ACPAtmospheric Chemistry and PhysicsACPAtmos. Chem. Phys.1680-7324Copernicus GmbHGöttingen, Germany10.5194/acp-15-9765-2015Sensitivity of the recent methane budget to LMDz sub-grid-scale physical parameterizationsLocatelliR.robin.locatelli@lsce.ipsl.frBousquetP.SaunoisM.ChevallierF.https://orcid.org/0000-0002-4327-3813CressotC.Laboratoire des Sciences du Climat et de l'Environnement (LSCE), Gif sur Yvette, FranceR. Locatelli (robin.locatelli@lsce.ipsl.fr)1September201515179765978025February201521April201517July201512August2015This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://acp.copernicus.org/articles/15/9765/2015/acp-15-9765-2015.htmlThe full text article is available as a PDF file from https://acp.copernicus.org/articles/15/9765/2015/acp-15-9765-2015.pdf
With the densification of surface observing networks and the development of
remote sensing of greenhouse gases from space, estimations of methane
(CH4) sources and sinks by inverse modeling are gaining additional
constraining data but facing new challenges. The chemical transport model (CTM)
linking the flux space to methane mixing ratio space must be able to
represent these different types of atmospheric constraints for providing
consistent flux estimations.
Here we quantify the impact of sub-grid-scale physical parameterization
errors on the global methane budget inferred by inverse modeling. We use the
same inversion setup but different physical parameterizations within one
CTM. Two different schemes for vertical diffusion, two
others for deep convection, and one additional for thermals in the planetary
boundary layer (PBL) are tested. Different atmospheric methane data sets are used as
constraints (surface observations or satellite retrievals).
At the global scale, methane emissions differ, on average, from 4.1 Tg CH4 per year due to the use of different sub-grid-scale
parameterizations. Inversions using satellite total-column mixing ratios
retrieved by GOSAT are less impacted, at the global scale, by
errors in physical parameterizations. Focusing on large-scale atmospheric
transport, we show that inversions using the deep convection scheme of
derive smaller interhemispheric gradients in methane
emissions, indicating a slower interhemispheric exchange. At regional scale,
the use of different sub-grid-scale parameterizations induces uncertainties
ranging from 1.2 % (2.7 %) to 9.4 % (14.2 %) of methane emissions
when using only surface measurements from a background (or an
extended) surface network. Moreover, spatial distribution of methane
emissions at regional scale can be very different, depending on both the
physical parameterizations used for the modeling of the atmospheric
transport and the observation data sets used to constrain the inverse
system.
When using
only satellite data from GOSAT, we show that the small biases found in
inversions using a coarser version of the transport model are actually
masking a poor representation of the stratosphere–troposphere methane
gradient in the model. Improving the stratosphere–troposphere gradient
reveals a larger bias in GOSAT CH4 satellite data, which largely
amplifies inconsistencies between the surface and satellite inversions. A simple
bias correction is proposed. The results of this work provide the level of
confidence one can have for recent methane inversions relative to physical
parameterizations included in CTMs.
Introduction
Inverse modeling techniques are a way to derive sources and sinks of methane
using atmospheric measurements as constraints. Today, large uncertainties
still affect the recent methane budget estimated by inverse modeling, even
at the global scale. For example, estimated methane
sources between 526 and 569 TgCH4 year-1 during the
2000–2009 period. The two major causes of uncertainties of methane
inversions are the limited and uneven coverage of atmospheric observations
and the errors made in the representation of atmospheric transport. However,
the increasing number of satellite data retrieving greenhouse gas atmospheric
columns and the densification of surface networks in space and time gradually
solve the issue related to atmospheric observations. Consequently, the
quality of the representation of atmospheric transport becomes the leading
issue in order to improve estimations by inverse modeling. Indeed, inverse
modeling requires a model to link methane emissions to methane mixing ratios
in the atmosphere. Such a model is generally a chemical transport model (CTM)
or a chemistry–climate model (CCM). Then, an atmospheric inversion scheme is
applied to greenhouse gas observations to derive the optimal methane source–sink scenario which satisfactorily fits both atmospheric observations,
given a CTM or CCM; a prior scenario of sources and sinks; and errors for
observations, model and emission scenarios . The optimal
character of such approaches assumes that these errors are properly estimated
in magnitude and are unbiased. Indeed, inversions are largely sensitive to
any sort of bias impacting simulated or measured methane mixing ratios. These
biases may be related to the CTM and/or the observation data sets, and they
directly perturb the optimization of methane fluxes by inverse modeling.
Biases in measurements, especially in satellite retrievals, are very likely
. For
example, the first release of SCIAMACHY data in 2005 was largely biased
producing very large tropical emissions . A major
revision has been done to the SCIAMACHY satellite retrievals
based on a revisit of the spectroscopic parameters
for methane, but inversions using SCIAMACHY retrievals still need to carry
out large bias corrections up to several tens of ppb
. Systematic errors in CTMs also have
significant impacts on inverse estimates. In , it was
shown that transport model errors are responsible for an uncertainty of
27 TgCH4 year-1 in the estimations of methane fluxes by
inverse modeling at the global scale. Moreover, showed
that stratosphere–troposphere exchanges are systematically too fast in the
version of LMDz (Laboratoire de Météorologie Dynamique model with Zooming
capability) using a low vertical resolution (19 levels), which could largely
impact the estimation of gas fluxes, like N2O, whose stratospheric
mixing ratios influence tropospheric mixing ratios. Furthermore, following
, showed that a bad representation of
the interhemispheric exchange in an ensemble of state-of-the-art CTMs can
explain most of the discrepancies in the global methane fluxes derived by
inverse modeling using these different CTMs.
Inconsistencies in inversions due to CTM errors may have multiple origins:
vertical/horizontal resolution, meteorological fields used to nudge
horizontal winds, sub-grid-scale physical parameterizations, advection
schemes, numerical methods, etc. Among the different contributions to CTM
errors, the quality of vertical mixing appears to be a key point to improve
. In the vertical, in global models,
transport processes such as planetary boundary layer (PBL) mixing or deep
convection have to be parameterized, due to them being on sub-scales of the model grid.
Here, we propose to assess the impact of different parameterizations of
sub-grid-scale transport on the inverted methane emissions for the year 2010.
Consequently, we run an ensemble of inversions using different versions of
the LMDz model. These LMDz versions differ only by the physical
parameterizations they use. Two parameterizations of vertical diffusion
, one parameterization of the thermals
and two deep convection schemes
are tested in three different versions of
LMDz. As the impact of model parameterizations can be different when either
assimilating surface data or satellite column data, we evaluate this impact
for three observational systems: two surface networks and one data set of
GOSAT retrievals.
As a result, this paper is not to be taken as an assessment of the global and
regional methane budget for 2010 but more as a study on the sensitivity of
this budget to atmospheric transport errors. In the following,
Sect. presents the setup of the ensemble of inversions
performed. The consistency between surface-based and satellite-based
inversions is then presented and a bias correction is proposed for the
satellite data (Sect. ). The impacts of the different
parameterizations used are then analyzed through the estimates of methane
emissions at the global scale (Sect. ) and at regional scales
(Sect. ).
Setup of variational methane inversionsPYVAR-LMDz-SACS
The PYVAR-LMDz-SACS (PYthon VARiational–Laboratoire de Météorologie
Dynamique model with Zooming capability–Simplified Atmospheric Chemistry
System) system is based on a variational
data assimilation system to derive the optimal state of CH4 fluxes
given CH4 observations and a background estimate of CH4
fluxes. Variational data assimilation involves minimizing a cost function
J, which is a measure of both the discrepancies between measurements and
simulated mixing ratios and between the background fluxes and the estimated
fluxes, weighted by their respective uncertainties, expressed in the
covariance matrices R (observations) and B (prior
fluxes), defined as follows:
J(x)=(y-Hx)TR-1(y-Hx)+(x-xb)TB-1(x-xb),
where x is the
state vector that contains the variables to be optimized during the inversion
process. In PYVAR, methane fluxes are optimized over 8-day periods in all
the grid cells of the model. The vector xb represents the prior
state of x. Likewise, the vector y contains the observations
of CH4. B is the prior error covariance matrix, and its
characteristics are mainly based on the work of :
its diagonal is filled in with the variances set to 70 % of the square of the maximum
of emissions over the nine model grid cells, which contain and surround each grid cell during each
month;
its off-diagonal terms of B (covariances) are based on correlation e-folding lengths (500 km over land and 1000 km over
sea);
no temporal correlations are considered in the B matrix.
Furthermore, the prior information included in the B matrix have several origins:
CH4 anthropogenic emissions are based on EDGAR v4.2-FT2010 estimates,
CH4 biomass burning emissions are based on the GFED3 inventory,
wetland emissions are based on the personal communication of J. O. Kaplan (2007)
.
The CH4 initial concentration is also optimized with a 1σ
uncertainty of ±10 ppb in the inverse process in order to not alias its
error into flux errors. However, only the initial CH4 atmospheric
columns at each grid cell are optimized for saving valuable computational
time. The R matrix accounts for all errors contributing to
mismatch between measurements and simulated CH4 mixing ratios.
