Introduction
Dichloromethane, CH2Cl2, is a short-lived chlorocarbon of mainly
(up to 90 %, Montzka et al., 2011b) anthropogenic origin. Its main
applications include use in paint strippers, degreasers and solvents; in
foam production and blowing applications; as a chemical feedstock; and as an
agricultural fumigant (Montzka et al., 2011b). The contribution
from natural sources (mainly biomass burning and an oceanic source) is
uncertain. Simmonds et al. (2006) obtained a good model fit to their
observations using a 10 % combined oceanic and biomass burning source,
although they showed that a stronger terrestrial source could support
natural emissions of up to 30 %. However, recent field measurements of
biomass burning plumes have indicated that this source is likely to be
smaller than previously estimated (Simpson et al., 2011). With an
atmospheric lifetime of around 5 months (Montzka et al.,
2011b), CH2Cl2 displays significant atmospheric spatial variations
and temporal trends. Its seasonal cycle is mainly due to reaction with the
OH radical, with maxima in late winter/early spring and corresponding minima
in late summer or early autumn (Cox
et al., 2003). There are no discernible seasonal variations in emissions or
end uses (Gentner et al., 2010; McCulloch and
Midgley, 1996). Significantly higher concentrations are observed in the
Northern Hemisphere (NH, Southern Hemisphere = SH) due to the predominant
industrial source. A NH : SH mole fraction ratio of 2.7 has been reported for
the lower troposphere
(Simmonds
et al., 2006).
Short-lived chlorocarbons, including CH2Cl2, contribute to
stratospheric chlorine and its cycling with stratospheric ozone. Their
current contribution is minor – Laube et al. (2008) found that at 15.2 km (the level of zero radiative heating)
1.4 % of chlorine from organic compounds was from short-lived
chlorocarbons, of which half was from CH2Cl2. This level is
important because air parcels at or above this level point are likely to be
transported to the stratosphere. However, current and projected decreases of
longer-lived anthropogenic chlorocarbons (such as CH3CCl3,
CCl4, halons and CFCs) could mean a greater relative importance of
shorter-lived chlorocarbons with respect to stratospheric chlorine,
especially if their atmospheric abundances were to increase. Due to its
predominantly anthropogenic source CH2Cl2 is susceptible to
changes in industrial emissions. CH2Cl2 is also of concern as it
is also a toxic air pollutant and possible carcinogen and, as such, is
regulated by national and European Union Law, for example the Solvent
Emissions Directive, 1999/13/EC (E.C.S.A., 2007).
The earliest reported NH atmospheric measurements of CH2Cl2 were
made in the mid-1970s and observed concentrations of 35 ± 19 ppt
(Cox et al., 1976). A range of global measurements in the 1980s
and 1990s (many of which will be discussed further throughout this
paper and are included in Table 2, see also Simmonds et al. (2006) for
an in-depth discussion of many of the observations during this period)
showed a consistent picture of peaking concentrations, with an average of
∼ 30–40 ppt around 1990, followed by a decline linked to
decreasing industrial use of CH2Cl2 (McCulloch et al.,
1999). Measurements made between 1989 and 1996 at Alert, Canadian Arctic,
observed a decline of around -4 % (-1.8 ppt) per year
(Gautrois et al., 2003). Long-term measurements
(1995–2004) at Mace Head, Ireland demonstrated a decline in CH2Cl2
pollution events since measurements began in 1995, although this decline had
stabilised in the later years of the data set (Simmonds et al., 2006). In the
SH, Advanced Global Atmospheric Gases Experiment (AGAGE) atmospheric
measurements began at Cape Grim in 1998 and reported mean CH2Cl2
concentrations for 1998–2000 of 8.9 (±0.2) ppt
(Cox et al., 2003). These early
measurements were supported by firn records which indicated that SH
CH2Cl2 concentrations increased from 1–2 ppt at the beginning of
the record (pre-1940) to 9 ppt around 1990
(Trudinger et al., 2004). Due to the
lack of industrial emissions in the SH the rapid decline in atmospheric
concentrations seen in the NH was not observed in the AGAGE Cape Grim time
series (Simmonds et al., 2006).
In recent years increasing CH2Cl2 concentrations have been
observed in both the NH and SH. Montzka et al. (2011b) reported
an increase of around 8 % between 2007 and 2008, based on updated AGAGE
data from Simmonds et al. (2006). There was no corresponding increase in
CHCl3, 70 % of which is believed to be of natural origin
(Worton et al., 2006). The increase was
also noted in Montzka et al. (2011a, see
their Supplementary Information) whose time series of CH2Cl2
between 1995 to 2009 shows increasing atmospheric concentrations in recent
years. CARIBIC (Civil Aircraft for the Regular Investigation of the
atmosphere Based on an Instrument Container) CH2Cl2 measurements
up to the end of 2012 provide the opportunity to investigate this increase
from a global time series perspective and may help improve our understanding
of the recent changes in atmospheric CH2Cl2.
Methods
The CARIBIC platform and whole-air samples
CARIBIC centres on a large air-freight container accommodating a range of
scientific equipment which is deployed monthly aboard a commercial passenger
aircraft departing from Germany for up to four consecutive long-haul
flights. Details of both CARIBIC phases can be found on our website,
caribic-atmospheric.com. CARIBIC phase 1 (CARIBIC1) operated between 1997
and 2002 aboard a Boeing 767 departing for several global
destinations from either Düsseldorf or Munich airport. Whole-air samples
were collected using twelve 21 L stainless steel tanks pressurised to 17 bar.
Details of CARIBIC1, including the range of other measurements made, can be
found in Brenninkmeijer et al. (1999). Halocarbon data are
available for 1998–2002.
Between 2003 and 2005 a new container was developed and this system was
deployed aboard a Lufthansa Airbus 340–600 departing from Frankfurt Airport.
CARIBIC phase 2 (CARIBIC2) began in 2005 and, at the time of writing, is
still in operation. Samples are taken en route to destinations across the
globe with flights occurring approximately monthly. Two whole-air samplers
consisting of fourteen 2.7 L glass flasks collect 28 air samples for halocarbon,
non-methane hydrocarbon (NMHCs) and greenhouse gas measurements at
pre-determined intervals during the flight, mainly within cruising altitudes
of around 10–12 km. Filling times are between 30–90 s, averaging 45 s or 10 km of the flight path. Further air sampler information can be found in
Baker et al. (2010) and
Schuck et al. (2009). The fully automated CARIBIC2 system contains a range
of other sampling equipment, including, but not limited to, equipment for
the in situ or post-flight analysis of ozone (O3), carbon monoxide
(CO), aerosols and water vapour. Further information can be found in
Brenninkmeijer
et al. (2007).
Halocarbon analysis
During both CARIBIC1 and CARIBIC2 air samples were sent to the University of
East Anglia (UEA, UK) for halocarbon analysis via gas chromatography mass
spectrometry (GCMS). During CARIBIC1 subsamples were removed from the main
canisters into electropolished stainless steel cans and sent to UEA. For
CARIBIC2 the whole-air sampling units were sent directly to UEA for
analysis. During CARIBIC1 two separate GCMS systems were used. The first was
an Agilent/HP 5890A GC coupled to a double-focusing, tri-sector mass
spectrometer (VG/Micromass Autospec). Each 200 mL air sample was dried by
passing through magnesium perchlorate (MPC) before being trapped in a
previously evacuated stainless steel loop filled with 100 µm glass
beads and immersed in liquid argon (-186 ∘C). The bulk of the air
passed into an evacuated stainless steel flask where the pressure change,
and hence sample volume, was measured with a capacitance manometer (Edwards
Barocel). The MPC trap was shown to have no effect on the measured
CH2Cl2 concentration. Halocarbons were separated on a
60 m × 0.53 mm (1.5 µm film thickness) DB5 capillary column (J&W), with helium
carrier gas (2 mL min-1) and a temperature program of -20 ∘C
(2 min) rising to 220 ∘C, at a rate of 15 ∘C min-1. The mass
spectrometer was operated in selected ion mode (voltage switching) using
electron ionisation (EI). Each air sample was analysed at least twice, with
a working standard analysed before and after each sample pair to allow
correction for small changes in instrument response.
