δO2/N2 (per meg)
500
400
300
200
100
(a)
390
380
370
360
(b)
δ13C (‰)
-7.5
-8.0
-8.5
-9.0
APO (per meg)
-9.5
(c)
400
300
200
100
0
(d)
2000
2001
2002
2003
2004
Figure 4.2: Comparison of (a) O2 /N2 ratio, (b) CO2 mixing ratio, (c) δ 13 C of CO2 , and (d) atmospheric
potential oxygen (APO) at Jungfraujoch (circles, solid line) and Puy de Dôme (crosses, dashed line). The lines
are harmonic t curves through the data.
−1.9 ± 0.2 (R2 = 0.77). This could be explained by the enhanced inuence of the land
biosphere compared to oceanic inuences on δ O2 /N2 in spring and summertime.
4.3.2
δ 13 C of Atmospheric CO2
The seasonal variations of δ 13 C are in opposite phase with CO2 , with depleted values in
summer and autumn and enriched values in winter and spring. The mean peak-to-peak
amplitudes are 0.45 % at Jungfraujoch and 1.00 % at Puy de Dôme. No signicant long-term
trends can be seen in these records. From the two-component mixing approach according to
Keeling [1958, 1961] we calculated the δ 13 C signature of the source CO2 . In Figure 4.4
the correlation of detrended δ 13 C and the detrended inverse CO2 mixing ratio is shown.
The mean carbon isotope ratios of source CO2 obtained by geometric mean regression are
−25.2 ± 1.5 % (R2 = 0.81) at Jungfraujoch and −28.8 ± 0.8 % (R2 = 0.90) at Puy de
Dôme. The larger standard error of the intercept at Jungfraujoch is due to the smaller CO2
range compared to Puy de Dôme. The mean value at Jungfraujoch of −25.2 % is comparable
with the source signature of −24.8 ± 1.4 % observed for free tropospheric air from aircraft
sampling over Orléans, France, by Levin et al. [2002]. The δ 13 C source signatures also showed a
seasonal dierence with isotopically depleted values in autumn and winter of −29.6±2.4 % and
4.3. RESULTS AND DISCUSSIONS
63
500
δO 2/N2 (per meg)
400
300
200
100
Jungfraujoch
Puy de Dôme
0
360
370
380
CO2 (ppm)
390
400
Figure 4.3: Detrended O2 /N2 versus CO2 mixing ratio at Jungfraujoch (solid line) and Puy de Dôme (dashed
line). The geometric mean regression slopes for O2 changes expressed in units equivalent to CO2 yield −2.1 ±
0.2 mol O2 /mol CO2 for Jungfraujoch and −2.2 ± 0.2 mol O2 /mol CO2 for Puy de Dôme.
δ13C of CO2 (‰)
-7.5
Jungfraujoch
Puy de Dôme
-8.0
-8.5
-9.0
-9.5
0.0025
0.0026
0.0027
1/CO2 (ppm-1)
0.0028
Figure 4.4: Keeling plot from the detrended data of the δ 13 C of CO2 and the inverse CO2 mixing ratio.
The mean δ 13 C source signatures are −25.2 ± 1.5 % at Jungfraujoch (solid line) and −28.8 ± 0.8 % at Puy
de Dôme (dashed line).
−31.4 ± 1.1 % and enriched values in spring and summer of −24.4 ± 1.9 % and −26.0 ± 0.6 %
at Jungfraujoch and Puy de Dôme, respectively. In wintertime, when the CO2 source is likely
dominated by fossil fuel combustion, the isotopic signal of the source is more depleted than
in summer, when biological processes dominate. Similar seasonal dierences in the order of
46 % have also been found at other continental mid-northern latitude sites [Bakwin et al.,
1995; Levin et al., 2002; Ramonet et al., 2002; Demény and Haszpra , 2002].
64
4. O2 , CO2 AND δ 13 C FROM JUNGFRAUJOCH AND PUY DE DÔME
4.4 Summary and Conclusions
We present O2 /N2 , CO2 and δ 13 C records from two remote sites in central Europe, Jungfraujoch and Puy de Dôme, for a period of four and three years, respectively. After we became
aware of storage eects being crucial for highly precise O2 /N2 ask measurements [Sturm
et al., 2004b], we improved the sampling and analysis methods and storage drift corrections
were applied to the older data. Distinct seasonal cycles and superimposed long-term trends
are observed in all measured parameters. O2 :CO2 correlations and carbon isotope ratios show
at both sites seasonal variations in the sources of CO2 . The atmospheric O2 /N2 measurements
provide new information, which will help to better infer regional sources and sinks in models.
Carbon ux estimates based on O2 measurements are most robust when averaged over
many years, because of inter-annual variations of oceanic uxes. Longer observational records
of δ O2 /N2 are therefore needed to provide reliable constraints on the European carbon budgets.
Flask sampling at Jungfraujoch and Puy de Dôme will continue within the framework of the
succeeding EU project CarboEurope-IP. Additionally, recently installed O2 and CO2 analyzers
at the high-altitude site Jungfraujoch will provide continuous records of O2 and CO2 and give
new insights in source apportionment of atmospheric CO2 .
Acknowledgments
This work was supported by the Swiss National Science Foundation, in particular the R'equip program,
and the EU projects AEROCARB, MILECLIM and ALPCLIM. We thank P. Nyfeler for technical assistance. We acknowledge the International Foundation High Altitude Research Stations Jungfraujoch
and Gornergrat (HFSJG), which made it possible to take ask samples at Jungfraujoch. We also
thank the custodians of the research station Jungfraujoch for their support.
REFERENCES
65
References
Bakwin, P. S., P. P. Tans, C. Zhao, W. Ussler, and E. Quesnell (1995), Measurements of carbon dioxide
on a very tall tower, Tellus, 47B (5), 535549.
Baltensberger, U., H. W. Gäggeler, D. T. Jost, M. Lugauer, M. Schwikowski, and E. Weingartner (1997), Aerosol climatology at the high-alpine site Jungfraujoch, Switzerland, J. Geophys. Res.,
102 (D16), 19,70719,715.
Battle, M., M. L. Bender, P. P. Tans, J. W. C. White, J. T. Ellis, T. Conway, and R. J. Francey
(2000), Global Carbon Sinks and Their Variability Inferred from Atmospheric O2 and δ 13 C, Science,
287, 24672470.
Bender, M. L., P. P. Tans, T. J. Ellis, J. Orchardo, and K. Habfast (1994), A high precision isotope
ratio mass spectrometry method for measuring the O2 /N2 ratio of air, Geochim. Cosmochim. Acta,
58 (21), 47514758.
Ciais, P., P. P. Tans, J. W. C. White, M. Trolier, R. J. Francey, J. A. Berry, D. R. Randall, P. J. Sellers,
J. G. Collatz, and D. S. Schimel (1995), Partitioning of ocean and land uptake of CO2 as inferred
by δ 13 C measurements from the NOAA Climate Monitoring and Diagnostics Laboratory Global Air
Sampling Network, J. Geophys. Res., 100 (D3), 50515070.
Demény, A., and L. Haszpra (2002), Stable isotope composition of CO2 in background air and at
polluted sites in hungary, Rapid Commun. Mass Spectrom., 16 (797-804).
Gloor, M., S.-M. Fan, S. Pacala, and J. Sarmiento (2000), Optimal sampling of the atmosphere for
purpose of inverse modelling: a model study, Global Biogeochem. Cycles, 14 (1), 407428.
Keeling, C. D. (1958), The concentration and isotopic abundances of atmospheric carbon dioxide in
rural areas, Geochim. Cosmochim. Acta, 13, 322334.
Keeling, C. D. (1961), The concentration and isotopic abundances of carbon dioxide in rural and
marine air, Geochim. Cosmochim. Acta, 24, 277298.
Keeling, R., and S. Shertz (1992), Seasonal and interannual variations in atmospheric oxygen and
implications for the global carbon cycle, Nature, 358, 723727.
Keeling, R., R. Najjar, M. Bender, and P. Tans (1993), What atmospheric oxygen measurements can
tell us about the global carbon cycle, Global Biogeochem. Cycles, 7 (1), 3767.
Leuenberger, M., P. Nyfeler, H. Moret, P. Sturm, and C. Huber (2000a), A new gas inlet system for
an isotope ratio mass spectrometer improves reproducibility, Rapid Commun. Mass Spectrom., 14,
15431551.
Leuenberger, M., P. Nyfeler, H. Moret, P. Sturm, A. Indermühle, and J. Schwander (2000b), CO2
concentration measurements on air samples by mass spectrometry, Rapid Commun. Mass Spectrom.,
14, 15521557.
Leuenberger, M., M. Eyer, P. Nyfeler, B. Stauer, and T. F. Stocker (2003), High-resolution δ 13 C
measurements on ancient air extracted from less than 10 cm3 of ice, Tellus, 55B (2), 138144.
Levin, I., P. Ciais, R. Langenfelds, M. Schmidt, M. Ramonet, K. Sidorov, N. Tchebakova, M. Gloor,
M. Heimann, E.-D. Schultze, N. N. Vygodskaya, O. Shibistova, and J. Lloyd (2002), Three years of
trace gas observations over the EuroSiberian domain derived from aircraft sampling a concerted
action, Tellus, 54B (5), 696712.
66
4. O2 , CO2 AND δ 13 C FROM JUNGFRAUJOCH AND PUY DE DÔME
Lugauer, M., U. Baltensberger, M. Furger, H. W. Gäggeler, D. T. Jost, M. Schwikowski, and H. Wanner
(1998), Aerosol transport to the high Alpine sites Jungfraujoch (3454 m a.s.l.) and Colle Gnifetti
(4452 m a.s.l.), Tellus, 50B (1), 7692.
Machta, L., and E. Hughes (1970), Atmospheric Oxygen in 1967 to 1970, Science, 168, 15821584.
Manning, A. C. (2001), Temporal variability of atmospheric oxygen from both continuous measurements and a ask sampling network: Tools for studying the global carbon cycle, Ph.D. thesis, University of California, San Diego, California, U.S.A.
Nakazawa, T., M. Ishizawa, K. Higuchi, and N. Trivett (1997), Two curve tting methods applied to
CO2 ask data, Environmetrics, 2, 197218.
Pépin, L., M. Schmidt, M. Ramonet, D. E. J. Worhty, and P. Ciais (2001), A new Gas Chromatographic
Experiment to Analyze Greenhouse Gases in Flask Samples and in Ambient Air in the Region of Saclay,
Notes des Activités Instrumentales, Institut Pierre-Simon Laplace, http://www.ipsl.jussieu.fr.
Ramonet, M., P. Ciais, I. Nepomniachii, K. Sidorov, R. E. M. Neubert, U. Langendörfer, D. Picard,
V. Kazan, S. Biraud, M. Gusti, O. Kolle, E.-D. Schultze, and J. Lloyd (2002), Three years of aircraftbased trace gas measurements over the Fyodorovskoye southern taiga forest, 300 km north-west of
Moscow, Tellus, 54B (5), 713734.
Sellegri, K., P. Laj, F. Peron, R. Dupuy, M. Legrand, S. Preunkert, J.-P. Putaud, H. Cachier, and
G. Ghermandi (2003), Mass balance of free tropospheric aerosol at the Puy de Dôme (France) in
winter, J. Geophys. Res., 108 (D11), 4333, doi:10.1029/2002JD002747.
Severinghaus, J. P. (1995), Studies of the terrestrial O2 and carbon cycles in sand dune gases and in
Biosphere 2, Ph.D. thesis, Columbia University, New York, U.S.A.
Stephens, B. B., R. F. Keeling, M. Heimann, K. D. Six, R. Murnane, and K. Caldeira (1998), Testing
global ocean carbon cycle models using measurements of atmospheric O2 and CO2 concentration,
Global Biogeochem. Cycles, 12 (2), 213230.
Sturm, P. (2001), Entwicklung eines neuen Einlasssystems für die massenspektrometrische Messung
des O2 /N2 Verhältnisses, Master's thesis, Physics Institute, University of Bern, Bern, Switzerland.
Sturm, P., M. Leuenberger, J. Moncrie, and M. Ramonet (2004a), Atmospheric O2 , CO2 and δ 13 C
observations from aircraft sampling over Grin Forest, Perthshire, UK, J. Geophys. Res., submitted.
Sturm, P., M. Leuenberger, C. Sirignano, R. E. M. Neubert, H. A. J. Meijer, R. Langenfelds, W. A.
Brand, and Y. Tohjima (2004b), Permeation of atmospheric gases through polymer O-rings used in
asks for air sampling, J. Geophys. Res., 109, D04309, doi:10.1029/2003JD004073.
Chapter 5
Atmospheric O2, CO2 and δ 13C
Observations from Aircraft Sampling
over Grin Forest, Perthshire, UK
P. Sturm1 , M. Leuenberger1 , J. Moncrie2 , and M. Ramonet3
Submitted to Journal of Geophysical Research, December 2004
Abstract
Regular vertical aircraft sampling have been performed in the lower troposphere above Grin
Forest, near Aberfeldy, Perthshire, UK, (56◦ 37'N, 3◦ 47'W) between February 2003 and May
2004 for analysis of O2 /N2 , CO2 and δ 13 C of CO2 . We sampled asks between 800 m and
3100 m a.s.l. The peak-to-peak amplitude of the seasonal cycle of O2 /N2 decreases from
171 per meg at 800 m to 113 per meg at 3100 m. Furthermore the seasonal cycle is shifted
from low to high altitudes with a lag of about 1 month. The same features are observed for
CO2 with a decrease in the peak-to-peak amplitude of the seasonal cycle from 17.6 ppm at
800 m to 11.4 ppm at 3100 m. The vertical proles show decreasing O2 /N2 ratios in summer
and increasing O2 /N2 ratios in wintertime with increasing sampling height, due to surface
exchange of oxygen with the land biosphere. The O2 :CO2 exchange ratios for selected vertical
proles vary between −1.5 and −2.4 mol O2 /mol CO2 .
1
Climate and Environmental Physics, Physics Institute, University of Bern, Sidlerstrasse 5, 3012 Bern,
Switzerland
2
School of GeoSciences, The University of Edinburgh, Mayeld Road, Edinburgh EH9 3JU, United Kingdom
3
Laboratiore des Science du Climat et de l'Environnement, UMR CEA-CNRS, CE Saclay, Orme des
Merisiers, 91191 Gif sur Yvette, France
5. O2 , CO2 AND δ 13 C OVER GRIFFIN FOREST
68
5.1 Introduction
The atmospheric burden of carbon dioxide (CO2 ) is increasing as a result of anthropogenic
emissions from fossil fuel combustion and land use changes. However, only about half of the
CO2 emitted by burning of fossil fuel accumulates in the atmosphere. The remaining amount
of the CO2 is taken up by the oceans and the land biosphere. Determining this ocean-land
partition and its temporal variation is essential for understanding ecosystem carbon storage
and ocean uptake of carbon dioxide. Several dierent methods have been used to experimentally constrain this partition. One approach involves measuring changes of the oxygen
(O2 ) concentration in air. O2 and CO2 are inversely linked by photosynthesis, respiration,
and combustion. However, because of their dierent solubilities, atmospheric O2 is decoupled
from CO2 for processes related to the oceanic carbon cycle. Thus, precise measurements of
atmospheric O2 , along with concurrent CO2 measurements, can be used for estimates of global
oceanic and land biotic carbon sinks [Keeling and Shertz , 1992; Keeling et al., 1993; Bender
et al., 1994; Battle et al., 2000].
Another independent approach makes use of the isotopic composition of atmospheric CO2 ,
which contains information about the magnitude and distribution of sources and sinks of
carbon. During terrestrial photosynthesis the uptake of carbon favors the lighter 12 CO2 against
13 CO , whereas the discrimination between the two isotopes in oceanic carbon uptake is very
2
small [Ciais et al., 1995]. Thus, δ 13 C of atmospheric CO2 can also be used to separate uxes
of atmospheric CO2 between the terrestrial biosphere and the ocean.
An increasing number of global sampling networks are measuring CO2 and associated
tracers. However, uncertainties in estimating the terrestrial and oceanic carbon uptake are
still partly caused by insucient data coverage. The network of observing stations can be
improved particularly by vertical proles in addition to ground based monitoring stations
[Gloor et al., 2000]. Several measurement programs of vertical gradients of CO2 and other
trace gases exist, but there are only very few measurements of vertical O2 /N2 gradients so far
[Langenfelds , 2002].
In the framework of the EU project AEROCARB O2 /N2 ask sampling of vertical proles
over Grin Forest, Perthshire, UK, started in 2003. The samples have been analyzed for
O2 /N2 ratio, CO2 mixing ratio and δ 13 C of CO2 at the Physics Institute, University of Bern
(PIUB). We describe the ask sampling and analysis methods and report on seasonal variations
in the vertical structure of O2 /N2 and CO2 . Twelve vertical proles of O2 /N2 and CO2 are
presented and discussed together with δ 13 C of CO2 of three selected proles.
