Large tundra methane burst during onset of freezing
Mikhail Mastepanov1, Charlotte Sigsgaard2, Edward J. Dlugokencky3, Sander Houweling4,5, Lena Ström1,
Mikkel P. Tamstorf6 & Torben R. Christensen1

Terrestrial wetland emissions are the largest single source of the
greenhouse gas methane1. Northern high-latitude wetlands contribute
significantly to the overall methane emissions from wetlands,
but the relative source distribution between tropical and
high-latitude wetlands remains uncertain2,3. As a result, not all the
observed spatial and seasonal patterns of atmospheric methane
concentrations can be satisfactorily explained, particularly for
high northern latitudes. For example, a late-autumn shoulder is
consistently observed in the seasonal cycles of atmospheric methane
at high-latitude sites4, but the sources responsible for these
increased methane concentrations remain uncertain. Here we
report a data set that extends hourly methane flux measurements
from a high Arctic setting into the late autumn and early winter,
during the onset of soil freezing. We find that emissions fall to a
low steady level after the growing season but then increase significantly
during the freeze-in period. The integral of emissions during
the freeze-in period is approximately equal to the amount of
methane emitted during the entire summer season. Three-dimensional
atmospheric chemistry and transport model simulations of
global atmospheric methane concentrations indicate that the
observed early winter emission burst improves the agreement
between the simulated seasonal cycle and atmospheric data from
latitudes north of 606 N. Our findings suggest that permafrost-associated
freeze-in bursts of methane emissions from tundra regions
could be an important and so far unrecognized component of the
seasonal distribution of methane emissions from high latitudes.

Methane emissions from permafrost dominated tundra regions
are well documented5–7 and also recognized as considerable contributors
to the dynamics of high-latitude atmospheric methane concentrations
8,9. The scale and dynamics of growing-season methane
emissions from tundra settings have been documented mostly
through flux measurements made with low time resolution using
manual chambers5,6,10 together with some at higher time resolution
taken only during the growing season7,11,12. Here we report a data set
that extends hourly CH4 flux measurements from a high Arctic setting
into the frozen season. The measurement site is located in
Zackenberg Valley, northeast Greenland, 74.30uN 21.00u W. Six
automated chambers provided flux measurements once per hour,
in a typical fen area dominated by graminoids Eriophorum scheuchzeri,
Dupontia psilosantha and Arctagrostis latifolia. Methane concentration
in the chambers was measured by a laser off-axis integratedcavity
output spectroscopy analyser (Fast Methane Analyser, Los
Gatos Research). The instrument sensitivity is better than 10 p.p.b.;
time resolution of the primary concentration data is 1 s.
As part of the field season of the 2007 International Polar Year, the
Zackenberg research station was kept open two months longer than
normal. This gave us a chance to observe autumn and early-winter
fluxes, which showed some surprisingly high emissions (Fig. 1;
Supplementary Table 1). This very high and variable flux happened
when the active layer was gradually freezing, so CH4 that had accumulated
in this layer was probably squeezed out through the frost
action. This feature has not been observed in studies at lower latitudes,
possibly because the permafrost bottom is necessary to prevent
CH4 from diffusing downwards. The autumn fluxes varied greatly
over small distances (chambers were less than 1m apart), probably
because peat and vegetation structure provided pathways for emission
to the atmosphere. A late-autumn increase in methane emissions
was observed in one of the early tundra flux studies13, but it
lacked the time resolution needed to quantify the relative importance
for the annual flux budget.
The observed growing season emission dynamics are comparable
to earlier work at the same6,7 and at similar tundra sites12. Integrated
summer season emissions, roughly 4.5 gCH4m–2 for the season, also
match well with previous estimates for the same climatic and ecosystem
setting6,7.
Emissions decreased during September until they reached the presumed
low winter emission level (Fig. 1). However, at the onset of soil
freeze-in, a substantial increase in emissions was observed and was
sustained for several weeks, corresponding to the time required for a
complete freeze-in of the entire soil and root zone profile. Freeze-in
emissions were much more variable than summer emissions. Peak
emissions during the freeze-in period in individual chambers reached
levels of 112.5 mgCH4m22 h21, which to our knowledge are the
highest rates reported from tundra ecosystems (excluding hotspot
emissions from thermokarst lakes14), and they appear at a time when
previous assumptions would put tundra emissions at a negligible
level (see Supplementary Information for further discussion).