The VG Autospec system was used from the start of analysis at UEA in June
1998 until December 1999. In 1999 a new instrument (“Entech”) was purchased
by UEA and became the main instrument for CARIBIC sample analysis. This
system consisted of an Agilent 6890 GC and 5973 quadrupole MS. With this
system pre-concentration was achieved using a commercial, fully automated,
three-stage pre-concentrator (Entech Instruments, model 7100). This system was
used throughout the rest of CARIBIC1 and CARIBIC2. The Entech
pre-concentrator employs multiple traps to remove water (Trap 1), CO2
(Trap 2) and to cryo-focus the sample prior to injection into the GC (Trap
3). Typically, between 800–1000 mL of air are trapped at 100 mL min-1
onto a 1/8” (external diameter, OD) stainless steel trap (Trap 1) packed
with glass beads and held at -150 ∘C. The contents of Trap 1 are then
swept onto Trap 2, consisting of 1/8” OD stainless steel packed with Tenax
adsorbent and held at -40 ∘C. Trap 3 cryo-focuses the sample on a fused
silica lined stainless steel tube (1/32” OD). Until 2010 a DB-5 capillary
column (J&W Scientific, 105 m × 0.32 mm ID, 1.5 µm film thickness)
was used for separation. In 2010 the column was changed to an Agilent
GC-GasPro column (30 m × 0.32 mm). The temperature program used with the
DB-5 column was 30 ∘C for 8 min rising to 220 ∘C at a rate of 10 ∘C min-1. The temperature program used with the GasPro column is
-10 ∘C for 2 min rising to 200 ∘C at a rate of 10 ∘C min -1. The MS is operated in selected ion mode using EI
at 70 eV. The system allows for the unattended analysis of up to 16 samples,
interspersed with equal volume aliquots of a working standard analysed at
regular intervals. Each sample is normally analysed only once and, as the
response of the quadrupole analyser is more stable than the Autospec, the
working standard is analysed less frequently.
To assist with the transition between the VG Autospec and the Entech
system parallel analysis was conducted for two flights in July 2000.
Agreement between the two systems was excellent. Of the 24 samples analysed
on both systems all but five had a difference of less than ±1 ppt
(corresponding to a difference of < 3 % standard deviation,
σ, or less than the precision of these instruments). For the five
remaining samples the difference was less than ± 2 ppt. The
CH2Cl2 samples that were analysed on both systems were treated in
the following manner. If the difference was less than ± 1 ppt
(3 %σ) the values were averaged and the variation between the two
measurements incorporated into error bars plotted with these values. Where
the difference was greater than ± 1 ppt the VG Autospec value was
selected based on the better precision of this instrument with respect to
CH2Cl2.
To provide additional support to the CARIBIC2 data set, three flights, one
a year between 2009 and 2011, were also analysed on a highly sensitive Waters
Autospec magnetic sector GCMS. This system is the direct replacement of the
VG Autospec described above and, whilst a number of minor modifications
have been made to the analytical procedure (see
Laube et al., 2010), the system is
essentially the same. Where the Entech and Autospec values agreed within
±1σ (based on replicate Autospec measurements) the values
were combined. As with CARIBIC1, these values all agreed within ±1 ppt. For the remaining samples the values from the higher precision Autospec
system have been used. The limit of detection for all three analytical systems
was 0.1 ppt or better.
For CH2Cl2 the UEA calibration is tied to the 2003 GCMS
gravimetric scale of the Global Monitoring Division of the Earth System
Research Laboratory of the National Oceanic and Atmospheric Administration
(NOAA-ESRL-GMD) in Boulder, CO, USA. A number of calibrated, high-pressure
whole-air samples collected at Niwot Ridge (a remote site near Boulder) were
acquired between 1994 and 2009. These were used for the propagation of
mixing ratios to all CARIBIC measurements. Further details on this procedure
can be found in the Supplement. The CH2Cl2 data are
reported on the latest (2003) NOAA-ESRL calibration scale. NOAA do not
provide an absolute accuracy on their calibrated gas standards but, in a
recent international comparison exercise (IHALACE), the mean of the
CH2Cl2 calibration scales from the three independent calibration
laboratories was found to have a standard deviation of ± 9 % (Hall
et al., 2014). In Sect. 3.1, CARIBIC data are compared to the long-term
CH2Cl2 record from Mace Head (53.3∘ N, 9.9∘ W,
42 m a.s.l.) measured by NOAA-ESRL-GMD. These data are obtained
from regularly collected flasks samples analysed by GCMS. Sampling at Mace
Head is done in a manner to characterise only air that is arriving from the
clean air sector, specifically when wind direction is between 180
to 320∘ and the wind speed is greater than 4 m s-1. For
further information see Montzka et al. (2011a) and http://www.esrl.noaa.gov/gmd/hats/gases/CH2Cl2.html. A
comparison between NOAA and UEA calibration scales is discussed in the
Supplementary Information which provides a comparison of data from Cape
Grim, a ground-based site sampled by both groups. Cape Grim samples analysed
by both groups compare well (65 % agree within the respective 1σ
standard deviations), with no apparent offset or change in the relationship
between both groups' results over time (NOAA / UEA ratio of 1.02 ± 0.06).
Throughout this paper we refer to the dry air mole fraction of
CH2Cl2 as “concentrations” to increase the accessibility and
readability of this paper.
The analytical precision during CARIBIC1 was 0.9 % for the VG Autospec
(based on repeat analysis of randomly selected samples, 1998–2002) and
2.4 % for the Entech (based on repeat analyses of the working standard,
1999–2002). During CARIBIC2 the Entech system was managed by several
operators and the analytical precision was calculated for each of these
periods, again based on repeat analysis of randomly selected samples or
repeat analysis of the working standards. Average precision was 3.42 %
between May 2005 and September 2006, 4.0 % between October 2006 and October 2009,
5.5 % between November 2009 and October 2012 and 3.3 % in November and
December 2012. Average precision for the Autospec system during CARIBIC2 was
0.48 %. The final data set used in this study is from 1998–2002 and
2005–2012.
Ancillary measurements – CO, O3 and back-trajectories
With a typical cruise altitude of 10–12 km CARIBIC intercepts air of both
tropospheric and stratospheric origin. Data were labelled to indicate if
they were of mainly tropospheric or stratospheric origin based on a chemical
definition of the tropopause. O3 is measured in situ onboard the
CARIBIC platform (see Sprung and Zahn, 2010) and therefore provides a
measure of upper troposphere/lower stratosphere (UTLS) structure with a
temporal and spatial resolution more suited to the discrete whole-air
samples than parameters derived from meteorological analyses, such as
potential vorticity. Samples were classed as being predominantly
stratospherically influenced if the integrated O3 mixing ratio for that
sampling period was above a seasonal threshold determined by Eq. (), a
method derived from CARIBIC data by Zahn and Brenninkmeijer (2003), confirmed by Thouret et al. (2006) and used as
part of CARIBIC halocarbon analysis by Wisher et al. (2014):
O3tropopausein ppbv=97+26sin2πDay of Year-30365.
A detailed discussion of O3 as a chemical marker for the structure of
the UTLS is provided by Zahn and Brenninkmeijer (2003) and
Sprung and Zahn (2010). Briefly, the extratropical O3 chemical
tropopause is observed around 100 ppbv O3 and can be seen in changes in
the relationship between O3 and tropospheric tracers such as CO and
acetone. Above the chemical tropopause, a compact “mixing line” between
O3 and, for example, CO, denotes the mixing of tropospheric and
stratospheric air in the extratropical tropopause layer (ExTL). The ExTL
extends up to a maximum of 400–500 ppbv O3, above which lies the
lowermost stratosphere (LMS). Figure 1 shows which samples were classed as
being of predominantly tropospheric origin and which were stratospheric. As
tropospheric trends in CH2Cl2 form the focus of this
investigation, stratospherically influenced samples (which comprised between
∼ 6–40 % of samples, depending on the region) were excluded
from the bulk of the discussion for each region. Vertical profiles
incorporating stratospheric samples are discussed in Sects. 3.4 and 3.5.