5.2 Sampling Sites and Methods
5.2.1
Flight Location
The sampling location is at 56◦ 37'N, 3◦ 47'W close to the town of Aberfeldy in Perthshire,
UK. The exact location is overhead Grin Forest, a CarboEuroFlux site since 1997. Forest
cover at Grin is typical for this part of Scotland (about 17 % of the country is forested
and at Grin, about 75 % of the land surface is under the canopy). Grin is 97.3 % Sitka
Spruce (Picea sitchensis), 2.1 % Douglas Fir (Pseudotsuga menziesii) and 0.6 % birch (Betula
pendula). The remaining 25 % of Grin is open space: roads, streams and ridges, dominated
by heather and grasses. The forest leaf area index (LAI) is about 8, and where the canopy is
closed, the understorey is quite spares and dominated by mosses. The canopy height is about
5.2. SAMPLING SITES AND METHODS
69
8.5 m. The site is at an elevation of 350 m a.s.l. and the regional topography is dominated by
the valley of the River Tay. The Grin catchment has varied relief, extending from 50 m a.s.l.
to 600 m a.s.l.; the forest is in a broad northwest facing valley with a mean slope of about 9 %.
The countryside around Grin Forest is typical of much of the central highlands of Scotland
with a mixture of land use types from forestry to heather moorland, rough grazing for sheep
and many lochs of varying size. Small towns and villages are the norm and the nearest large
city is Glasgow, some 100 km to the south-west.
5.2.2 Aircraft and Sampling Protocol
The aircraft usually employed for the air sampling campaigns is a Piper Arrow IV (G-WEND)
operated by Tayside Aviation and own from Perth Airport at Scone. The inlet tube for the
ask-based sampling system is a 6 mm Dekabon tube which is extended about 15 cm out the
pilot's storm window via a specially made adapter with holes for tube access. The tube inlet is
located suciently far from the cockpit window that it is outside the inuence of the boundary
layer attached to the aircraft skin and thus away from any possible exhaust inuence from the
pull-propeller on this type of aircraft. The inlet tube is connected directly to the sampling
unit which sits on the passenger seat in the rear of the aircraft. As part of the AEROCARB
project, ights are scheduled to occur about every 20 days; in practice, they are dependent
on suitable weather conditions for ying and for maximizing the chance that all sampling
levels can be own. An element of bias can creep in that the majority of ights will have
been done in fair-weather, probably anticyclonic conditions. Figure 5.1 shows the general
location of Grin in relation to the rest of the UK and the back trajectories for a selection
of the dates reported in this paper. The back trajectories have been calculated using the
HYSPLIT model [Draxler and Hess , 1997] as modied by Zahorowski [2003]. According to an
analysis of the Jenkinson Daily Weather Types [Salmon , 1998] for this part of the UK, about
2/3rds of the time the circulation types come from a relatively restricted range of circulation
type viz anticyclonic, cyclonic, linear south-westerly, westerly, southerly and north-westerly, in
decreasing magnitude. Figure 5.2 shows the seasonal variability in daily back trajectories for
the years 20022003 inclusive, based on the Grin location; the relative scarcity of trajectories
from the European continent over this part of the UK is noticeable, particularly in the summer
months.
5.2.3 Flask Sampling
After takeo from Perth Airport, the aircraft is own directly to Grin Forest in a steady
climb to 10,000 feet and this can take up to 30 minutes depending on outside air temperature
since that aects the rate of climb on aircraft such as the Piper Arrow. Upon reaching the
rst sampling height, a pair of asks are ushed at about 3.5 L min−1 for 5 minutes and then
sampled for one minute to an absolute pressure of 1000 hPa. The asks are then sealed and the
aircraft descends to the next sampling level and the process is repeated. The air samples were
collected into preconditioned 0.5 L glass asks, equipped with two Viton O-ring high-vacuum
valves at both ends (Louwers Hapert, Netherlands). A schematic diagram of the sampling
unit is shown in Figure 5.3. Drying of the air was performed with magnesium perchlorate
(Mg(ClO4 )2 ), placed in a U-shaped glass tube and sealed at each end by glass wool plugs.
The asks were ushed with a pump (KNF Neuberger, Switzerland, N86KNDC with EPDM
diaphragm) located upstream of the asks and the overpressure was regulated by a modied
70
5. O2 , CO2 AND δ 13 C OVER GRIFFIN FOREST
Figure 5.1: Back trajectories for selected ights discussed in this paper and based on the Grin eld site
(Key: red = 800 m; yellow = 1600 m; light blue = 2100 m; dark blue = 3100 m).
KNF relief valve, which was adapted so that its working pressure is independent of the ambient
air pressure. A sample pressure which is constant at all sampling heights and similar to the
pressure where the asks are stored before analysis is required for precise O2 /N2 measurements
to prevent storage related artifacts like permeation of gases through the asks' valves [Sturm
et al., 2004b]. Eight 0.5 L asks for the purpose of O2 /N2 analysis could be transported in
the aircraft. From February 2003 to February 2004 duplicate ask samples were taken at the
four dierent altitudes of 3100, 2100, 1600 and 800 m a.s.l. Because the pair dierences of
these samples were with very few exceptions within the measurement precision, we abandoned
duplicate sampling at all levels in favor of more height levels. From March 2004 ask samples
were taken at six dierent heights (3100, 2600, 2100, 1600, 1100, 800 m a.s.l.), with duplicate
asks at the lowest and highest level and single asks at the four levels in-between.
5.2. SAMPLING SITES AND METHODS
71
Figure 5.2: Seasonal variability in daily back trajectories (calculated at local noon) for the years 20022003
inclusive and based on Grin Forest.
5.2.4 Flask Analysis
Atmospheric O2 is reported in units of per meg (where 1 per meg = 0.001 %) according to
δO2 /N2 = [(O2 /N2 )sample /(O2 /N2 )reference − 1] × 106 . Because the mole fraction of O2 in dry
atmospheric air is about 20.95 % [Machta and Hughes , 1970], it follows that a change of 1 part
per million (ppm) O2 is equivalent to 4.8 per meg in δO2 /N2 .
We use an isotope ratio mass spectrometer (IRMS) for determining the oxygen content in
an air sample, following the method described by Bender et al. [1994]. To achieve the precision
required in global carbon cycle studies for O2 /N2 measurements (< 5 per meg), the conventional dual inlet system of the IRMS with metal bellows was modied. We have developed
a new inlet system, which is based on an open split design as known from gas chromatography/mass spectrometry (GC/MS) systems [Leuenberger et al., 2000a; Sturm , 2001]. Samples
are analyzed relative to a working gas, which in turn is calibrated against reference gases
from four high pressure cylinders. All results are reported on our local PIUB scale, based on
cylinder LK560962. For our samples storage drift is expected to be negligible and no storage
drift correction was applied to the data.
At PIUB the CO2 mixing ratio was inferred from δ CO2 /N2 measured by mass spectrometry, simultaneously with δO2 /N2 . The production of N2 O from excited N2 and O2 molecules
or fragments of those in the ion source of the IRMS yields a signal for the mass to charge ratio
44, which is in the same magnitude as the m/z 44 signal originating from the CO2 of the air
sample [Leuenberger et al., 2000b]. This large m/z 44 background varies mainly depending
on the time that passed since the last vacuum breakdown of the ion source occurred. Hence,
it was determined at the end of each measurement day using zero CO2 air. With the known
5. O2 , CO2 AND δ 13 C OVER GRIFFIN FOREST
72
Air
Intake
Drying Unit
Toggle
Valve
Flush
Pump
Flasks
Flush
Flowmeter
Relief
Valve
Toggle
Valve
Press.
Gauge
Figure 5.3: Schematic diagram of the sampling unit used for vertical prole sampling.
CO2 mixing ratio of the working standard and the measured background correction one can
then convert the δ CO2 /N2 into a CO2 mixing ratio. CO2 mixing ratios are reported on the
WMO CO2 mole fraction scale. Primary standards from NOAA/CCGG, Boulder, CO, USA,
are used to calibrate the working and secondary standards. The accuracy of the CO2 data is
estimated to be better than ±0.5 ppm. Flask samples have also been collected for Laboratiore
des Sciences du Climat et de l'Environnement (LSCE) simultaneously with the ask samples
for PIUB. At LSCE these ask samples were analyzed among other species for CO2 mixing
ratio by gas chromatography [Pépin et al., 2001]. Figure 5.4 shows the dierence of the CO2
measurements between PIUB and LSCE for eight ights in 2003 as a function of the CO2
mixing ratio measured by PIUB. The dierences in the CO2 mixing ratio of the samples from
24 June 2003 were remarkably larger than for the other dates. The cause of this discrepancy
is unknown. The mean dierence between the remaining CO2 measurements of PIUB and
LSCE is −0.15 ± 0.23 ppm, which is satisfactory considering the measurement precision of the
mass spectrometric method.
The δ 13 C of CO2 was determined with GC/MS and a syringe method, which uses only
about 1 mL STP of air. This technique is adapted from δ 13 C analysis of very small air amounts
extracted from ice as it was developed at PIUB [Leuenberger et al., 2003]. The air sample is
injected in a helium stream where a preconcentration system separates the CO2 cryogenically
from the air. The small CO2 amount is then released into a low helium stream which forces
the CO2 via an open split device to the mass spectrometer. The precision is about ±0.1 %,
compared to ±0.02% using a conventional CO2 extraction. However, our principle is fast and
we can do several replicates. A N2 O correction of −0.23 % is applied to the data. Carbon
isotopic compositions are expressed on VPDB-CO2 scale.
The mean pair dierence of the duplicate samples is 5.1 per meg for O2 /N2 , 0.12 ppm for
CO2 and 0.06 % for δ 13 C. These pair reproducibilities are compatible with the precision of
the individual measurements.
CO2 difference PIUB – LSCE (ppm)
5.3. RESULTS AND DISCUSSIONS
73
2
1
0
-1
-2
364
368
372
376
CO2 mixing ratio (ppm)
380
Figure 5.4: Dierence of CO2 mixing ratio measurements between Physics Institute, University of Bern
(PIUB) and Laboratiore des Sciences du Climat et de l'Environnement (LSCE) for eight ights in 2003. The
open circles denote the samples of 24 June 2003. The mean dierence of the remaining CO2 measurements
(black dots) is −0.15 ± 0.23 ppm
5.3 Results and Discussions
The O2 /N2 ratio and CO2 mixing ratio data from aircraft sampling at Grin are shown in
Figure 5.5. All sampling altitudes between 800 and 3100 m a.s.l. are included. The CO2
data measured by LSCE extend the CO2 record to three complete years. Smoothed time
series of δO2 /N2 and CO2 were constructed by tting a quadratic polynomial and two annual
harmonics to the data. The seasonal cycle observed in δO2 /N2 reached a minimum in the
middle of April and a maximum in the middle of August, while the CO2 varied seasonally
in opposite phase with δO2 /N2 . Thus, the gradual decrease during autumn and winter is
followed by a faster increase between April and August for δO2 /N2 and vice versa for CO2 .
The peak-to-peak amplitudes derived from the t were 113 per meg for δO2 /N2 and 13.3 ppm
for CO2 .
In Figure 5.6 the data are sorted by the dierent height levels. During summer ights,
the lower levels may lie within the atmospheric boundary layer (ABL), whereas the 3100 m
level represents as a rst order approach the large-scale background situation of the lower
troposphere above the ABL. The tted curves show the seasonal cycle amplitudes of δO2 /N2
(upper panel) and CO2 (middle panel) for four altitudes (800, 1600, 2100 and 3100 m a.s.l.).
These seasonal amplitudes decrease monotonically from about 17.6 ppm at 800 m a.s.l. to
about 11.4 ppm at 3100 m a.s.l. for CO2 and from 171 to 113 per meg for δO2 /N2 , respectively
(Table 5.1). Remarkably, the O2 :CO2 ratios of the respective amplitudes are nearly constant
for all levels with values between −2.02 and −2.10 mol O2 /mol CO2 . There is also a discernible
phase shift of the seasonal cycle of δO2 /N2 and CO2 from low to high altitudes with a lag
of about 1 month between 800 and 3100 m, in accordance with ndings of Sidorov et al.
[2002]. These features are qualitatively consistent with the seasonal cycle being forced mainly
by oceanic and land biotic surface exchange and damping of the seasonal amplitude as it is
5. O2 , CO2 AND δ 13 C OVER GRIFFIN FOREST
CO2 (ppm)
δO2/N2 (per meg)
74
350
300
250
200
385
380
375
370
365
360
2001
2002
2003
2004
Figure 5.5: O2 /N2 ratio (upper panel) and CO2 mixing ratio (lower panel) from aircraft sampling at Grin
(56◦ 37'N, 3◦ 47'W) measured by PIUB (lled circles) and LSCE (open circles). Measurements from all sampling
altitudes between 800 and 3100 m a.s.l. are shown. The lines are harmonic t curves through the data.
Height
(m a.s.l.)
peak-to-peak amplitude
CO2
O2 /N2
APO
(ppm) (per meg) (per meg)
800
1600
2100
3100
17.6
17.0
13.9
11.2
171
167
140
113
73
68
62
58
Table 5.1: Peak-to-peak amplitudes of the seasonal cycle of CO2 , O2 /N2 and APO for dierent sampling
heights. The amplitudes are derived from the harmonic t curves.
propagated vertically into the troposphere.
To distinguish between land and ocean processes aecting the air masses over Grin, we
computed the atmospheric potential oxygen (APO) [Stephens et al., 1998], dened as
APO = δO2 /N2 +
1.1
∆CO2
0.2095
(permeg).
(5.1)
∆CO2 is the dierence in the CO2 mixing ratio of the sample from an arbitrary reference gas,
in ppm, the factor 1/0.2095 converts CO2 from mixing ratio (ppm) to per meg units and −1.1
is the O2 :CO2 exchange ratio for land biotic photosynthesis and respiration [Severinghaus ,
1995]. If we neglect minor inuences by CH4 and CO oxidation, variations in APO can then
be caused only by air-sea exchange of CO2 , O2 and N2 and by combustion of fossil fuels. An
O2 :CO2 exchange ratio of approximately −1.4 is generally assumed for fossil fuel combustion
[Manning , 2001]. The lower panel of Figure 5.6 shows the APO for the dierent height levels.
5.3. RESULTS AND DISCUSSIONS
75
δO 2/N2 (per meg)
800m
1100m
1600m
2100m
2600m
3100m
800m
1600m
2100m
3100m
350
300
250
200
CO 2 (ppm)
385
380
375
370
APO (per meg)
365
350
300
250
200
2003
2004
Figure 5.6: O2 /N2 ratio (upper panel), CO2 mixing ratio (middle panel) and atmospheric potential oxygen
(APO, lower panel) for the dierent sampling heights. Solid lines are harmonic t curves through the respective
data.
The peak-to-peak amplitudes of the seasonal cycle of APO decrease from about 73 per meg at
800 m to about 58 per meg at 3100 m (Table 5.1). Based on this calculation about half of the
seasonal cycle of δO2 /N2 can be attributed to land biogenic exchange (δO2 /N2 minus APO).
The magnitude of these APO amplitudes agrees well with other maritime stations at similar
latitudes, for example Cold Bay, Alaska, at 55◦ N which shows an APO peak-to-peak amplitude
of about 75 per meg [Keeling et al., 1998]. According to the harmonic ts the maxima of APO
lag the δO2 /N2 maxima by about 23 weeks, whereas no consistent picture can be seen for
the minima, most probably due to the limited seasonal data coverage.
The twelve vertical proles from February 2003 to May 2004 are shown in Figure 5.7.
Summertime δO2 /N2 proles reveal decreasing O2 /N2 ratios with increasing sampling height,
reecting O2 net sources at the ground from photosynthesis. In winter, δO2 /N2 is observed
to increase with altitude due to a net O2 sink from soil respiration and fossil fuel combustion.
On 15 February 2003 and 12 February 2004 the O2 /N2 ratio reaches the maximum between
1600 and 2100 m a.s.l. with again a depleted O2 /N2 value at the highest sampling level. The
5. O2 , CO2 AND δ 13 C OVER GRIFFIN FOREST
76
15 Feb 03
06 Mar 03
24 Jun 03
15 May 03
Altitude (m a.s.l.)
δO 2 /N2 (per meg)
δO2 /N2 (per meg)
δO 2 /N2 (per meg)
δO2 /N2 (per meg)
260 280 300 320 240 260 280 300 240 260 280 300 220 240 260 280 300
3500
3000
2500
2000
1500
1000
500
374
376 378
CO2 (ppm)
380 376
04 Aug 03
382 376
378 380
CO 2 (ppm)
378 380
CO2 (ppm)
382 370 372 374 376 378
CO 2 (ppm)
02 Oct 03
14 Aug 03
Altitude (m a.s.l.)
δO2 /N2 (per meg)
δO2 /N2 (per meg)
δO2 /N2 (per meg)
300 320 340 360 380 320 340 360 380 280 300 320 340
3500
δO 2 /N2 (per meg)
280 300 320
3000
2500
2000
1500
1000
500
364 366 368 370 372
CO 2 (ppm)
364 366 368
CO 2 (ppm)
11 Nov 03
δO 2 /N2 (per meg)
240 260 280 300
3500
Altitude (m a.s.l.)
