Atmospheric CH4
Atmospheric CH4 values simulated using prior fluxes (prior atmospheric
CH4) increase continuously during 2000–2012, and quickly exceed
observed atmospheric CH4 levels, especially in the NH (Figs. 2, 3). The
seasonal cycle of prior atmospheric CH4 values agrees poorly with the
observations, with a positive bias from winter to summer in the NH and around
the end of each year in the Southern Hemisphere (SH) (Fig. 2). Furthermore,
prior atmospheric CH4 values are negatively biased compared to the
observations in the SH during 2002–2004 (Fig. 2). This is likely due to an
underestimation in the prior emissions in the SH. Posterior atmospheric
CH4 values generally match the observations to a level close to the
expected model–data mismatch, indicating a proper choice of observation
covariance. A seasonal bias remains in the NH (especially in L62T), and
the decrease in atmospheric CH4 in the SH around 2002–2004 also remains
in the posterior, although shorter in duration and of smaller magnitude than
in the prior (Fig. 2). The negative bias in posterior atmospheric CH4
around the equator remains unresolved throughout the study period in all
inversions, and mainly originates from the sites Bukit Koto Tabang, Indonesia
(BKT), (-25 to -27 ppb), and Mt. Kenya, Kenya (MKN) (-18 to -23 ppb). The
posterior atmospheric CH4 values are especially low relative to
observations during June–October. The bias became smaller when CH4
emissions were increased in the South American tropical mTC region, although
this led to compensating fluxes and mismatches with observations elsewhere
(not shown). Posterior emissions for the South American tropical region
(mTC3) remain similar to the prior, and the inversion does not significantly
decrease the uncertainty of the prior emission estimates in this mTC (see
Sect. 3.4.4 and 4.2).
Agreement between simulated CH4 and surface observations is slightly
better in L78T and L62G than in L62T (Fig. 2), as indicated by
the root mean square error (RMSE), which is about 0.5 ppb smaller. In
addition, the biases in annual amplitude are about 1–2 ppb smaller. The
negative bias in the SH from 2002 to 2004 is seen in all inversions, but is
most prominent in L62T. Although the difference in the average RMSE is
small, it is significant as it is calculated from all the observations
assimilated in the study period. In addition, differences are significant
when the ensemble distributions of posterior atmospheric CH4 are
considered. The spread (1 standard deviation (SD)) of ensembles is less
than 5 ppb for most sites and less than 1 ppb for MBL sites, mostly located
in the SH.
Further evidence of poorer performance in L62T than in other runs is
seen in its global fluxes. L62T produced the smallest total global
emission estimates for 2002–2004, which in turn led to the largest increase
in the total global emission estimates from 2001–2006 to 2007–2012. Based
on previous studies (e.g. Bergamaschi et al., 2013; Bousquet et al., 2006;
Bruhwiler et al., 2014; Fraser et al., 2013), the increases in L78T and
L62G are more reasonable (see Sect. 3.4.1). The differences in RMSE and
bias between the latter inversion estimates are small near 30∘ N,
where many observations are located. However, the RMSE and bias in L78T
are about 1 and 2 ppb smaller at high northern latitudes
(60–75∘ N), and about 3 and 6 ppb larger around the equator
(EQ–15∘ N) than in L62G, respectively. Moreover, low
atmospheric CH4 values in the SH during 2002–2004 are not as prominent
in the prior when the G2000 convection scheme is used (Fig. 2), probably due
to enhanced transport between the NH and SH in L62G. Mean Chi-squared
statistics (Michalak et al., 2005) of the observations are typically between
0 and 2, and follow normal distributions (not shown), which again indicates
that the MDM estimates are appropriate at most of the sites.
In contrast to the prior, the growth rate (GR) of posterior XCH4 does
not change strongly before 2007, but increases after 2007 (Fig. 3). All
inversions show an increase in XCH4 by about 6 ppb yr-1 after
2007, with some seasonal and interannual variations (Fig. 3). The timing of
the change in posterior XCH4 GR is in line with the GR calculated from
the global network of NOAA MBL observations (Dlugokencky et al., 2011) and
with the retrieved XCH4 GR at Park Falls (Fig. 3). This indicates that
the GR of prior XCH4 is too large throughout 2000–2012 (see also
Fig. 2), and this can only result from overestimated emissions or
underestimated loss of CH4. Note that the NOAA MBL observations compared
in Fig. 3 are calculated from surface observations.
Performance of inversion L62G at European in situ observation
sites, whose data were assimilated in the model, and at the locations of
four aircraft campaigns. The campaign locations are marked with stars.
Aircraft observations were used for evaluation. The colour of the marker for
the in situ observation site is determined by the RMSE of observed and
simulated posterior atmospheric CH4 values divided by the pre-defined
MDM. The radius of each circle provides the correlation between observed and
simulated posterior atmospheric CH4 values, where a larger radius
corresponds to weaker correlation. Thick grey lines identify the mTC
borders.
Evaluation with TCCON and GOSAT XCH4
XCH4 provides additional information about the spatial distribution of
atmospheric CH4. TCCON and GOSAT XCH4 retrievals were not
assimilated in the inversions, so the following comparisons also allow an
assessment of model performance at independent locations and times.