CARIBIC measurements of CO were used during the analysis of CH2Cl2
measured on flights to South Africa and India (Sects. 3.2 and 3.3
respectively). Details of CO measurements can be found in
Brenninkmeijer et al. (1999) for CARIBIC1 and
Scharffe et al. (2012) for CARIBIC2. For comparison with
the whole-air samples the CO values (produced every 2 s) were integrated
over the sampling period of each whole-air sample. Back-trajectory analyses
for CARIBIC flights are provided by the Royal Netherlands Meteorological
Institute (KNMI) – further details can be found at
http://knmi.nl/samenw/campaign_support/CARIBIC/ or in
Scheele et al. (1996). The trajectory model used European Centre
for Medium range Weather Forecasting (ECMWF) data at a 1∘ × 1∘ resolution to calculate both 5-day back-trajectories at
3 min intervals along the flight track and 8-day back-trajectories for the
collection interval of each whole-air sample. During the early CARIBIC
flights ECMWF “first guess” fields were used to calculate the back-trajectories, changing to re-analysis data after September 2000.
CARIBIC whole-air samples analysed for CH2Cl2
between 1998–2002 and 2005–2012. The grey points show all the
CH2Cl2 samples collected by CARIBIC during these periods. The
subset of data used in this study are also shown: (1) the samples within the
European region described in Sect. 3 are marked by the black box, (2)
flight routes to Africa, India and Central America are shown by coloured
samples, see inset legend, and (3) samples within the tropical region are
delineated by the dashed lines. More details of the regions used in this
study can be found in Table 1.The identification of samples as tropospheric
or stratospheric is described in Sect. 2.
Results and discussion
Between 1998 and 2012 CARIBIC flights covered a substantial area of the
global free troposphere (Fig. 1). However, for the purpose of this
investigation several regions were selected and only the data from these
flights will be discussed. These regions and the rationale behind their
selection are described here. Firstly, a European region within a box
spanning 40 to 55∘ N and -10 to 20∘ E (Frankfurt
Airport = 50.03∘ N, 8.57∘ E) was selected as it is
the area with the greatest temporal coverage. Secondly, routes to South
Africa, India and across the North and Central Atlantic were also chosen as
these routes were traversed by CARIBIC over multiple years, allowing changes
over time to be observed in these regions. Finally, samples collected in the
tropical region covering 25∘ S to 25∘ N were used to
investigate the concentration of CH2Cl2 in air masses with the
potential to enter the tropical tropopause layer (TTL). Further details of
these five case studies can be found in Table 1 and they are highlighted in
Fig. 1. Throughout the paper mean values prefaced by ± refer to
the 1σ standard error associated with that mean value.
Long-term time series of
CH2Cl2 measured over Europe
The CH2Cl2 time series of European CARIBIC and NOAA Mace Head data
can be seen in Fig. 2a. A fairly consistent seasonal cycle is observed
inter-annually in the boundary layer air samples from Mace Head whereas the
CARIBIC data show greater variability. This variability in the CARIBIC data
is mainly because these samples represent a wide variety of air masses
sampled over a large area (Fig. 2c) compared to clean-sector air sampled at
Mace Head (Sect. 2.2). Analysis of back-trajectories indicates that air
sampled by CARIBIC over Europe originates from a large NH geographical
region, including industrial areas where high emission “pollution” events
may occur as well as contrasting regions where pristine tropospheric air
masses are sampled. In contrast, the Mace Head site commonly samples clean
sector air. Although a previous study involving data collected by aircraft
at an average altitude of 4 km (Miller et al., 2012)
observed seasonality in atmospheric concentrations of CH2Cl2, we do
not see a strong seasonal pattern at 10–12 km in our more sporadic data set.
Further analysis, discussed in subsequent sections, will highlight the
importance of strong source regions (e.g. India and Southeast Asia) on
observed CH2Cl2 concentrations in the mid- and upper troposphere.
A summary of CARIBIC data used in this study.
Region
Temporal data coverage
Number of tropospheric (stratospheric∗) samples
Annual
Monthly
Europe
Data collected 1998–2002 and 2005–2012 Min. coverage: n=1 in 1999 Max. coverage: n=16 in both 2009 and 2011
All months covered
123 (87)
Africa
2000, 2009–2011. Individual flights:
Mainly NH winter
140 (21)
2000
May, Jul, Dec
32 (4)
2009
Mar, Oct
47 (8)
2010
Nov, Dec
22 (6)
2011
Jan, Feb, Mar
39 ()
India
Data collected 1998–2001, 2008, 2011 and 2012. The summer monsoon (July, Aug and Sept) was sampled in 1998, 1999, 2000 and 2008. In other years samples were only taken outside of the monsoon season.
295 (134)
Monsoon
Jul, Aug, Sept
105 (38)
Non-monsoon
Rest of year
190 (95)
North and Central Atlantic
Data collected 2001–2002, 2007, 2009–2012
All months covered
282 (108)
Tropics
All data within ± 25∘ of equator. All years included.
All months covered
539 (34)
∗ Stratospheric samples have been excluded from the analysis of most
regions, see Sect. 2.
(a) European CH2Cl2 time series from June 1998
to December 2012. Where a sample was analysed multiple times an error bar is
given based on the variation between these measurements, see Sect. 2.
(b)
The mean value of all CH2Cl2 values taken within this region for
each year (“annual tropospheric value”), shown separately for CARIBIC and
NOAA Mace Head data, error bars are 1σ. (c) Geographical distribution
of CARIBIC samples within the European region 40–55∘ N and -10–20∘ E. Frankfurt Airport (CARIBIC2 base) and the NOAA sampling site
at Mace Head are also shown. Samples are coloured by year.
CH2Cl2 descriptive statistics for regions included in this study and a comparison
to data from the existing literature.
Study and Region
Time perioda
Valuesb/ ppt
This paper
Europe CARIBIC data Mace Head NOAA data
(a)c 1998–2002 (b) 2009–2012 Increase (a→b) (a) 1998–2002 (b) 2003–2004 (c) 2009–2012 (d) 2011–2012 Increase (a→c) Increase (a→d)
x¯=24.6,σ= 4.3, n=21 x¯=38.6,σ8.4, n=49 14 x¯=32.9,σ= 5.4, n=28, x¯=33.6,σ= 5.1, n=48, x¯=45.7,σ= 6.1, n=88 x¯=47.0,σ= 6.5, n=45 13 13.5
Africa Above 30∘ N Below 30∘ N
2000 2009–2011 2000 2009–2011 Increase
x¯=21.7,σ= 1.3, n=3 x¯=34.2,σ= 7.6, n=23, x¯=15.8,σ= 3.1, n=29 x¯=22.7,σ= 5.1, n=85 7
India Summer monsoon (Jul–Sept) Non-monsoon months
1998–2000 2008 Increase (a) 1998–2000 (b) 2008 (c) 2011–2012 Increase (a→b) Increase (a→c)
x¯=21.5,σ= 3.8, n=56 x¯=36.4,σ= 9.4, n=50 15 x¯=20.1,σ= 4.9, n=81 x¯=30.4,σ9.7, n=62 33.4, σ= 15.6, n=30 10 13
North and Central Atlantic
2000–2002 2009–2011
x¯=23.2,σ3.6, n=89 x¯=32.0,σ7.8, n=180
Other aircraft studies
TTL (Schauffler et al., 1993)
1991–1992
x¯=14.9,σ= 1.1, n=12
Tropical Indian Ocean 1.2–12.5 km altitude (Scheeren et al., 2002)
1999
x¯=29, SD = 12, n=71
ASM outflow, E. Mediterranean 6–13 km alt. (Scheeren et al., 2003)
2001
x¯=23,σ= 3
Canada & Greenland, commonly 0.8–4.7 km but up to 12 km alt. (Simpson et al., 2011)
2008
x¯=35.8 ± 2.9
Tropics (0 ± 25∘ ), 345–350 K θ band (HIPPO, Wofsy et al., 2012)
2009–2011
x¯=26.3, R=15.87-49.83, n=20
Ground-based
Atlantic cruise 45∘ N–30∘ S (Koppmann et al., 1993) SH NH
1989
x¯=18 ± 4 x¯=36 ± 6 ppt
Alert, Canada (Gautrois et al., 2003)
1989–1996
x¯=47.2 ± 2, x̃=45.8 r=24.2-71.6
Cape Grim, Tasmania (Cox et al., 2003)
1998–2000
x¯=8.9 ± 0.2
Chinese cities (Barletta et al., 2006) Background Urban
2001
x¯=28,σ= 4 x¯=226,σ= 232
a Time period should be viewed alongside Table 3 which shows the
seasonal distribution of samples.