R is usually considered a diagonal matrix because considering
covariance dramatically slows down the optimization and the knowledge about
these covariances is too poor. The main contributions to variances are
instrumental and model errors. In surface-based inversions, instrumental
errors are considered equal to 3 ppb and model errors are computed at each
site as the residual standard deviation (RSD) of the measurements on a smooth
curve fitting them. The RSD at each site is considered a proxy of the
transport model errors. Previous studies using PYVAR-LMDz-SACS have used this
approach . In satellite-based
inversions, GOSAT retrieval random errors are estimated to be about 0.6 %
of satellite measurements and a transport model error of
1 % of the observation values is added according to the results of
on tuning of error statistics. H is the
observation operator that projects the state vector x into the
observation space. H is represented here by the offline version of
LMDz complemented by a simplified chemistry module (SACS) to represent the
main reactions of the oxidation chain of methane . Here,
OH and O(1D) fields are prescribed. They come from a
full-chemistry simulation of the chemistry–climate model LMDz-INCA
. The different characteristics of the OH field
used here (for example, the global mean concentration is
11.5 × 105 molec cm-3 between surface and 100 hPa) are in
the range of the current knowledge on the radical hydroxyl (see the ACCMIP
experiment; ). No inter-annual variability is applied to
the OH field in this study.
The iterative minimizing process implies calculating the gradient of the cost
function, which is implemented using the adjoint technique, iteratively
solved with the M1QN3 algorithm developed by
until the gradient norm gets reduced by more than 99 %.
We also propose an
analysis of the posterior flux uncertainties based on the work of
. Indeed, our inversion setup is similar to that of
, in which error statistics in methane
inversions were inferred from different satellite and surface measurements, using
the same transport model. Consequently, we apply here the uncertainties
reductions found in to our inversions to provide an
estimate of the posterior flux uncertainties (see
Sect. ).
Three different versions of LMDz: LMDz-TD, LMDz-SP and LMDz-NP
LMDz is the general circulation model (GCM) of the IPSL Earth system model
. Here we use an offline version of LMDz
implemented in the variational inverse system described in
Sect. . The computation of air mass fluxes used by the
offline model is performed using the full LMDz GCM nudged on the analyzed
horizontal components of the wind from ERA-Interim . Only the
mass balance equation is solved within the variational system, based on the
stored air mass fields. In the following, LMDz refers to the offline version
of the GCM embedded in the variational system.
In this study, we use three different versions of LMDz using different
physical parameterizations (LMDz-TD, LMDz-SP and LMDz-NP) to simulate the
atmospheric transport. LMDz-TD uses the physical parameterizations included
in the original version of the inverse system of :
vertical diffusion is parameterized by a local approach from
, and deep convection processes are parameterized by the
scheme. LMDz-SP uses also a local approach to
parameterize vertical diffusion, but the scheme
parameterizes deep convection. LMDz-NP uses a combination of the
scheme and the thermal plume model of
to simulate atmospheric mixing in the boundary layer. Atmospheric transport
by deep convection is parameterized according to .
The horizontal resolution of these three different versions of LMDz is
3.75∘×1.875∘, and the vertical discretization has 39
layers. Some results coming from an old version of LMDz-TD using 19 layers
and a horizontal resolution of 3.75∘×2.5∘ are also
presented in Sect. .
More details on the characteristics of these three
versions of LMDz concerning the modeling of atmospheric transport and the
impact of the different versions of LMDz on the methane concentrations
can be found in . Briefly, LMDz-TD is characterized by a
low variability in the PBL due to an overestimation of the PBL mixing. It
leads to an underestimation of the strong gradient observed near sources. On
the other hand, LMDz-NP simulates the diurnal cycle of the PBL quite well.
The modeling of large-scale atmospheric transport has also been improved
with the deep convection scheme. The interhemispheric
(IH) exchange, which is known to be too fast in LMDz-TD, agrees better with
the indirectly measured IH exchange when using the
scheme, as is done in LMDz-SP and LMDz-NP. However, the
convection scheme is still used in the scientific community, which justifies
its inclusion as well in this work. Finally, it is important to remember that
these three different versions of LMDz simulate different CH4
lifetimes . LMDz-NP derives a lifetime 0.2 years longer than
LMDz-TD, which consequently contributes to uncertainties in the estimation of
methane sources and sinks by inverse modeling based on these different
versions of LMDz.
Three different observation data sets
Different observation data sets exist to constrain methane atmospheric
inversions. Surface observations have been assimilated for years, mostly for
background or coastal locations. However, more and more continuous and/or
continental sites have appeared in the recent years, which largely increase
the space and time density of surface observations. These observations are
precise and accurate although unevenly distributed in space and time at the
surface. Since 2003, satellite data for total methane weighted columns also
exist, largely increasing spatiotemporal coverage of observations but at the
cost of a lower precision of individual measurements.
In this study, two surface observation data sets (Sect. ) and
one GOSAT data set (Sect. ) are used to constrain
our inversions.
Two surface observation data sets
Location of the surface stations in the “background” (red circles only) and “extended” network (blue and red circles).
Two networks of surface stations have been used in the different inversions
of this study: the “background” and the “extended” networks. Red circles
in Fig. represent the location of surface stations in
the “background” network. The “background” (BG) network is mainly
representative of “background” air since most of the surface stations of
this network are located far away from the main methane sources. The
“extended” (EXT) network is an extension from the background network, where
24 stations are added to the “background” network (blue circles in
Fig. ). These additional stations have been selected for
their continental footprint, closer to methane emissions than most of the
background sites and therefore providing more direct information on methane
emissions. However, being closer to emission areas, and generally located
inland, they show more variable mixing ratios and are more sensitive to
transport errors . In the following, we use BG and EXT
to respectively refer to surface measurements in the background and extended
configuration of the surface network.
Inversions using these surface observations data sets have been run between
2006 and 2012, but we mainly present results for 2010 to be consistent with
the satellite inversions.
One satellite data set: GOSAT
Methane total weighted columns retrieved by GOSAT are also used in
our study to constrain methane inversions. Version 4.0 of the TANSO-FTS
XCH4 proxy retrievals performed at the University of Leicester
are used with associated averaging kernels and a priori
profiles. In the “proxy” method, it is considered that the CO2 and
CH4 spectral absorption bands are close enough to assume that light
path perturbations affecting CO2 total-column mixing ratio are
similar to those affecting CH4 total-column mixing ratio. Thus, the
ratio between the measured CH4 and CO2 vertical mixing ratio
is not affected by any perturbations due to aerosol scattering and clouds.
Consequently, the total column of CH4 (XCH4) is computed
according to XCH4=[CH4]meas[CO2]meas×XCO2mod, where [CH4]meas and
[CO2]meas are respectively the CH4 and CO2
measured mixing ratio, and XCO2mod is a model-derived estimate of
XCO2 coming from
In the following this data set is referred to as PR-LEI, which stands for
“Proxy-Leicester”.
Data from July 2009 to June 2011 are used in the inverse procedure to extract
the inferred methane fluxes for the year 2010 with limited time-side effects.
Consistency between surface-based and satellite-based inversions
The use of total-column CH4 retrievals from satellite is fundamental
for global inversions as it provides constraints within regions not sampled
by surface stations. In particular, satellite data provide information in
tropical regions, which are known to largely contribute to the global methane
budget and where few surface measurements are available. However, random and
systematic errors may be significant in satellite data sets. For example,
and have shown that SCIAMACHY
satellite retrievals were usable in methane inversions only if a bias
correction algorithm was added. also showed
inconsistencies between surface and GOSAT inversions, which could
be explained by space- or time-dependant biases in the GOSAT retrievals.
Discrepancies in the modeling of methane
vertical transport in the atmosphere could be another reason for this. Here, using the different versions of
the LMDz model, we estimate the inconsistencies between surface-based and
satellite-based inversions and we investigate the impact of the
representation of vertical transport on these inconsistencies.
Latitudinal distribution of the bias between simulated methane
mixing ratio using an optimized flux distribution coming from a
satellite-based inversion and methane mixing ratio measured at different
surface stations.
Four inversions are performed using GOSAT data without any bias correction
using the three versions of LMDz (LMDz-TD, LMDz-SP, LMDz-NP as described in
Sect. ) and the former 19-layer model version (LMDz-19)
related to LMDz-TD . The optimized atmosphere is then
sampled at surface stations and compared to surface observations for the four
different versions of the LMDz model (Fig. ). Methane surface
mixing ratios simulated from optimized fluxes using GOSAT
retrievals do not fit methane mixing ratios directly measured at surface
stations (Fig. ). The different 39-layer versions of the LMDz
model show a bias of about +40 ppb, with a small latitudinal dependency.
This means that, at the surface, the optimized atmospheric methane
concentrations seen by GOSAT are on average 40 ppb higher than
the observed concentrations. Such a bias can be due to satellite retrievals
and/or transport model errors. For comparison, found
lower but still significant global mean biases of +6.9 and +16.9 ppb in
two different GOSAT inversions. Little information is given on the latitudinal
distribution of these biases, even though they also seem slightly larger in
the Southern Hemisphere than in our study. After analyzing the scaling of the
optimized initial condition, we found that around 15–20 ppb of the 40 ppb
is explained by an increase in the initial condition, the rest being
explained by an increase in the CH4 emissions.