25 Oct 03
12 Feb 04
δO 2 /N2 (per meg)
220 240 260
370 372 374
CO 2 (ppm)
372 374 376
CO2 (ppm)
29 Mar 04
27 May 04
δO2 /N2 (per meg)
δO 2 /N2 (per meg)
160 180 200 220 240 220 240 260 280
3000
2500
2000
1500
1000
500
374 376 378
CO2 (ppm)
376
378 380
CO2 (ppm)
382
380 382 384 386
CO 2 (ppm)
374
376 378
CO2 (ppm)
380
Figure 5.7: Twelve vertical proles of δO2 /N2 (black dots) and CO2 mixing ratio (grey squares) from ask
samples.
CO2 proles show the same pattern with gradients opposite to δO2 /N2 .
In Figure 5.8 the O2 /N2 ratios are plotted against the CO2 mixing ratio for seven proles
with a vertical CO2 gradient of > 1.5 ppm. The slopes calculated by geometric mean regression
are given in Table 5.2. It has to be remembered that these regression slopes are calculated
based on only 4 to 6 data points and are therefore not very well constrained. However, they
are all larger than the land biota O2 :CO2 exchange ratio of −1.1 mol O2 /mol CO2 and the
globally averaged fossil fuel O2 :CO2 combustion ratio of −1.4 mol O2 /mol CO2 . This points
again to a strong oceanic inuence on the vertical O2 /N2 variations or natural gas being the
5.3. RESULTS AND DISCUSSIONS
77
24 Jun 03
04 Aug 03
320
300
280
260
374 376 378 380
CO 2 (ppm)
300
280
260
240
360
340
320
300
364 366 368 370 372
CO 2 (ppm)
29 Mar 04
27 May 04
380
360
340
320
362 364 366 368 370
CO 2 (ppm)
δO 2/N2 (per meg)
260
240
220
δO 2/N2 (per meg)
240
280
δO 2/N2 (per meg)
400
220
370 372 374 376 378
CO 2 (ppm)
12 Feb 04
14 Aug 03
380
δO 2/N2 (per meg)
δO 2/N2 (per meg)
δO 2/N2 (per meg)
320
δO 2/N2 (per meg)
15 Feb 03
220
200
180
160
200
376 378 380 382
CO 2 (ppm)
380 382 384 386
CO 2 (ppm)
280
260
240
220
374 376 378 380
CO 2 (ppm)
Figure 5.8: The O2 /N2 ratios versus the CO2 mixing ratio for seven proles with a vertical CO2 gradient of
> 1.5 ppm. The slopes calculated by geometric mean regression and expressed in units of mol O2 /mol CO2 are
listed in Table 5.2. The molar O2 :CO2 exchange ratios of −1.1 (dashed line) and −1.4 (dashed-dotted line)
are shown for clearness in the last diagram of 27 May 2004.
Flight date
O2 :CO2 ratio
(mol O2 /mol CO2 )
15 Feb 03
24 Jun 03
04 Aug 03
14 Aug 03
12 Feb 04
29 Mar 04
27 May 04
1.8 ± 0.5
1.8 ± 0.3
1.5 ± 0.2
2.2 ± 0.3
2.4 ± 0.3
2.1 ± 0.1
2.1 ± 0.1
Table 5.2: O2 :CO2 ratio of the seven vertical proles shown in Figure 5.8.
predominant fuel type, with a molar O2 :CO2 ratio of about −2.0. The seasonal covariance of
O2 and CO2 gives −1.9 ± 0.1 mol O2 /mol CO2 , which is comparable with the mean seasonal
O2 :CO2 ratio at Jungfraujoch, Switzerland, and Puy de Dôme, France [Sturm et al., 2004a],
indicating that this value is representative on a large, probably continental scale.
The δ 13 C of CO2 was determined for selected vertical proles with CO2 gradients large
enough to resolve δ 13 C variations. Generally, the δ 13 C observations mirrored the CO2 mixing
ratios, with values at the lower altitude levels being enriched compared to the higher levels
in summer (24 June 2003 and 04 August 2003) and depleted with respect to the top level in
winter/spring (29 March 2004). The δ 13 C signature of source CO2 was calculated from the
two-component mixing approach according to Keeling [1958, 1961]. The correlation of δ 13 C
and the inverse CO2 mixing ratio is shown in Figure 5.9. The mean carbon isotope ratios
5. O2 , CO2 AND δ 13 C OVER GRIFFIN FOREST
78
24 Jun 03
δ13C (‰)
-7.8
-8.0
-8.2
-8.4
0.00264
0.00268
1/CO2 (ppm-1)
0.00272
04 Aug 03
δ13C (‰)
-7.5
-7.7
-7.9
-8.1
0.00270
0.00274
1/CO2 (ppm-1)
29 Mar 04
δ13C (‰)
-8.1
-8.3
-8.5
-8.7
0.00258
0.00262
1/CO2 (ppm-1)
0.00266
Figure 5.9: Keeling plots from δ 13 C of CO2 and the inverse CO2 mixing ratio for three selected proles.
The δ 13 C source signatures are −20.6 ± 2.4 % (24 Jun 03), −29.1 ± 5.9 % (04 Aug 03) and −30.8 ± 3.5 %
(29 Mar 04).
of source CO2 obtained by geometric mean regression are −20.6 ± 2.4 % on 24 June 2003,
−29.1 ± 5.9 % on 04 August 03 and −30.8 ± 3.5 % on 29 March 2004. The rather large
uncertainties of these values are partly due to the small CO2 range covered by a given vertical
prole and are comparable with results of similar vertical prole studies [Lloyd et al., 2001;
Ramonet et al., 2002]. Furthermore, the source signatures likely represent weighted values
from dierent sources (ecosystem exchange, anthropogenic emissions, air-sea uxes), resulting
in a larger scatter in the Keeling plots. If the air samples from the dierent altitude levels
originate from the atmospheric boundary layer as well as from the free troposphere, the basic
5.4. SUMMARY AND CONCLUSIONS
79
assumption of this analysis, namely a single CO2 source or sink, which is mixed to a constant
background concentration, may then not hold true.
5.4 Summary and Conclusions
Twelve vertical aircraft proles in the lower troposphere from ask sampling were carried
out over Grin Forest, Perthshire, UK between February 2003 and May 2004. The data
show the amplitude of the seasonal cycle in O2 /N2 to be about 171 per meg at 800 m and
about 113 per meg at 3100 m. The seasonal cycle amplitude of the CO2 mixing ratio decreases from about 17.6 ppm at 800 m to 11.4 ppm at 3100 m. The vertical structure of O2 /N2
shows gradients of up to 60 per meg with decreasing concentrations in summer and increasing concentrations in wintertime, and concurrent variations in CO2 and δ 13 C. The O2 :CO2
exchange ratio of the vertical prole measurements indicates a strong oceanic inuence on vertical O2 /N2 variations. This is further supported by trajectory analyses suggesting that the
primary origin of air masses is over the North Atlantic. Atmospheric O2 /N2 proles provide
an important additional piece of information, which will help to better infer the ocean versus
land partitioning of carbon uxes in models. To successfully use such data in model inversions
in order to derive carbon ux estimates, high demands on measurement accuracy as well as
on inter-laboratory comparability are made. A recently initiated international intercomparison program for O2 /N2 measurements strives to merge the existing O2 /N2 calibration scales
at dierent measurement laboratories. Additionally, much longer observational records are
needed to characterize the inter-annual variability of carbon uxes. At Grin Forest, vertical
prole sampling for O2 /N2 and CO2 will continue in the framework of the succeeding EU
project CarboEurope-IP.
Acknowledgments
This work was supported by the Swiss National Science Foundation, in particular the R'equip program,
and the EU projects AEROCARB, MILECLIM and ALPCLIM. We thank P. Nyfeler for technical
assistance. J. M. acknowledges the nancial support of the AEROCARB project funded under the
EU Framework V programme Contract No. EVK2-1999-00013 and we also acknowledge the assistance
given by the CFI at Tayside Aviation, Mr Sandy Torrance. We thank Max Priestman for performing
the trajectory analyses as part of his BSc Undergraduate Project in Ecological Science, 2004.
80
5. O2 , CO2 AND δ 13 C OVER GRIFFIN FOREST
References
Battle, M., M. L. Bender, P. P. Tans, J. W. C. White, J. T. Ellis, T. Conway, and R. J. Francey
(2000), Global Carbon Sinks and Their Variability Inferred from Atmospheric O2 and δ 13 C, Science,
287, 24672470.
Bender, M. L., P. P. Tans, T. J. Ellis, J. Orchardo, and K. Habfast (1994), A high precision isotope
ratio mass spectrometry method for measuring the O2 /N2 ratio of air, Geochim. Cosmochim. Acta,
58 (21), 47514758.
Ciais, P., P. P. Tans, J. W. C. White, M. Trolier, R. J. Francey, J. A. Berry, D. R. Randall, P. J. Sellers,
J. G. Collatz, and D. S. Schimel (1995), Partitioning of ocean and land uptake of CO2 as inferred
by δ 13 C measurements from the NOAA Climate Monitoring and Diagnostics Laboratory Global Air
Sampling Network, J. Geophys. Res., 100 (D3), 50515070.
Draxler, R. R., and G. D. Hess (1997), Description of the Hysplit-4 modelling system, Tech. Rep. ERL
ARL-224, NOAA Tech. Mem., 24p.
Gloor, M., S.-M. Fan, S. Pacala, and J. Sarmiento (2000), Optimal sampling of the atmosphere for
purpose of inverse modelling: a model study, Global Biogeochem. Cycles, 14 (1), 407428.
Keeling, C. D. (1958), The concentration and isotopic abundances of atmospheric carbon dioxide in
rural areas, Geochim. Cosmochim. Acta, 13, 322334.
Keeling, C. D. (1961), The concentration and isotopic abundances of carbon dioxide in rural and
marine air, Geochim. Cosmochim. Acta, 24, 277298.
Keeling, R., and S. Shertz (1992), Seasonal and interannual variations in atmospheric oxygen and
implications for the global carbon cycle, Nature, 358, 723727.
Keeling, R., R. Najjar, M. Bender, and P. Tans (1993), What atmospheric oxygen measurements can
tell us about the global carbon cycle, Global Biogeochem. Cycles, 7 (1), 3767.
Keeling, R. F., B. B. Stephens, R. G. Najjar, S. C. Doney, D. Archer, and M. Heimann (1998), Seasonal
variations in the atmospheric O2 /N2 ratio in relation to the kinetics of air-sea gas exchange, Global
Biogeochem. Cycles, 12 (1), 141163.
Langenfelds, R. L. (2002), Studies of the global carbon cycle using atmospheric oxygen and associated
tracers, Ph.D. thesis, Univ. of Tasmania, Hobart, Tasmania, Australia.
Leuenberger, M., P. Nyfeler, H. Moret, P. Sturm, and C. Huber (2000a), A new gas inlet system for
an isotope ratio mass spectrometer improves reproducibility, Rapid Commun. Mass Spectrom., 14,
15431551.
Leuenberger, M., P. Nyfeler, H. Moret, P. Sturm, A. Indermühle, and J. Schwander (2000b), CO2
concentration measurements on air samples by mass spectrometry, Rapid Commun. Mass Spectrom.,
14, 15521557.
Leuenberger, M., M. Eyer, P. Nyfeler, B. Stauer, and T. F. Stocker (2003), High-resolution δ 13 C
measurements on ancient air extracted from less than 10 cm3 of ice, Tellus, 55B (2), 138144.
Lloyd, J., R. J. Francey, D. Mollicone, M. R. Raupach, A. Sogachev, A. Arneth, J. N. Byers, F. M.
Kelliher, C. Rebmann, R. Valentini, S.-C. Wong, G. Bauer, and E.-D. Schultze (2001), Vertical proles,
boundary layer budgets, and regional ux estimates of CO2 and its 13 C/12 C ratio and for water vapor
above a forest/bog mosaic in central Siberia, Global Biogeochem. Cycles, 15 (2), 267284.
Machta, L., and E. Hughes (1970), Atmospheric Oxygen in 1967 to 1970, Science, 168, 15821584.
REFERENCES
81
Manning, A. C. (2001), Temporal variability of atmospheric oxygen from both continuous measurements and a ask sampling network: Tools for studying the global carbon cycle, Ph.D. thesis, University of California, San Diego, California, U.S.A.
Pépin, L., M. Schmidt, M. Ramonet, D. E. J. Worhty, and P. Ciais (2001), A new Gas Chromatographic
Experiment to Analyze Greenhouse Gases in Flask Samples and in Ambient Air in the Region of Saclay,
Notes des Activités Instrumentales, Institut Pierre-Simon Laplace, http://www.ipsl.jussieu.fr.
Ramonet, M., P. Ciais, I. Nepomniachii, K. Sidorov, R. E. M. Neubert, U. Langendörfer, D. Picard,
V. Kazan, S. Biraud, M. Gusti, O. Kolle, E.-D. Schultze, and J. Lloyd (2002), Three years of aircraftbased trace gas measurements over the Fyodorovskoye southern taiga forest, 300 km north-west of
Moscow, Tellus, 54B (5), 713734.
Salmon, M. (1998), Trajectory data from the Climatic Research Unit (CRU), University of East Anglia,
http://www.cru.uea.ac.uk/cru/data/lwt.htm, (accessed 27th March 2004).
Severinghaus, J. P. (1995), Studies of the terrestrial O2 and carbon cycles in sand dune gases and in
Biosphere 2, Ph.D. thesis, Columbia University, New York, U.S.A.
Sidorov, K., A. Sogachev, U. Langendörfer, J. Lloyd, I. L. Nepomniachii, N. N. Vygodskaya,
M. Schmidt, and I. Levin (2002), Seasonal variability of greenhouse gases in the lower troposphere
above the eastern European taiga (Syktyvkar, Russia), Tellus, 54B (5), 735748.
Stephens, B. B., R. F. Keeling, M. Heimann, K. D. Six, R. Murnane, and K. Caldeira (1998), Testing
global ocean carbon cycle models using measurements of atmospheric O2 and CO2 concentration,
Global Biogeochem. Cycles, 12 (2), 213230.
Sturm, P. (2001), Entwicklung eines neuen Einlasssystems für die massenspektrometrische Messung
des O2 /N2 Verhältnisses, Master's thesis, Physics Institute, University of Bern, Bern, Switzerland.
Sturm, P., M. Leuenberger, and M. Schmidt (2004a), Atmospheric O2 , CO2 and δ 13 C measurements
from the remote sites Jungfraujoch, Switzerland, and Puy de Dôme, France, J. Geophys. Res., submitted.
Sturm, P., M. Leuenberger, C. Sirignano, R. E. M. Neubert, H. A. J. Meijer, R. Langenfelds, W. A.
Brand, and Y. Tohjima (2004b), Permeation of atmospheric gases through polymer O-rings used in
asks for air sampling, J. Geophys. Res., 109, D04309, doi:10.1029/2003JD004073.
Zahorowski, W. (2003), Trajectory, batch processing of backward trajectories. User's manual, ANSTO,
Australia.
Chapter 6
Continuous Observations of CO2,
222Rn, O /N , Ar/N and Stable
2
2
2
Isotopes of CO2, O2, N2 and Ar at
Bern, Switzerland
Abstract
A one-year time series of continuous atmospheric CO2 measurements from Bern, Switzerland
is presented. O2 /N2 and Ar/N2 ratios as well as stable carbon and oxygen isotopes of CO2
and δ 29 N2 , δ 34 O2 and δ 36 Ar were measured periodically in a continuous way during a one
year period. Additionally, the 222 Rn activity was measured during three months in the winter
2004. Using the correlation from short term uctuations of CO2 and 222 Rn we estimated
a mean CO2 ux density between February 2004 and April 2004 in the region of Bern of
90 ± 40 tC km−2 month−1 . The continuous observations of carbon dioxide and associated tracers shed light on diurnal and seasonal patterns of the carbon cycle in an urban atmosphere.
There is considerable variance in nighttime δ 13 C and δ 18 O of source CO2 throughout the
year, however, with generally lower values in winter compared to summertime. The O2 :CO2
oxidation ratio during the nighttime build-up of CO2 varies between −0.96 and −1.54. Furthermore, Ar/N2 measurements showed that artifacts like thermal fractionation at the air
intake are relevant for high precision measurements of atmospheric O2 .
6.1 Introduction
We present continuous records of atmospheric CO2 and associated tracers measured in the
city of Bern, Switzerland. At this urban site, anthropogenic CO2 emissions (e.g. car exhausts,
heating) mix with the background and biogenic CO2 components, which are inuenced by
photosynthesis and respiration. Local sources and sinks in the catchment area of Bern and
changing meteorological conditions are expected to lead to large short-term variations of the
observed tracers. In order to be able to interpret and apportion these observations highresolution measurements of multiple tracers are needed.
Stable carbon and oxygen isotopes in atmospheric CO2 can be used to measure the size
of the CO2 uxes and to discriminate between the various processes in the carbon cycle.