b x¯= mean, x̃= median, σ= standard
deviation, R= range and n= number of samples.
c See Sect. 3.1 for a description of the different time periods
selected for Europe.
The trend in European observations of CH2Cl2 is shown in Fig. 2b.
Error bars represent the 1σ variation associated with the mean of
all tropospheric samples taken within each year (hereafter referred to as
the annual tropospheric value). As seen in Fig. 2a, only a small number of
NOAA samples were collected in the first few years of the data set. Due to
this small sample size, biases, for example the influence of seasonality,
could be introduced (see Table 3 for seasonal distributions). This adds an
additional, unquantified uncertainty to these annual values. To account for
this, data from Mace Head can be compared to data collected at other NOAA NH
sites such as Barrow, Alaska. Data from Barrow show a very similar pattern
to those from Mace Head and support the trend seen at Mace Head (data not
shown but available from
http://www.esrl.noaa.gov/gmd/hats/gases/CH2Cl2.html
and published in Montzka et al., 2011a).
Determining trends for the first few years of the database is hard due to
the reduced data coverage. However, NOAA data covers the whole year from
2003 onward (Fig. 2a).
The NOAA Mace Head data show no trend between 2003 and 2006 (linear fit is
displayed in Fig. 2b) with a steadily increasing trend from 2006. Despite
the 1σ annual error bars being relatively large, due to the seasonal
variation seen at the Mace Head boundary layer site, a linear fit for the
NOAA data between 2006 and 2012 shows a strong positive correlation (see Fig. 2b, r2=0.97). The increase between the mean of all values collected
within the first 5 years (1998–2002) and the final 4 years (2009–2012)
of the NOAA data set was from 32.9 ppt (σ=5.4, sample size, n=28)
to 45.7 ppt (σ=6.1,n=88), an increase of ∼ 13 ppt. A test of the robustness of the trend is to compare the change from
2003–2004 (the first period with coverage across the entire year) to
2011–2012. During this period CH2Cl2 increased from 35.6 ppt
(σ=5.1,n=48) to 47.0 ppt (σ=6.5,n=45), an increase
of 13.5 ppt. CARIBIC data broadly mirror the increasing trend between
2006–2012, bearing in mind the more sporadic nature of the data set and the
wider distribution of air mass sources, discussed above. The increase
between the mean of all values collected between 1998–2002 and 2009–2012 was
also similar to that seen at Mace Head, increasing from 24.6 ppt (σ=4.3,n=21) to 38.6 ppt (σ=8.4,n=49), an increase of 14 ppt.
Whilst the overall trends are similar, mean annual values are higher at Mace
Head (Fig. 2b). This is likely to be because the NOAA samples were collected
at a lower altitude than the CARIBIC samples. Calibration scales between
NOAA and UEA compare well, as described in Sect. 2 and in the Supplement. Vertical profiles of CH2Cl2 are discussed in
Sects. 3.4 and 3.5.
Flight-based measurements in the NH were made
by Simpson et al. (2011)
(comparative data from the literature is outlined in Table 2 for all
regions). Although they flew up to 12 km their flight altitudes were
generally lower than CARIBIC (0.8–4.7 km). They reported a 2008 summer
average of 35.8 ± 2.9 ppt over Canada and Greenland. CARIBIC mean
values for the 3 summers around this time (2007, 2008 and 2009) were around
35, 27 and 37 ppt respectively, however as sample sizes were small (n=6, 4
and 6 respectively) we combined these three summer periods to obtain a mean
for summers 2007–2009 of 33.9 ± 2.2 ppt. Our value is similar to that
of Simpson et al. (2011), the higher values over industrial Europe possibly
offsetting some of the decrease we would expect with altitude.
Flights to Africa – investigating biomass burning emissions and
NH : SH gradients
Flights to South Africa (Table 1) allow us to investigate the NH : SH gradient
in CH2Cl2, see Fig. 3a. A strong latitudinal gradient is observed,
in accordance with a strong industrial NH source of
CH2Cl2. The increase is largest over the northern section
of the flight route, which crosses Europe. We have a limited data set (n=3)
for the section of the flight route that crosses Europe (here defined as
> 30∘ N to provide a clear delineation between samples
taken over Africa and those taken within our European box, see Fig. 1) and
so we do not wish to quantify an increase over time, but from the limited
data available it is similar to that seen for our European box in Sect. 3.1. During the previous peak in CH2Cl2 concentrations (around
1990, see Sect. 1) a difference of ∼ 18 ppt between NH and
SH average concentrations was observed in the Atlantic region by
Koppmann et al. (1993) (Table 2). Whilst our
results do not provide full SH coverage, and so cannot be used to estimate a
NH : SH ratio, they do show an increasing latitudinal variation indicative of
increasing NH industrial activity with respect to CH2Cl2.
(a) Latitudinal distributions of CH2Cl2 observed
during flights to South Africa where colour = year (see inset colour bar,
colour scale is consistent with Figs. 2, 4 and 6). Where multiple
measurements of the same sample have been made 1σ error bars are
given, see Sect. 2. (b) Annual tropospheric values (see Fig. 2) with
1σ error bars. These average values have been split into above and
below 30∘ N.
Annual tropospheric values are shown in Fig. 3b, separated into samples
taken above and below 30∘ N. The 44 % increase seen at
latitudes < 30∘ N is smaller than that seen over Europe,
although still statistically significant (Mann-Whitney test at p < 0.001). Further details of the concentrations
observed above and below 30∘ N are provided in Table 2. Inferring year-on-year trends is
difficult given the varying data coverage between years. However, the
increase seen between 2009 and 2011 (Fig. 3b), along with the European
data set (Fig. 2), suggest that concentrations continue to increase into the
2010s.
Despite the importance of biomass burning with respect to atmospheric trace
gas emissions over Africa (Roberts et al., 2009) no correlation
(r=0.14,p > 0.05) was observed between CH2Cl2 and the
common combustion tracer CO. Enhancements of CO, which commonly peak near
the equator in CARIBIC data (Umezawa et al., 2014), are predominantly from
biomass burning sources. In contrast to the latitudinal distribution of CO,
CH2Cl2 decreased constantly from north to south (Fig. 3).
Observations of CH2Cl2 along the CARIBIC flight track to South
Africa appear to be dominated by a strong NH source and subsequent decline
towards lower latitudes, with little impact from biomass burning. This
observation fits with a recent study which saw no evidence for
CH2Cl2 emissions in boreal biomass burning plumes, suggesting that
previous calculations of CH2Cl2 emissions from biomass burning
(e.g. Rudolph et al., 1995) were overestimates
(Simpson et al., 2011). Whilst
emissions from boreal and tropical forest fires may differ, recent analysis
of air samples collected during flights over biomass burning events in the
Brazilian rainforest also showed no significant fire emissions of
CH2Cl2 (A. Wisher, UEA, personal communication, 2014).