Earlier studies have indicated the possibility of a spring burst from
trapped methane during the winter15,16. We have early-season flux
data from Zackenberg for 2006 (M. Mastepanov et al., manuscript in
preparation) showing that spring emissions amounted to less than
2% of summer emissions (Fig. 1 insert; Supplementary Table 2), with
summer emissions being very similar for 2006 and 2007
(Supplementary Tables 1 and 2). Emissions of methane during spring
from this type of tundra environment are therefore not considered as
a major contributor to annual methane emissions.
To investigate the potential importance of the observed methane
emissions during freezing of the permafrost surface layer at large
scales, we carried out model simulations of atmospheric transport
and compared them with observations. Model-simulated methane
concentrations were sampled at the times and locations when measurements
were taken at selected background monitoring sites of the
NOAA Earth System Research Laboratory’s cooperative air sampling
network4. Average seasonal cycles were constructed from air samples
collected over the 4-year simulation period. Furthermore, background
sites were averaged into two latitudinal bands: 25–55uN
1GeoBiosphere Science Centre, Physical Geography and Ecosystems Analysis, Lund University, So¨lvegatan 12, 22362, Lund, Sweden. 2Institute of Geography and Geology, University of
Copenhagen, Øster Voldgade 10, DK-1350 Copenhagen, Denmark. 3NOAA Earth System Research Laboratory, 325 Broadway, Boulder, Colorado 80305, USA. 4SRON Netherlands
Institute for Space Research, Sorbonnelaan 2, 3584 CA Utrecht, The Netherlands. 5Institute for Marine and Atmospheric Research Utrecht (IMAU), Utrecht University, Princetonplein
5, 3584 CC Utrecht, The Netherlands. 6National Environmental Research Institute, University of Aarhus, Frederiksborgvej 399, 4000 Roskilde, Denmark.
Vol 456|4 December 2008| doi:10.1038/nature07464
628
© 2008 Macmillan Publishers Limited. All rights reserved
and 55–85uN (see Methods for a list of sites and where to access the
CH4 data). The averages represent 30-day running means of all samples
(either modelled or measured), calculated in 5-day intervals.
There is a very reasonable agreement between the reference emission
scenario (SC1) and the measurements (Fig. 2). At mid latitudes
of the Northern Hemisphere, both model scenarios nicely reproduce
the seasonal amplitude, although the phase lags the measurements by
about a month in the second part of the year. At high northern
latitudes the differences between model and measurements are more
pronounced, highlighting a deficiency of the reference model in
simulating the timing of the concentration increase from summer
to winter. Interestingly, the largest deviations occur in October when
the unrepresented emissions from permafrost are highest. The difference
between the two model simulations confirms that the influence
of the simulated permafrost emissions is considerable and does
improve the simulated seasonal cycle. Significant differences remain
between model SC2 and the measurements, but it should be kept in
mind that the underlying parameterization is only a preliminary
extrapolation of the actual flux measurements. Therefore, once additional
information on permafrost freeze-in emissions become available,
confirmation of our model results is needed on the basis of a
more sophisticated emission parameterization. Nevertheless, these
results show that CH4 emissions from the freezing active layer in
permafrost areas may be an important missing process that limits
model performance at high northern latitudes.
We also investigated whether there has been a change in the shape
of the seasonal cycle in recent years by comparing observed seasonal
cycles for the periods 1992–95 and 2002–05. The results
(Supplementary Fig. 3) demonstrate that both seasonal cycles (25–
55uN and 55–85u N) were remarkably constant over these periods,
indicating that the signature of permafrost emissions in the observed
seasonal cycle is not a recent phenomenon.