The similarity of biases derived by LMDz-TD, LMDz-SP and LMDz-NP
(Fig. ) highly suggests that sub-grid-scale parameterizations
of vertical transport only play a minor role in inconsistencies between
surface and simulated (based on satellite retrievals constraints) methane
mixing ratio. found similar results performing different
sensitivity tests to explain inconsistencies between surface-based and
satellite-based inversions. As a result, we can conclude that
parameterizations of deep convection and diffusion are likely not the cause
of these inconsistencies.
Interestingly, we find a very different result with the 19-layer version of
LMDz (LMDz-19). Indeed, LMDz-19 derives a smaller bias (up to +15 ppb in
the high latitudes of the Southern Hemisphere decreasing to down to
-10 ppb in the Northern Hemisphere). LMDz-19 differs from LMDz-TD only by
a coarser vertical resolution. Therefore, a higher vertical resolution seems
to degrade the bias of GOSAT inversions, despite the improvement in the
large-scale transport shown in for this new version of
LMDz.
Vertical profiles of methane mixing ratio in ppb (lines with stars)
for LMDz-39 (red) and LMDz-19 (black) and compared with HALOE climatology
(green). These profiles have been averaged between 60∘ N and
60∘ S for 2010. Only the contribution (in percent) of each
layer of the satellite retrievals to the total column (lines with crosses) is
also shown for the two versions of the model. The computation of the model
contributions to the total column does not account for the averaging kernel
of the satellite retrievals
.
In order to understand this large difference between the two vertical
resolutions of the LMDz model, we compare the simulated vertical profiles of
methane mixing ratios using LMDz-TD with 19 (LMDz-19) and 39 (LMDz-39)
vertical levels (Fig. ). Both simulated profiles use
the corresponding optimized methane fluxes derived by inversions using the
same atmospheric constraints (GOSAT PR-LEI). Figure
shows that the CH4 profile is very sensitive to the vertical
resolution. The largest differences are found in the stratosphere: LMDz-19
simulates much higher stratospheric methane mixing ratios compared to
LMDz-39. However, and consistent with mass balance, LMDz-19
tropospheric mixing ratios are smaller than LMDz-39. As found in
, the two versions of the model have very different
abilities to reproduce stratosphere–troposphere exchange (STE). STE is
particularly fast in LMDz-19 compared to LMDz-39, due to a coarser resolution
inducing more vertical diffusion. This induces stronger methane mixing ratio
in the stratosphere in LMDz-19. One could think that LMDz-19 simulates a more
consistent methane vertical distribution than LMDz-39 as biases in
Fig. are smaller for LMDz-19 than for LMDz-39. However, we
have compared the modeled methane mixing ratio vertical gradients with the
climatology from the HALOE instrument , and we have found
extremely similar gradients between LMDz-39 and HALOE data. Indeed, the
methane gradient between 200 and 3 hPa is 2.2, 5.5 and 5.3 ppb hPa-1
for LMDz-19, LMDz-39 and HALOE, respectively. As a result, we find that the
relative contribution of each vertical layer to the total column is very
different in LMDz-39 and LMDz-19 (lines with cross markers in
Fig. ). Stratospheric (tropospheric) layers in LMDz-39
contribute much less (more) than in LMDz-19. Consequently, the inverse system
derives lower methane fluxes with LMDz-19 to simulate a lower tropospheric
methane mixing ratio, compensating for the over-contribution of the stratospheric
methane mixing ratio to the total column. The relatively small bias found in
the validation of LMDz-19 satellite-based fluxes by surface measurements is
unfortunately due to an inadequate representation of the
troposphere–stratosphere methane mixing ratio gradient. LMDz-39 derives a
stronger bias between simulated and surface measurements, but it can be assert
that this bias is not due to errors in the modeling of the
troposphere–stratosphere gradient, which is improved compared to LMDZ-19. One
way to sort out this issue would be to compare to another model such as TM5
model (with 25 vertical levels), which was found to simulate a slower STE
than LMDz-19 in and infers a global mean bias of +6.9 and
+16.9 ppb depending on the GOSAT data set version.
Overall, to determine the reason for such biases in satellite inversions
still needs more attention on the model side, but most probably also on the
data side.
To overcome these model–data inconsistencies, satellite-based inversions are
performed in two steps. Firstly, we run inversions using GOSAT data without
adding any bias corrections. Secondly, we remove the latitudinal bias found
when we compute the difference of the concentrations simulated at each
surface stations using the optimized methane fluxes coming from the first
inversion with the surface observations considered unbiased. One could
argue that the constant part of the bias (∼ 40 ppb) could be absorbed
by the initial conditions. This is indeed partly the case for 15–20 ppb.
But the remaining bias translates into additional surface emissions, which
justifies removing the bias and performing a second inversion.
In the following, in addition to surface-based inversions, we focus on and
present only results associated with these two-step satellite-based
inversions.
Impact of physical parameterizations on global methane fluxes
Methane flux estimates (in Tg CH4 year-1) for 2010
at the global scale for each inversion (surface inversions using the
background and the extended networks, and inversions using proxy products
provided by the University of Leicester relative to GOSAT). Inversions
using LMDz-TD, LMDz-SP and LMDz-NP as CTM are plotted in red,
green and blue, respectively.
Figure displays the sensitivity of the global
methane budget to physical parameterizations by showing the global methane
estimates from nine inversions using the three different versions of the
model (Sect. ) and the three different data sets
(Sect. ). Using the BG, EXT, and PR-LEI data sets to
constrain methane inversions, we find that the spread (maximum - minimum) in
derived methane emissions due to changes in physical parameterizations is
respectively 2.7, 7.5, and 2.1 Tg CH4 in 2010. This
represents respectively 0.5, 1.4 and 0.4 % of methane global emissions. However, these
spreads are lower than when considering changes in
atmospheric observations: using TD, SP and NP model versions with the
different data sets, we find spreads of 5.0, 6.7 and 10.3 Tg CH4.
Therefore, the choice of atmospheric observations gives more spread in the
results than changing model parameterizations, at least in our case. This
is understandable, as changing the physical parameterization only accounts for
part of the total transport uncertainty. Indeed, the parameterization-based
spreads of 2.1 to 7.5 Tg CH4 are much lower than the
27 Tg CH4 year-1 found in the pseudo-experiment of
, which was assumed as an estimate for “total”
transport model errors. “Total” here refers to all the possible causes of
transport errors. The use of different physical parameterizations within the
same CTM integrated in the same inverse system has a significant impact on
global methane emissions, although logically smaller than using different
CTMs as done in . Indeed, here we only test a few
parameterizations of the vertical transport in one model. Transport models
can also differ in their horizontal resolution and horizontal advection, in
their meteorological forcings and the way they constrain atmospheric
transport, and in the coupling between their different characteristics.
The largest spread (7.5 Tg CH4) for one given data set is found for
the EXT inversions. This is especially due to the EXT-NP inversion, which
estimates global methane emissions of 539.8 Tg CH4 in 2010 compared
to 532.3 and 533.3 Tg CH4 for EXT-TD and EXT-SP, respectively. In
particular, this large estimation is due to a specific region: China. The
impact of the parameterizations on the methane flux estimates for China is
further discussed in Sect. .
We find that, at the global scale, the spread in GOSAT inversions
(2.1 Tg CH4) is lower than both BG (2.7 Tg CH4) and EXT
(7.5 Tg CH4) surface-based inversions. First, sub-grid-scale
parameterizations in CTMs mainly impact the modeling
of vertical transport. An inaccurate representation of methane vertical
distribution has larger impacts on simulated mixing ratios at the surface
than on the simulated total column. Indeed, a simulation of surface methane
mixing ratios, which takes place at a specific level of the atmosphere, could
miss or underestimate a methane plume if, for example, methane is transported
too quickly into the upper atmosphere. However, the simulated total
column would not miss this methane plume since it would stay in the
atmospheric column, even if the methane plume is simulated at a wrong level.
Secondly, surface sources induce weaker signatures in the total-column
amounts than in surface concentrations , which could result
in a smaller sensitivity of the inverse system to the total column than to
surface measurements.
Discrepancies in global methane estimates derived by inverse modeling are
usually largely explained by large-scale characteristics of the modeling of
interhemispheric (IH) exchanges. For example, the overestimation of the
north–south gradient in methane mixing ratios in the a priori simulations
of the TM5 model have been assumed to be caused by too slow IH exchanges in
TM5 . Furthermore, in
, LMDz-TD (using a coarser horizontal and vertical
resolutions than the version of LMDz-TD used here) is in the range of CTMs
simulating a too fast IH exchange, which has been shown to induce a positive
(negative) bias in methane emissions in the Northern (Southern) Hemisphere
after inversion.
Annual hemispheric methane fluxes (Tg CH4 year-1) for
the common year of simulation (2010).
In order to investigate the representation of IH exchange in our inversions,
we present in Table the methane estimates in the Northern and
the Southern Hemispheres, as well as the IH methane emission gradient for the common
year (2010) of the different inversions. Whatever the constraints used in our
inversions, the hemispheric differences simulated by LMDz-TD are larger
compared to those simulated by LMDz-SP. Indeed, BG, EXT, and PR inversions
using LMDz-TD derive hemispheric differences of 21.2, 31.5 and
1.0 Tg CH4 higher, respectively, than in inversions using LMDz-SP. These results
confirm the conclusion of , who showed that LMDz-SP
simulates IH exchange slower than LMDz-TD based on an analysis of
SF6 simulations. Indeed, LMDz-SP simulating slower IH exchange
finds, on average, higher (smaller) methane mixing ratios in the Northern
(Southern) Hemisphere than in LMDz-TD. In response, the inverse system using
LMDz-SP derives smaller (higher) methane emissions in the Northern (Southern)
Hemisphere to fit the observed mixing ratio. This leads to a smaller
hemispheric difference in methane emissions compared to what the inverse
system derives when it uses LMDz-TD.