Emissions of
CH2Cl2 from India investigated
during the Indian monsoon
CARIBIC data collected during flights to India in 2008 have previously been
used to demonstrate the impact of the Indian/Asian Summer Monsoon (ASM) on
UTLS trace gas concentrations
(Baker
et al., 2011, 2012; Schuck et al., 2010). As these studies reported elevated
concentrations of many trace gases linked to the persistent convection and
anticyclonic flow of the ASM, we have divided flights to India into “monsoon”
(July–September, inclusive) and “non-monsoon” (rest of year) for this
study (Tables 1 and 2). Latitudinal distributions of CH2Cl2 along
routes to Indian destinations are shown in Fig. 4, displayed with the oldest
flights at the top of the plot.
As the ASM was a particular focus of CARIBIC during 2008 we begin our
discussion of Fig. 4 with these flights. A pronounced difference in the
CH2Cl2 distribution along the latitudinal flight track can be seen
between monsoon and non-monsoon months. During the 2008 monsoon (Fig. 4d)
concentrations between ∼ 25–40∘ N are elevated
compared to non-monsoon months (Fig. 4c). During non-monsoon months a
relatively flat latitudinal distribution of CH2Cl2 is observed
along the majority of the flight path with some elevated concentrations at
latitudes less than 20∘ N. Analysis of back-trajectories
indicates that these elevated samples (Fig. 4c) probed air that had recently
been at low altitude over Southeast Asia. The pattern during the monsoon
season is consistent with previous CARIBIC studies, referenced above, which
reported elevated concentrations of NMHCs, methane and other compounds
within ∼ 25–40∘ N due to interception of air masses
with influence from the continental boundary layer.
Latitudinal distributions of CH2Cl2 observed
during flights to India for non-monsoon months on the left and monsoon
months (July, August, September) on the right where colour = year (see
inset colour bar; colour scale is consistent with Figs. 2, 4 and 6). Where
multiple measurements of the same sample have been made 1σ error
bars are given, see Sect. 2.
The difference between monsoon and non-monsoon months can also be observed
in the earlier (CARIBIC1) data, although the monsoonal elevation between
∼ 25 and 40∘ N (Fig. 4b) is superimposed on a
north–south latitudinal gradient more clearly seen outside of the monsoon
season (Fig. 4a). This north–south gradient is similar to that seen in the
data from flights to South Africa. An important feature of Fig. 4 is the
shift in the dominant latitudinal feature over time. In the 1998–2001 period
a north–south gradient, suggesting low CH2Cl2 emissions from
India, is clear. In contrast, very high concentrations at low latitudes are
observed in later flights conducted in 2008 and 2011–2012 (Fig. 4e). High
values at lower latitudes are in contrast with results for Africa (Sect. 3.2) and to Central America (Sect. 3.4). These results suggest a shifting
latitudinal profile and an increase in emissions within the Indian region.
One 2011 flight in particular showed exceedingly high levels of
CH2Cl2 (Fig. 4e). Analysis of back-trajectories indicate that these air
masses originated from a low altitude over India and Southeast Asia. This
region is discussed further in Sect. 3.5.
Increases were calculated for the period between 1998–2000 and 2008 and
between 1998–2000 and 2011–2012. These are provided in Table 2 for both
monsoon and non-monsoon months. However, as was illustrated in our
discussion of Fig. 4 (previous paragraph), during the non-monsoon months we
may sample air masses that originate from outside the Indian region. During
the monsoon months air masses within the monsoon anticyclone are much more
isolated (full details provided below) and so the increases during the
monsoon period are more likely to represent changing CH2Cl2
emissions from India and its neighbours. The increase between the 1998–2000
and 2008 monsoon periods was 15 ppt (69 %, further details in Table 2).
Measurements of air masses from the Indian and South Asian region were made
by Scheeren et al. (2002, 2003) during two
campaigns in 1999 and 2001. Their observations averaged 29 (σ=12)
ppt in 1999 and 23 ppt (σ=3) in 2001 (see Table 2). Their 2001
average, in particular, corresponds well to our early measurements over
India which averaged ∼ 20–22 ppt.
As the strong convection associated with the ASM quickly elevates air masses
from the Indian continental boundary layer and then isolates them within the
monsoon anticyclone, UT mixing ratios over India during the monsoon are
closely coupled to boundary layer emissions
(Baker
et al., 2011, 2012; Rauthe-Schöch, et al., 2015; Schuck et al., 2010).
This makes the ASM an ideal case study for calculating emission estimates
using the CARIBIC data set. As the majority of CH2Cl2 emissions are
industrial we assume that emissions do not change during the monsoon and so the
increase in UT concentrations seen during this period can be attributed
wholly to meteorological changes. This assumption is justified based on
findings by Gentner et al. (2010) who
showed an absence of seasonality in CH2Cl2 emissions based on
measurements made in California and a study by McCulloch and Midgley (1996)
who also reported an absence of seasonality based on their analysis of data
on the global industrial use of CH2Cl2. Previous analyses of
meteorological parameters during 2008
(Rauthe-Schöch, et al., 2015;
Schuck et al., 2010) have demonstrated that the monsoon anticyclone was
present in July–September. For emission estimates we take all tropospheric
samples where both CO and CH2Cl2 were measured and which were
collected < 40∘ N
(for further
explanation and justification of this method see Baker et al., 2011; Schuck
et al., 2010) in July–September 2008, a total of 35 samples.
Emission estimates are often calculated using ratios whereby the compound of
interest is compared to a compound with which it correlates and for which
emissions are quantified, in this case CO. The emission estimate is based on
the slope of the linear correlation between the two tracers using Eq. (2)
(where ECH2Cl2 and ECO are the emission estimates for
CH2Cl2 and CO respectively):
ECH2Cl2=ECO×ΔCH2Cl2ΔCO.
The CH2Cl2 vs. CO correlation within the monsoon (< 40∘ N, July–September 2008) can be seen in Fig. 5a. The correlation
has a statistically significant (Pearson's correlation coefficient,
p < 0.05) r value of 0.62. Correlations for individual months are
also shown in Fig. 5a. No statistical difference (Fisher's z test with a
z-crit. value of 0.05) exists between the slopes for individual months,
allowing us to use the slope of the correlation for the whole ASM period for
our emission estimate. Table 4 includes ΔCH2Cl2 / Δ CO values from this study as well as a range of published values from a
variety of sources. Lower ratios are seen for wildfire and biomass burning
plumes, with higher ratios (more similar to the ones we observed) for urban
(likely industrial) emissions.
Seasonality of available CARIBIC data for regions included
in this study.a Black squares show that samples
were taken during that month and the number refers to the number of
samples.b
Periodc
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Europe, CARIBIC
1998–2001
3
6
2
4
1
2009–2012
7
3
12
20
24
5
4
13
9
8
5
1
Mace Head NOAA
1998–2001
1
1
1
1
3
4
1
3
1
2003–2004
4
4
4
4
4
4
4
3
5
3
5
4
2009–2012
6
9
6
8
10
8
7
7
7
5
8
7
Africa > 30∘ N
2000
1
1
1
2009–2011
3
3
8
4
4
2
Africa < 30∘ N
2000
9
9
11
2009–2011
10
10
31
18
6
10
Atlantic/Central America
2001–2002
11
20
17
9
11
10
11
2009–2011
26
31
19
13
20
21
18
8
10
19
a India is not included as focus is on summer monsoon and so
seasonality throughout the whole year is not relevant. Tropical region not
included due to reduced seasonality in the tropics. Samples sizes for both
these regions can be found in Tables 1 and 2.b Samples are an average of two flask samples (NOAA) or at least two
analyses of sample (UEA).c Relates to any discrete time period mentioned in the text or Tables 1
and 2.
A comparison of enhancement ratios from this study (air
mass age corrected) and existing literature.