84
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
Photosynthetic uptake of CO2 , plant and soil respiration, and fossil fuel burning lead to
carbon and oxygen isotope signals of atmospheric CO2 , which can be used as a tracer at
various temporal and spatial scales [Friedli et al., 1987; Keeling et al., 1989; Ciais et al.,
1997a]. Hesterberg [1990] has measured δ 13 C and δ 18 O of CO2 in Bern during a one-year
period in 1989. His measurements of ask samples collected twice a week showed seasonal
dierences in the δ 13 C and δ 18 O signals.
Another approach involves measuring changes of the atmospheric O2 concentration. O2
and CO2 are inversely coupled by photosynthesis, respiration and combustion. However, the
dierent processes have dierent O2 :CO2 exchange ratios and thus can be distinguished from
each other. The high precision of O2 measurements that is necessary to constrain such carbon
uxes, has given new insights into gas handling procedures and fractionation eects [Bender
et al., 1994; Keeling et al., 1998; Langenfelds , 2002]. Though continuous on-line measurements circumvent any storage related eects as observed in some ask sampling programs
(see Chapter 3), they are still susceptible to diusive fractionation processes. Fractionation
of O2 /N2 at the intake as well as at tees have rst been observed by Manning [2001] and
have been attributed to thermal diusion. Yet, the exact cause of the fractionation and the
points at which it can occur in the ow path of air are still not well established. Thermal
diusion results from temperature gradients. Heavier molecules generally accumulate in the
colder region hence leading to concentration changes [Severinghaus et al., 1996; Chapman and
Cowling , 1970].
Radon-222 is a radioactive noble gas with a half-life T1/2 of 3.82 days. It is produced in
all soils as part of the natural uranium-radium α-decay series and it emanates into the soil air
and diuses to the atmosphere where it is diluted by atmospheric transport and radioactive
decay. The 222 Rn ux from ocean surfaces is about two orders of magnitude smaller than
from continents [Wilkening and Clements , 1975]. Because 222 Rn emissions from soils turned
out to be rather homogeneous in a restricted region and relatively constant in time, 222 Rn is
a useful tracer to parameterize transport and dilution in the atmospheric boundary layer.
This chapter rst summarizes sampling and analysis techniques and reports on tests we
have performed to assess fractionation eects at the air intake. Results for CO2 , its stable isotopes, O2 /N2 and 222 Rn over roughly a one-year period are described and possible mechanisms
for these observations are discussed.
6.2 Sampling and Analysis Techniques
6.2.1
Sampling Site
The city of Bern (about 127'000 inhabitants) is situated on the Swiss Plateau. The measurements were made at the Physics Institute, University of Bern (PIUB), which is located on the
eastern edge and about 20 m above the city center. The building is surrounded by residential
and urban areas. The air is collected from the roof of the building (46◦ 57'04N, 7◦ 26'20E,
575 m a.s.l.) about 15 m above local ground. Meteorological measurements were made by the
Institute for Applied Physics, University of Bern. The weather station is located at the same
height about 10 m away from the air intake. The sample air is sucked through a ∼ 40 m long
6 mm outer diameter Dekabon tube into our laboratory. We use a diaphragm pump (KNF
Neuberger, Switzerland, N86KNDC with EPDM diaphragm). The ow rate is between 100
and 300 mL min−1 depending on which instruments are connected to the air stream. The air is
dried cryogenically at −70 ◦ C. The total volume from the air intake to the analyzers of about
750 mL restricts the time resolution of the measurements to about 5 min.
6.2. SAMPLING AND ANALYSIS TECHNIQUES
6.2.2
85
CO2 Mixing Ratio
The CO2 mixing ratio was measured by non-dispersive infrared adsorption (NDIR) technique.
In the beginning the CO2 measurements were performed by a S710 UNOR CO2 analyzer
(SICK MAIHACK GmbH, Germany). From March 2004 a LI-7000 CO2 /H2 O analyzer (LICOR, USA), was used. The ow rate of the sample gas is about 100 mL min−1 and every
minute the mean CO2 mixing ratio is recorded. The CO2 data are reported on the WMO
CO2 mole fraction scale. Primary standards from NOAA/CCGG, Boulder, CO, USA, are
used to calibrate the working and secondary standards. However, the CO2 mixing ratio of
these primary standards lies in the range of 192 to 363 ppm. The calibration of our CO2
scale above 363 ppm is therefore based on extrapolation. Still, the accuracy of the CO2 data
is estimated to be better than ±0.5 ppm for mixing ratios below 400 ppm and better than
±1 ppm for 400450 ppm.
6.2.3
δ 13 C and δ 18 O of CO2
The carbon and oxygen isotopes of CO2 were determined with GC/MS in a semi-continuous
way. Every 12 min an air parcel of about 0.5 mL STP was cryogenically trapped in a glass
capillary. The small air amount is then released into a low helium stream (1 mL min−1 ). This
gas stream is additionally split into three similar uxes entering capillaries of dierent lengths.
A multi-port valve handles the ow path such that the three gas portions are injected one
after another via an open split device to an isotope ratio mass spectrometer (DELTAplus XL,
Thermo Electron, Bremen, Germany), where the m/z ratios 45/44 and 46/44 of the CO2 are
measured [Leuenberger et al., 2003]. The precision of this method estimated by the pooled
standard deviations of the triplicate measurements is about ±0.10 % for δ 13 C and ±0.13 %
for δ 18 O. A N2 O correction of −0.23 % is applied to the δ 13 C data. Carbon and oxygen
isotopic compositions are expressed on VPDB-CO2 scale.
δ 18 O measurements from ask sampling often face additional experimental problems, due
to the risk of isotopic exchange of CO2 with water, that may occur anywhere in the sample
treatment from the moment of sampling until the input of the sample in the mass spectrometer [Gemery et al., 1996]. For example, isotope exchange during ask storage of CO2 with
water, that permeates through the ask seals (see Chapter 3), interferes with any real atmospheric signal. Our ask measurements of δ 18 O from Jungfraujoch, Puy de Dôme and Grin
(Chapter 4 and 5) are therefore believed not to represent the true isotopic composition of atmospheric CO2 . Therefore, the advantage of the continuous analysis method used here, apart
from the high time resolution, is that such storage related eects can largely be circumvent.
6.2.4 Elemental and Isotopic Ratios of Air
The elemental ratios δO2 /N2 and δAr/N2 as well as the isotopic ratios δ 29 N2 , δ 34 O2 and
δ 36 Ar are analyzed by an isotope ratio mass spectrometer (DELTAplus XP, Thermo Electron,
Bremen, Germany) and expressed in the δ -notation as per meg deviation from our local PIUB
reference gas. A glass capillary at a tee takes about 0.2 mL min−1 of the ow to the gas
inlet system [Leuenberger et al., 2000; Sturm , 2001] of the IRMS. One measurement comprises
eight standard/sample cycles and takes about 12 min. Hence, one data point represents a
mean concentration of the last 12 min.
86
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
δAr/N2 (per meg)
-1520
-1560
-1600
-1640
-1680
8
12
16
20
24
28-Apr-04
30-Apr-04
2-May-04
Temperature (°C)
4
4-May-04
Figure 6.1: Diurnal variations of δAr/N2 (top) and outdoor temperature (bottom). Note the inverted axis
of the temperature.
6.2.5
222
Rn activity
The specic 222 Rn activity is measured by an alpha-decay detector (Alphaguard 2000 Pro,
Genitron Instruments, Frankfurt, Germany). The instrument was placed on the roof of the
PIUB building, about 10 m next to the air intake. Using digital signal processing for pulse
shape analysis the detection limit of the detector in a 10 min measuring interval is about
3 Bq m−3 [Lehmann et al., 2004].
6.3 Results and Discussion
6.3.1
Temperature Dependent Fractionation at the Air Intake
The Ar/N2 ratio can be measured simultaneously with O2 /N2 and is a useful tracer to reveal
fractionation eects. Only the temperature dependence of the gas solubility in seawater leads
to seasonal variations in air-sea uxes and small changes in atmospheric Ar/N2 ratio [Keeling
et al., 2004]. On diurnal timescales, however, the atmospheric Ar/N2 ratio is expected to
be constant, because no biogeochemical processes inuence these inert gases. However, our
Ar/N2 measurements revealed large variability. Diurnal variations of δAr/N2 and outdoor
temperature are shown as an example in Figure 6.1. The outdoor temperature was measured
by a HOBO H8 data logger (Onset Computer Corporation, MA, USA) placed at the bottom
of the intake pole. The air intake is a Dekabon tube with 4 mm inner diameter (ID) and the
ow rate was about 250 mL min−1 . The higher the air temperature is, the lower the δAr/N2
gets. To assess the causes of the observed δAr/N2 variations and to better quantify this
eect, we conducted tests with dierent intake tubes and sampling ows. In addition to the
Dekabon tube with ow rates of 250 mL min−1 and 35 mL min−1 , also a stainless steel tube
with 0.8 mm ID and a ow rate of 155 mL min−1 was used. The correlation of δAr/N2 and
6.3. RESULTS AND DISCUSSION
87
-1400
δAr/N2 (per meg)
-1600
-1800
-2000
0.8mm ID, 150ml/min
4mm ID, 250ml/min
4mm ID, 35ml/min
-2200
5
10
15
20
Temperature (°C)
25
30
Figure 6.2: Correlation of δAr/N2 and outdoor temperature for dierent types of air intake.
Temperature sensitivity
of δAr/N2 (per meg/°C)
0
-4
-8
-12
-16
-20
0
1
2
3
Flow velocity (m/s)
4
5
6
Figure 6.3: Temperature sensitivity of δAr/N2 depending on the gas velocity at the air intake for the three
experiments of Figure 6.2.
outdoor temperature for dierent types of air intakes is shown in Figure 6.2. Remarkably, the
temperature records lag the δAr/N2 variations by 90 to 150 min. This is probably due to a slow
response of the temperature logger used for these tests and the fact that the temperature sensor
was not exposed to sunlight in contrast to the air intake. This time shift was applied in the
calculations of Figure 6.2 to obtain the best correlation. The temperature sensitivities obtained
by geometric mean regression are −17.5 ± 0.6 per meg/◦ C (R2 = 0.70), −7.2 ± 0.2 per meg/◦ C
(R2 = 0.71) and −3.6 ± 0.2 per meg/◦ C (R2 = 0.51) for the 4 mm ID/35 mL min−1 , 4 mm
ID/250 mL min−1 and 0.8 mm ID/155 mL min−1 experiments, respectively. As shown in
Figure 6.3, the temperature sensitivities of δAr/N2 mainly depend on the gas velocity at the
air intake.
Variations of the laboratory temperature can also potentially inuence the δAr/N2 measurements. Especially in summer there is a diurnal cycle of the laboratory temperature with
88
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
amplitudes of 23 ◦ C. However, the most striking feature of the diurnal temperature variations
in the laboratory is a rapid drop of about 3 ◦ C at midnight caused by the air-conditioning.
Because in these experiments no change in δAr/N2 can be observed at midnight (Figure 6.1),
the variations in δAr/N2 are indeed mainly caused by fractionation at the air intake. This
supposition was further conrmed by actively heating the intake tube, which resulted in large
δAr/N2 deviations. An explanation is that during the day especially when the sun heats the
black coating of the Dekabon tube, there builds up a small temperature gradient between the
intake tube and the surrounding air. This leads to thermal diusion with preferential accumulation of the lighter molecules in regions with higher temperatures. A thermal diusion
factor for Ar and N2 of 0.071 [Grew and Ibbs , 1952] would lead to a steady state fractionation of 240 per meg/◦ C. However, a steady state is not achieved at the intake because of
the continuous ow of gas. Even though, the lower the ow velocity the more the air can
approach a steady state. The slope of the correlation plot of δAr/N2 versus δO2 /N2 for the
4 mm ID/35 mL min−1 and the intake heating experiments gives 3.9 ± 0.1 and is in good
accordance with what is expected from thermal fractionation [Grew and Ibbs , 1952; Keeling
et al., 2004].
The isotopic ratios δ 29 N2 , δ 34 O2 and δ 36 Ar show also small variations that are correlated
with the temperature, providing compelling evidence of diusive fractionation. However, the
signal-to-noise ratio relative to measurement precision is much higher for Ar/N2 than for δ 29 N2 ,
δ 34 O2 or δ 36 Ar, because Ar/N2 is more sensitive to mass-dependent fractionation processes
owing to the comparatively large mass dierence between Ar and N2 .
Additional tests with sample air from a high pressure cylinder showed that there is also a
measurable inuence of the laboratory temperature on δAr/N2 . A cylinder was placed outside
the laboratory where only small and not abrupt temperature variations occur to exclude any
fractionation related to the cylinder or the pressure regulator. Then, the measured δAr/N2
showed to be positively correlated with the laboratory temperature (in contrast to the negative temperature sensitivity for fractionation at the intake). Dierent sources of thermal
fractionation inside the laboratory may lead to these eects: a) The cold trap which is partly
immersed in silicon oil at −70 ◦ C. Because of the relatively large volume (∼ 150 mL) and
the large temperature gradient (∼ 90 ◦ C) thermal diusion is likely to occur inside this cold
trap. Changing temperature gradients due to varying room temperatures could therefore lead
to thermal eects. b) Temperature dependent fractionation at tees [Manning , 2001], and c)
Fluctuations of the working gas due to thermally induced eects at the high-pressure gas
cylinders.
Experiments showed that the thermal fractionation at the intake could be reduced if instead
of Dekabon other types of tubing are used. Intakes both made of transparent plastic and
stainless steel signicantly reduced this eect, presumably because of a smaller inuence of
solar heating. However, thermal fractionation could also be observed on days with overcast
sky. Shading of the intake from sunlight can therefore only reduce but not eliminate this
eect. High ow velocities at the intake either by large sampling ows or by intake tubes with
small inner diameters may be most helpful for reducing thermal diusion at the intake.
6.3.2
The CO2 Record
Statistics of wind direction and wind speed from October 2003 to September 2004 is shown
in Figure 6.4 (data courtesy of the Institute of Applied Physics, University of Bern). The
prevalent local wind directions are the northern and western wind sector. High wind speeds
6.3. RESULTS AND DISCUSSION
89
0
315
Wind speed
(m/s)
<=1
>1 - 2
>2 - 3
>3
45
270
90
0%
2%
4%
225
6%
135
180
Figure 6.4: Statistics of wind direction and wind speed from October 2003 to September 2004 (data courtesy
of the Institute of Applied Physics, University of Bern).
0
315
45
270
90
380 390 400
CO2 (ppm)
410
225
420
430
135
180
Figure 6.5: Mean CO2 mixing ratio as a function of wind direction.
occur with winds coming from north-east to east (4590 ◦ ) and from west to south-west (225
270 ◦ ), which are the predominant mesoscale wind directions on the Swiss Plateau.
The mean CO2 mixing ratio as a function of wind direction is shown in Figure 6.5. It
was rst suspected that the CO2 measurements in the wind sector between 130◦ and 200◦ are
inuenced by a ventilating pipe from the building located south (170◦ ) and about 20 m away
from the air intake. However, there is a good correlation between CO2 and 222 Rn from this
wind sector, indicating that the high CO2 mixing ratios are mainly due to the low average wind
speed for this wind direction. Generally, no signicant correlation between wind direction and
CO2 mixing ratio was found indicating that no distinct CO2 sources are in the immediate
90
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
vicinity of the sampling site.
An overview of atmospheric records of CO2 , 222 Rn, δO2 /N2 , δ 13 C and δ 18 O of CO2 at Bern
between October 2003 and November 2004 is given in Figure 6.6. The diurnal variability is
much larger than the seasonal variability and is mainly caused by local sources and sinks and
by diurnal changes of atmospheric mixing conditions in the boundary layer. CO2 mixing ratios
were highest in the wintertime with nighttime maximum values reaching more than 500 ppm.
During atmospheric inversion events, characterized by persistent fog in autumn and winter,
the CO2 mixing ratio was clearly above 400 ppm over a period of several days. Afternoon
values in the spring and summer were commonly close to the background value as measured
for example at Jungfraujoch (see Chapter 4).
The 222 Rn activity was measured during a three month period between 27 January and 26
April 2004. The specic activity ranged from 020 Bq m−3 . Elemental ratios (O2 /N2 , Ar/N2 )
and carbon and oxygen isotopes of CO2 (δ 13 C, δ 18 O) could only be measured periodically,
especially on weekends when no other applications and measurements were running on the
analyzers. The O2 /N2 , δ 13 C and δ 18 O showed also large diurnal variations corresponding to
changes in CO2 mixing ratio.
Figure 6.7 shows the mean diurnal cycles of the CO2 mixing ratio averaged for the months
October 2003 to September 2004. The diurnal variations show for all months a minimum in
the afternoon followed by an increase towards the maximum in the early morning hours. The
amplitudes are largest in summer (5560 ppm peak-to-peak) and smallest is winter (1525 ppm
peak-to-peak). Outstanding are the months November and December 2003 with daily mean
CO2 mixing ratios of 440 and 437 ppm, respectively. This is due to frequent and persistent
atmospheric inversion situations with low wind speeds where CO2 accumulates in the boundary
layer. In all other months the daily mean CO2 mixing ratio is between 393 ppm and 410 ppm.