For many TCCON sites in the NH, the XCH4 in L62T and L78T is
slightly lower than observed, but the trend and seasonal variability are
generally well captured. However, the 2007–2012 trends at Izaña (Spain),
Park Falls (USA) and Lamont (USA) are much stronger than in the retrievals
(Fig. 6). Since the emission estimates at similar latitudes would affect the
XCH4 estimates, this could be an effect of the strongly increasing
northern temperate emission estimates after 2007 (Sect. 3.4.2). The RMSE
between the estimates and retrievals is smallest in L62G at all sites,
except at Garmisch, Germany (Table 4). Garmisch is a mountain site (altitude
734 m a.s.l.), and the mean of observed XCH4 is statistically
significantly lower than at nearby sites, e.g. Karlsruhe, Germany, and
Bialystok, Poland (Figs. 6, S5).
For the SH TCCON sites, a strong negative bias is found in all inversions
(Figs. 6, S5). Agreement is especially poor for Wollongong, which has the largest
RMSE (more than 30 ppb) among all TCCON sites in all inversions (Table 4).
As the site is located in the city of Wollongong, where the influence of
local emissions is high, it is difficult for models to reproduce XCH4
well (Fraser et al., 2013). The comparison with the nearest in situ site,
Cape Grim, Australia (CGO) shows that the negative bias is much smaller (-6
to -11 ppb) compared to Wollongong (-32 to -35 ppb), and the
correlation with the retrievals is high (> 0.85). In addition, the
negative bias in XCH4 is much smaller (-12 to -15 ppb) at
background site Lauder, New Zealand (LAU) and the correlation at the LAU in situ
site is again strong (> 0.85) in all inversions. The disagreement at
Darwin is probably due to little constraint of the emissions. Although in
situ observations at Gunn Point, Australia (GPA) were assimilated, the
inversion probably did not benefit significantly from these observations
because data were available only after mid-2010, and the MDM was set high
(75 ppb). Furthermore, emissions from the tropics also affect the XCH4
estimates in Australia. Our emission estimates for the tropics
(30∘ S–30∘ N) are about 10–20 Tg CH4 yr-1
smaller than the estimates by Houweling et al. (2014), for example. When the
prior emission estimates for the South American tropical region (mostly
between 15∘ S–15∘ N) were increased (see Sect. 3.1),
agreement in the SH improved (not shown). The comparison with GOSAT XCH4
also supports the finding from the comparison with the TCCON retrievals,
showing a mean negative bias of 13 ppb in the SH (Fig. S6). We currently do
not have sufficient information to correct the errors that affected the SH
XCH4 in our system, or to identify the exact cause.
Observed and simulated daily mean XCH4 at TCCON sites.
Root mean square error (RMSE) between TCCON and model XCH4
with averaging kernel applied (ppb). The inversion with the smallest
posterior RMSE is marked in bold.
Site names
Coordinates
Prior
Posterior
Latitude
Longitude
L62T, L78T
L62G
L62T
L78T
L62G
Eureka, Canada
80.05∘ N
86.42∘ W
80.2
78.6
13.6
13.9
8.8
Sodankylä, Finland
67.37∘ N
26.63∘ E
85.1
82.5
13.3
13.2
11.3
Bialystok, Poland
53.23∘ N
23.03∘ E
75.5
75.6
17.2
17.4
10.4
Karlsruhe, Germany
49.10∘ N
8.44∘ E
86.4
87.8
12.7
13.4
11.2
Garmisch, Germany
47.48∘ N
11.06∘ E
86.8
88.1
11.7
12.1
15.3
Park Falls, WI, USA
45.95∘ N
90.27∘ W
65.5
66.9
13.9
15.7
10.6
Indianapolis, IN, USA
39.86∘ N
86.00∘ W
83.5
85.1
11.9
13.6
8.7
Lamont, OK, USA
36.60∘ N
97.49∘ W
69.5
73.3
17.0
19.6
12.4
Pasadena, CA, USA (Caltech1)
34.14∘ N
118.13∘ W
78.6
88.2
14.3
16.6
11.0
Pasadena, CA, USA (JPL2)
34.12∘ N
118.18∘ W
41.5
45.9
26.6
27.9
17.9
Pasadena, CA, USA (JPL3)
34.12∘ N
118.18∘ W
75.3
80.1
24.1
25.4
16.3
Saga, Japan
33.24∘ N
130.29∘ E
80.1
85.6
26.2
26.8
18.6
Izaña, Tenerife, Spain
28.30∘ N
16.50∘ W
74.8
80.8
11.9
12.8
10.0
Ascension Island
7.92∘ S
14.33∘ W
51.5
57.0
26.8
26.2
21.7
Darwin, Australia
12.42∘ S
130.89∘ E
29.1
32.5
28.3
26.9
25.4
Réunion, France
20.90∘ S
55.49∘ E
44.5
48.3
27.1
25.5
24.7
Wollongong, Australia
34.41∘ S
150.88∘ E
25.0
29.4
36.6
34.4
34.0
Lauder, New Zealand (120HR)
45.04∘ S
169.68∘ E
17.9
22.6
23.6
21.4
20.2
Lauder, New Zealand (125HR)
45.04∘ S
169.68∘ E
38.8
44.6
23.4
21.2
20.7
1 California Institute of Technology, 2012. 2 Jet Propulsion Laboratory,
2007–2008.