Source
ΔCH2Cl2/ΔCO / mol mol-1
India, summer monsoon period (this study)a
1998
1.0 × 10-4 (±4.3 × 10-5) r= 0.69
1999
1.6 × 10-4 (±6.4 × 10-5) r= 0.53
2000
2.5 × 10-4 (±3.7 × 10-5) r= 0.86
2008
4.0 × 10-4 (8.7 × 10-5) r= 0.62
Other studiesb
Biomass burning, Africa savanna, ground-based,19911
2.5 × 10-5 (error = 26 %) r= 0.65
Wildfires, Cape Grim, ground-based, 1998-20042
< 1–6 × 10-7
Asian pollution outflow, Bay of Bengal, boundary layer flights, 19993
4.4 × 10-5 (±4.7 × 10-5) r= 0.39
Urban, California, ground-based, 20054
3.1 × 10-4 (±3.0 × 10-5) r= 0.58=0.66
Urban, USA, boundary layer flights, 20045
2.4 × 10-4 (1.8 × 10-4–2.9 × 10-4) r= 0.56–0.83
Urban, Mexico, ground-based, 20065
1.9 × 10-4 (1.1 × 10-4–2.9 × 10-4) r= 0.43–0.81
a CARIBIC fits are orthogonal distance regression fits using IGOR Pro
software.
b described by: emission source, location, sampling location, year of
study
1 (Rudolph et al., 1995);
2 (Simmonds
et al., 2006); 3 (Scheeren et al., 2002);
4 (Gentner et al., 2010);
5 (Millet et al., 2009).
Correlation plots of CH2Cl2 and integrated CO
(see Sect. 2) for (a) the 2008 summer monsoon period (coloured by
individual months) and (b) the early years of the CARIBIC India data set
(coloured by individual years). Further details in Table 4.
Before discussing our emission estimates we provide details of the
assumptions and potential errors associated with this method and our
treatment of these factors. Firstly, this method assumes that the two
compounds share a common, dominant source and/or that emissions are
co-located. Whilst CH2Cl2 is of predominantly industrial origin
(see Sect. 1) with emissions likely to be dominated by areas of heavy
anthropogenic influence (e.g. cities), CO has a more diffuse source. It has a
large combustion source which, in India, is dominated by the burning of
biofuels and biomass (Dickerson et al.,
2002). Despite this, we believe that CO provides the best option for emission
estimates in this region. On the scale of a regional emission estimate, CO
and CH2Cl2 sources are co-located: both show strong signatures
from the Indian subcontinent where it is known that air masses sampled
within the monsoon anticyclone have likely originated from. CO emissions are
also well quantified, and comparisons between CO and anthropogenic
chlorocarbons, including CH2Cl2, have also been conducted in
several other studies including
Gentner
et al. (2010); Millet et al. (2009); Simmonds et al. (2006) and Palmer et al. (2003) (Table 4)
Schuck et al. (2010) discussed the use of SF6 as a tracer. However, its extremely
patchy distribution (strong point sources) results in a poor correlation
with CH2Cl2 and a poorer representation of the Indian monsoon
plume.
To further support the suitability of the CH2Cl2–CO ratio for
estimating CH2Cl2 emissions we describe two analyses which
demonstrate that the variability we observe is due to recent emissions, as
opposed to variations in transport time or route prior to sampling. Firstly,
we compared the ΔCH2Cl2 / ΔCO value from our sample
set (n=35) to the ΔCH2Cl2 / ΔCO value obtained
from a smaller data set based on the method used in Baker et al. (2011).
Baker et al., when performing emission estimates for the same CARIBIC 2008
monsoon data set, minimised the influence of variability with respect to
processing or other transport effects by selecting a data set that included
only those samples whose back-trajectories indicated low-level (pressure > 600 hPa) contact within the monsoon anticyclone in the previous
5 days (n=15). Comparing this subset of samples to our full sample set
gives a very similar ΔCH2Cl2 / ΔCO and r value. The
ΔCH2Cl2 / ΔCO and r values for our data set are shown
in Table 4 and for the Baker et al. (2011) subset the age-corrected slope was
4.9×10-4 (±8.7×10-5) and r=0.67. The similarity between
the two values suggests that the correlation observed in our data set (Fig. 5) is influenced by local emissions and not differences in transport times
or source regions. This is supported by a second method in which we compared
the CH2Cl2 vs. CO correlation for the ASM samples with the
correlation calculated for samples taken within the same 14–40∘ N
latitudinal band but along flight routes to Africa (Sect. 3.2) and across
the Atlantic (Sect. 3.4). The correlation for the Africa and Atlantic
flights are much weaker and do not show the same dynamical range as the
correlation for the India data. This supports our assumption that samples
taken during the ASM provide a unique correlation that represents local
emissions due to the rapid convection and isolation that occurs within the
monsoon system.
Secondly, there are errors and assumptions associated with the measured
emission ratio. This includes the assumption that the emission ratio
measured by CARIBIC is similar to that at the source, i.e. it has not been
affected by dilution and/or photochemical/chemical loss processes. We
believe this assumption to be valid with respect to dilution based on
analysis conducted by Baker et al. (2011) on the same CARIBIC data set as
used in this paper. Baker et al. (2011) reported an i-butane / n-butane
ratio in the ASM of 0.77 ± 0.07 pptv pptv-1, suggesting that the
invariability of this ratio provided evidence of minimal dilution. A broader
investigation of the ASM by Randel and Park (2006) using back-trajectory models found that 70 % of parcels initialised within the
anticyclone were still there after 10 days.
Dilution is also considered in our discussion of the validity of our ratio
with respect to the transport time between emission source and sampling
(previous paragraph), assuming that variations in transport time lead to
variations in the degree of mixing, and in the bootstrapping error analysis
of the ΔCH2Cl2 / ΔCO regression line (subsequent
paragraph). With respect to photochemical loss processes we assume transport
times from the boundary layer to our sampling altitude of around 4 days
based on Baker et al. (2011). Within this
time, CH2Cl2, with a lifetime of around 5 months, does not
experience large losses. However, the lifetime of CO (∼ 2 weeks in mid-latitude summer (Scharffe et al., 2012
referencing Warneck, 1988) is short enough that concentration changes are
likely to have occurred during this time and so we age-correct our emission
ratios with respect to CO using Eq. (3), a method used by both
Baker et al. (2011) and Scheeren et al. (2002). Here, the emission ratio at time 0, ER0, is related to the
emission ratio at time t, ERt, by accounting for the change in time,
Δt (4 days), the reaction rates, k, of CO and CH2Cl2 with
OH at 298 K and the average concentration of OH predicted at 20∘ N and 500 hPa (estimated uncertainty of ±25 %). Both kCO at
2.1 × 10-13 cm3 molec-1 s-1 and
〈[OH]〉
at 2.48 × 106 molec cm-3 are taken from
Baker et al. (2011, and refs. within) and
kCH2Cl2 at 1.1 × 10-13 cm3 molec-1 s-1 from
Villenave et al. (1997). This procedure leads to a correction in
our emission ratio for the 2008 monsoon season of around -8 %:
ER0=ERte(kCO-kCH2Cl2)〈[OH]〉Δt.
Also associated with the emission ratio are errors arising during the
calculation of the ΔCH2Cl2 / ΔCO slope. These errors
arise from two sources: (1) uncertainties in the analytical measurements of
both CH2Cl2 and CO (see Sect. 2) and (2) uncertainties
associated with using a slope calculated from a discrete set of samples to
calculate a regional emission estimate. The errors associated with (1) are
small compared to those associated with (2), see Sect. 2.2, and so we use
(2), calculated using a bootstrapping procedure, to set bounds on our
emission estimates. Using the Wood (2003) bootstrapping
procedure we resampled, with replacement, our CH2Cl2 and CO
data sets 10 000 times, each time calculating ΔCH2Cl2 / ΔCO. The output from the resampling procedure
provides a probability distribution for the slope of CH2Cl2 / CO,
allowing us to understand how dependent ΔCH2Cl2 / ΔCO may be on the sampled data and allowing us to provide an idea of the
potential variation in ΔCH2Cl2 / ΔCO. The
bootstrapping procedure has been used to calculate a possible range of
emission values to aid the comparison between years. In the following text
this ±1σ range is given in brackets following each emission
estimate.