The flux measurements and atmospheric transport model results
presented here are likely to be of a general nature, as there is nothing
unique or artificial about this study site. It is situated in one of the
most pristine environments in the world (the National Park of northeast
Greenland) and there is no reason why such a physical mechanism
should not happen everywhere that there are similar
ecosystems. This study benefited from the unique opportunity
through the International Polar Year effort to keep the Zackenberg
1 July
4
8
12
16
20
Flux (mg CH4 m–2 h–1) 0
24
0
+2
+4
+6
+8
+10
–4
–2
Temperature (ºC)
+12
–6
11 July
0
2
Methane flux, daily average
Soil temperature at 5 cm depth
Soil temperature at 10 cm depth
Soil temperature at 15 cm depth
Soil temperature (climate station) at 10 cm depth
10 June 20 June 30 June 10 July 20 July 30 July 9 Aug. 19 Aug. 29 Aug. 8 Sept. 18 Sept. 28 Sept. 8 Oct. 18 Oct.
Date (2007)
28 Oct.
Date (2006)
21 July
Figure 1 | Full-season methane emission and soil temperature. Soil
temperatures at three depths shown as a coloured area between daily
minimum and daily maximum values (5, 10 and 15 cm depth as red, green
and blue). The arrows of the same colour show the date of freezing of each
horizon. Soil temperatures from the nearby climate station (light blue) are
shown for the period when on-site data are lacking. Site-average fluxes are
shown as daily mean values averaged over six individual chambers. The error
bars show standard error of mean between the chambers. The lower inserted
panel shows early-season emission in 2006 during the corresponding period
relative to the date of snowmelt in 2007 (yellow arrows indicate date of
snowmelt in the two years). The onset of the second emission peak coincides
with freezing of the upper horizon and continues to reach a maximum when
soil freezes down to 215 cm.
3 6 9 12 3
Month
–20
0
CH4 25–55º N (p.p.b.)
–40
–20 20
0
CH4 55–85º N (p.p.b.)
–40
20
Figure 2 | Comparison of measured and model-simulated latitudinally
averaged seasonal cycles of methane. Black, measurements with 2 sigma
uncertainty intervals; red, the model simulation using the reference
scenario; green, the model simulation including a representation of
additional emissions from freezing permafrost (see text).
NATURE| Vol 456|4 December 2008 LETTERS
629
© 2008 Macmillan Publishers Limited. All rights reserved
station open for longer than usual, and thus to observe a phenomenon
that has most likely been missed in other measurements around
the circumpolar north because of the difficulties of maintaining flux
measurements into the frozen season at remote high-emitting wet
tundra sites. If the fluxes measured at Zackenberg are applied to all of
0.8831012m2 of wet meadow tundra17 (disregarding possible similar
emissions from mesic tundra which covers even greater areas), it
will amount to a pulse of ,4TgCH4 from the highest latitudes at
what was previously thought to be an inactive time of year in terrestrial
ecosystems. This is in agreement with a corresponding estimate
based on the three-dimensional modelling which amounts to
3.9 TgCH4 (see Methods). This does not greatly increase emission
estimates from high northern latitudes, but it revises our view of the
seasonal distribution of known emissions.

METHODS SUMMARY
Methane emissions were measured by an automatic chamber method; flux was
calculated from the increase in the chamber CH4 concentration, corrected by air
temperature and pressure. Erroneous measurements (for instance, during strong
southerly winds that tend to cause improper closing of the chambers) were
filtered out and no artificial corrections or gap filling were applied.
Global methane concentrations were simulated using an atmospheric chemistry
and transport model, which includes a dedicated representation of the
methane cycle. Calculations were performed with and without a parameterization
of methane emissions from freezing permafrost. Simulated concentrations
were compared with high precision methane measurements representative of the
background conditions at mid to high northern latitudes.
Full Methods and any associated references are available in the online version of
the paper at www.nature.com/nature.
Received 25 April; accepted 18 September 2008.
1. Mikaloff Fletcher, S. E., Tans, P. P., Bruhwiler, L. M., Miller, J. B.&Heimann, M.CH4
sources estimated from atmospheric observations of CH4 and its 13C/12C isotopic
ratios: 1. Inverse modeling of source processes. Glob. Biogeochem. Cycles 18,
GB4004 (2004).