3 Jet Propulsion Laboratory, 2011–2012.
Root mean square error (RMSE) between GOSAT and model XCH4
with averaging kernel applied (ppb). The inversions with the smallest RMSE
are marked in bold.
Region (mTC)
Prior
Posterior
L62T,
L62G
L62T
L78T
L62G
L78T
Global (1–20)
68.5
68.5
9.5
9.7
5.1
Europe (11–14)
94.1
94.1
11.5
12.1
16.3
North American boreal (1)
94.0
94.0
11.2
11.7
15.3
North American temperate (2)
87.1
87.1
10.1
11.3
11.7
South American tropical (3)
54.8
54.8
23.0
22.7
19.8
South American temperate (4)
48.3
48.3
17.4
15.9
16.0
Northern Africa (5)
80.5
80.5
7.8
9.8
8.9
Southern Africa (6)
49.0
49.0
18.2
17.3
16.3
Eurasian boreal (7)
96.4
96.4
12.2
12.9
17.5
Asian temperate (8)
90.0
90.0
10.5
12.2
10.2
Asian tropical (9)
87.8
87.8
22.7
23.9
17.3
Australia (10)
48.2
48.2
15.4
13.7
13.4
South-west Europe (11)
90.6
90.6
12.5
12.9
15.8
South-east Europe (12)
93.4
93.4
13.8
14.7
18.7
North-west Europe (13)
93.5
93.5
15.0
16.0
19.1
North-east Europe (14)
93.0
93.0
12.6
13.5
17.5
Ocean (16–20)
60.1
60.1
13.7
13.0
9.3
Mean emission estimates and their uncertainties before and after
2007 (Tg CH4 yr-1). The uncertainties are 1 standard deviation of
ensemble distributions. Prior uncertainties are from inversion L62T and
L62G. The L78T has larger prior uncertainties in all regions due
to its setup. For other regions, see the Supplement. Emission estimates after 2007 that
are more than 1 Tg CH4 yr-1 larger than those before 2007 are
marked in bold.
Region (mTC)
Total
Anthropogenic
Biospheric
Before 2007
After 2007
Before 2007
After 2007
Before 2007
After 2007
Global (1–20)
Prior
532.9 ± 86.7
566.0 ± 102.6
313.0 ± 80.7
350.5 ± 97.5
172.8 ± 31.6
171.8,± 31.8
L62T
507.0 ± 45.1
526.3 ± 43.7
287.0 ± 36.4
314.9 ± 34.5
172.8 ± 28.7
167.7 ± 28.7
L78T
508.2 ± 62.0
526.3 ± 60.9
311.4 ± 50.2
326.0 ± 49.7
149.7 ± 45.1
156.6 ± 44.1
L62G
509.1 ± 45.9
527.6 ± 44.0
287.9 ± 37.4
312.2 ± 34.8
174.1 ± 28.8
171.7 ± 28.9
Europe (11–14)
Prior
56.2 ± 14.2
55.0 ± 14.5
45.4 ± 13.6
45.0 ± 14.1
9.8 ± 3.9
9.0 ± 3.5
L62T
54.2 ± 10.4
51.5 ± 10.5
46.8 ± 10.3
43.8 ± 10.5
6.4 ± 2.7
6.8 ± 2.5
L78T
53.3 ± 13.3
53.3 ± 13.3
45.1 ± 13.4
45.1 ± 13.5
7.2 ± 3.6
7.1 ± 3.4
L62G
59.7 ± 10.6
58.5 ± 10.7
50.9 ± 10.6
49.1 ± 10.7
7.7 ± 2.7
8.4 ± 2.5
North American temperate (2)
Prior
42.0 ± 20.5
41.9 ± 20.5
33.2 ± 20.3
32.9 ± 20.3
7.7 ± 3.0
7.8 ± 3.0
L62T
49.2 ± 7.7
51.9 ± 6.8
41.8 ± 7.7
45.1 ± 7.0
6.3 ± 2.7
5.7 ± 2.6
L78T
48.4 ± 9.2
48.1 ± 6.8
42.2 ± 9.4
43.1 ± 7.3
5.1 ± 3.7
3.8 ± 3.5
L62G
55.6 ± 8.4
59.1 ± 7.5
47.4 ± 8.4
51.3 ± 7.7
7.2 ± 2.7
6.6 ± 2.7
South American temperate (4)
Prior
40.0 ± 14.9
42.8 ± 16.0
23.2 ± 13.1
25.5 ± 14.4
14.2 ± 7.0
14.5 ± 6.9
L62T
49.4 ± 14.6
63.3 ± 14.9
28.0 ± 12.9
39.9 ± 13.5
18.8 ± 6.9
20.6 ± 6.7
L78T
51.9 ± 24.6
66.0 ± 24.7
33.6 ± 22.5
46.4 ± 23.0
15.7 ± 9.8
16.9 ± 9.9
L62G
46.0 ± 14.6
58.8 ± 15.0
26.3 ± 12.9
37.9 ± 13.5
17.0 ± 6.9
18.2 ± 6.8
Asian temperate (8)
Prior
142.4 ± 72.7
164.7 ± 89.8
106.2 ± 72.1
129.3 ± 89.3
34.2 ± 9.6
33.4 ± 9.5
L62T
76.3 ± 24.2
83.7 ± 20.1
36.9 ± 25.0
50.1 ± 20.7
37.4 ± 6.5
31.5 ± 6.1
L78T
66.8 ± 28.7
80.6 ± 24.2
48.4 ± 26.6
54.8 ± 23.2
16.4 ± 24.7
23.8 ± 22.5
L62G
78.2 ± 25.2
81.0 ± 19.9
37.8 ± 26.1
44.2 ± 20.6
38.5 ± 6.9
34.8 ± 6.4
Asian tropical (9)
Prior
67.7 ± 15.8
70.8 ± 16.6
30.6 ± 8.7
35.7 ± 9.8
31.1 ± 13.2
31.3 ± 13.3
L62T
67.5 ± 14.3
68.3 ± 14.7
32.0 ± 8.4
35.1 ± 9.3
29.6 ± 12.1
29.4 ± 12.1
L78T
69.2 ± 27.8
67.5 ± 28.8
32.2 ± 23.0
32.5 ± 24.7
31.1 ± 19.6
31.3 ± 19.7
L62G
63.2 ± 14.3
65.1 ± 14.8
29.8 ± 8.4
32.8 ± 9.4
27.4 ± 12.2
28.5 ± 12.2