CO emissions for the Indian region are taken from the Emission Database for
Global Atmospheric Research (EDGAR) v. 4.2 (JRC & PBL,
2009). We include emissions from the following countries; Bangladesh,
Bhutan, India, Sri Lanka, the Maldives, Nepal and Pakistan. EDGAR emissions are
provided per year and are split into categories including various industrial
and domestic processes, transport and biomass and biofuel burning.
Baker et al. (2012) used the Global Fire
Emission Database (GFED, v.3.2, van der Werf et
al., 2010) to show that CO emitted from biomass burning was greatly reduced
during the monsoon, accounting for around 0.5 % of total annual CO fire
emissions during 2008. To account for this reduction, the EDGAR biomass
burning emissions were corrected for the effect of the monsoon using the
GFED data and the method in Baker et al. (2012). As we had no evidence that anthropogenic Indian CO emissions had
seasonality we divided these emissions evenly throughout the year. We
believe any errors arising from this assumption are likely to be within the
general errors associated with the EDGAR emissions (see below), in
particular due to the dominance of burning as a source of CO in India
(Dickerson et al., 2002). This method gave an average monthly emission
during the 2008 monsoon of 4.2 Tg CO month-1. Maximum errors on the
EDGAR CO database are given as up to ±50 %
(Olivier et al., 1999), likely reduced by our additional
use of the GFED database. We do not consider the given error on the EDGAR
data further as the main objective of these emission estimates is to provide
a comparison of CH2Cl2 emissions over time and we assume that this
error remains constant throughout the EDGAR database.
Using the 2008 CH2Cl2 / CO slope (Table 4) and the EDGAR CO
emissions of 4.2 Tg CO month-1 gives an emission estimate of 1.7
(1.3–2.1) Gg CH2Cl2 month-1 from the Indian region. As
industrial sources of CH2Cl2 have no seasonality we assume that this
emission rate is constant over the year and so estimate that 20.3
(15.8–24.8) Gg of CH2Cl2 were emitted from the Indian region in
2008. The most recent estimate of global emissions is 515 ± 22 Gg yr-1 given in Montzka et al. (2011b) which is based on top-down estimates from
Simmonds
et al. (2006) from data collected between 1999 and 2003. Considering the
caveat that global emissions are likely to have increased since this figure
was published, our estimate for emissions from the Indian region in 2008 is
roughly 5 % of the global total. These estimates are discussed further in Sect. 3.6.
CH2Cl2 emissions from the Indian subcontinent were also estimated
for the 1998, 1999 and 2000 ASM seasons using the same analysis described
above for 2008. Figure 5b shows the CH2Cl2 vs. CO correlation for
1998–2000, coloured by year. No significant difference (Fisher's z test with
a z-crit. value of 0.05) was observed between the three years but we
consider the three years individually to provide similar data sets for
comparison with the 2008 data set (with respect to sample size and length of
sampling period, see Table 2). EDGAR monthly CO emissions, modified to
account for reduced burning during the monsoon, as described above, were
4.1, 4.2 and 4.2 Tg CO month-1 for 1998, 1999 and 2000 respectively. As
CO has anthropogenic sources one may expect its emissions to have increased
over time, however the monthly emissions for 1998–2000 are similar to those
for 2008. To investigate this result we compared the EDGAR data to three
previous studies. EDGAR monthly emissions compare well to those of
Fortems-Cheiney et al. (2011) who reported relatively stable CO emissions
from South Asia between 2000 and 2010. EDGAR monthly emissions also compare
well to the GIS-based emission estimate of Dalvi et al. (2006) and the air pollutant emission inventory of Streets et al. (2003) who
estimated that CO emissions from India in 2000 were 69 and 63 Tg
respectively, similar to the 66 Tg annual emission estimated from the EDGAR
database for 2000. We use only the EDGAR data in the subsequent emission
estimates for consistency with both our 2008 emission estimate as well as
previous studies, referenced above. The resulting annual CH2Cl2
emissions are estimated at 4.9 (2.7–7.2) Gg yr-1 in 1998, 7.9
(5.1–10.8) Gg yr-1 in 1999 and 12.6 (10.8–14.4) Gg yr-1 in 2000.
Our emission estimates suggest that emissions have increased significantly
over time, from a range of 3–14 Gg in the late 1990s to around 16–25 Gg in
2008.
CH2Cl2 measured
during flights across the Atlantic to Central America
The final flight route with the temporal resolution needed for identifying
CH2Cl2 trends is across the Atlantic to Central America (Cuba,
the Dominican Republic and northern Venezuela). Figure 6 shows the
distribution of CH2Cl2 against both latitude and longitude
sampled on flights to these destinations. The gradient along the flight
tracks, shown in Fig. 6 as average values binned for every 5∘
latitude and 10∘ longitude, show very little variation in the
early years of the data set (2001–2002). For example, the average
CH2Cl2 value in the 5∘ latitude bins varied between
20 and 25 ppt along the entire transect. The mean CH2Cl2 concentration
was 23.8 (σ 3.9, n=9) ppt between 50 and 55∘ N and 19.5
(σ 2.6, n=8) ppt between 20 and 25∘ N. The latitudinal
gradient increases over time and can be seen in the binned mean values for
2009–2011 which is to be expected as the northern section of the flight
route crosses western Europe (Sect. 3.1). Due to the smaller number of
samples collected at the far ends of the transect (e.g. n=5 and n=3 for
the far northern and southern bins during the 2009–2011 flights) we do not
quantify the gradient along the flight track. However, it is still less
pronounced than the latitudinal gradients observed en route to Africa or
India, likely due to the influence of clean Atlantic air.
(a) Latitudinal and (b) longitudinal distributions of
CH2Cl2 along flight routes across the North and Central Atlantic
to Central America where colour = year (see inset colour bar, colour scale
is consistent with Figs. 2, 4 and 6). Average, x¯, values for
5∘ latitude and 10∘ longitude bins are shown for
2001–2002 and 2009–2011 (see Sect. 3.4), error bars are the 1σ
variation within these bands. (c) Annual tropospheric values (see Fig. 2)
with 1σ error bars.
Profiles of CH2Cl2 relative to O3 from
samples collected on flights across the North and Central Atlantic to
Central America. Median values are for 50 ppb O3 bins between 0–100 ppbv and 100 ppbv O3 bins above this, error bars are 1σ. The
coloured band highlights the region between 400–500 ppbv O3 discussed
in Sect. 3.4. The dashed line represents 30 ppt of CH2Cl2 (see Sect. 1), provided as a visual marker to illustrate the shift over time to
higher concentrations of CH2Cl2.
Profiles of CH2Cl2 relative to potential
temperature for samples taken within the latitude range 0∘ ± 25∘. Median (error bars are range) values for 5 K bins are
overlaid in black. Colour represents flight route, as shown by the inset
colour bar.
Annual tropospheric values (Fig. 6c) show an increase in time, as discussed
in previous sections, although the magnitude of the increase is much smaller
and within the levels of variation observed. Unlike previous sections, where
we had defined spatial and temporal regions (e.g. Europe, Indian monsoon)
for which to calculate CH2Cl2 increases, this flight route can
provide an idea that we see CH2Cl2 increases, albeit small (from
23.2 ppt (σ=3.6) to 32.0 ppt (σ=7.8)), even in cleaner
air masses from over the Atlantic and Central America. There is a lack of
CH2Cl2 measurements in the Central American region with which to
compare the CARBIC data.
Quasi-vertical profiles of CH2Cl2 along this route can also
provide information on changes in CH2Cl2 over time. In Fig. 7
profiles of CH2Cl2 are plotted as a function of O3, as
described in Sect. 2. To investigate changes over time, Fig. 7 shows
CH2Cl2–O3 profiles for 2000–2002, 2009–2010 and 2011–2012.