Apart from November and December 2003, all maxima lie in the range of 417434 ppm, whereas
the minima are between 372 and 402 ppm.
6.3.3
δ 13 C of CO2 , δ 18 O of CO2 and O2 /N2 Measurements
Figure 6.8 shows an example for typical diurnal cycles of CO2 , δO2 /N2 , δ 13 C and δ 18 O records
between 26 April and 01 May 2004. The δO2 /N2 , δ 13 C and δ 18 O measurements mirror the
CO2 variations. The inuence of thermal fractionation on δO2 /N2 was corrected using the
δAr/N2 measurements. In a rst step, the biogenic and anthropogenic components of δO2 /N2
variations were removed by subtracting the CO2 record scaled by the observed O2 :CO2 ratio. Variations in the residual δO2 /N2 are then expected to represent thermal fractionation
eects. Secondly, the fractionation ratio of δO2 /N2 and δAr/N2 was determined by the correlation between the residual δO2 /N2 and δAr/N2 . The slope of this correlation varies between
1.7 and 3.9 (R2 = 0.40.8), which likely reects mixed inuences of laboratory and outdoor temperature fractionation or dierent ow conditions in the inlet system depending on
the measurement setup. With this O2 /N2 -Ar/N2 fractionation ratio, the δAr/N2 variations
are then used to nally subtract the thermally induced δO2 /N2 variations from the original
δO2 /N2 to obtain the corrected δO2 /N2 . In Figure 6.8 the original (shaded line) as well as
the corrected δO2 /N2 (black line) are shown.
A representative Keeling plot of δ 13 C and δ 18 O and the corresponding O2 CO2 correlation
during the night of 27/28 April 2004 is shown in Figure 6.9. Only nighttime values (18:00
06:00 local time (LT)) have been used to match the Keeling plot assumptions, i.e. a constant
background CO2 concentration and a constant isotopic signature of the source or sink, as ac-
6.3. RESULTS AND DISCUSSION
91
CO2 (ppm)
520
480
440
400
222
Rn (Bq m-3)
360
20
15
10
5
δO2/N2 (per meg)
0
200
0
-200
-400
δ13C (‰, VPDB)
-600
-8
-10
-12
-14
δ18O (‰, VPDB)
-16
2
0
-2
-4
-6
1-Oct-03
1-Jan-04
Figure 6.6: Atmospheric records of CO2 ,
2003 and November 2004.
1-Apr-04
222
1-Jul-04
1-Oct-04
Rn, δO2 /N2 , δ 13 C and δ 18 O of CO2 at Bern between October
92
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
460
Oct 03
Nov 03
Dec 03
Jan 04
Feb 04
Mar 04
Apr 04
May 04
Jun 04
Jul 04
Aug 04
Sep 04
CO2 (ppm)
440
420
400
380
360
0
3
6
9
12
15
Time (UTC)
18
21
24
Figure 6.7: Mean diurnal cycles of CO2 for the months October 2003 to September 2004.
curately as possible. Still, the δ 13 Csource , δ 18 Osource and O2 :CO2 ratios mostly represent a ux
weighted average of more than one source and/or sink. Varying proportions of CO2 sources
containing distinct isotope ratios violate the assumptions of the 2-ended mixing model, and
are more common with oxygen than carbon isotopes, causing poorer relationships between
δ 18 O and 1/CO2 . In Figure 6.10 the nighttime buildup of CO2 was used to derive δ 13 Csource ,
δ 18 Osource and the O2 :CO2 ratio daily by the Keeling plot intercept method. The correlations
were calculated by geometric mean regression and error bars are the standard deviation of
the slopes. Only nights with more than 25 measurements, correlation coecients larger than
R2 = 0.9 for δ 13 C and O2 :CO2 and larger than R2 = 0.7 for δ 18 O were considered. The
nighttime δ 13 Csource varies between −33 % and −26 % with generally lower values in the
winter months than during the rest of the year. This likely represents the larger inuence of
fossil fuel combustion in wintertime. The large variability from one night to the next may be
caused by dierent weather conditions resulting in advection of dierent air masses or changes
in the CO2 source distribution. A similar picture can be seen for the nighttime δ 18 Osource .
The values range from −26 % to −3 % with the depleted (more negative) signatures again
occurring in wintertime and highly enriched (more positive) values in the spring and summer.
This seasonal variation of δ 18 Osource is somewhat larger than the variations of about −21 %
to −11 % observed by Pataki et al. [2003] in a similar study. The δ 18 O of atmospheric CO2
is a signal dominated by CO2 exchange with the terrestrial biosphere [Ciais et al., 1997a,b;
Keeling , 1995]. Fractionation of the oxygen isotopes of CO2 occurs in plants owing to differential diusion of C18 O16 O and C16 O16 O and to isotope eects in oxygen exchange with
chloroplast water. The higher δ 18 O values in summer compared to winter are probably caused
by strong photosynthetic activity and exchange of 18 O with leaf water in the plants, which is
generally enriched in δ 18 OH2 O if compared to the ground water due to evapotranspiration.
Furthermore, CO2 from combustion has a δ 18 O value similar to the δ 18 O of atmospheric O2
at −18 % [Kroopnick and Craig , 1972]. Thus, a larger fossil fuel CO2 component in winter
6.3. RESULTS AND DISCUSSION
93
CO2 (ppm)
520
480
440
400
δO2/N2 (per meg)
360
200
0
-200
δ13C (‰, VPDB)
-400
-8
-10
-12
δ18O (‰, VPDB)
-14
2
1
0
-1
26-Apr-04
27-Apr-04
28-Apr-04
29-Apr-04
30-Apr-04
1-May-04
Figure 6.8: Example of continuous CO2 , δO2 /N2 , δ 13 C and δ 18 O records between 26 April and 01 May 2004.
For the δO2 /N2 the original (shaded line) as well as the corrected values (black line) are shown (see text for
explanation).
leads to decreased δ 18 O values.
For a possible interpretation of the observed O2 :CO2 ratios we consider a simple model.
If we assume that the diurnal variations can be described by biogenic and fossil fuel uxes
of carbon and oxygen in the catchment area of the sampling site, then the atmospheric mass
balance for CO2 and O2 can be written
∆CO2 = F + B
(6.1)
∆O2 = αF F + αB B,
(6.2)
and
where ∆CO2 and ∆O2 are the observed changes in the atmospheric CO2 and O2 concentration, F and B are the uxes of carbon to the atmosphere due to fossil fuel combustion
94
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
-8
δ13C of CO2 (‰, VPDB)
-8.5
y = 7231(110) x - 27.4(0.3)
R2 = 0.9874
-9
-9.5
-10
-10.5
-11
27/28 April, 2004
2
y = 4247(222) x - 9.6(0.6)
R2 = 0.7948
y = -5.00(0.05) x + 2170(18)
R2 = 0.9931
300
δO2 /N2 (per meg)
δ18O of CO2 (‰, VPDB)
400
1.5
1
0.5
200
100
0
0
-0.5
-100
0.0022
0.0023 0.0024 0.0025 0.0026
1/CO2 (ppm -1)
0.0027
380
400
420
CO2 (ppm)
440
Figure 6.9: Representative Keeling plot of carbon and oxygen isotope ratios and O2 :CO2 oxidation ratio
during the night (18:0006:00 local time) of 27/28 April 2004. The equations shown are derived from geometric
mean regression (uncertainty in parentheses).
and the terrestrial biosphere, respectively (positive for release to the atmosphere). The coecients αF and αB are average O2 :CO2 exchange ratios for fossil fuel and land biota. We
use αF = −1.4 mol O2 /mol CO2 [Manning , 2001] and αB = −1.1 mol O2 /mol CO2 [Severinghaus , 1995]. By combining equations (6.1) and (6.2) and assuming that F and B are constant
over the considered time period, we can estimate from the observed O2 :CO2 exchange ratio
(∆O2 /∆CO2 ) the proportion of the biogenic ux in relation to the fossil fuel ux
B
∆O2 /∆CO2 − αF
=−
.
F
∆O2 /∆CO2 − αB
(6.3)
The measured nighttime O2 :CO2 oxidation ratios from April 2004 to November 2004 are
between −0.96 and −1.54 (lowest panel in Figure 6.10). For nighttime air sampling it is
assumed that respired CO2 is added to the atmosphere and both uxes F and B are positive.
This would lead to O2 :CO2 ratios between −1.1 and −1.4. However, almost half of all values
and especially the summer values are between −0.96 and −1.1. This could only be explained
with a biogenic CO2 sink up to four times as strong as the fossil fuel source, which is obviously
not the case, since the CO2 concentration is increasing and not decreasing during the night.
The same considerations can also be applied to the δ 13 Csource data, where no conicting
picture can be seen. The measured δ 13 Csource are in the range between an assumed δ 13 C of
the biospheric component of about −26 % and the more depleted values of the fossil fuel
component. One possibility to explain these low O2 :CO2 ratios would be an overestimation
6.3. RESULTS AND DISCUSSION
95
δ13Csource (‰, VPDB)
-26
-28
-30
-32
-34
δ18Osource (‰, VPDB)
-5
-10
-15
-20
-25
-30
O2 :CO2 oxidation ratio
(mol O2 /mol CO2 )
-1.6
-1.4
-1.2
-1
-0.8
1-Jan-04
1-Mar-04
1-May-04
1-Jul-04
1-Sep-04
1-Nov-04
Figure 6.10: δ 13 C and δ 18 O of source CO2 calculated from Keeling plots, and O2 :CO2 oxidation ratios.
of the span of our CO2 scale by more than 10 %. Comparison of our CO2 data with CO2
measurements from Laboratiore des Science du Climat et de l'Environnement, CE Saclay,
France, (Chapter 5) lead to the conclusion that this is highly unlikely. An underestimation of
the δO2 /N2 span of this magnitude can not be ruled out a priori, though. Since international
intercomparison programs for O2 /N2 measurements are being initiated only now, we do not
have any independent validation of our O2 /N2 scale yet. Measurement eects related to the
mass spectrometric technique like cross contamination (see Section 2.4.2) could potentially
lead to an underestimation of the δO2 /N2 span. However, no such eects have been reported
by other laboratories also measuring δO2 /N2 by mass spectrometry so far. Another alternative
to explain the observed O2 :CO2 ratios is that processes with O2 :CO2 exchange ratios of about
−1.0 play a major role in nighttime build-up of CO2 . Stephens et al. [2001] have also reported
on O2 :CO2 relationships that are considerably lower than theory would suggest, but potential
atmospheric or physiological origins for such relationships are unknown.
96
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
δ36Ar (per meg)
δ34O2 (per meg)
δ29N2 (per meg)
40
20
0
-20
-40
80
40
0
-40
-80
400
200
0
-200
-400
1-May-04
1-Jul-04
1-Sep-04
1-Nov-04
Figure 6.11: Measurements of the isotopic ratios δ 29 N2 , δ 34 O2 and δ 36 Ar at Bern between April 2004 and
November 2004
6.3.4
δ 29 N2 , δ 34 O2 and δ 36 Ar of Air
Measurements of the isotopic composition of N2 , O2 and Ar are shown in Figure 6.11. These
isotopic ratios are as a rst approximation constant within measurement precision (Table 2.1).
The small variations of the mean appearing in Figure 6.11 are probably due to changes in mass
spectrometer performance. Owing to the relatively high sampling rate of 5 measurements per
hour, there are also diurnal variations detectable. However, because of the many gaps in
the available record and the poor signal-to-noise ratio it is dicult to quantify these eects.
As already mentioned in Section 6.3.1, thermal diusion fractionation at the air intake is
considered to be the dominant cause for such diurnal variations. The thermal diusion factors
α of 29 N2 −28 N2 , 34 O2 −32 O2 and 36 Ar−40 Ar at 20 ◦ C are about 0.0045, 0.0099 and 0.0137,
respectively [Lang , 1999]. With α = 0.071 for Ar−N2 [Grew and Ibbs , 1952], an assumed
variation in δAr/N2 of 200 per meg due to thermal diusion would correspond to variations in
δ 29 N2 , δ 34 O2 and δ 36 Ar of about 13, 28 and 39 per meg, respectively. This might be consistent
with the observed variations in δ 29 N2 and δ 34 O2 . For δ 36 Ar the measured amplitude of the
variations are estimated to be larger (about 150 per meg) and not in opposite direction as is
expected for a mass dependent fractionation, because δ 36 Ar denotes the ratio of the lighter
(36 Ar) to the heavier (40 Ar) isotope, in contrast to the other isotopic ratios. Therefore, δ 36 Ar
is probably also inuenced by other unidentied measurement artifacts.
6.3. RESULTS AND DISCUSSION
97
360
270
180
90
0
17
18
19
20
21
March 2004
22
380
16
12
8
4
Wind speed (m s-1)
6
5
4
3
2
1
0
400
Rn (Bq m -3)
20
16
12
8
4
0
222
Wind speed (m s-1)
380
420
23
5
4
3
2
1
0
Rn (Bq m -3)
400
440
222
420
0
360
270
180
90
0
1
2
3
4
5
April 2004
6
Wind direction (°)
CO2 (ppm)
460
440
Wind direction (°)
CO 2 (ppm)
460
7
Figure 6.12: Examples of hourly mean values of CO2 and
222
Rn, together with local wind speed and wind
direction between March 17 and 23, 2004 (left panel) and between April 1 and 7, 2004 (right panel).
The δ 18 O of atmospheric O2 (δ 34 O2 ) is aected by photosynthesis and respiration in the
carbon/oxygen cycle and by the hydrological cycle, similar to the oxygen isotopic composition
of CO2 . Could we potentially detect biogeochemical variations of δ 18 O of atmospheric O2 in
our record? The δ 18 O of atmospheric oxygen is enriched by 23.5 % relative to average ocean
water [Kroopnick and Craig , 1972], which is known as the Dole eect. The δ 18 O produced
by photosynthesis is similar to that of the water from which the oxygen isotopes originate,
whereas respiration fractionates by about −20 % relative to atmospheric O2 [Guy et al.,
1993]. The δ 18 O of leaf water is elevated by 4 to 8 % compared to oceanic water due to
evapotranspiration [Dongmann , 1974; Farquhar et al., 1993]. If we assume that a diurnal
increase in atmospheric δO2 /N2 of 500 per meg is driven by the input of photosynthetic O2
that is about 20 % lower in δ 18 O than atmospheric O2 , we would expect a change in δ 18 O of
O2 of 10 per meg. The seasonal variability of O2 /N2 (∼ 150 per meg) leads to an even smaller
signal in δ 18 O of O2 (∼ 3 per meg). Detecting such small variations is thus very dicult with
current mass spectrometric measurements.
6.3.5
222
Rn Tracer Method to Estimate Regional CO2 Emissions
Examples of hourly mean values of CO2 and 222 Rn, together with local wind speed and
wind direction between 17 and 23 March 2004 and between 1 and 7 April 2004 are shown
in Figure 6.12. During the rst days of these periods wind speeds are low with frequently
changing directions. The trace gas concentration records show a typical diurnal pattern caused
by nighttime inversion situations. On 19 March and 4 April, respectively, winds change to
westerly directions with persistently high wind speeds. The CO2 and 222 Rn concentrations
are then close to continental background level [Schmidt et al., 1996]. The correlation between
222 Rn and CO will be used to estimate CO uxes for the catchment area of the sampling
2
2
site.
98
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
The 222 Rn measurements can be used to infer CO2 emission by using a simple onedimensional approach. This method has been used for other greenhouse gases and at different sites and is described in detail by Schmidt et al. [2001]. Assuming that each trace gas
is released to the atmosphere at a constant rate j i and that it accumulates in a well-mixed
boundary layer of height H(t), the short-term change in concentration ∆ci (t) is then:
∆ci (t)
j
= i .
∆t
H(t)
Since H(t) is the same for
j CO2 = j Rn
(6.4)
222 Rn
and CO2 , we can eliminate H(t) by combining both tracers
∆cCO2
.
∆cRn
(6.5)
The CO2 ux can thus be calculated from the measured slope between CO2 and 222 Rn variations and the 222 Rn exhalation rate. The radioactive decay of 222 Rn in the atmosphere during
a typical nighttime inversion situation lasting about 12 hours leads to a net 222 Rn loss of about
5 %. Therefore, a mean correction factor of 0.95 is applied when estimating 222 Rn-based CO2
uxes.
Daily CO2 /222 Rn correlations were determined using geometric mean regression. To derive
a mean value for the period from February to April 2004 only those days were included for
which more than 12 hourly mean values existed and which showed a correlation coecient
larger than R2 = 0.4. 33 days (39 %) satisfy this criterion. The mean CO2 /222 Rn slope is
5.2 ± 1.7 ppm/Bq m−3 . This is larger than estimates from other European sites. Schmidt
et al. [1996] have measured CO2 222 Rn slopes at Schauinsland, Germany, in the range of 1.8
to 3.5 ppm/Bq m−3 for the winter months. A wintertime estimation for western Europe from
the Mace Head record by Biraud et al. [2000] gives a slope of 1.42.1 ppm/Bq m−3 for all
selected events and 4.76.8 ppm/Bq m−3 for polluted events.