Concerning LMDz-NP, results are slightly different. In surface inversions,
the hemispheric differences simulated by LMDz-NP are also smaller than those
simulated by LMDz-TD, even if the difference is smaller between LMDz-NP and
LMDz-TD than between LMDz-SP and LMDz-TD. However, these results are in
agreement with the study of , which shows that the
thermal plume model implemented in LMDz-NP was responsible for a faster IH
exchange in LMDz-NP than in LMDz-SP. Thus, it is not surprising to simulate a
hemispheric difference of 6.0 Tg CH4 (BG inversions) and
25.6 Tg CH4 (EXT inversions) higher in LMDz-NP than in LMDz-SP.
Moreover, the larger difference in EXT compared to BG inversions can be
explained by the higher number of stations located closer to methane sources,
where the thermal plume model strongly affects the boundary layer mixing
.
However, the higher hemispheric difference simulated by PR-LEI-NP was not
expected from the study of . Indeed, PR-LEI-NP simulates
a hemispheric difference of 262.0 Tg CH4, which is surprisingly
higher than PR-LEI-TD (249.9 Tg CH4). Large methane emissions are
derived in tropical regions for the year 2010. These regions are across the
Equator and experience important vertical mixing during the year (e.g.,
monsoon in India). Therefore, they are sensitive to the parameterization of
this transport. A small but incorrect repartition of emissions between the
Northern and Southern Hemisphere can strongly affect the hemispheric
difference computed here. Moreover, satellite inversions generally derive
stronger methane emissions in the tropics than surface-based inversions
. For example,
found a shift in the emissions from the extra-tropics
to the tropics of 50±25 Tg CH4 year-1. Thus, one can
expect that the hemispheric difference can be changed in satellite-based
inversions because emissions can easily be attributed to the Southern or Northern
Hemisphere. However, surface inversions do not have enough
constraints to derive accurate tropical emissions which are expected to be
high due to strong wetlands and biomass burning methane emissions. The
“missing” amount of methane emissions in tropical regions derived by surface
inversions are generally shifted to the extra-tropics, which lead to less
ambiguous definition of the interhemispheric gradient since emissions are
clearly attributed to one of the two hemispheres. Moreover, we expect that
PR-LEI-NP would simulate a smaller hemispheric difference for a year without
such large emissions in the tropics .
Overall and across the different data sets assimilated, the largest spread in
methane global emission estimations due to parameterization errors reaches
7.5 Tg CH4 year-1, representing 1 % of the total global of
methane emissions. The choice of the deep convection scheme has a significant
impact on the relative distribution of methane emissions between
extra-tropical and tropical regions because deep convection strongly impacts
large-scale atmospheric transport. Versions of LMDz using the deep convection
of , like LMDz-SP and LMDz-NP, produce a smaller
interhemispheric gradient in methane emissions, improving one of the PYVAR
inverse system's flaws identified in and
. Among data sets, the impact of parameterization
uncertainties on methane emission estimations is smaller when using satellite
total-column data compared to surface observations, suggesting that errors
related to the modeling of vertical transport have less impact on
estimations when considering total-column data.
Impact of physical parameterizations on regional methane flux estimates
Estimations of methane fluxes (in Tg CH4 year-1) for
11 regions in three versions of the model (LMDz-TD, LMDz-SP and LMDz-NP) and
based on three different observation constraints (BG, EXT and PR-LEI). Prior
error bars are overplotted on grey bars, which represent the prior
estimations. Posterior error bars are plotted for LMDz-TD inversions only, and
they have been estimated based on the work of . See
Sect. for more information on the analysis of posterior
uncertainties. NAB: North America boreal; NATm: North America temperate;
SATr: South America tropical; SATm: South America temperate; Afr: Africa;
EurB: Eurasia boreal; SEAs: Southeast Asia; Aus: Australia; Eur: Europe;
Chi: China; Ind: India
Figure gives a representation of methane flux
estimations derived by the nine inversions for 11 continental regions in
2010. Estimations using LMDz-TD (LMDz-SP and LMDz-NP) are plotted with
different colors (see figure's legend). The prior estimates for each region
are plotted in grey with the associated prior error bars. Posterior error
bars are plotted for LMDz-TD inversions only, based on the posterior
(residual) errors computed by with the same setup and
same model (see Sect. ). Comparing emission estimates
derived by the different inversions allows us to quantify the impact of
sub-grid-scale parameterizations on regional inverted estimates
(Sect. ) and to assess its significance compared to total
model errors, to residual errors returned by the inversion, and to the choice
of observation data set (Sect. ).
Quantification of the impact of physical parameterizations on regional methane flux estimatesSurface-based inversions (BG and EXT data sets)
In the BG configuration of surface-based inversions (the first three
bars for each region), larger differences are found between inversions using
different deep convection schemes than between inversions using different
parameterizations of boundary layer mixing. In tropical regions, where deep
convection is predominant, like in South America tropical, Southeast Asia or
India, it is expected that BG-SP and BG-NP, which both use the deep
convection scheme of , derive similar estimates, while
BG-TD, which uses the deep convection scheme of , derives
slightly different estimations. For example, BG-SP and BG-NP derive
estimates of 79.0 and 79.2 Tg CH4 year-1, respectively, in
Southeast Asia, compared to 76.5 Tg CH4 year-1 for BG-TD.
However, these changes remain small compared to the residual uncertainties
(plotted for LMDz-TD inversions in Fig. ). For
extra-tropical regions, where deep convection is less directly predominant,
like in North America temperate, Europe or China, the representation of
interhemispheric exchange can have large impacts on regional estimates.
Indeed, BG inversions are mainly constrained by remote stations (see
Sect. ), where simulated concentrations are largely impacted
by the representation of large-scale transport (like interhemispheric
exchange). However, as mentioned in Sect. , the deep
convection scheme of has improved the representation of
interhemispheric exchange in LMDz. Thus, LMDz-SP and LMDz-NP both using the
scheme derive similar estimates in regions like North
American temperate, Europe or China. Flux estimations for boreal regions
(like North America boreal and Eurasia boreal) are also strongly dependent on
the modeling of large-scale atmospheric transport since they are far from
the main sources of methane. Thus LMDz-SP and LMDz-NP logically derive
similar estimates in these two boreal regions.
When the extended data set (EXT inversions) is combined with the thermal plume model
and the scheme, LMDz-NP appears to have large impacts.
Indeed, this scheme plays a key role in the mixing in the boundary layer and
can produce large differences in methane mixing ratio simulated for stations
located close to high methane sources as in the EXT network. Thus, large
impacts are found in China (five stations have been added close to China in the
EXT network), where EXT-NP derives emissions of
74.7 Tg CH4 year-1, significantly higher compared to 68.2 and
67.2 Tg CH4 year-1 for EXT-SP and EXT-TD, respectively.
tropical regions (like in South America tropical) are also affected by the
thermal plume model, even if the reasons are less obvious than in China,
although the thermals play an important role at the base of deep convection
layers . However, there are still very few stations
constraining tropical emissions in the EXT network.
GOSAT-based inversions (PR-LEI data sets)
However, in
PR-LEI inversions, GOSAT data bring strong constraints in tropical
regions, where methane sources are supposed to be large. Thus, it is not surprising to
see large impacts on tropical region estimates in satellite-based inversions
due to the implementation of the thermal plume model (see the rightmost three
bar plots for each region in Fig. ). Indeed, PR-LEI-TD
and PR-LEI-SP derive methane emissions of 68.3 and
67.0 Tg CH4 year-1 in Southeast Asia, while PR-LEI-NP
derives methane emissions of 80.7 Tg CH4 year-1 in the same
region, out of the uncertainty bounds given for PR-LEI-TD inversion. In South
America, PR-LEI-TD and PR-LEI-SP derive larger methane emissions (68.4 and
67.7 Tg CH4 year-1) compared to PR-LEI-NP, which derives
methane emissions of 62.0 Tg CH4 year-1. As a consequence,
satellite-based inversions derive different spatial distribution in methane
emissions between the different tropical regions (compared to surface-based
inversions), although their total methane emissions in the tropics remain
close.
Quantification for all regions
Spreads of regional methane flux (maximum - minimum of CH4
emissions) in BG, EXT and PR-LEI inversions due to changes in physical
parameterizations. The spread is expressed in percent and
in Tg CH4 year-1. The numbers are relative to the common year
of inversions, which is 2010. However, average spreads between 2007 and 2011
are also shown for BG and EXT inversions (inside the parentheses) since
surface-based inversions have been run for several years.