Spring peaks seen in GOSAT XCH4 in global, ocean and the Asian tropical
mTC region point to an important role of the vertical mixing scheme, which
are well captured in L62G, but not in L62T and L78T
(Figs. 7, S6). The difference is statistically significant considering the
ensemble distribution. Monthly emission estimates in L62G are generally
larger than in L62T and L78T during November–April, especially in
the northern-latitude temperate regions (35–60∘ N, Fig. S7). This
suggests that winter emissions in the northern latitude temperate regions,
enhanced in the model by faster vertical mixing around the surface, play an
important role to reproduce the XCH4 seasonal cycle in the tropics well.
Although GOSAT retrievals are valuable for evaluating model performance, it
is important to keep in mind that the satellite retrievals do not always
agree with ground-based TCCON retrievals. GOSAT XCH4 has been evaluated
against TCCON retrievals, but biases in the GOSAT products remain, especially
in the latitudinal gradient (Yoshida et al., 2013). This is probably one of
the reasons for the positive model bias in the NH compared to GOSAT
(Fig. S6). Furthermore, the seasonal amplitude of GOSAT XCH4 is much
smaller than that of the posterior estimates, especially in the SH (Fig. S6).
This is not in line with the TCCON comparison (Figs. 6, S5), which suggests
that disagreement with GOSAT XCH4 in the latitudinal gradient and the
seasonal amplitude may not only be due to problems in the inversions.
Global GOSAT and simulated regional 10-day mean XCH4.
Prior and posterior annual emission estimates for global, Asian
temperate and Asian tropical regions. Shaded areas are prior uncertainties,
and vertical bars illustrate L62T posterior uncertainties. The
uncertainties are 1 standard deviation of ensemble distributions. Note
different ranges on the y axes.
Anomalies of 12-month moving averages of monthly mean emission
estimates from five sources. Note that ocean emissions are only from natural
sources, i.e. anthropogenic emissions over the ocean are included in
anthropogenic emission. Zero levels shown by black lines are the mean of the
2000–2012 moving averages.
Emission estimates
Global
Our posterior mean total global emission estimate for 2000–2012 is
517 ± 45 Tg CH4 yr-1 with an increasing trend of
3 Tg CH4 yr-1 (Table 6, inversion L62G). Posterior mean
total global emissions for 2000–2012 are approximately
29 Tg CH4 yr-1 smaller than the prior (Table 6), although the
posterior estimates are within the range of prior uncertainties
(±93 Tg CH4 yr-1). Posterior mean total global emission
estimates from inversions L62T, L78T and L62G agree well, and
are in line with previous studies, e.g. Bousquet et al. (2006) and Fraser et
al. (2013). The main differences in the long-term mean are that anthropogenic
mean annual emission estimates in L78T are more than
10 Tg CH4 yr-1 larger than in L62T and L62G, which are
compensated by smaller biospheric emissions (Fig. 8). This change in
long-term mean flux is not robust in the L78 configuration, as the
uncertainty is large.
All inversions show an increase in posterior mean total global emissions from
before 2007 to after 2007 by 18–19 Tg CH4 yr-1 (Table 6), which
is much smaller than the increase in prior emissions of
33 Tg CH4 yr-1. The increase in posterior emissions during
2000–2010 is 15–16 Tg CH4 yr-1 and this agrees well with
previous studies by Bergamaschi et al. (2013) and Bruhwiler et al. (2014) for
example, who estimated an increase of about 16–20 Tg CH4 yr-1.