The 222 Rn ux emitted over continents is not strictly uniform, but depends mainly on
the soil type and the hydrological conditions. There is no regional map of observed 222 Rn
emissions available, but 222 Rn ux measurement at dierent sites in Germany showed an
average ux of about 50 Bq m−2 h−1 , corresponding to 0.7 atoms cm−2 s−1 [Dörr and Münnich ,
1990; Schmidt et al., 2001]. The uncertainty of the 222 Rn exhalation rate is estimated as
±25 % [Schmidt et al., 2001]. Using this estimation of the 222 Rn exhalation rate one obtains a
mean CO2 ux density during the period February/March/April 2004 for the Bern region of
11.0 ± 4.5 mmole m−2 h−1 or 95 ± 39 tC km−2 month−1 . This CO2 ux includes both biogenic
and fossil fuel uxes. The two main sources of uncertainty for this estimate, namely the 222 Rn
ux and the CO2 /222 Rn correlation amount to an overall uncertainty of the CO2 ux estimate
of ±41 %. To reduce this uncertainty the exact source areas of both CO2 and 222 Rn should
be known. This would require a explicit transport model, which can resolve the spacial and
temporal patterns of the CO2 and 222 Rn sources.
6.4 Summary and Outlook
This chapter summarizes results from the rst year of on-going continuous measurements of
CO2 , its stable isotopes and O2 /N2 in Bern. Concurrent Ar/N2 measurements revealed that
diusive fractionation due to thermal gradients at the air intake is an important modifying
process for high precision O2 /N2 measurements. The CO2 , O2 /N2 , δ 13 C and δ 18 O of CO2
6.4. SUMMARY AND OUTLOOK
99
data show strong diurnal and seasonal cycles. The diurnal variations are modulated by surface
uptake and release by vegetation and soils, emissions from fossil fuel combustion, and by the
diurnal development of the atmospheric boundary layer. Both stable carbon isotopes showed
depletion in the winter and enrichment in the summer due to changes in the proportions of
fossil fuel combustion and biogenic respiration at dierent times of the year. Additionally,
222 Rn was used to estimate a mean CO ux density in the catchment area of the sampling
2
site. As the measurements go on and more data are available the CO2 isotope and mixing ratio
data can be used to quantify with a mass balance calculation the proportional contribution of
each component to the total CO2 source. We also expect to observe inter-annual variations
of the seasonal cycle, and changes in CO2 mixing ratio relative to background sites, such as
Jungfraujoch. Comparison of boundary layer mixing ratios with background mixing ratios
should help to improve our understanding of atmosphere/surface exchange of CO2 on the
continental scale.
100
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
References
Bender, M. L., P. P. Tans, T. J. Ellis, J. Orchardo, and K. Habfast (1994), A high precision isotope
ratio mass spectrometry method for measuring the O2 /N2 ratio of air, Geochim. Cosmochim. Acta,
58 (21), 47514758.
Biraud, S., P. Ciais, M. Ramonet, P. Simmonds, V. Kazan, P. Monfray, S. O'Doherty, T. G. Spain,
and G. S. Jennings (2000), European greenhouse gas emissions estimated from continuous atmospheric
measurements and radon 222 at Mace Head, Ireland, J. Geophys. Res., 105 (D1), 13511366.
Chapman, S., and T. G. Cowling (1970), The Mathematical Theory of Non-Uniform Gases, Cambridge
Univ. Press, Cambridge.
Ciais, P., A. S. Denning, P. P. Tans, J. A. Berry, D. A. Randall, G. J. Collatz, P. J. Sellers, J. W. C.
White, M. Trolier, H. A. J. Meijer, R. J. Francey, P. Monfray, and M. Heimann (1997a), A threedimensional synthesis study of δ 18 O in atmospheric CO2 , 1. Surface uxes, J. Geophys. Res., 102 (D5),
58575872.
Ciais, P., P. P. Tans, A. S. Denning, R. J. Francey, M. Trolier, H. A. J. Meijer, J. W. C. White, J. A.
Berry, D. A. Randall, G. J. Collatz, P. J. Sellers, P. Monfray, and M. Heimann (1997b), A threedimensional synthesis study of δ 18 O in atmospheric CO2 , 2. Simulations with the TM2 transport
model, J. Geophys. Res., 102 (D5), 58735883.
Dongmann, G. (1974), The contribution of land photosynthesis to the stationary enrichment of
in the atmosphere, Radiat. Environ. Biophys., 11, 219225.
18
O
Dörr, H., and K. O. Münnich (1990), 222 Rn ux and soil air concentration proles in West-Germany.
Soil 222 Rn as tracer for gas transport in the unsaturated soil zone, Tellus, 42B, 2028.
Farquhar, G. D., J. Lloyd, J. A. Taylor, L. B. Flanagnan, J. P. Syvertsen, K. T. Hubick, S. C. Wong,
and J. R. Ehleringer (1993), Vegetation eects on the isotope composition of oxygen in atmospheric
CO2 , Nature, 363, 439443.
Friedli, H., U. Siegenthaler, D. Rauber, and H. Oeschger (1987), Measurements of concentration,
13
C/12 C and 18 O/16 O ratios of tropospheric carbon dioxide over Switzerland, Tellus, 39B (1-2), 8088.
Gemery, P. A., M. Trolier, and J. W. C. White (1996), Oxygen isotope exchange between carbon dioxide and water following atmospheric sampling using glass asks, J. Geophys. Res., 101 (D9), 14,415
14,420.
Grew, K. E., and T. L. Ibbs (1952), Thermal Diusion in Gases, Cambridge Univ. Press, Cambridge.
Guy, R. D., M. L. Fogel, and J. A. Berry (1993), Photosynthetic Fractionation of the Stable Isotopes
of Oxygen and Carbon, Plant Physiol., 101, 3747.
Hesterberg, R. (1990), Das Kohlendioxid und seine stabilen Isotope in Atmosphäre und Boden, Master's thesis, Physics Institute, University of Bern, Bern, Switzerland.
Keeling, C. D., R. B. Bacastow, A. F. Carter, S. C. Piper, T. P. Whorf, M. Heimann, W. G. Mook,
and H. Roelozen (1989), A three-dimensional model of atmospheric CO2 transport based on observed
winds: 1. Analysis of observational data, in Aspects of Climate Variability in the Pacic and the
Western Americas, Geophys. Monogr. Ser., vol. 55, edited by D. H. Peterson, pp. 165236, AGU,
Washington D.C.
Keeling, R. (1995), The atmospheric oxygen cycle: The oxygen isotopes of atmospheric CO2 and O2
and the O2 /N2 ratio, Rev. Geophys., Supplement, 12531262.
REFERENCES
101
Keeling, R. F., B. B. Stephens, R. G. Najjar, S. C. Doney, D. Archer, and M. Heimann (1998), Seasonal
variations in the atmospheric O2 /N2 ratio in relation to the kinetics of air-sea gas exchange, Global
Biogeochem. Cycles, 12 (1), 141163.
Keeling, R. F., T. Blaine, B. Paplawsky, L. Katz, C. Atwood, and T. Brockwell (2004), Measurement
of changes in atmospheric Ar/N2 ratio using a rapid-switching, single-capillary mass spectrometer
system, Tellus, 56B (4), 322338.
Kroopnick, P., and H. Craig (1972), Atmospheric Oxygen: Isotopic Composition and Solubility Fractionation, Science, 175 (4017), 5455.
Lang, C. (1999), Bestimmung und Interpretation der Isotopen- und Elementverhältnisse von Luft aus
polaren und alpinen Eisbohrkernen, insbesondere zur Temperaturrekonstruktion unter Ausnutzung des
Eekts der Thermodiusion, Ph.D. thesis, Physics Institute, University of Bern, Bern, Switzerland.
Langenfelds, R. L. (2002), Studies of the global carbon cycle using atmospheric oxygen and associated
tracers, Ph.D. thesis, Univ. of Tasmania, Hobart, Tasmania, Australia.
Lehmann, B. E., B. Ihly, S. Salzmann, F. Conen, and E. Simon (2004), An automatic static chamber
for continuous 220 Rn and 222 Rn ux measurements from soil, Radiation Measurements, 38, 4350,
doi:10.1016/j.radmeas.2003.08.001.
Leuenberger, M., P. Nyfeler, H. Moret, P. Sturm, and C. Huber (2000), A new gas inlet system for
an isotope ratio mass spectrometer improves reproducibility, Rapid Commun. Mass Spectrom., 14,
15431551.
Leuenberger, M., M. Eyer, P. Nyfeler, B. Stauer, and T. F. Stocker (2003), High-resolution δ 13 C
measurements on ancient air extracted from less than 10 cm3 of ice, Tellus, 55B (2), 138144.
Manning, A. C. (2001), Temporal variability of atmospheric oxygen from both continuous measurements and a ask sampling network: Tools for studying the global carbon cycle, Ph.D. thesis, University of California, San Diego, California, U.S.A.
Pataki, D. E., D. R. Bowling, and J. R. Ehleringer (2003), Seasonal cycle of carbon dioxide and its
isotopic composition in an urban atmosphere: Anthropogenic and biogenic eects, J. Geophys. Res.,
108 (D23), 4735, doi:10.1029/2003JD003865.
Schmidt, M., R. Graul, H. Sartorius, and I. Levin (1996), Carbon dioxide and methane in continental
Europe: a climatology, and 222 Radon-based emission estimates, Tellus, 48B (4), 457473.
Schmidt, M., H. Glatzel-Mattheier, H. Sartorius, D. E. Worthy, and I. Levin (2001), Western European
N2 O emissions: A top-down approach based on atmospheric observations, J. Geophys. Res., 106 (D6),
55075516.
Severinghaus, J. P. (1995), Studies of the terrestrial O2 and carbon cycles in sand dune gases and in
Biosphere 2, Ph.D. thesis, Columbia University, New York, U.S.A.
Severinghaus, J. P., M. L. Bender, R. F. Keeling, and W. S. Broecker (1996), Fractionation of soil
gases by diusion of water vapor, gravitational settling, and thermal diusion, Geochim. Cosmochim.
Acta, 60 (6), 10051018.
Stephens, B., P. Bakwin, P. Tans, and R. Teclaw (2001), Measurements of atmospheric O2 variations
at the WLEF tall-tower site, in Sixth International Carbon Dioxide Conference, Extended Abstracts,
vol. I, pp. 7880, Tohoku Univ., Sendai, Japan.
Sturm, P. (2001), Entwicklung eines neuen Einlasssystems für die massenspektrometrische Messung
des O2 /N2 Verhältnisses, Master's thesis, Physics Institute, University of Bern, Bern, Switzerland.
102
6. CO2 ,
222 Rn,
O2 /N2 AND ASSOCIATED TRACERS AT BERN
Wilkening, M. H., and W. E. Clements (1975), Radon 222 From the Ocean Surface, J. Geophys. Res.,
80 (27), 38283830.
Chapter 7
Development of Continuous O2 and
CO2 Analyzer Systems
7.1 Introduction
One limitation of the mass spectrometric method for O2 /N2 analysis is that the size of the
instruments prohibits their eld use. The temporal and spatial resolution of O2 /N2 measurements is limited to that which is feasible through ask sampling. Additionally, eects related
to the storage of air samples in asks (see Chapter 3) potentially complicate the interpretation
of O2 /N2 data.
Recently, Manning et al. [1999] have developed a continuous O2 analyzer based on paramagnetic properties of O2 , which has a precision comparable or better than conventional
methods. Lueker et al. [2001, 2003] used such an analyzer, for example, to reveal stoichiometric O2 :CO2 ratios of wildre emissions or air-sea uxes of O2 resulting from coastal upwelling.
Paramagnetic analyzers are not suited, however, for measurements from a moving platform
because of vibration and motion sensitivity. A less motion sensitive vacuum ultraviolet adsorption instrument was developed at the same time by Stephens [1999]. The instrument detects
changes in O2 by the adsorption of vacuum ultraviolet radiation and has been deployed on
ships in the equatorial Pacic and Southern Ocean [Stephens et al., 2003]. Stephens et al.
[2001] have also adapted a commercial detector based on electrochemical cells to make continuous O2 measurements. This technique, although somewhat less precise, was successfully
deployed on a tall tower in a forest ecosystem.
In the framework of the ongoing (20042008) European CarboEurope-IP project we decided to complement our existing ask sampling program at the high altitude site, Jungfraujoch, Switzerland (3580 m a.s.l., 46◦ 33'N, 7◦ 59'E), with continuous in-situ O2 and CO2 measurements. Our approach for an analyzer at Jungfraujoch was to combine the paramagnetic
and electrochemical technique for O2 analysis and to connect this system with a non-dispersive
infrared adsorption (NDIR) CO2 analyzer. The gas handling scheme of the analyzers was basically adapted from Manning [2001].
A second O2 and CO2 analyzer system dedicated for making measurements during long distance ights in a passenger aircraft was developed in the framework of the European CARIBIC
project. This airborne analyzer is an integral part of an instrument container consisting of
several trace gas and aerosol analyzers and ask sampling devices. Our instrument is based
on electrochemical cells for O2 analysis and a NDIR analyzer for CO2 measurements.
104
7. CONTINUOUS O2 AND CO2 ANALYZERS
Both instruments have been developed and tested in the course of this thesis. At the
time of writing, the Jungfraujoch analyzer system is being installed on site and test ights
within the CARIBIC project are being performed. Regular ights are scheduled in 2005. This
chapter presents an overview of the layout of the gas handling and analysis system of the two
continuous O2 and CO2 instruments.
7.2 Paramagnetic and Electrochemical Oxygen Sensors
The paramagnetic oxygen sensor measures the oxygen concentration in a gas by utilizing the
paramagnetic properties of O2 . Molecular oxygen has by far the highest magnetic susceptibility compared to other constituents of air [Manning et al., 1999], and changes in magnetic
susceptibility of air are therefore dominated by variations in O2 . The measurement principle is
described in detail by Kocache [1986]. Basically, the sensor contains a cell where a small glass
dumbbell lled with nitrogen is suspended in a strong non-uniform magnetic eld. Paramagnetic molecules are attracted to regions with higher magnetic elds. This attraction results
in a pressure gradient across the cell, which creates a torque on the dumbbell. A feedback
control adjusts an electric current which produces an electromagnetic force counteracting the
torque, thereby keeping the dumbbell in its original position. The position of the dumbbell is
detected optically using a light emitting diode, a mirror attached to the dumb bell and a photodiode. The regulating current is essentially proportional to the paramagnetic susceptibility
and hence to the O2 partial pressure of the gas owing through the cell.
We use a Parox 1000 paramagnetic oxygen sensor from MBE AG, Switzerland. After rst
tests it became clear that the noise of the internal electronics of this sensor makes it impossible
to achieve the required precision, although we could control temperature, pressure and ow
rate of the gas in the cell to a very high degree. In collaboration with the manufacturer of
the sensor we succeeded step by step to reduce the electronic noise by almost three orders of
magnitude. With the improved sensor we are convinced to possess now an ideal instrument
for making high precision O2 measurements.
The paramagnetic analyzer is well-suited for remote eld-stations, but can not be used for
mobile measurement campaigns such as on airplanes due to its motion sensitivity. A motioninsensitive method based on the electrochemical properties of oxygen are the galvanic cells.
They are lead-oxygen batteries, consisting of a lead anode, a oxygen cathode made of gold,
and a weak acid electrolyte. Oxygen molecules enter the electrochemical cell through a nonporous Teon membrane, diuse in the acid electrolyte, and are reduced at the gold cathode.
The current generated between the electrodes is directly proportional to the partial pressure
of oxygen at the sensing surface of the cell.
We have tested oxygen sensors from two dierent manufacturers, the KE-25 from Figaro,
Japan, and the Max-250 from Maxtec, U.S.A.. They both have the same specications,
but showed to have varying internal noise levels. In general, Max-250 sensors appeared to
have lower noise signals that the KE-25, but there was also a considerable dierence in the
performance between individual sensors. The measured root-mean-square (rms) noise was
between 15 and 100 per meg at 1 s resolution.
7.3. O2 AND CO2 ANALYZER SYSTEM DESIGN
105
7.3 O2 and CO2 Analyzer System Design at Jungfraujoch,
Switzerland
Figure 7.1 shows the gas handling schematic of the continuous O2 and CO2 analyzer system at
Jungfraujoch, Switzerland. A diaphragm pump (PM12906-86, KNF Neuberger) pulls sample
air at about 100 mL min−1 through the gas inlet. For drying the sample air passes through a
180 mL glass cold trap, immersed in silicon oil and maintained at −80 ◦ C by a cryogenic chiller
(FC100D, Kinetics). An electrically operated 4-port switching valve (Valco Instruments Co.
Inc.) chooses between this sample air and four calibration gases. The calibration gases are
connected to the sample line via electronic 2-way valves (ET-2-24, Clippard Europe S.A.).