2. Dlugokencky, E. J. et al. Atmospheric methane levels off: Temporary pause or a
new steady-state? Geophys. Res. Lett. 30, 1992 (2003).
3. Miller, J. B. et al. Airborne measurements indicate large methane emissions from
the eastern Amazon basin. Geophys. Res. Lett. 34, L10809 (2007).
4. Dlugokencky, E. J., Steele, L. P., Lang, P. M. & Masarie, K. A. The growth rate and
distribution of atmospheric methane. J. Geophys. Res. 99, 17021–17043 (1994).
5. Reeburgh, W. S. et al. A CH4 emission estimate for the Kuparuk River basin,
Alaska. J. Geophys. Res. 103, 29005–29013 (1998).
6. Christensen, T. R. et al. Trace gas exchange in a high-arctic valley 1. Variations in
CO2 and CH4 flux between tundra vegetation types. Glob. Biogeochem. Cycles 14,
701–713 (2000).
7. Friborg, T., Christensen, T. R., Hansen, B. U., Nordstroem, C. & Soegaard, H. Trace
gas exchange in a high-arctic valley 2. Landscape CH4 fluxes measured and
modeled using eddy correlation data. Glob. Biogeochem. Cycles 14, 715–723
(2000).
8. Gedney, N., Cox, P. M. & Huntingford, C. Climate feedback from wetland methane
emissions. Geophys. Res. Lett. 31, L20503 (2004).
9. Bousquet, P. et al. Contribution of anthropogenic and natural sources to
atmospheric methane variability. Nature 443, 439–443 (2006).
10. Christensen, T. R. Methane emission from Arctic tundra. Biogeochemistry 21,
117–139 (1993).
11. Fan, S. M. et al. Micrometeorological measurements of CH4 and CO2 exchange
between the atmosphere and subarctic tundra. J. Geophys. Res. 97, 16627–16644
(1992).
12. Corradi, C., Kolle, O., Walter, K., Zimov, S. A. & Schulze, E.-D. Carbon dioxide and
methane exchange of a north-east Siberian tussock tundra. Glob. Change Biol. 11,
1910–1925 (2005).
13. Whalen, S. C. & Reeburgh, W. S. A methane flux time series for tundra
environments. Glob. Biogeochem. Cycles 2, 399–409 (1988).
14. Walter, K. M., Zimov, S. A., Chanton, J. P., Verbyla, D. & Chapin, F. S. III. Methane
bubbling from Siberian thaw lakes as a positive feedback to climate warming.
Nature 443, 71–75 (2006).
15. Hargreaves, K. J., Fowler, D., Pitcairn, C. E. R. & Aurela, M. Annual methane
emission from Finnish mires estimated from eddy covariance campaign
measurements. Theor. Appl. Climatol. 70, 203–213 (2001).
16. Tokida, T. et al. Episodic release of methane bubbles from peatland during spring
thaw. Chemosphere 70, 165–171 (2007).
17. Bliss, L. C. & Matveyeva, N. V. in Arctic Ecosystems in a Changing Climate: An
Ecophysiological Perspective (eds Chapin, F. S. III, Jefferies, R. L., Reynolds, J. F.,
Shaver, G. R. & Svoboda, J.) 59–89 (Academic, 1992).
Supplementary Information is linked to the online version of the paper at
www.nature.com/nature.
Acknowledgements This work was carried out under the auspices of the GeoBasis
programme and part of the Zackenberg Ecological Research Operations funded by
the Danish Ministry of the Environment and the ISICaB project funded by the
Commission for Scientific Research in Greenland (KVUG). ASIAQ–Greenland
Survey provided climate data. The work was also supported by the Swedish
Research Councils VR and FORMAS. We thank P. Bergamachi (JRC) and J.-F.
Meirink (KNMI) for providing the TM5 model setup. T. Tagesson helped with the
field work in Zackenberg. We are grateful for comments on earlier versions of this
manuscript from A. Lindroth and B. Christensen.