Samples within each of these 2-year periods are distributed evenly across
many months and so it is unlikely that seasonal bias plays a role in the
changes observed over time. Between 400 and 500 ppbv O3 (n=4 for each 2-year period) the median CH2Cl2 value increased from 11 ppt in
2001–2002 to 16.0 ppt in 2011–2012. However, due to the low sample number
and the high variability this increase is within the uncertainties of these
averages.
Vertical profiles of CH2Cl2 in the tropics
Air in the TTL may move quasi-horizontally into the ExTL or the LMS, or
vertically into the stratospheric overworld. Despite the fact that only a
portion of the air from the TTL moves into the free stratosphere, for
short-lived species rapid convective transport to the TTL followed by ascent
to the free stratosphere is the most efficient transport pathway to the
stratosphere (Law et al., 2007). With a common cruise altitude
of 10–12 km, CARIBIC flies at the lower edge of the TTL, which is commonly
defined as covering a potential temperature, θ, region of between
around 345 K (∼ 12 km) and 380 K (∼ 17 km). For
this reason, we present only an upper limit on the changing input of
CH2Cl2 into this important region.
All data sampled within the tropical latitude band of 25∘ S to 25∘ N
(the distribution of which can be seen in Fig. 1, with further
information in Table 1) were plotted as quasi-vertical profiles relative to
θ in Fig. 8. A clear increase in the magnitude of high
CH2Cl2 “pollution” events can be seen. Because of this skew,
median values and the range (min–max) are given for each 5 K altitude bin.
In 1998–2002 the median CH2Cl2 concentration between 345–350 K was
18.1 (13.4–25.0, n=20) ppt. The median value within this altitude bin
between 2009–2012 was 23.2 ppt with a range spanning 12.4 to 90.4 ppt
(n=97). The increase in median values is modest due to the inclusion of
the 2009–2010 data (see Sect. 3.4). However, despite the variability there
is a statistically significant (Mann–Whitney test, p < 0.05)
difference between the CH2Cl2 concentration observed between
345 and 350 K.
With the exception of 2009–2010, see Sect. 3.4, CH2Cl2
concentrations observed in this vertical band increased over time, reaching
26.8 (12.4–90.4, n=63) ppt in 2011–2012. Other measurements of
CH2Cl2 within this region are sparse. In 1991–1992
Schauffler et al. (1993) measured a mean CH2Cl2
concentration of 14.9 (σ=1.1, n=12) in the TTL between 15.3 and 17.2 km (366–409 K). Schauffler's average is lower than that seen in the early
CARIBIC data, although this is likely to be due to the fact that (1) their
measurements were taken a few years earlier than CARIBIC; (2) their
measurements were taken at a higher altitude; and (3) their suggestion that
their mean value was biased low due a high degree of mixing with
stratospheric air during their sampling period. Point (3) demonstrates the
influence that stratospheric mixing may have in this region and may explain
the low values we observed in the 2009–2010 period (see also Sect. 3.4).
Between 2009 and 2011 tropical CH2Cl2 measurements were made by
the HIAPER Pole-to-Pole Observations (HIPPO) project
(Wofsy et al., 2012). The HIPPO database contains 20
samples taken within the latitude band 0∘ ± 25∘
and within the 345–350 K θ band. These values are reported on the
NOAA scale and are therefore comparable with CARIBIC data. The HIPPO
results, an average CH2Cl2 value of 26.3 (15.9–49.8) ppt, compare
well to the CARIBIC results discussed above.
The data in Fig. 8 have been coloured by sampling route, thus providing a
rough indication of possible air mass source regions. Of interest is the
group of high values in Fig. 8e sampled on the one flight made to Bangkok
and Kuala Lumpur at the end of 2012. With the rise of industrial activity in
Asia it is likely that emissions of industrial solvent emissions have also
increased. Studies in China have shown exceedingly high ground-level
concentrations of CH2Cl2. For example, a 2001 study in 45
different Chinese cities by Barletta et al. (2006) saw an urban average of 226 ppt (σ232) and individual
occurrences of up to 3 ppb. It is possible that high levels are also emitted
in other industrial parts of Asia, although there are little, if any,
ground-based measurements to support this.
Potential causes for increasing CH2Cl2
One likely contributor to the increase in CH2Cl2 is the
increasing use of hydrofluorocarbons (HFCs) as replacements for
ozone-depleting CFCs and HCFCs, the production and consumption of which are
strictly controlled by the Montreal Protocol. Specifically, CH2Cl2
is used in the production of difluoromethane, also known as HFC-32
(Ramanathan et al., 2004). HFC-32 is used in combination with
HFC-125 to make the refrigerant R410A, a direct replacement for HCFC-22. It
is estimated that about 96 % of HFC-32 emissions are in the NH
(McCulloch, 2004), where the majority of production and
consumption of this HFC is likely to occur. Recent analysis of archived and
AGAGE air samples shows that HFC-32 has increased from around 0.7 ppt, when
the first measurements were made in 2004, to around 6.2 ppt in 2012, with
the growth rate reaching 17 % yr-1 in recent years
(O'Doherty
et al., 2014; Montzka et al., 2011b). As HFCs do not deplete stratospheric
ozone they are not controlled by the Montreal Protocol. However, they are
potent greenhouse gases and, as such, are covered by European legislation
controlling their production, consumption and emission. This legislation is
likely to reduce CH2Cl2 emissions from HFC production in Europe in
the coming years. In contrast, it is expected that much of the future demand
for HFCs is likely to come from developing countries
(Velders et al., 2009). The use of air
conditioning systems is growing rapidly in India (e.g. NRDC,
2013); it is the world's third largest consumer of CH2Cl2
(IHS, 2014) and HFC-32 production plants have opened in recent
years (Daikin, 2012). A rapidly expanding air conditioning industry
and increased consumption of CH2Cl2 in India could at least partly
explain the occurrence of high CH2Cl2 observations in the latter
years of the CARIBIC data set (Sect. 3.3 and Fig. 4e). A shift in the main
consumers and emitters of HFCs is supported by O'Doherty et al. (2014) who
suggest that East Asian emissions are underestimated in some inventories
(e.g. EDGAR) and that emissions from East Asia are growing in importance.
There are other uses for CH2Cl2 which could be contributing to
the increasing atmospheric concentrations. Industrial sources include use in
office (plastic) materials and electronics (Bin Babar and
Shareefdeen, 2014; Kowalska and Gierczak, 2012), the production and use of
which is increasing in developing nations such as India and those in Southeast Asia. A CH2Cl2 source from municipal waste disposal
(Majumdar and Srivastava, 2012) may be of particular importance for
India where mismanagement of waste disposal has been found to lead to high
levels of fugitive volatile organic compound emissions from waste disposal
sites. CH2Cl2 is also used by the pharmaceutical industry in drug
preparation, where its use may be increasing as a replacement for CCl4
which is regulated by the Montreal Protocol (UNEP CAP, 2009).
The likely sources of increased CH2Cl2 emissions over the past
decade suggest that India might be an increasingly important source of
industrial CH2Cl2 emissions, as seen in the CARIBIC data set,
although its emissions are still small on a global scale. We estimate
CH2Cl2 emissions from the Indian region in 2008 to be in the
region of 20.3 (15.8–24.8) Gg. This is similar to an estimate of 24 (16–33) Gg yr-1 for 2005 USA emissions calculated by
Millet et al. (2009). The latest
global estimate provided by the WMO (Montzka et al., 2011b) gave global
emissions of 515 Gg yr-1 for 1999–2003 based on Simmonds et al., 2006).
Both the India and USA emissions are small fractions of this total (which is
a lower limit due to the time frame it was based upon and the increasing
emissions over time) suggesting that other regions contribute significantly.
Significant growth in industrial production and consumption of HFCs in Asia,
in particular in China, was projected by
Velders et al. (2009), suggesting
that these regions may be or may become important source regions of
CH2Cl2 and warrant further study.