A second 4-port switching valve selects between the sample air/calibration gas stream or the
working gas stream. The gas stream not passing through the analyzer at any given time
is vented through an exhaust outlet. The air stream to be analyzed passes next a 15 µm
lter to remove any remaining particles. A pressure regulator (PR50A15Z2 Combo Pressure
Regulator, VICI) serves as a rst coarse pressure control. It primarily attenuates pressure
pulsation from the pump. To ensure that all calibration gases, working gas and sample air
are dried to the same dew point, the gas stream passes through a second 30 mL glass cold
trap. After passing a 2 µm lter the gas stream enters the O2 analyzer box. This custommade aluminium box houses the pressure control and the paramagnetic and electrochemical O2
sensors. The box is covered by exible heaters (Minco EC AG) and brous insulation material.
A microprocessor controller (dTRON 16.1, Jumo) adjusts the temperature measured by a
PT100 temperature sensor inside the box to 40±0.02 ◦ C. A fan mounted inside the aluminium
box ensures a homogenous temperature. Pressure control of the O2 analyzers is established
with a ±1 hPa dierential pressure manometer (Baratron 220D, MKS). This sensor type has
an internal temperature control, which, however, is switched o for this application, because
of the external temperature stabilization of the aluminium box. The reference pressure for the
sensor is supplied by a 1 L stainless steel cylinder (Swagelok) lled to about 200 hPa above
atmospheric pressure. A PID control module (Type 250E, MKS) maintains a zero pressure
dierential across the manometer by adjusting a solenoid-actuated control valve (Type 248,
MKS). This valve vents excess air to the atmosphere and ensures a constant pressure and ow
rate in the O2 analyzers. The dierential pressure is controlled to a precision of ±0.002 hPa.
Finally, the gas stream rst enters the paramagnetic O2 sensor and then the electrochemical
sensors. Four identical galvanic cells (Max-250, Maxtec) are mounted in series to improve
measurement precision. The small dierences in output voltage of the galvanic cells that have
to be resolved (typically 0.6 mV/% O2 ) require very stable electronic components. The lownoise ampliers of the cells are therefore mounted in the analyzer box directly next to the
cells.
The gas next travels from the O2 analyzer box to the CO2 analyzer. This is a commercially
available NDIR analyzer (S710 UNOR, SICK MAIHACK) housed in a 19 case. The reference cell of the CO2 analyzer is constantly ushed at a ow rate of about 10 mL min−1 with
a reference gas from a high pressure cylinder. The outlet pressure of both the CO2 reference
gas and the sample/calibration gas is adjusted by a second absolute and dierential pressure
control unit. The absolute pressure is measured on the sample side by a absolute pressure
transmitter (dTrans p30, Jumo) and the pressure dierence between sample and reference side
by a 10 hPa dierential pressure manometer (Type 225AD, MKS). Two microprocessor controllers (dTRON 16.1, Jumo) adjust on both sides a Fluistor microvalve (NC-1500, Redwood
Working Gas
Cold Trap
-80 °C
Pump
4-way Valve
Filter
Pressure
Regulator
Filter
Calibration Calibration
Gas 2
Gas 3
4-way Valve
CO2
Reference
Gas
Bypass
Vent
Controller
Px
Pr
Pressure
Gauge
Temperature controlled box
Fill Port
Pressure
Reference
Controller
Fluistor
Controller
Fluistor
40 °C
Fuel Cell
O2
Sensors
Px
Pr
Differential
Pressure Gauge
Paramagnetic
O2 Sensor
NDIR
CO2 Analyzer
Figure 7.1: Gas handling schematic of the continuous O2 and CO2 analyzer system at Jungfraujoch, Switzerland.
Filter
Sample
Air in
Calibration
Gas 1
SICK MAIHAK S710
Flow Restriction
(Capillary)
Flow Sensor
Solenoid Valve
Needle Valve
Sample
Air out
106
7. CONTINUOUS O2 AND CO2 ANALYZERS
7.4. THE CARIBIC O2 ANALYZER
107
Microsystems) in order to keep the pressures constant. The absolute pressure on the sample
side is maintained with a precision of ±0.03 hPa at about 100 hPa above ambient pressure,
and the zero dierential pressure is precise to ±0.05 hPa. The ow rates are monitored by
dierent mass ow meters (Type 179, MKS and AWM3100V, Honeywell).
The analyzer system is completely computer controlled. A LabVIEW program handles
data acquisition from the paramagnetic, the electrochemical, and the NDIR analyzer as well
as all valve switching and the acquisition of other system parameters such as ow rates,
temperatures, pressures in the analyzers and in the high pressure cylinders.
7.4 The CARIBIC O2 Analyzer
7.4.1 The CARIBIC Project
The European project Civil Aircraft for Regular Investigation of the Atmosphere Based on
an Instrument Container (CARIBIC) aims at studying chemical and physical processes in the
upper troposphere and lowermost stratosphere [Brenninkmeijer et al., 1999]. The fundamental idea is to use passenger aircraft for making measurements during long distance ights.
An airfreight container with scientic instruments is placed in the forward cargo bay of an
Airbus A340-600 from Lufthansa. The container houses analyzers for in-situ measurements
of trace gases and aerosols and devices for collecting air and aerosol samples for subsequent
laboratory analyses. Using a passenger aircraft as a measurement platform allows to regularly survey atmospheric composition over large distances at relatively low cost. The Climate
and Environmental Physics division of the Physics Institute, University of Bern, undertook
the commitment to measure atmospheric O2 concentrations. These measurements will constitute additional important information to explore the contribution to spatial and temporal
variations from dierent processes in the global carbon cycle.
7.4.2 Analyzer System Design
Although the CO2 mixing ratio is already measured by another dedicated analyzer in the
CARIBIC instrument container and our task in the project is restricted to O2 measurements,
we decided to incorporate a NDIR CO2 analyzer in our instrument, too. Atmospheric O2
measurements are most valuable with concurrent CO2 measurements and the simultaneous
analysis of CO2 in the same gas stream may provide an important diagnostic tool for O2
measurements.
Figure 7.2 shows the gas handling and analysis schematic of the analyzer. The analyzer
system is similar to the instrument at Jungfraujoch, but had to be adapted to the specic
requirements for the use in a passenger aircraft. First of all, the entire instrument containing
the O2 sensors, the CO2 analyzer, the pumping and drying unit, the pressure and ow control
system as well as all electronic components had to be assembled in a EMC-shielded case, with
dimensions of 45 cm×27 cm×40 cm. The working and calibration gases for both the O2 and
CO2 analyses are contained in three 2 L high pressure cylinders. Due to a shortage of space in
the instrument container we can not employ more calibration gases during the ight. Another
diculty was that the inlet and outlet pressure during a ight is changing between 200
1000 hPa depending on the aircraft altitude. In order to get a constant inlet pressure the pump
capacity is adjusted by a Fluistor microvalve downstream the diaphragm pump (PM 2019686.3, KNF Neuberger). The switching between the sample, working and calibration gases
Fluistor
Pump
Calibration
Gas 2
Working
Gas
Filter
Drying Unit
Mg(ClO4)2
Pressure
Regulator
Figure 7.2: Gas handling schematic of the CARIBIC analyzer system.
Sample
Air in
Filter
Controller
Flow Sensor
Solenoid Valve
Needle Valve
Calibration
Gas 1
Bypass
Vent
Controller
Px
Pr
Fill Port
Flow
Restriction
(Capillary)
Temperature Controlled Box
40 °C
Pressure
Reference
Differential
Pressure Gauge
Fuel Cell
O2 Sensors
Reference Cell
LI-6262
CO2 Analyzer
Sample Cell
CO2 Absorber
(Ascarite)
Air out
108
7. CONTINUOUS O2 AND CO2 ANALYZERS
7.5. OUTLOOK
109
is performed by electronic 2-way valves (ET-2-24, Clippard Europe S.A.). The drying unit
consists of a stainless steel tube containing about 50 g of magnesium perchlorate (Mg(ClO4 )2 ).
Again, four galvanic cells (Max-250, Maxtec) are used simultaneously. The pressure control
of the cells is achieved by an electronic controller (Type 250E, MKS) which maintains a zero
pressure dierential across the pressure transducer (Baratron 223, MKS) by adjusting the
nearby solenoid-actuated control valve (Type 248, MKS).
For CO2 analysis we use a LI-6262 NDIR analyzer (LI-COR). Because we can not employ
a separate CO2 reference gas for the reference cell of the analyzer and in order to save an
additional dierential pressure control between sample and reference cell, the CO2 analyzer is
operated in serial mode, meaning that the same gas stream is injected into the sample and
reference cell in series, with the CO2 being removed before it enters the reference cell. The
CO2 mixing ratio of the gas in the reference cell is thus zero and the analyzer measures the
absolute CO2 mixing ratio in the sample cell. The removal of CO2 after the gas has passed
the sample cell is done chemically by Ascarite (sodium hydroxide-coated silica), a carbon
dioxide absorbent, lled in a stainless steel tube. Magnesium perchlorate is placed at the
exit end of the Ascarite to take up water produced as a by-product of the absorption process.
Finally, the outlet pressure is controlled by an electronic pressure controller (Type 640, MKS).
The computer hardware consists of a modular industrial PC (FBCube bus, Syslogic), which
includes in a 6 slot housing the processor, analog and digital inputs and outputs, the power
supply and a CompactFlash memory card to store the data. The electronic system and the
programming of the software is designed according to the guidelines specied by the engineers
of the CARIBIC project.
7.5 Outlook
Continuous O2 measurement programs improve our understanding of the carbon cycle and of
various terrestrial and biogeochemical processes at high temporal resolution. The new O2 and
CO2 analyzer system at the remote site Jungfraujoch will provide continuous records of O2
and CO2 and give new insights in the partitioning of oceanic and land biotic inuences on the
air masses arriving at this site. Additionally, processes occurring on hourly, daily, seasonal,
and annual timescales can be studied.
High precision atmospheric O2 measurements with the CARIBIC O2 analyzer exhibit an
additional challenge, because the instrument has to comply with special restrictions related
to the available space in the aircraft for the whole instrument including calibration gases.
But we are condent that with thorough experimental testing and possibly some ne-tuning
improvements the instrument can yield valuable oxygen data. Recommendations by the World
Meteorological Organisation [2005] for optimizing the carbon observing system include aircraft
ights over under-sampled areas, such as tropical South America, Africa or South East Asia,
and ights in high altitudes to cope with strong convective mixing up to 10 km in the tropics.
Our analyzer on board of the passenger aircraft will measure for the rst time precise O2
concentrations in the upper troposhere/lower stratosphere and over large, intercontinentalscale areas, and it will also complement measurements from vertical proling in the lower
troposphere.
110
7. CONTINUOUS O2 AND CO2 ANALYZERS
References
Brenninkmeijer, C. A. M., P. J. Crutzen, H. Fischer, H. Güsten, W. Hans, G. Heinrich, J. Heintzenberg, M. Hermann, T. Immelmann, D. Kersting, M. Maiss, M. Nolle, A. Pitscheider, H. Pohlkamp,
D. Schare, K. Specht, and A. Wiedensohler (1999), CARIBIC Civil Aircraft for Global Measurement of Trace Gases and Aerosols in the Tropopause Region, Journal of Atmospheric and Oceanic
Technology, 16 (10), 13731383, doi:10.1175/1520-0426.
Kocache, R. (1986), The measurement of oxygen in gas mixtures, J. Phys. E: Sci. Instrum., 19,
401412.
Lueker, T. J., R. F. Keeling, and M. K. Dubey (2001), The Oxygen to Carbon Dioxide Ratios observed
in Emissions from a Wildre in Northern California, Geophys. Res. Lett., 28 (12), 24132416.
Lueker, T. J., S. J. Walker, M. K. Vollmer, R. F. Keeling, C. D. Nevison, and R. F. Weiss (2003),
Coastal upwelling air-sea uxes revealed in atmospheric observations of O2 /N2 , CO2 and N2 O, Geophys. Res. Lett., 30 (6), 1292, doi:10.1029/2002GL016615.
Manning, A. C. (2001), Temporal variability of atmospheric oxygen from both continuous measurements and a ask sampling network: Tools for studying the global carbon cycle, Ph.D. thesis, University of California, San Diego, California, U.S.A.
Manning, A. C., R. F. Keeling, and J. P. Severinghaus (1999), Precise atmospheric oxygen measurements with a paramagnetic oxygen analyzer, Global Biogeochem. Cycles, 13 (4), 11071115.
Stephens, B., P. Bakwin, P. Tans, and R. Teclaw (2001), Measurements of atmospheric O2 variations
at the WLEF tall-tower site, in Sixth International Carbon Dioxide Conference, Extended Abstracts,
vol. I, pp. 7880, Tohoku Univ., Sendai, Japan.
Stephens, B. B. (1999), Field-based Atmospheric Oxygen Measurements and the Ocean Carbon Cycle,
Ph.D. thesis, University of California, San Diego, California, U.S.A.
Stephens, B. B., R. F. Keeling, and W. J. Paplawsky (2003), Shipboard measurements of atmospheric
oxygen using a vacuum-ultraviolet absorption technique, Tellus, 55B, 857878.
World Meteorological Organisation (2005), Recommendations of the 12th WMO/IAEA Meeting of
Experts on Carbon Dioxide Concentration and Related Tracer Measurement Techniques, Toronto,
Canada, 1518 Sep. 2003, in press.
Appendix A
About Straight Line Regression
Models when Both Variables are
Subject to Error
When computing regression lines between two variables, we usually use a linear least squares
tting technique, often available as standard functions in commercial software packages. Fitting by least squares assumes an independent x variable with no or negligible errors and a
dependent y variable, which is subject to error. The regression line is then obtained by minimizing the sum of the squares of the y -deviations from the linear model.
However, for some regression problems, such as when computing O2 :CO2 oxidation ratios or
1/CO2 δ 13 C correlations (Keeling plot), both x and y variables are subject to error. In these
cases the situation gets by far more complicated and commonly no universal tting techniques
are available. There are many approaches for such problems, but they are all limited to special applications. Depending on the characteristics of the data and the scientic objective,
dierent statistical methods can be applied. Due to the many dierent methods in use, the
literature on this subject is somewhat controversial. The purpose of this appendix is to briey
review dierent linear regression techniques that can be used when both variables are subject
to error.
In statistics, functional and structural regression models are distinguished. Problems
where the true points lie precisely on the line are called functional models. The scatter of the
data is only due to measurement errors. Problems where the data (in absence of measurement
error) have intrinsic scatter, that is the true points are scattered about the line, are called
structural models.
A.1 Functional Regression Models
Functional regression models are variously called errors-in-variables models or measurement
error models. To dene the measurement error model, we assume that
(Xi , Yi ),
i = 1, ..., n,
(A.1)
are observed random variables with underlying true values
(xi , yi ),
i = 1, ..., n.
(A.2)
112
A. REGRESSION WITH ERROR IN BOTH VARIABLES
We also assume that δi and ²i are random errors associated with xi and yi , respectively, and
that
Xi = xi + δi
(A.3)
Yi = yi + ²i .
(A.4)
Finally, we restrict to linear functions and assume that
yi = α + βxi
(A.5)
Yi = α + βXi + (²i − βδi ),
(A.6)
or
where α and β are two unknown parameters.
A.1.1
Ordinary Least Squares
If we can directly observe xi or if δi ¿ ²i , we can use ordinary least squares. In the ordinary
least squares (OLS) problem the estimator of (α, β) is the (α̂, β̂) for which the sum of squared
vertical deviations of the points (Yi , xi ) from the line yi = α + βxi is minimized, i.e. (α̂, β̂) is
the solution of
min
n
X
α,β
(Yi − α − βxi )2 .
(A.7)
i=1
This gives
Pn
β̂ =
(x − x)(Yi −
i=1
Pn i
2
i=1 (xi − x)
Y)
=
SxY
Sxx
(A.8)
and
(A.9)
α̂ = Y − β̂x,
where x and Y denote the arithmetical mean of xi and Yi , respectively. The correlation
coecient R is dened as
SxY
R= √
,
(A.10)
Sxx SY Y
P
where SxY = ni=1 (xi − x)(Yi − Y ) and Sxx and SY Y accordingly.
If we are unable to observe xi directly and instead observe the sum Xi = xi + δi , the
situation is not so simple. If we still apply the ordinary least squares method to the data, the
estimators of the model parameters (α, β) are no longer unbiased [Fuller , 1987; Draper and
Smith , 1998; Bruzzone and Moreno , 1998]. Under the assumption that the errors are normally
and independently distributed
δi ∼ N I(0, V (δ)),
²i ∼ N I(0, V (²)),
(A.11)
and that xi , δi and ²i are uncorrelated
C(xi , δi ) = C(δi , ²i ) = 0,
(A.12)
A.1. FUNCTIONAL REGRESSION MODELS
113
it follows that
E(β̂) = β
V (x)
,
V (x) + V (δ)
(A.13)
with E and V denoting the expected value and the variance of the corresponding variable,
respectively. The bias is negative, which means that β will be underestimated on the average
and the magnitude of the bias is depending on the relative sizes of V (x) and V (δ). The bias
arises from the fact that Xi is not independent of the error (²i − βδi ) in equation (A.6).