The increase in total global emissions is dominated by the anthropogenic
sources in both posterior and prior, and again the increase in the posterior
(15–28 Tg CH4 yr-1) is much less than in the prior EDGAR v4.2
FT2010 inventory (37 Tg CH4 yr-1) (Fig. 9, Table 6). The
posterior anthropogenic emission estimates from 2003–2005 to 2007–2010
increase by 15–23 Tg CH4 yr-1, which agrees well with
Bergamaschi et al. (2013) who estimated an increase of
14–22 Tg CH4 yr-1. However, the increase in anthropogenic
emission estimates is larger than reported by Bruhwiler et al. (2014) who
found an increase of around 10 Tg CH4 yr-1 from 2000–2005 to
2007–2010. The differences between the inversions are partly due to
different time periods used, but also due to the use of different sets of
observations and prior fluxes. Bergamaschi et al. (2013) used SCIAMACHY
satellite-based retrievals and NOAA observations, whereas Bruhwiler et
al. (2014) used in situ NOAA discrete and Environment and Climate Change
Canada (ECCC) continuous observations. Our study is also based on in situ
observations, but includes more discrete and continuous observations globally
than the previous two studies. Therefore, estimates from our study could
potentially contain important additional information from observations other
than those from NOAA and ECCC. In regard to prior emissions, this study and
Bergamaschi et al. (2013) used EDGAR v4.2 inventory estimates (the estimates
are similar although slightly different versions were used), while Bruhwiler
et al. (2014) used a constant prior from EDGAR v4.2 for 2000. Although
Bergamaschi et al. (2013) found a significant increase in anthropogenic
emissions in the constant-prior inversion, the increase was slightly smaller
than in their inversions with the trend included in the prior. This could
have caused the smaller trend estimated by Bruhwiler et al. (2014), compared
to this study.
Biospheric emission estimates in the L62T and L62G inversions after
2007 are slightly smaller than before 2007 (-5 to
-2 Tg CH4 yr-1), following the prior
(-1 Tg CH4 yr-1). In contrast, L78T shows an increase
(+7 Tg CH4 yr-1). The increase is driven by much smaller
biospheric emission estimates in the L78T inversion before 2007, mainly
due to significantly smaller biospheric emissions in the temperate Asian
region (discussed in Sect. 3.4.3). The small negative trend in biospheric
emissions in L62T and L62G is in line with the finding by
Bergamaschi et al. (2013). Here, it is again important to note that
interannual variability in the CH4 sink, which could also influence
total emissions to the atmosphere, is not included in this study.
Northern Hemisphere boreal regions and Europe
In this section, results for the following mTCs are presented: North American
boreal region (mTC1), Eurasian boreal region (mTC7), and Europe (mTC11–14).
Posterior anthropogenic emissions for Europe as a whole (mTC11–14) are
similar to the prior (L62T, L78T) (Table 6), but shifts in the
relative contributions to total European emissions from different parts of
Europe occurred. Posterior emissions are larger than the prior in southern
Europe (south-west Europe (mTC11) and south-east Europe (mTC12)), whereas the
posterior is smaller than the prior in north-east Europe (mTC14) in all
inversions (Table S1). Most of the increase in southern Europe and the
reduction in north-east Europe are due to anthropogenic emissions. Observed
atmospheric CH4 during winter at many of the in situ sites in northern
Europe can be good indicators of anthropogenic signals, because emissions
from biogenic sources are small during winter. Posterior atmospheric CH4
at these sites during winter agrees well with observations, which would
indicate that the posterior anthropogenic emissions are reasonable. Southern
Europe is only a small source of biospheric emissions, so most of the
atmospheric signals captured at the in situ sites in the region are from
anthropogenic sources. In southern Europe, posterior atmospheric CH4
values at some sites in France, Spain and Italy have a strong positive bias
(> 10 ppb), which exceeds the ensemble standard deviations, although the
correlations between observed and posterior CH4 are strong (0.8 or
larger). The posterior atmospheric CH4 values at other sites in
south-east Europe are not overestimated, but the correlations are often
weaker. This suggests that the inversion did not find a solution that matches
all the observations equally, because of an incorrect distribution in the
prior within the optimization region. It could also imply that some
measurements had local influence that the model could not represent or that
the MDM was too small for a few sites. However, the Chi-squared statistics at
European sites showed no indication that MDM was too small. Evaluation with
aircraft observations shows that vertical transport of CH4 in Europe is
generally good, but evaluation data were only available from central Europe,
i.e. we cannot exclude the problem of mixing in the atmosphere elsewhere.
Posterior anthropogenic emissions for north-west Europe are similar to the
prior. This finding is in line with Bergamaschi et al. (2015), who estimated
the anthropogenic emissions in north-west European countries to be similar to
the EDGAR v4.2 estimates and larger than the emissions reported in
UNFCCC (2013).
For biospheric emission estimates, differences between prior and posterior
emissions are negligible in southern Europe (Table S1), whereas the reduction
in the posterior is clear in northern Europe (north-west and north-east
Europe) (Fig. S8). A reduction in biospheric emissions is also
estimated for the North American boreal region (Fig. S8). This suggests that
the prior biospheric emissions in boreal regions are too large, which results
in larger prior atmospheric CH4 values than observed. The interannual
variability in the posterior emissions also does not follow the prior. An
increase in the posterior biospheric emissions is found for
50–90∘ N in 2006, followed by a decrease until 2010, which is not
prominent in the prior. Most of the 2006 increase is from the North American
boreal region. This finding does not agree with previous studies, e.g.
Bousquet et al. (2011), who found little increase in high northern latitude
wetland emissions in 2006. Instead, a significant increase in emissions was
found in 2007 in their study. However, observations from specific locations
support our findings, although the representativeness of a regional-scale
signal is questionable. Moore et al. (2011) reported that 2006 was a warm and
wet year at Mer Bleue bog in Canada (45.41∘ N, 75.48∘ W),
and for the period 2004–2008, the highest autumn CH4 emissions were
observed in 2006. The posterior biospheric emission estimates for north-east
Europe in 2006 are about 60 % smaller than the prior estimate in all
inversions. Drewer et al. (2010) found that CH4 emissions in September
in Lompolojänkkä fen in Finland (67.60∘ N,
24.12∘ E) were larger in 2006 than in 2007 due to heavy rain.