BG EXT PR-LEI %Tg yr-1%Tg yr-1%Tg yr-1North America boreal5.00.79.41.02.20.3(4.5)(0.6)(7.7)(0.9)North America temperate3.71.310.23.87.02.9(3.4)(1.2)(9.9)(3.7)South America tropical5.63.310.46.19.86.4(7.2)(4.0)(9.1)(5.1)South America temperate9.03.115.35.310.13.8(8.7)(3.0)(12.5)(4.3)Africa1.30.83.52.11.81.1(1.2)(0.7)(2.7)(1.6)Eurasia boreal9.41.913.12.67.51.5(9.4)(1.9)(14.2)(2.8)Southeast Asia3.12.42.92.317.212.4(4.0)(3.1)(3.8)(2.9)Australia5.50.29.60.42.60.1(6.6)(0.3)(9.8)(0.4)Europe4.61.918.17.49.64.8(9.0)(3.9)(13.7)(5.8)China10.88.017.012.86.33.4(8.6)(6.2)(10.5)(7.5)India2.20.78.83.112.84.1(4.8)(1.5)(6.4)(2.2)Middle East2.50.610.02.89.42.7(2.7)(0.7)(9.2)(2.4)
More quantitatively, Table summarizes the spread
(difference between the maximum and the minimum of methane emission
estimations) in BG, EXT and PR-LEI inversions due to changes in physical
parameterizations. The spread is expressed in percent and in
Tg CH4 year-1. The numbers are relative to the common year of
inversions, which is 2010. However, average spreads between 2007 and 2011 are
also shown for BG and EXT inversions (inside the parentheses in
Table ) since surface-based inversions have been run for
several years (2006–2012). First, one can notice that the spreads (in
percent) at regional scales caused by changes in sub-grid-scale
parameterizations appear larger compared to what was found at the global
scale (see Sect. ). Indeed, at regional scales, spreads range
from 1 to 11 % (0.2 to 8.0 Tg CH4 year-1), 3 to 18 % (0.4
to 12.8 Tg CH4 year-1) and 2 to 17 % (0.1 to
12.4 Tg CH4 year-1) for BG, EXT and PR-LEI
inversions, respectively. Across the networks, the largest spreads (in
Tg CH4 year-1) are found in China, Southeast Asia, Europe,
South America tropical and South America temperate. Furthermore, spreads in
surface-based inversions are larger in EXT compared to BG configuration of
the surface network, similar to what we found at the global scale. Indeed,
constraints added in EXT inversions are located closer to large methane
mixing ratio gradients, which makes EXT inversions more sensitive to the
modeling of the boundary layer mixing. Yet, skills of the different LMDz
versions (LMDz-TD, LMDz-SP and LMDz-NP) to simulate PBL mixing can be highly
different . Thus, the different configurations of the
inverse system induce larger spreads in EXT compared to BG inversions. On
average, the mean regional spread is 5, 11 and 8 % for BG, EXT
and PR-LEI inversions, respectively. This gives a mean error of 8 % at regional scales
considering the three types of inversions.
Significance of the impact of physical parameterizations at regional scales
A question arising from the previous section is the significance of the
spread due to physical parameterizations compared to others errors: total
transport error, residual error from the inversion and spread in methane
emissions due to the choice of the observing network. Thus, we compare here
the impact of physical parameterization uncertainties on methane inversions
with these three sources of errors.
Physical parameterizations versus total transport error
Similar to results found for the global scale, spreads at regional scales
when using different parameterizations in LMDz are smaller than spreads
between inversions using different atmospheric transport models as in
. Indeed, in , spreads between
inversions using different CTMs range from 23 % for Europe to 48 % for
South America, with an average of 33 %. Here, we found that the mean error
in methane estimates due to parameterization errors is 8 % at regional
scales (Sect. ). Consequently, if one assumes that the
spread given in is representative of the total error
due to the modeling of transport, errors related to physical
parameterizations explain, on average 24 % (corresponding to the ratio of
8/33) of the total transport model errors, but it can reach more than
50 % in some specific regions. Therefore, the different parameterizations
used within LMDz span relatively more of the transport error at regional
scales than at the global scale.
Physical parameterizations versus residual errors
One can compare the spreads in methane emission estimates due to
parameterization uncertainties in different regions in
Fig. to the amplitudes of posterior (residual)
uncertainties estimated by . A spread in methane inversion
estimates that is larger than the amplitude of the posterior uncertainties is
found for three regions only: South America tropical, Europe and China.
Consequently, one can interpret that the impact of parameterization
uncertainties in these three regions is meaningful. An improvement of the
representation of sub-grid processes could significantly increase the
confidence in methane emission estimates in these regions. In the other
regions, the spreads in inverted estimates are found to be similar to (Eurasia boreal,
North America temperate, South America temperate) or smaller than (Africa and
India) the amplitudes of posterior uncertainties, indicating that
changes due to the use of different parameterizations are less meaningful for
these regions.
Physical parameterizations versus the choice of observation
networks
Spreads (maximum - minimum) in regional methane estimates (in Tg CH4 year-1) due to the choice
of the observation data set (blue error bars) and due to the choice of the model version (green error bars). The total spreads in the nine inversions run in this work are plotted in red.
As already mentioned regarding the global scale, the spread in methane inverted
regional estimates is in general smaller when one considers inversions
constrained by the same observation data set (Fig. ).
Comparing the spread in inversions using a same observation data set but
different versions of LMDz (green bar plots in Fig. )
with inversions using the same version of LMDz but different observation
data sets (blue bar plots) shows that the choice of observation data set has a
larger impact on inverted regional estimates. Spreads in inversions based on
different versions of LMDz are similar (or slightly smaller) to spreads in
inversions based on different observation data sets in South America
temperate, Africa, Eurasia boreal, India and Australia. But, in all the
others regions, the spreads in inversions based on different observation
data sets are much larger. This is especially obvious in Europe, China and South
America tropical. Indeed, information provided by the different data sets is
different but can also be differently interpreted by the different versions
of LMDz. A good example of this assessment is found in tropical regions for
inversions constrained by the PR-LEI data set. The inverse system based on
LMDz-NP derives a very different spatial distribution: larger emissions in
Southeast Asia but lower emissions in South America tropical compared to
inversions based on LMDz-TD and LMDz-SP.
Overall, we find that, if the choice of the observation data set used to
constrain the inversion generally remains a larger source of uncertainty than
the change in physical parameterizations for one given data set, both can
generate significantly different spatial distributions of methane emissions,
especially within the tropics and for Europe and China. For example, the
partition of methane emissions within the tropics can shift from Southeast
Asia to South America and methane emission estimates derived by BG and PR-LEI
inversions differ by 15 and 22 Tg CH4 year-1 in Europe and
China, respectively (Fig. ).
Conclusions
This study presents the sensitivity of the recent methane budget estimated by
the PYVAR inversion system to different LMDz sub-grid-scale physical
parameterizations for vertical transport, combined with different observation
data sets. Three versions of LMDz (LMDz-TD, LMDz-SP and LMDz-NP) have been
used within the PYVAR system to simulate atmospheric transport of methane
emitted at the surface. Three methane observation data sets (two surface
data sets and one GOSAT data set) have been assimilated to constrain
these different atmospheric inversions. Finally, the comparison between these
nine inversions quantifies the impact of LMDz sub-grid-scale parameterizations
on methane inverted estimates.
Satellite-based inversions generate atmospheric mixing ratios, which are
inconsistent with surface observations when no bias correction is applied.
These inconsistencies are not related to the physical parameterizations of
the vertical transport for LMDz. The relative agreement between methane
concentrations simulated by the former version of LMDz and GOSAT data masks a
poor representation of the methane gradient at the tropopause in the former
version of the LMDz model. However, our results based on different
new versions of LMDz, properly reproducing the vertical gradient of methane
in the upper troposphere/lower stratosphere, suggest a large bias in the
GOSAT data (∼ 40 ppb), with a small latitudinal dependency.
This bias is corrected to analyze and compare the different inversions
performed.
At the global scale, we find that the spread due to physical parameterization
uncertainties is 0.5, 1.4 and 0.4 % of global methane emissions. Moreover,
the analysis of the north–south gradient in inferred emissions confirms that
the deep convection scheme improves the representation of
IH exchange, as already mentioned in .
At regional scales, the spreads due to physical parameterization are larger
than at the global scale (5–11 %) but remain generally lower than the
spread due to the choice of observation data set and always lower (24 % on
average) than the total uncertainty due to the modeling of atmospheric
transport as estimated by . Comparison of these spreads to
the residual regional inversion uncertainties given in
indicates that they are significant for a few key regions of the methane
cycle: South America, China and Europe. Indeed, changing the
parameterizations can lead to significantly different spatial partitions of
methane emissions for (or between) these regions. One important additional
result is that the thermal plume model combined with the vertical diffusion
scheme of implemented in LMDz-NP largely impact regional
estimations, especially when considering atmospheric constraints located
close to high methane sources (like in tropical regions for satellite-based
inversions).
After the quantification of transport model errors in global and regional
methane flux estimates based on a TransCom intercomparison
and the evaluation of new parameterizations in LMDz to
simulate trace gas concentrations , this paper goes one
step further in the understanding and the quantification of causes and
impacts of model errors in methane inversions. In these different studies,
precisions have been given on the degree of confidence in the global and
regional methane estimations using inverse modeling relative to transport
model errors. At the global scale, the impact of transport model errors
(5 % of global methane emissions) and physical parameterizations errors
(0.8 %) is acceptable. However, the picture is different at regional
scales, with a possible significant impact of transport errors on the
attribution of regional methane emission estimated by inverse modeling. A
striking finding of our work is the possibility to change the partition of
methane tropical emissions depending on the combination of observation
data set and physical parameterizations used, even after correcting probable
biases in satellite data sets. Our work provides elements to understand why
the different inversions recently published lack regional consistency with
each other in their attribution of the renewed increase in atmospheric methane
since 2007. Our results indicate a need for more work to be carried out to improve transport models in order to reduce transport errors. A way to
achieve this is, for one, to strengthen collaborations between experts in atmospheric
dynamics and experts in tracer transport and, for another, to develop
measurement campaigns and use specific tracers in order to better evaluate
transport models on the other hand. Finally, inconsistencies between
surface-based and satellite-based inversions need to receive more attention
in the future.