A.1.2 Measurement Error Model
Given some additional information, for example the relative magnitude of the error variances
associated with the two variables, we can obtain the parameter estimation for the model of
equation (A.6) [Fuller , 1987]. The maximum likelihood estimation of (α, β) with C(δ, ²) = 0
and the ratio
λ=
V (²)
V (δ)
(A.14)
known, leads to
£
¤1/2
2
SY Y − λSXX + (SY Y − λSXX )2 + 4λSXY
β̂ =
,
2SXY
α̂ = Y − β̂X.
(A.15)
(A.16)
In many parameter estimation problems, condence intervals for the estimated parameters
are needed. In the linear ordinary least square case, these can be obtained exactly. For the
measurement error model an asymptotic form of the covariance matrix for the estimated
parameters can be derived. The result for the variances of α and β is
"
#
β̂ 2
β̂ 2 SXY Z
(SY Y − β̂SXY )Z
−
V̂ (β̂) =
+
(SY Y − β̂SXY )2 ,
(A.17)
2
λ
(n − 1)λ2
SXY
β̂
V̂ (α̂) =
Z =
Z
2
+ X V̂ (β̂),
n
n h
i2
X
1
Yi − Y − β̂(Xi − X) .
(n − 2)
(A.18)
(A.19)
i=1
This variance is applicable only in cases of large samples and normal residuals. Therefore,
reliable error analysis for small samples or nonnormal residuals often requires resampling techniques like bootstrap (random sampling of n points of the observed data set with replacement).
The estimator of β for model (A.6) can also be interpreted in terms of the least squares
method. In the errors-in-variables situation an observation can deviate from the true line in
both the horizontal and vertical direction. If ²i and δi are independent, the squared statistical
distance from the observation to the line is
[statistical distance]2 =
²2i
δ2
+ i .
V (²) V (δ)
(A.20)
114
A. REGRESSION WITH ERROR IN BOTH VARIABLES
The x value for the point on the line that is the smallest statistical distance from (Xi , Yi ) is
(see Figure A.1)
ẍi =
β+
Yi − α
V (²) 1
V (δ) β
(²) Xi
+ VV (δ)
β
(A.21)
.
The corresponding y value is
(A.22)
y¨i = α + β ẍi
Using the least squares criterion, that is minimizing the sum of squared statistical distances
from the observed points (Xi , Yi ) to the associated point on the line (ẍi , y¨i ) leads again to
(A.15) and (A.16) for the estimators of β and α.
Equation (A.15) is derived under the assumption that λ is known, which in general is not
the case. This solution can therefore only be used if a reasonably accurate estimate of λ can
be made.
For the more general case, where the measurement errors are dierent for each point and
each variable (λi 6= constant), weighted regressions can be used [Barker and Diana , 1974; Press
et al., 1992]. The regression coecients α̂ and β̂ are obtained by minimizing the quantity χ2
χ2 (α, β) =
n
X
(Yi − α − βXi )2
i=1
V̂ (²i ) + β 2 V̂ (δi )
.
(A.23)
The solution to this problem can be found by iterative numerical calculation (e.g. using the
algorithm texy by Press et al. [1992]).
A means to assess the goodness of t, i.e. whether the model is appropriate or not, is to
calculate the probability q that an accurate model would give a value for χ2 greater than the
calculated. The probability distribution of χ2 is the chi-square distribution for n − 2 degrees
of freedom. Therefore, q is 1 − p, where p is the chi-square cumulative density function for
the value χ2 (α̂, β̂) and n − 2 degrees of freedom. A probability q close to zero indicates a
poor t between the data and model, either because the model is wrong or because the data
uncertainty is underestimated. Most likely the scatter about the line can then not be explained
by measurement error only.
Note that these models actually should only be used with scale-free variables or if both
variables were measured on the same scale, because these slopes will change when one of the
variables is expressed in dierent units.
A.1.3
Orthogonal Distance Regression
When λ = 1, the solution (A.15) denes the line that minimizes the sum of squares of perpendicular deviations from the line (see Figure A.1). This situation is commonly called orthogonal
distance regression (ODR) or major axis regression. It is important to notice that orthogonal
distance regression is appropriate only when V (²) = V (δ) (and for scale-free variables).
Weighted orthogonal distance regression, which accounts for the unequal precision of the
observations Xi and Yi by appropriate weights, gained some popularity through the publicly
available software package ODRPACK [Boggs et al., 1992]. However, a shortcoming of this
method is that it does not converge to the OLS solution if V (²) À V (δ). Therefore, it only
produces reasonable results if λi ≈ 1 even with weights included in the calculation.
A.2. STRUCTURAL REGRESSION MODELS
115
OLS
ODR
MEM
GMR
Figure A.1: Ordinary least squares (OLS): the vertical deviations of the data points (dots) from the line are
minimized. Orthogonal distance regression (ODR): the perpendicular deviations are minimized. Measurement
error model (MEM): statistical distances dependent on the ratio of measurement variances are minimized
(shown here for V (δ) > V (²)). Geometric mean regression (GMR): shaded areas are minimized.
A.1.4 Geometric Mean Regression
If λ = SY Y /SXX , the estimators are
β̂ = (SY Y /SXX )1/2 ,
α̂ = Y − β̂X,
(A.24)
which is the geometric mean regression (GMR) or reduced major axis regression. The estima−1
tor β̂ is the geometric mean of the quantities b1 = SXY /SXX and a−1
1 = (SXY /SY Y ) , where
b1 and a1 are the slopes in ordinary least squares ts of Y versus X (Y = b0 + b1 X ) and X
1/2 is a compromise
versus Y (X = a0 + a1 Y ), respectively. The geometric mean β̂ = (b1 /a−1
1 )
value lying in between the two slopes. Comparison with equation (A.8) and (A.10) shows that
β̂GMR =
β̂OLS
.
R
(A.25)
In Figure A.1 another interpretation of the geometric mean regression can be seen. It is the
minimum of the sum of the areas obtained by drawing horizontal (parallel to the X -axis) and
vertical (parallel to the Y -axis) lines from each data point. The symmetry of the solution in
respect to interchange of the X and Y axes is obvious.
A.2 Structural Regression Models
If there is no understanding of the nature of the scatter about a linear relationship and
if measurement error is relatively unimportant, various unweighted least-squares regression
lines can be calculated [Isobe et al., 1990; Babu and Feigelson , 1992]. In situations where the
classication of variables in dependent and independent is unclear or arbitrary, a regression line
treating the variables symmetrically should be used. Symmetrical models are the orthogonal
distance regression (ODR), which minimizes the perpendicular distances; the OLS bisector, the
line which bisects the angle formed by the two lines OLS(Y|X), which minimizes the residuals in
Y, and OLS(X|Y), which minimizes the residuals in X; the geometric mean regression (GMR);
116
A. REGRESSION WITH ERROR IN BOTH VARIABLES
and the OLS mean, the arithmetic mean of the two lines OLS(Y|X) and OLS(X|Y). It has
to be emphasized that these methods give results that are theoretically dierent from each
other, and are not dierent estimates of the same quantity. Unless there is additional prior
knowledge regarding the data or the scientic question being asked, there is no statistically
better line than another. Recommendations regarding the choice of regression method seem
to be controversial. The most accurate regression can often only be assessed via simulation
studies.
A.3 Bivariate Correlated Errors and Intrinsic Scatter
An extension of the OLS method is the bivariate correlated errors and intrinsic scatter (BCES)
method, which is designed for data with intrinsic scatter in addition to measurement error
[Akritas and Bershady , 1996]. The model of equation (A.5) is adapted to
(A.26)
yi = α + βxi + ei ,
where ei is assumed to have zero mean and nite variance and represents the intrinsic scatter.
Also, this estimator makes it possible to have correlated errors between both variables and
to consider the case when the magnitudes of the measurement errors depend on the measurements. The estimated regression parameters are
Pn
Pn
Ĉ(δi , ²i )
i=1 (Xi − X)(Yi − Y ) −
,
α̂ = Y − β̂X.
(A.27)
β̂ =
Pn
Pn i=1
2
i=1 (Xi − X) −
i=1 V̂ (δi )
Other versions of this model, namely the BCES bisector and BCES mean regression can be
dened in terms of BCES(Y|X) and BCES(X|Y). Estimates of the variances of β̂ and α̂ for
large samples can also be calculated:
ξˆi =
(Xi − X)(Yi − β̂Xi − α̂) + β̂ V̂ (δi ) − Ĉ(δi , ²i )
V̂ (Xi ) − V̂ (δ)
ζ̂i = Yi − β̂Xi − Xξi ,
n
1 X ˆ ˆ2
V̂ (β̂) =
(ξi − ξ) ,
n2
,
(A.28)
(A.29)
(A.30)
i=1
V̂ (α̂) =
n
1 X
(ζ̂i − ζ̂)2 .
n2
(A.31)
i=1
Most observations contain some measurement error. Moreover, correlations between variables for physical or biogeochemical systems often have intrinsic scatter, which reects the
complex dependencies of these systems. The BCES method is valid in general and therefore
probably provides the least-biased estimates of regression slopes and variances. However, it
can only be used if the estimates of C(δi , ²i ) and V (δi ) are reasonably accurate.
A.4 Keeling Plots
In terrestrial carbon cycle research Keeling plots are widely used for relating changes in CO2
and δ 13 C to the isotopic signature of a source or sink of atmospheric CO2 . A prevalently
A.4. KEELING PLOTS
117
used technique for Keeling plot analysis is geometric mean regression [Pataki et al., 2003;
Miller and Tans , 2003], not least because of its simplicity (equation (A.25)). If the two basic
assumptions of the Keeling plot method hold true, i.e. only one source or sink component is
mixed to a constant background concentration and the isotope ratios of these two components
are constant, a functional model may be appropriate. Then, the GMR model may be a good
alternative if the ratio of measurement error to the data range is the same for CO2 and δ 13 C.
However, a drawback of GMR is that no calculations (apart from bootstrap methods) are
available for constructing condence intervals of the parameters. On the other hand, if not
all the scatter about the line can be explained by measurement error, for example because of
multicomponent mixing or varying isotope signals, GMR may also be an accurate structural
model.
As Miller and Tans [2003] point out, for precise CO2 and δ 13 C measurements the dominant
source of error in the Keeling plot calculations is the appropriateness of the model to the data
and not the data uncertainty. It is essential to be aware of the underlying assumptions of
the dierent regression models and the nature of the measured data to nd the best match
between model and data.
118
A. REGRESSION WITH ERROR IN BOTH VARIABLES
References
Akritas, M. G., and M. A. Bershady (1996), Linear regression for astronomical data with measurement
errors and intrinsic scatter, The Astrophysical Journal, 470, 706714.
Babu, G. J., and E. D. Feigelson (1992), Analytical and Monte Carlo Comparisons of Six Dierent
Linear Least Squares Fits, Communications in Statistics - Simulation and Computation, 21 (2), 533
549.
Barker, D. R., and L. M. Diana (1974), Simple Method for Fitting Data when Both Variables Have
Uncertainties, Amer. J. Phys., 42, 224227.
Boggs, P. T., R. H. Byrd, J. E. Rogers, and R. B. Schnabel (1992), User's reference guide for ODRPACK version 2.01 software for weighted orthogonal distance regression, National Institute of Standards and Technology, NISTIR 4834.
Bruzzone, H., and C. Moreno (1998), When errors in both coordinates make a dierence in the tting
of straight lines by least squares, Measurement Science and Technology, 9, 20072011.
Draper, N. R., and H. Smith (1998), Applied Regression Analysis, Wiley Series in Probability and
Statistics, 3rd ed., John Wiley & Sons, New York.
Fuller, W. A. (1987), Measurement Error Models, Wiley Series in Probability and Statistics, John
Wiley & Sons, New York.
Isobe, T., E. D. Feigelson, M. G. Akritas, and G. J. Babu (1990), Linear Regression in Astronomy. I.,
The Astrophysical Journal, 364, 104113.
Miller, J. B., and P. P. Tans (2003), Calculating isotopic fractionation from atmospheric measurements
at various scales, Tellus, 55B (2), 207214.
Pataki, D. E., J. R. Ehleringer, L. B. Flanagan, D. Yakir, D. R. Bowling, C. J. Still, N. Buchmann,
J. O. Kaplan, and J. A. Berry (2003), The application and interpretation of Keeling plots in terrestrial
carbon cycle research, Global Biogeochem. Cycles, 17 (1), 1022, doi:10.1029/2001GB001850.
Press, W. H., S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery (1992), Numerical Recipes in C:
The Art of Scientic Computing, 2nd ed., Cambridge University Press.
Publications
Leuenberger, M., P. Nyfeler, H. Moret, P. Sturm, and C. Huber (2000a), A new gas inlet
system for an isotope ratio mass spectrometer improves reproducibility, Rapid Commun. Mass
Spectrom., 14, 15431551.
Leuenberger, M., P. Nyfeler, H. Moret, P. Sturm, A. Indermühle, and J. Schwander (2000b),
CO2 concentration measurements on air samples by mass spectrometry, Rapid Commun. Mass
Spectrom., 14, 15521557.
Sturm, P., M. Leuenberger, C. Sirignano, R. E. M. Neubert, H. A. J. Meijer, R. Langenfelds,
W. A. Brand, and Y. Tohjima (2004a), Permeation of atmospheric gases through polymer Orings used in asks for air sampling, J. Geophys. Res., 109, D04309, doi:10.1029/2003JD004073.
Sturm, P., M. Leuenberger, and M. Schmidt (2004b), Atmospheric O2 , CO2 and δ 13 C measurements from the remote sites Jungfraujoch, Switzerland, and Puy de Dôme, France, J.
Geophys. Res., submitted.
Sturm, P., M. Leuenberger, J. Moncrie, and M. Ramonet (2004c), Atmospheric O2 , CO2
and δ 13 C observations from aircraft sampling over Grin Forest, Perthshire, UK, J. Geophys.
Res., submitted.
Acknowledgments
Many thanks to. . .
. . . Markus Leuenberger, supervisor of this thesis, for his excellent support, his constant enthusiasm and optimism, and his wealth of ideas. I am also grateful to him for having been
able to take part at several conferences and meetings throughout Europe.
. . . Thomas Stocker, head of the Climate and Environmental Physics (KUP) division and
advisor of this thesis, for oering me to work in the exciting eld of climate research.
. . . Peter Nyfeler, for all the practical solutions to technical and mechanical problems and his
help and patience with many annoyances in the lab.
. . . Hanspeter Moret for his help with electronic parts and LabVIEW programming, and HansJürg Lüthi, for glass-making.
. . . Christof Huber, Marc Filot, Francesco Luca Valentino, Marc Eyer, my fellow PhD students
in our lab, for the enjoyable collaboration and their assistance with some ask analyses.
. . . Bernhard Lehmann for the
222 Rn
measurements.
. . . Doris Rätz for her perfect organization of the KUP oce and ecient handling of any
administrative problems.
. . . Urs Beyerle, Reto Knutti, and Stefan Zoller for maintaining and improving the computer
system.
. . . Lorenz Martin for providing me with the meteorological data from the Institute of Applied
Physics, University of Bern.
. . . All members of KUP for the hours we spent drinking coee, eating birthday cakes, having
Apéros, skiing, swimming in the Aare, and much more.
. . . My parents for their support during my entire education.
. . . All the people contributing to this thesis in any conceivable, even non-scientic, way.
Curriculum Vitae
Patrick Sturm, born February 21, 1976, in Unterseen, Switzerland.
Education
19831987
Primary School Wilderswil.
19871991
Secondary School Wilderswil.
19911992
Secondary School Interlaken.
19921996
Gymnasium Interlaken, Matura Type C (Natural Sciences).
19962001
Studies in Physics at the University of Bern. Subsidiary subjects:
Mathematics and General Ecology.
3/20001/2005
Research assistant at the research division Climate and Environmental Physics, Physics Institute, University of Bern, Switzerland.
3/2001
Diploma in Physics at the research division Climate and Environmental Physics, Physics Institute, University of Bern, Switzerland. Thesis
Advisor: Prof. Dr. T. F. Stocker.
1/2005
Ph.D. at the research division Climate and Environmental Physics,
Physics Institute, University of Bern, Switzerland. Thesis Advisors:
PD Dr. M. Leuenberger and Prof. Dr. T. F. Stocker.
Conference attendances
3/2000
CarboEurope Meeting, Torgiano, Italy.
3/2001
Aerocarb First General Meeting, Autrans, France.
3/2002
Aerocarb Second General Meeting, Budapest, Hungary.
11/2002
Aerocarb Third General Meeting, Barcelona, Spain.
1/2004
Aerocarb Fifth General Meeting, Paris, France.
1/2004
1st CarboEurope-IP Integrated Project Meeting, Spoleto, Italy.
3/2004
SIBAE-BASIN Conference, Interlaken, Switzerland.
1/2005
2nd CarboEurope-IP Integrated Project Meeting, Dublin, Ireland.