However, the summer of 2006 was dry with low emissions and snow had already
started to fall by the end of September, cutting the emission season short
with below zero (∘C) temperatures. As such, mean annual CH4
emissions from the fen were lower in 2006 than in 2007. The high prior
emissions in September–October 2006 could be due to a bias in precipitation
(excluding snow) and temperature in meteorological data from the Climatic
Research Unit (CRU), University of East Anglia, UK (Mitchell and Jones,
2005), which was used as an input for the LPX-Bern model. CRU precipitation
and temperature at Lompolojänkkä and the mTC14 average are larger
than the observations at Lompolojänkkä during autumn 2006. The
posterior summer biospheric emissions in 2007 are nearly twice as large as
the prior. The posterior shows high emissions in July, but the LPX-Bern
estimates are low during the summer and autumn at Lompolojänkkä and
in mTC14 on average. This could be due to problems in the wetland fraction or
in the precipitation dependence. CRU precipitation in 2007 is high in early
summer and extremely heavy in July at Lompolojänkkä and in mTC14 on
average, which is in line with Drewer et al. (2010). Although the seasonal
cycle of the precipitation is well captured in CRU, if the peatland soil is
already saturated with water in early summer, CH4 emissions would not
have increased with additional high summer precipitation. For north-west
Europe, similar results are found; posterior biospheric emissions are low in
summer–autumn 2006 and high in summer 2007, compared to the prior. The CRU
meteorology again agrees well with measurements at Stordalen mire in northern
Sweden (68.20∘ N, 19.03∘ E) for example, where the measured
emissions (Jackowicz-Korczyński et al., 2010) also support the posterior
estimates more than the prior.
Differences in emissions between the T1989 and the G2000 convection schemes
are prominent in all northern boreal regions and Europe. Posterior emissions
in L62G are larger than in L62T and L78T throughout
2000–2012. The estimated prior surface atmospheric CH4 values in these
regions are lower when the G2000 scheme is used. This indicates that the
stronger vertical transport in the G2000 reduces the surface CH4
abundance faster than the T1989 scheme and led to larger posterior
emissions. We cannot conclude which convection scheme is more suitable for
northern boreal regions and Europe based only on the posterior atmospheric
CH4 of those regions, but the agreement with the model-independent
aircraft and TCCON retrievals are better in the inversion using the G2000
scheme than in others using the T1989 scheme. Also, van der Veen et al. (2013)
found that G2000 more accurately represented vertical transport based on
simulations of atmospheric SF6. Note that the number of available GOSAT
retrievals, which agree better with the inversion results using T1989
scheme, is limited for northern Europe, and the retrieval bias (Yoshida et al.,
2013) makes the independent information less reliable.
Northern Hemisphere temperate regions
In this section, results for North American (mTC2) and Asian (mTC8)
temperate regions are presented.
Posterior total emissions for the North American temperate region are larger
than prior emissions in all inversions (Fig. S8, Table 6). The main
contribution to the increase in total regional emissions is from
anthropogenic emissions. Posterior mean anthropogenic emissions for
2000–2001 are closer to the prior, but nearly 10 Tg CH4 yr-1
larger than the prior for 2004–2012 (Fig. S8). The trend during 2000–2012
is not significant in the prior or in the posterior, although the posterior
shows an increase of 0.5 Tg CH4 yr-1 during 2000–2012. The
estimated growth rate is similar to the estimates reported by Bruhwiler et
al. (2014), but only about one third of that reported by Turner et
al. (2016). Our evaluation shows that the trend in posterior XCH4
matches well with the GOSAT and TCCON retrievals regionally and at sites in
the USA, e.g. Park Falls and Oklahoma (Figs. 6, S5, S6). In this study,
emissions were optimized by region, and there was only one scaling factor
for anthropogenic emission estimates for the North American temperate region.
Therefore, it is not possible to study the differences in the emissions trend
on the eastern and western sides of the North American temperate region, as
in Turner et al. (2016). However, this study suggests that a large increase
in local emissions is not necessary to reproduce the increasing atmospheric
CH4 trend. Long-range transport plays a more important role than the
local emissions.
A negative correlation is found between mean posterior anthropogenic and
biospheric emissions for the North American temperate region, i.e.
anthropogenic emissions increased when biospheric emissions decreased. This
is an effect of the inversion not being able to separate biospheric and
anthropogenic emissions based on the current observational network. In situ
observation sites in this area are mostly close to anthropogenic emission
sources, so the interannual variability found in biospheric emission
estimates may not represent the real variability.
The Asian temperate region has large anthropogenic and biospheric emissions
(Table 6). Anthropogenic emissions are responsible for most of the increase
in the prior regional and total global emission estimates after 2007.
However, prior anthropogenic emissions in this mTC are reduced by more than
half in the posterior (Fig. 8, Table 6). Moreover, the increase in posterior
anthropogenic emissions for 2000–2012 is not as strong as in the prior
(Fig. 8, Table 6). The significant reduction in anthropogenic emissions from
prior to posterior estimates for 2002–2010 is driven by observations from
two continental sites in Korea; Anmyeon-do (AMY, data available for
2000–2012) and Gosan (GSN, data available for 2002–2011). Small values of
MDM were initially assigned and thus the sites had a large impact on the
regional flux estimates. When MDMs for those sites are set to 1000 ppb,
thereby reducing their influence in the inversion (referred to as
L62T-K, L78G-K), the estimated total emission in this mTC is about
30 Tg CH4 yr-1 larger and in better agreement with Bruhwiler et
al. (2014) and Bergamaschi et al. (2013) for example.