Acknowledgements
This work is supported by the DGA (Direction Générale de l'Armement) and
CEA (Centre à l'Energie Atomique et aux Energies Alternatives).
We would like to thank Vanessa Sherlock, Yi Yin, Frédéric Hourdin and
Catherine Rio for fruitful discussions. We would like also thank Sébastien
Leonard for his IT support.
The authors wish to thank R. Parker and H. Boesch (EOS, University of
Leicester) for providing the GOSAT Proxy XCH4 data set. This product was
developed partly with funding from the ESA GHG-CCI project and the UK National
Centre for Earth Observation.
We acknowledge the contributors to the World Data Centre for Greenhouse Gases
for providing their data of methane and methyl chloroform atmospheric mole
fractions. The authors thank in particular A. J. Gomez-Pelaez (AEMET), R.
Prinn (AGAGE), R. Weiss (AGAGE), P. Krummel (CSIRO), D. Worthy (EC), S.
Piacentino (ENEA), Y. Fukuyama (JMA), Y. Tohjima (NIES), E. Dlugokencky
(NOAA) and K. Uhse (UBA). Moreover, the authors wish to thank the respective
funding organizations/institutions for their long-term support of these
measurement programs.Edited by: P. Jöckel
ReferencesBergamaschi, P., Krol, M., Dentener, F., Vermeulen, A., Meinhardt, F., Graul,
R., Ramonet, M., Peters, W., and Dlugokencky, E. J.: Inverse modelling of
national and European CH4 emissions using the atmospheric zoom model TM5,
Atmos. Chem. Phys., 5, 2431–2460, 10.5194/acp-5-2431-2005, 2005.Bergamaschi, P., Frankenberg, C., Meirink, J. F., Krol, M., Dentener, F.,
Wagner, T.,
Platt, U., Kaplan, J. O., Körner, S., Heimann, M., Dlugokencky, E. J., and Goede, A.:
Satellite chartography of atmospheric methane from SCIAMACHY on
board ENVISAT: 2. Evaluation based on inverse model simulations,
J. Geophys. Res.-Atmos., 122, D02304, 10.1029/2006JD007268, 2007.Bergamaschi, P., Frankenberg, C., Meirink, J. F., Krol, M.,
Villani, M. G., Houweling, S., Dentener, F., Dlugokencky, E. J.,
Miller, J. B., Gatti, L. V., Engel, A., and Levin, I.:
Inverse modeling of global and regional CH4 emissions using
SCIAMACHY satellite retrievals,
J. Geophys. Res.-Atmos., 114, D22301, 10.1029/2009JD012287, 2009.Bergamaschi, P., Krol, M., Meirink, J. F., Dentener, F., Segers, A., van
Aardenne, J., Monni, S., Vermeulen, A. T., Schmidt, M., Ramonet, M., Yver,
C., Meinhardt, F., Nisbet, E. G., Fisher, R. E., O'Doherty, S., and
Dlugokencky, E. J.: Inverse modeling of European CH4 emissions
2001–2006,
J. Geophys. Res., 115, D22309, 10.1029/2010JD014180, 2010.Bergamaschi, P., Houweling, S., Segers, A., Krol, M., Frankenberg, C.,
Scheepmaker, R. A., Dlugokencky, E. J., Wofsy, S. C., Kort, E. A., Sweeney,
C., Schuck, T., Brenninkmeijer, C., Chen, H., Beck, V., and Gerbig, C.:
Atmospheric CH4 in the first decade of the 21st century: Inverse
modeling analysis using SCIAMACHY satellite retrievals and NOAA surface
measurements,
J. Geophys. Res.-Atmos., 118
7350–7369, 10.1002/jgrd.50480, 2013.Bousquet, P., Hauglustaine, D. A., Peylin, P., Carouge, C., and Ciais, P.:
Two decades of OH variability as inferred by an inversion of atmospheric
transport and chemistry of methyl chloroform, Atmos. Chem. Phys., 5,
2635–2656, 10.5194/acp-5-2635-2005, 2005.Bousquet, P., Ciais, P., Miller, J. B., Dlugokencky, E. J., Hauglustaine, D.
A., Prigent, C.,
Van der Werf, G. R., Peylin, P., Brunke, E.-G., Carouge, C., Langenfelds, R. L.,
Lathiere, J., Papa, F., Ramonet, M., Schmidt, M., Steele, L. P., Tyler, S. C., and White,
J.:
Contribution of anthropogenic and natural sources to atmospheric
methane variability,
Nature, 443, 439–443, 10.1038/nature05132, 2006.Bousquet, P., Ringeval, B., Pison, I., Dlugokencky, E. J., Brunke, E.-G.,
Carouge, C., Chevallier, F., Fortems-Cheiney, A., Frankenberg, C.,
Hauglustaine, D. A., Krummel, P. B., Langenfelds, R. L., Ramonet, M.,
Schmidt, M., Steele, L. P., Szopa, S., Yver, C., Viovy, N., and Ciais, P.:
Source attribution of the changes in atmospheric methane for 2006–2008,
Atmos. Chem. Phys., 11, 3689–3700, 10.5194/acp-11-3689-2011, 2011.Chevallier, F.: On the parallelization of atmospheric inversions of CO2
surface fluxes within a variational framework, Geosci. Model Dev., 6,
783–790, 10.5194/gmd-6-783-2013, 2013.Chevallier, F., Fisher, M., Peylin, P., Serrar, S., Bousquet, P., Bréon,
F.-M.,
Chédin, A., and Ciais, P.:
Inferring CO2 sources and sinks from satellite observations: Method
and application to TOVS data,
J. Geophys. Res.-Atmos., 110, D24309, 10.1029/2005JD006390, 2005.Chevallier, F., Bréon, F.-M., and Rayner, P. J.: Contribution of the
Orbiting Carbon Observatory to the estimation of
CO2 sources and sinks: Theoretical study in a variational data assimilation
framework,
J. Geophys. Res.-Atmos.,
112, D09307, 10.1029/2006JD007375, 2007.Chevallier, F., Feng, L., Bösch, H., Palmer, P. I., and Rayner, P. J.:
On the impact of transport model errors for the estimation of CO2 surface
fluxes from GOSAT observations,
Geophys. Res. Lett., 37, L21803, 10.1029/2010GL044652, 2010.Chevallier, F., Palmer, P. I., Feng, L., Boesch,
H., O'Dell, C. W., and Bousquet, P.:
Toward robust and consistent regional CO2 flux estimates from in-situ and
spaceborne measurements of atmospheric CO2,
Geophys. Res. Lett., 41,
1065–1070, 10.1002/2013GL058772, 2014.Cressot, C., Chevallier, F., Bousquet, P., Crevoisier, C., Dlugokencky, E.
J., Fortems-Cheiney, A., Frankenberg, C., Parker, R., Pison, I., Scheepmaker,
R. A., Montzka, S. A., Krummel, P. B., Steele, L. P., and Langenfelds, R. L.:
On the consistency between global and regional methane emissions inferred
from SCIAMACHY, TANSO-FTS, IASI and surface measurements, Atmos. Chem. Phys.,
14, 577–592, 10.5194/acp-14-577-2014, 2014.Dee, D. P., Uppala, S. M., Simmons, A. J., Berrisford, P., Poli, P.,
Kobayashi, S., Andrae, U., Balmaseda, M. A., Balsamo, G., Bauer, P.,
Bechtold, P., Beljaars, A. C. M., van de Berg, L., Bidlot, J., Bormann, N.,
Delsol, C., Dragani, R., Fuentes, M., Geer, A. J., Haimberger, L., Healy, S. B.,
Hersbach, H., Holm, E. V., Isaksen, L., Kallberg, P., Kohler, M., Matricardi,
M., McNally, A. P., Monge-Sanz, B. M., Morcrette, J.-J., Park, B.-K., Peubey,
C., de Rosnay, P., Tavolato, C., Thépaut, J.-N., and Vitart, F.: The
ERA-Interim reanalysis: configuration and performance of the data
assimilation system,
Q. J. Roy. Meteorol. Soc., 137, 553–597, 10.1002/qj.828, 2011.Emanuel K. A. L.: A Scheme for Representing Cumulus Convection in