Author Contributions T.R.C., M.P.T., M.M., C.S. and L.S. designed the field
research; M.M. designed, constructed and set up the automatic measurement
system in Zackenberg; C.S. operated the system and performed manual
measurements;M.M. performed data analysis; E.D. and S.H. provided atmospheric
CH4 data and designed and ran the atmospheric transport model experiments;
T.R.C., M.M., S.H. and E.D. drafted the manuscript.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. Correspondence and requests for materials should be
addressed to T.R.C. (Torben.Christensen@nateko.lu.se).
LETTERS NATURE| Vol 456|4 December 2008
630
© 2008 Macmillan Publishers Limited. All rights reserved
METHODS
Zackenberg site description. The Zackenberg Valley is situated at 74u 309 N,
21u 009W in the National Park of northeast Greenland. A research station
(Zackenberg Research Station) was established in 1997 and offers basic logistical
facilities (an airstrip, laboratories, satellite-based communication systems and so
forth) necessary for carrying out efficient research. The site has a short record of
meteorological observations, and 1996 was the first full year when basic
meteorological variables were registered continuously. Mean annual temperature
in the first 10-year period of the station ranged from 28.5 uC to
210.1 uC with July as the warmest month (mean monthly air temperature of
5.8 uC) and February as the coldest month (mean monthly air temperature of
222.4 uC). The average frost-free period during these 10 years was 35 days, lasting
from mid July to late August. In Daneborg, situated on the outer coast 22 km
southeast of Zackenberg and with a longer period of meteorological measurements,
the mean annual temperature for the period 1960–90 was 210.3 uC. The
warmest month, July, had a mean of 3.8 uC and February, the coldest month, had
a monthly mean of 217.6 uC. The valley is dominated by minerotrophic sedgegrass-
rich fens mixed with elevated areas of dwarf shrub heaths with Cassiope
tetragona and Salix arctica as dominant species. A slightly sloping fen area is the
main study area here. The peat layer in the fen is 20–30 cm thick, typical of high
arctic fen ecosystems18. Onset of peat accumulation has been 14C-dated to AD
1290–1390 in a neighbouring fen area, and the surrounding Little Ice Age nival
fans and nivation basins primarily contain organic material deposited from AD
1420 to 1500–158019.
The active layer depth specifically on the measurement site reached 50–56 cm
(near different chambers) before soil freezing in 2007. Despite the low temperatures,
the snow cover was mosaic until 20 October, and then was no more than
3 cm deep until the first snowstorm on 26 October.
A large body of background information from the Zackenberg Research
Station has recently been summarized in a book volume celebrating the first
10 years of activities at the research station18.
Methane flux measurements. Automatic chambers were deployed in August
2005 and the first seasonal data set was obtained in 2006 (3 July to 26 August). In
2007 an extended season was carried out (26 June to 25 October). Six chambers
have been aligned in a row from the periphery of the fen towards its central part.
The distance between the chambers is 30–80 cm. The chambers are made of
Plexiglas with aluminium corners. Each chamber is 60360 cm and about
30 cm height (depending on microtopography). The chamber lid stays open
for 55 min per hour and closes for five minutes for the measurements. Air is
mixed in a closed chamber by a fan; the same fan ventilates a chamber when it is
open. Air from the chamber passes through 30m of tubing (internal diameter
4 mm) to the analytical box and after the non-destructive analysis it goes back to
the chamber. The analytical box contains a methane analyser (Fast Methane
Analyser, Los Gatos Research), CO2 analyser (SBA-4, PP Systems) and solenoid
valves. The concentration data are collected at 1 Hz rate; data acquisition starts
three minutes before a chamber closes, continues for five minutes while it is
closed, and then two minutes after the chamber opens, so the full cycle of six
chambers takes one hour.
Although we did not make direct measurements of soil temperature and
humidity inside the chambers, to avoid extra disturbance, the visual control
does not give any evidence of the construction affecting the temperature and
water regime inside the chambers. Visible water table and the snow level (during
the snowfall and snowmelt) is the same inside and outside the chambers.