The increased Asian temperate emissions in simulations L62T-K and
L78G-K are mainly compensated by reduced fluxes in the Asian tropical
region (about 10 Tg CH4 yr-1 in L62, about
20–30 Tg CH4 yr-1 in L78) (Fig. 8), as well as in the
Eurasian boreal region, Europe, and the ocean. Only small changes are found
in regional emission trends, but the anthropogenic ocean emission estimates
in L62T-K and L78G-K increase less during 2009–2012 compared to
that in L62T and L78T. When the two Korean sites are excluded from
the inversion, the posterior biospheric emissions in the Asian temperate
region remain close to the prior. The interannual variability in total
emissions in L62T-K and L78G-K is smaller than that of L62T
and L78G for the Asian temperate region. It is rather unrealistic that
regional anthropogenic emissions change by more than
30 Tg CH4 yr-1 over 1 to 2 years as is the case in L62T,
L78T, and L62G. Fast growing economies, such as China and India are
located in the Asian temperate region, and there is no evidence that the
anthropogenic emissions decreased significantly during 2002–2010 in that
region. Total emission estimates for the Asian temperate region in
L62T-K and L78G-K are larger and more reasonable than in L62T
and L78T, and the ratio of anthropogenic to biospheric emission
estimates in L62T-K and L78G-K are more consistent with each other
than in L62T and L78T. This suggests that the L62T and
L78T posterior anthropogenic emissions and the L78T posterior
biospheric emissions for 2002–2010 are probably unreasonably low due to the
influence of the two Asian sites, AMY and GSN. Nevertheless, the posterior
emissions in L62T and L78T are lower than in the EDGAR v4.2 FT2010,
which is in agreement with previous studies (Pandey et al., 2016; Thompson et
al., 2015). The effect of the changes in the emission estimates (L62T-K
and L78G-K) to XCH4 is small, although a slight increase is found
globally. The agreements with GOSAT and TCCON XCH4 in L62T-K and
L78G-K are slightly better for mTCs and at sites where negative biases
are found in L62T and L78T (not shown).
Asian and South American tropical regions
In this section, results for the following regions are presented: South
American tropical (mTC3) and Asian tropical (mTC9).
The Asian tropical region also has large anthropogenic and biospheric
emissions. Prior estimates from both sources are about
30 Tg CH4 yr-1 each, and they are reduced slightly by the
inversions (Fig. 8, Table 6). Posterior estimates for biospheric and
anthropogenic emissions are lower than in Bruhwiler et al. (2014), who
estimated the anthropogenic emissions to be even larger than, and biospheric
emissions to be similar to, our prior. The L78T anthropogenic emission
estimates are lower than the prior estimates due to enhanced, and probably
unrealistic, interannual variability compared to the L62T and L62G
estimates (Fig. 8). This partly correlates with the strong interannual
variability in the Asian temperate region. For example, the increase in
anthropogenic emissions in L78T around 2002–2005 is due to a strong
decrease in emissions in the Asian temperate region. In the test cases,
L62T-K and L78G-K, interannual variability in both the Asian
temperate and Asian tropical regions is smaller than in L62T and
L78T (Fig. 8). However, annual anthropogenic emission estimates in
L78G-K are much lower than in L78T, and about
20 Tg CH4 yr-1 smaller than in L62G. This is partly due to
the differences in the convection schemes, which is also seen in the L62
configuration. However, it is mostly due to compensating effect of the
increased Asian temperate anthropogenic emissions that resulted from reducing
the influence of the observations at the Korean sites. Evaluation with surface in
situ observations shows that L62G atmospheric CH4 values agree best
with observations at BKT, where the inversions have a strong negative bias.
Nevertheless, large uncertainty remains in the estimates, so further
information, such as additional observations and prior information about the
emissions, is needed to better quantify emissions in this region.
The emission estimates for the South American tropical region are very
similar to each other (Fig. S8, Table S1). All posterior emissions are close
to the prior, and the uncertainty in the posterior is not reduced by the
inversions. This is due to a lack of observations assimilated within the
optimization regions in mTC3. Three stations (MEX, KEY, RPB) near the edge of
mTC3 were assimilated, but due to strong vertical transport, these
observations do not capture signals from tropical wetlands, which is the main
CH4 source from this mTC. Moreover, most of the assimilated observations
are samples from well-mixed air masses that represented a large volume of the
atmosphere. Therefore, the inversions could not satisfactorily constrain
emissions in the South American tropical region.
Africa and southern mid-latitudes
In this section, results for the following regions are presented: South
American temperate region (mTC4), northern Africa (mTC5), southern Africa
(mTC6) and Australia (mTC10).
Posterior total emissions in the South American temperate region increase
significantly during 2006–2009 in all inversions (Fig. S8), and there is no
correspondent decrease in other mTCs, e.g. the Asian temperate region. All
inversions point in the same direction, but the results are still debatable.