Large-Scale Models,
J. Atmos. Sci., 48, 2313–2329, 10.1175/1520-0469(1991)048<2313:ASFRCC>2.0.CO;2, 1991.
Enting, I. G.: Inverse problems in atmospheric constituent transport, CUP,
Cambridge, UK, 2002.Fortems-Cheiney A., Chevallier, F., Pison, I., Bousquet, P., Szopa, S.,
Deeter, M. N., and Clerbaux, C.: Ten years of CO emissions as
seen from Measurements of Pollution in the Troposphere (MOPITT),
J. Geophys. Res.-Atmos., 116, D05304, 10.1029/2010JD014416, 2011.Frankenberg, C., Meirink, J. F., van Weele, M., Platt, U., and Wagner, T.:
Assessing Methane Emissions from Global Space-Borne Observations,
Science, 308, 1010–1014, 10.1126/science.1106644, 2005.Frankenberg, C., Bergamaschi, P., Butz, A., Houweling, S.,
Meirink, J. F., Notholt, J., Petersen, A. K., Schrijver, H.,
Warneke, T., and Aben, I.:
Tropical methane emissions: A revised view from SCIAMACHY onboard ENVISAT,
Geophys. Res. Lett., 35, L15811, 10.1029/2008GL034300, 2008.Geels, C., Gloor, M., Ciais, P., Bousquet, P., Peylin, P., Vermeulen, A. T.,
Dargaville, R., Aalto, T., Brandt, J., Christensen, J. H., Frohn, L. M.,
Haszpra, L., Karstens, U., Rödenbeck, C., Ramonet, M., Carboni, G., and
Santaguida, R.: Comparing atmospheric transport models for future regional
inversions over Europe – Part 1: mapping the atmospheric CO2 signals,
Atmos. Chem. Phys., 7, 3461–3479, 10.5194/acp-7-3461-2007, 2007.Gilbert J.-C. and Lemaréchal, C.: Some numerical experiments with
variable-storage quasi-Newton
algorithms,
Math. Program., 45,
407–435, 10.1007/BF01589113, 1989.Grooß, J.-U. and Russell III, James M.: Technical note: A stratospheric
climatology for O3, H2O, CH4, NOx, HCl and HF derived from
HALOE measurements, Atmos. Chem. Phys., 5, 2797–2807,
10.5194/acp-5-2797-2005, 2005.Hourdin, F. and Armengaud, A.: The Use of Finite-Volume Methods for
Atmospheric Advection of Trace
Species, Part I: Test of Various Formulations in a General Circulation
Model,
Mon. Weather Rev., 127, 822–837, 10.1175/1520-0493(1999)127<0822:TUOFVM>2.0.CO;2, 1999.Hourdin, F., Couvreux, F., and Menut, L.: Parameterization of the Dry
Convective Boundary Layer Based on a
Mass Flux Representation of Thermals,
J. Atmos. Sci., 59, 1105–1123, 10.1175/1520-0469(2002)059<1105:POTDCB>2.0.CO;2, 2002.Hourdin, F., Musat, I., Bony, S., Braconnot, P.,
Codron, F., Dufresne, J.-L., Fairhead, L., Filiberti, M.-A., Friedlingstein, P., Grandpeix, J.-Y., Krinner, G.,
LeVan, P., Li, Z.-X., and Lott, F.:
The LMDZ4 general circulation model: climate performance and
sensitivity to parametrized physics with emphasis on tropical convection,
Clim. Dynam., 27, 787–813, 10.1007/s00382-006-0158-0, 2006.Hourdin, F., Foujols, M.-A., Codron, F., Guemas, V.,
Dufresne, J.-L., Bony, S., Denvil, S., Guez, L.,
Lott, F., Ghattas, J., Braconnot, P., Marti, O.,
Meurdesoif, Y., and Bopp, L.:
Impact of the LMDZ atmospheric grid configuration on the climate and
sensitivity of the IPSL-CM5A coupled model,
Clim. Dynam., 40, 2167–2192, 10.1007/s00382-012-1411-3, 2012a.Hourdin, F., Grandpeix, J.-Y., Rio, C., Bony, S., Jam, A., Cheruy, F.,
Rochetin, N., Fairhead,
L., Idelkadi, A., Musat, I., Dufresne, J.-L., Lahellec, A., Lefebvre, M.-P., and Roehrig,
R.:
LMDZ5B: the atmospheric component of the IPSL climate model with
revisited parameterizations for clouds and convection,
Clim. Dynam., 40, 2193–2222, 10.1007/s00382-012-1343-y, 2012b.Houweling, S., Kaminski, T., Dentener, F., Lelieveld, J., and Heimann, M.:
Inverse modeling of methane sources and sinks using the adjoint of a global
transport model,
J. Geophys. Res., 104, 26137, 10.1029/1999JD900428, 1999.Houweling, S., Dentener, F., and Lelieveld, J.: Simulation of preindustrial
atmospheric methane to constrain the global source strength of natural
wetlands,
J. Geophys. Res.-Atmos., 105, 17243–17255, 10.1029/2000JD900193, 2000.Houweling, S., Röckmann, T., Aben, I., Keppler, F., Krol, M., Meirink,
J. F.,
Dlugokencky, E. J., and Frankenberg, C.:
Atmospheric constraints on global emissions of methane from plants,
Geophys. Res. Lett., 33, L15821, 10.1029/2006GL026162, 2006.Houweling, S., Aben, I., Breon, F.-M., Chevallier, F., Deutscher, N.,
Engelen, R., Gerbig, C., Griffith, D., Hungershoefer, K., Macatangay, R.,
Marshall, J., Notholt, J., Peters, W., and Serrar, S.: The importance of
transport model uncertainties for the estimation of CO2 sources and sinks
using satellite measurements, Atmos. Chem. Phys., 10, 9981–9992,
10.5194/acp-10-9981-2010, 2010.Houweling, S., Badawy, B., Baker, D. F., Basu, S., Belikov, D.,
Bergamaschi, P., Bousquet, P., Broquet, G., Butler, T., Canadell, J. P., Chen, J., Chevallier, F., Ciais, P.,
Collatz, G. J., Denning, S., Engelen, R., Enting, I. G., Fischer, M. L., Fraser, A.,
Gerbig, C., Gloor, M., Jacobson, A. R., Jones, D. B. A., Heimann, M., Khalil, A.,
Kaminski, T., Kasibhatia, P. S., Krakauer, N. Y., Krol, M., Maki, T., Maksyutov, S.,
Manning, A., Meesters, A., Miller, J. B., Palmer, P. I., Patra, P., Peters, W., Peylin,
P., Poussi, Z., Prather, M., Randerson, J. T., Rckmann, T., Rdenbeck, C.,
Sarmiento, J. L., Schimel, D. S., Scholze, M., Schuh, A., Suntharalingam, P., Takahashi, T., Turn-
bull, J., Yurganov, L., and Vermeulen, A.: Iconic CO2 time series at risk,
Science, 337, 1038, 10.1126/science.337.6098.1038-b, 2012.Houweling, S., Krol, M., Bergamaschi, P., Frankenberg, C., Dlugokencky, E.
J., Morino, I., Notholt, J., Sherlock, V., Wunch, D., Beck, V., Gerbig, C.,
Chen, H., Kort, E. A., Röckmann, T., and Aben, I.: A multi-year methane
inversion using SCIAMACHY, accounting for systematic errors using TCCON
measurements, Atmos. Chem. Phys., 14, 3991–4012,
10.5194/acp-14-3991-2014, 2014.Kirschke, S., Bousquet, P., Ciais, P., Saunois, M., Canadell, J. G.,
Dlugokencky, E. J., Bergamaschi, P., Bergmann, D., Blake, D. R., Bruhwiler,
L., Cameron-Smith, P., Castaldi, S., Cheval- lier, F., Feng, L., Fraser, A., Heimann, M., Hodson, E. L.,
Houweling, S., Josse, B., Fraser, P. J., Krummel, P. B., Lamarque, J.-F., Langenfelds,
R. L., Le Qur, C., Naik, V., ODoherty, S., Palmer, P. I., Pison, I., Plummer, D., Poul-
ter, B., Prinn, R. G., Rigby, M., Ringeval, B., Santini, M., Schmidt, M., Shindell, D.
T., Simpson, I. J., Spahni, R., Steele, L. P., Strode, S. A., Sudo, K., Szopa, S., van
der Werf, G. R., Voulgarakis, A., van Weele, M., Weiss, R. F., Williams, J. E., and
Zeng, G.: Three decades of global methane sources and sinks,
Nat. Geosci., 6, 813–823, 10.1038/ngeo1955, 2013.Locatelli, R., Bousquet, P., Chevallier, F., Fortems-Cheney, A., Szopa, S.,
Saunois, M., Agusti-Panareda, A., Bergmann, D., Bian, H., Cameron-Smith, P.,
Chipperfield, M. P., Gloor, E., Houweling, S., Kawa, S. R., Krol, M., Patra,
P. K., Prinn, R. G., Rigby, M., Saito, R., and Wilson, C.: Impact of
transport model errors on the global and regional methane emissions estimated
by inverse modelling, Atmos. Chem. Phys., 13, 9917–9937,
10.5194/acp-13-9917-2013, 2013.Locatelli, R., Bousquet, P., Hourdin, F., Saunois, M., Cozic, A., Couvreux,
F., Grandpeix, J.-Y., Lefebvre, M.-P., Rio, C., Bergamaschi, P., Chambers, S.