The CH4 fluxes are calculated from the slope of concentration change in the
closed chamber; if the increase was not linear during five minutes of closure, the
most linear part of this time is arbitrarily chosen. The air temperature and
pressure for flux calculations are obtained from Zackenberg micrometeorological
station located about 1km from the site. We made additional measurements
of water table level, active layer depth, PAR, soil temperature and humidity next
to the chambers.
AtmosphericCH4 measurements and chemical transport model. Atmospheric
CH4 measurements are from weekly samples collected at sites in theNOAAEarth
System Research Laboratory’s cooperative global air sampling network4. We
determined methane dry-air mole fractions by gas chromatography with flame
ionization detection against the WMO CH4 mole fraction standard scale. Over
the period of this study, repeatability of the measurements (1s) was ,2 p.p.b.
Latitudinal averages contained the following sites: 25–55uN contained ‘mid’,
‘bme’, ‘bmw’, ‘uta’, ‘mhd’, ‘ask’, ‘nwr’, ‘pta’, ‘azr’, ‘izo’, ‘pocn30’ and ‘pocn25’,
and 55–85uN contained ‘alt’, ‘shm’, ‘brw’, ‘ice’, ‘zep’, ‘stm’, ‘cba’ and ‘sum’ (see
www.esrl.noaa.gov/gmd/ccgg/flask.html for a list of site codes and ftp://
ftp.cmdl.noaa.gov/ccg/ch4/flask/event for access to data).
We made atmospheric transport model calculations using the TM5 model20
for the period 2002–05, at a spatial resolution of 6u34u and 25 vertical sigmapressure
levels. Two scenarios of methane sources and sinks were applied: a
reference scenario (SC1) and a scenario including emissions from permafrost
freeze-in (SC2). The methane sources and sinks of SC1 correspond with the a
priori assumptions that were used in the inverse modelling calculations of ref. 21,
with the exception of wetlands. Wetland emissions were taken from ref. 22. and
rescaled to a global total of 175 TgCH4 yr21 and high-latitude (50–90u N) emissions
of 20 TgCH4 yr21. SC2 is the same as SC1 except for additional emissions
from freezing permafrost. For lack of any detailed information on this process,
we followed a highly simplified procedure, assuming that emission started when
the diurnal mean temperature dropped below22 uC and continued for a period
of 1 month. This process was only active in those model grid boxes that were
classified as continuous or discontinuous permafrost according to the CAPS
circumpolar permafrost map23. The annual emission of freezing permafrost
was assumed to be the same as the (summer time) wetland emission in each
model grid box for which the process is active. This procedure introduces an
additional source of 3.9 TgCH4 yr21, which moves from north to south during
autumn and reaches maximum global emissions in October.
18. Meltofte, H., Christensen, T. R., Elberling, B., Forchhammer, M. C. & Rasch, M.
(eds) High-Arctic Ecosystem Dynamics in a Changing Climate. Advances in Ecological
Research Vol. 40 (Elsevier, 2008).
19. Christiansen, H. H. et al. Holocene environmental reconstruction from deltaic
deposits in northeast Greenland. J. Quat. Sci. 17, 145–160 (2002).
20. Krol, M. C. et al. The two-way nested global chemistry-transport zoom model
TM5: algorithm and applications. Atmos. Chem. Phys. 5, 417–432 (2005).
21. Bergamaschi, P. et al. Satellite chartography of atmospheric methane from
SCIAMACHY on board ENVISAT. 2. Evaluation based on inverse model
simulations. J. Geophys. Res. 112, D02304 (2007).
22. Walter, B. P., Heimann, M. & Matthews, E. Modeling modern methane emissions
from natural wetlands. 2. Interannual variations 1982-1993. J. Geophys. Res. 106,
34207–34217 (2001).
23. Brown, J., Ferrians, O. J. Jr, Heginbottom, J. A. & Melnikov, E. S. Circum-arctic Map
of Permafrost and Ground-Ice Conditions. USGS Circum-Pacific Map Series CP-45
(scale 1:10,000 000) (US Geological Survey, 1997).
doi:10.1038/nature07464
© 2008 Macmillan Publishers Limited. All rights reserved