Observations assimilated within mTC4 before 2006 are from Ushuaia (USH) in
Argentina. Due to its location (54.85∘ S) having few local emission
sources, the purpose of the site is to sample well-mixed air that represents
a large volume of the atmosphere. Observations at Arembepe, Brazil (ABP) were
available during 2006–2009, and at Natal, Brazil (NAT) during 2010–2012.
These sites capture the well-mixed air in the tropics better than USH,
although most of the signals are from the Atlantic Ocean and not from the
land. Interannual variability in the tropics is probably better represented
by ABP and NAT observations, but it is questionable whether the variability
is driven by the observation signals from the South American temperate
region. Similar interannual variability was reported by Bruhwiler et
al. (2014), where ABP observations were assimilated (the NAT observations
were outside their study period), although the changes were not as
significant as in this study.
South American temperate is the only region where all inversions show a
significant increase in both anthropogenic and biospheric emissions
(Table 6). As mTC4 is mostly within 30∘ S–30∘ N, and most
of the emissions are located in the northern part of this mTC, the estimates
agree with Houweling et al. (2014) who found that most of the increase in
total global emissions was in the tropics and the extratropics. The increase
in emissions during 2005–2008 and the subsequent decrease (Fig. S8) was also
found in Basso et al. (2016), who suggested that biospheric emissions from
the east part of the Amazon basin were the main contributor to interannual
variability. Dlugokencky et al. (2011), using constraints from CH4
isotopic measurements, suggested emissions from the tropics were an important
contributor to the significant growth in atmospheric CH4 after 2007. The
isotopic measurements showed a decrease in the δ13C–CH4,
which would indicate that the increased emissions were probably from biogenic
sources. The inversions in this study have difficulty changing the ratio of
anthropogenic to biospheric emissions from the prior, which could be a reason
why the interannual variability of total emissions is optimized by changing
emissions from the major sources, i.e. anthropogenic sources. Therefore, interannual
variability of the posterior emissions is dominated by the contributions from
anthropogenic sources.
Posterior anthropogenic emissions in the northern African and southern African
mTCs are larger than the prior for all inversions, with somewhat different
interannual variability in the north and south (Fig. S8). Evaluation with in
situ observations in northern Africa shows that there is only a small bias in
the posterior atmospheric CH4 values (< 1 ppb in L62G). For
southern Africa, agreement with the in situ observations is good, except for
Mt. Kenya, Kenya (MKN) where a strong negative bias is found (see Sect. 3.1).
The correlation between the posterior and observed atmospheric CH4
values at MKN is strong (≥ 0.8), and the site is located at a high
altitude (> 3000 m a.s.l.), which implies that the bias may not be due
to small local emissions. On the other hand, vertical transport in the
tropics is strong, and MKN is located near a biospheric source area in
central Africa. Therefore, the negative bias could also be due to an
underestimation of emissions from wetlands in central Africa. Bruhwiler et
al. (2014) also reported an increase in the posterior estimates compared to
their prior in Africa, but the increase was mainly in biospheric emissions.
However, our interannual variability in anthropogenic emissions in northern
Africa is similar to their variability in central African biospheric emission
estimates. Therefore, the differences may partly be due to differences in the
prior: the ratios of prior anthropogenic to biospheric emissions in this
study and Bruhwiler et al. (2014) are almost reciprocals of each other, i.e.
our prior anthropogenic emissions are larger and biospheric emissions are
lower than in Bruhwiler et al. (2014). It is not possible to conclude from
this study which estimates better capture actual emissions, because the
estimates for Africa are not well constrained by the observations in either
study.
Posterior emissions for Australia in L78T are systematically larger than
in L62T and L62G throughout 2000–2012 (Fig. S8). The southernmost
coast of Australia and much of New Zealand are defined as “biospheric” land
in L62 configuration (Fig. S4), i.e. anthropogenic emissions in that
optimization region were not optimized in L62T and L62G. Since
biospheric emissions are a minor source and the posterior emissions changed
little from the prior in L78T, the “biospheric” land in the
land-ecosystem map may need to be changed to “anthropogenic” land for mTC10
to be able to optimize anthropogenic emissions better in L62T and
L62G.
Ocean
Prior anthropogenic ocean emissions are mainly located in the tropics
(mTC20), and the main differences between prior and posterior emissions are
also located in this mTC (Fig. S9). All posterior fluxes are
5–10 Tg CH4 yr-1 larger than the prior, especially before 2006
and during 2011–2012 (Fig. S9). However, it is questionable whether these
results are reasonable, since there is no indication that non-road
transportation and coastal anthropogenic emission estimates varied from
year to year as the inversion results show. It is more likely that ocean
regions were used to compensate for missing tropical land emissions. Indeed,
the estimates for the ocean were sensitive to the estimates in other regions
(not shown). Further investigation without optimizing anthropogenic ocean
emissions or using only natural ocean emissions as prior, i.e. excluding
non-road transport (ship and aircraft) emissions, would help us to better
understand the anthropogenic emission estimates over land. Note that the
prior biospheric emission estimates in mTC16-20 were not optimized. Prior
biospheric emissions around the coast were not zero, partly due to
differences in the definition of the coast in the mTC16-20 line in our mTC
map and the prior. Only limited information is available in regard to
biospheric emissions around coastlines, and as it is a minor source, it was
assumed that the inversion would not be able to optimize it.