D., Karstens, U., Kazan, V., van der Laan, S., Meijer, H. A. J., Moncrieff,
J., Ramonet, M., Scheeren, H. A., Schlosser, C., Schmidt, M., Vermeulen, A.,
and Williams, A. G.: Atmospheric transport and chemistry of trace gases in
LMDz5B: evaluation and implications for inverse modelling, Geosci. Model
Dev., 8, 129–150, 10.5194/gmd-8-129-2015, 2015.Louis, J. F.: A parametric model of vertical eddy fluxes in the atmosphere,
Bound.-Lay. Meteorol., 7, 187–202, 10.1007/BF00117978, 1979.Monteil, G., Houweling, S., Dlugockenky, E. J., Maenhout, G., Vaughn, B. H.,
White, J. W. C., and Rockmann, T.: Interpreting methane variations in the
past two decades using measurements of CH4 mixing ratio and isotopic
composition, Atmos. Chem. Phys., 11, 9141–9153,
10.5194/acp-11-9141-2011, 2011.Monteil, G., Houweling, S., Butz, A., Guerlet, S., Schepers, D., Hasekamp,
O., Frankenberg, C., Scheepmaker, R., Aben, I., and Röckmann, T.:
Comparison of CH4 inversions based on 15 months of GOSAT and SCIAMACHY
observations,
J. Geophys. Res.-Atmos., 118, 807–811, 10.1002/2013JD019760, 2013.Monteil, G., et al.: Identifying and quantifying biases in inverse modelling
of GOSAT XCH4 retrievals, Atmos. Chem. Phys., in preparation, 2015.Naik, V., Voulgarakis, A., Fiore, A. M., Horowitz, L. W., Lamarque, J.-F.,
Lin, M., Prather, M. J., Young, P. J., Bergmann, D., Cameron-Smith, P. J.,
Cionni, I., Collins, W. J., Dalsøren, S. B., Doherty, R., Eyring, V.,
Faluvegi, G., Folberth, G. A., Josse, B., Lee, Y. H., MacKenzie, I. A.,
Nagashima, T., van Noije, T. P. C., Plummer, D. A., Righi, M., Rumbold, S.
T., Skeie, R., Shindell, D. T., Stevenson, D. S., Strode, S., Sudo, K.,
Szopa, S., and Zeng, G.: Preindustrial to present-day changes in tropospheric
hydroxyl radical and methane lifetime from the Atmospheric Chemistry and
Climate Model Intercomparison Project (ACCMIP), Atmos. Chem. Phys., 13,
5277–5298, 10.5194/acp-13-5277-2013, 2013.Olivier, J. G. J. and Janssens-Maenhout, G.: CO2 Emissions from Fuel
Combustion,
IEA CO2 report 2012, Greenhouse-Gas Emissions, 2012.Parker, R., Boesch, H., Cogan, A., Fraser, A., Feng, L.,
Palmer, P. I., Messerschmidt, J., Deutscher, N., Griffith, D. W. T., Notholt, J., Wennberg, P. O., and
Wunch, D.:
Methane observations from the Greenhouse Gases Observing SATellite:
Comparison to ground-based TCCON data and model calculations,
Geophys. Res. Lett., 38, L15807, 10.1029/2011GL047871, 2011.Patra, P. K., Takigawa, M., Ishijima, K., Choi, B.-C., Cunnold, D.,
Dlugokencky, E. J., Fraser, P., Gomez-Pelaez, A. J., Goo, T.-Y., Kim, J.-S.,
Krummel, P., Langenfelds, R., Meinhardt, F., Mukai, H., O'Doherty, S., Prinn, R. G., Simmonds, P., Steele, P., Tohjima, Y., Tsuboi, K., Uhse, K.,
Weiss, R., Worthy, D., and Nakazawa, T.:
Growth Rate, Seasonal, Synoptic, Diurnal Variations and Budget of Methane in
the Lower Atmosphere,
J. Meteorol. Soc. Jpn., 87, 635–663, 10.2151/jmsj.87.635, 2009.Patra, P. K., Houweling, S., Krol, M., Bousquet, P., Belikov, D., Bergmann,
D., Bian, H., Cameron-Smith, P., Chipperfield, M. P., Corbin, K.,
Fortems-Cheiney, A., Fraser, A., Gloor, E., Hess, P., Ito, A., Kawa, S. R.,
Law, R. M., Loh, Z., Maksyutov, S., Meng, L., Palmer, P. I., Prinn, R. G.,
Rigby, M., Saito, R., and Wilson, C.: TransCom model simulations of CH4
and related species: linking transport, surface flux and chemical loss with
CH4 variability in the troposphere and lower stratosphere, Atmos. Chem.
Phys., 11, 12813–12837, 10.5194/acp-11-12813-2011, 2011.Patra, P. K., Krol, M. C., Montzka, S. A., Arnold, T., Atlas, E. L., Lintner,
B. R.,
Stephens, B. B., Xiang, B., Elkins, J. W., Fraser, P. J., Ghosh, A., Hintsa, E. J.,
Hurst, D. F., Ishijima, K., Krummel, P. B., Miller, B. R., Miyazaki, K., Moore, F. L.,
Muhle, J., O'Doherty, S., Prinn, R. G., Steele, L. P., Takigawa, M., Wang, H. J., Weiss, R. F.,
Wofsy, S. C., and Young, D.:
Observational evidence for interhemispheric hydroxyl-radical
parity,
Nature, 513, 219–223, 10.1038/nature13721, 2014.Pison, I., Bousquet, P., Chevallier, F., Szopa, S., and Hauglustaine, D.:
Multi-species inversion of CH4, CO and H2 emissions from surface
measurements, Atmos. Chem. Phys., 9, 5281–5297, 10.5194/acp-9-5281-2009,
2009.Pison, I., Ringeval, B., Bousquet, P., Prigent, C., and Papa, F.: Stable
atmospheric methane in the 2000s: key-role of emissions from natural
wetlands, Atmos. Chem. Phys., 13, 11609–11623,
10.5194/acp-13-11609-2013, 2013.Randerson, J. T., van der Werf, G. R., Giglio, L., Collatz, G. J., and
Kasibhatla, P. S.: Global Fire Emissions Database, Version 3 (GFEDv3.1),
available at: http://daac.ornl.gov/, from Oak Ridge National Laboratory Distributed Active
Archive Center, Oak Ridge, Tennessee, USA, 10.3334/ORNLDAAC/1191, 2013.Rayner, P. and O'Brien, D.: The utility of remotely sensed CO2
concentration data in surface source inversions,
Geophys. Res. Lett., 28, 175–178, 10.1029/2000GL011912, 2001.Stephens, B. B., Gurney, K. G., Tans, P. P., Sweeney, C., Peters, W., Bruhwiler, L.,
Ciais, P., Ramonet, M., Bousquet, P., Nakazawa, T., Aoki, S., Machida, T., Inoue,
G., Vinnichenko, N., LLoyd, J., Jordan, A., Heimann, M., Shibistova, O., Langenfelds,
R. L., Steele, L. P., Francey, R. J., and Denning, A. S.: Weak northern and
strong tropical land carbon uptake from vertical profiles of atmospheric CO2,
Science, 316, 1732–17325, 10.1126/science.1137004, 2007.Stephens, B. B., Wofsy, S. C., Keeling, R. F., Tans, P. P., and Potosnak, M.
J.: The CO2 Budget and Rectification Airborne Study: Strategies
for Measuring Rectifiers and Regional Fluxes, in:
Inverse Methods in Global Biogeochemical Cycles, edited by: Kasibhatla, P., Heimann, M., Rayner, P., Mahowald, N., Prinn, R. G., and Hartley, D. E.,
American Geophysical Union, Washington DC, 10.1029/GM114p0311, 2013.
Szopa, S., Balkanski, Y., Schulz, M., Bekki, S., Cugnet, D., Fortems-Cheney,
A., Turquety, S., Cozic, A., Dandreis, C., Hauglustaine, D., Idelkadi, A.,
Lathière, J., Lefevre, F., Marchand, M., Vuolo, R., Yan, N., and Dufresne,
J.-L.: Aerosol and ozone changes as forcing for climate evolution between
1850 and 2100,
Clim. Dynam., 40, 2223–2250, 10.1007/s00382-012-1408, 2013.Tiedtke, M.: A Comprehensive Mass Flux Scheme for Cumulus Parameterization
in
Large-Scale Models,
Mon. Weather Rev., 117, 1779–1800, 10.1175/1520-0493(1989)117<1779:ACMFSF>2.0.CO;2, 1989.Yamada, T.: Simulations of Nocturnal Drainage Flows by a q2l Turbulence
Closure
Model,
J. Atmos. Sci., 40,
91–106, 10.1175/1520-0469(1983)040<0091:SONDFB>2.0.CO;2, 1983.Yver, C. E., Pison, I. C., Fortems-Cheiney, A., Schmidt, M., Chevallier, F.,
Ramonet, M., Jordan, A., Søvde, O. A., Engel, A., Fisher, R. E., Lowry,
D., Nisbet, E. G., Levin, I., Hammer, S., Necki, J., Bartyzel, J., Reimann,
S., Vollmer, M. K., Steinbacher, M., Aalto, T., Maione, M., Arduini, J.,
O'Doherty, S., Grant, A., Sturges, W. T., Forster, G. L., Lunder, C. R.,
Privalov, V., Paramonova, N., Werner, A., and Bousquet, P.: A new estimation
of the recent tropospheric molecular hydrogen budget using atmospheric
observations and variational inversion, Atmos. Chem. Phys., 11, 3375–3392,
10.5194/acp-11-3375-2011, 2011.