Introduction
The global stratospheric meridional circulation, also called the Brewer–Dobson
circulation (BDC), was recognized as a major component of the climate
system, which affects radiative forcing
and atmospheric circulation .
The increase in greenhouse gases, in particular the carbon dioxide (CO2)
concentration, affects the atmospheric temperature and wave propagation, which
increases the tropical upwelling mass flux
and therefore changes the BDC. Within the context of climate change, the stratospheric
circulation variability can also be diagnosed using the trace gas CO2.
CO2 is a useful tracer of atmospheric dynamics and transport because
of its long lifetime in the troposphere and stratosphere, where it has essentially
no sources or sinks since it is basically chemically inert in the free troposphere. The
only stratospheric source of CO2 is a small contribution from methane
oxidation that can reach up to 1 ppmv (i.e. parts per million per
volume) . CO2 is regularly exchanged between
four reservoirs: the biosphere (photosynthesis and respiration), the lithosphere
(soil and fossil pool), the hydrosphere (surface and deep ocean), and the atmosphere, with
a much longer residence time in the oceans and soil than in the atmosphere. These
exchanges are described as the carbon cycle. Anthropogenic emissions due primarily
to deforestation and biomass and fossil fuel burning have
systematically increased the mean CO2 and modified its seasonal cycle during
these last 2 decades .
With the influences of steady growth and seasonal variation, CO2 concentrations
in the atmosphere contain both monotonically increasing and periodic signals
that represent stringent tests of stratospheric transport and stratosphere–troposphere
exchange (STE) in models .
Despite its potential to increase global change by cooling the stratosphere and
warming the troposphere via the greenhouse effect, information on
stratospheric CO2 and its variability was sparse until recently.
In recent years, in situ aircraft and balloon campaigns were implemented in order to
measure a number of chemical tracers, including CO2. The in situ campaigns included
SPURT aircraft measurements in the upper troposphere and lower stratosphere (UTLS)
, CONTRAIL , and CARIBIC
. Although sporadic in time and space coverage,
these in situ measurements were used to analyse the BDC changes
, to validate chemistry-transport
models (CTMs) , and to diagnose
STE .
The stratospheric overworld circulation changes that affect the extratropical
UTLS were recently assessed by from
balloon-based measurements of SF6 (sulfur hexafluoride) and CO2. The stratospheric mean
age of the air, which is defined as the residence time of air parcels in the
stratosphere , was calculated
by from the in situ balloon measurements
of trace gases, with an idealized model to identify the natural variability
in the BDC and its significant linear trends. This study demonstrated the importance
of reconstructed in situ measurements to validate the stratospheric
representation in global CTMs and chemistry-climate models (CCMs).
In addition to the very localized in situ observations, which have high
spatial resolution, a large spatial and temporal coverage of CO2 is
obtained from space instruments such as the vertical nadir sounders TOVS
, AIRS , SCIAMACHY
, IASI , GOSAT , and recently OCO-2 . These spaceborne
instruments essentially measure total column CO2, weighted more by
the lower and mid-troposphere; hence, they provide limited information on the
upper troposphere and the stratosphere.
obtained 5 years of monthly mean CO2 vertical profiles by analysing
the ACE-FTS data . The ACE-retrieved CO2 shows
qualitatively good agreement with in situ observations for the 2004–2008
time period at the 50–60∘ N latitude bins. However, the retrieval
sensitivity is limited and averages need to be performed on a large number of
profiles.
Because of the limited observations, CTMs and Lagrangian transport models are a
complementary and useful framework for widely diagnosing the BDC and
representing
the global transport and distribution of long-lived species, such as CO2.
Previous studies using the two-dimensional CTM Caltech/JPL and
the three-dimensional CTM TM5 (Transport Model 5) were unable
to accurately represent the BDC. investigated the UTLS exchanges
in a three-dimensional transport model using the observed CO2 and SF6
distributions and concluded that major disagreements between the model and observations
occur in winter, where a too-strong BDC leads to some overestimates of the CO2,
and in boreal summer, where the vertical transport is too slow in the upper troposphere.
During autumn the models showed an unrealistic persistent reverse gradient
in the lower stratosphere, and during spring the transport processes through the
tropical tropopause were overestimated, inducing too-high CO2 values in the
lower stratosphere. Furthermore, many three-dimensional models are too diffusive
and/or have too-strong mixing when crossing the tropopause, which leads to an
underestimation of the amplitude of the seasonal cycle in the column of CO2
. suggested that the lack of a reliable representation
of stratospheric influence on CO2, e.g. intrusion and recirculation, could explain
part of the discrepancy between a CTM and observations.
The persistence of the inverted CO2 gradient noted in these models can
result in an underestimation of the exchange of air masses from the stratosphere
to the troposphere in mid-latitudes during autumn. Due to the short time period of
simulations, models also fail to calculate a reliable CO2 seasonal cycle as
well as STE. concluded that at least 3 years are required for
the surface CO2 to be transported into the upper troposphere and lowermost stratosphere (LMS) then moved
to the temperate and polar latitudes. In order to eliminate the spurious diffusivity effect,
Lagrangian or quasi-Lagrangian models, such as TRACZILLA and CLaMS
(Chemical Lagrangian Model of the Stratosphere) ,
were widely used to investigate transport properties. The combination of
these Lagrangian models with in situ observations to reconstruct chemical trace gas
distributions has significantly contributed to our understanding of the mixing effects
across the extratropical tropopause , the filamentary
structure in long-lived and short-lived species near the edge of the polar vortex
, the transport of long-lived species and CO from the tropical
troposphere to the stratosphere , and the processes that control
UTLS composition .
The small-scale variability of CO2, its strong gradients across the
tropopause, and the scarcity of suitable observations for validation purposes lead
to a challenging task for CTMs and CCMs in reconstructing its distribution
in the UTLS .
In this paper, our goal is to build a database of the monthly zonal mean distribution
of CO2 on a global scale extending from the upper troposphere to the stratosphere
using backward Lagrangian trajectories. This product can be used to validate CTMs;
CCMs, as a prior for inversion modelling; and to analyse features of the STE as well as the
stratospheric circulation and its variability.
The trajectory data set on which this work is based was used by
to study the age of air and its variability in the stratosphere. We refer to this previous
work for all related questions. The present study can also be seen as a further validation
of .
We reconstruct a global distribution of CO2 calculated over the time period
2000–2010 from a Lagrangian transport model driven by horizontal winds and diabatic
heating rates, representing the vertical velocity in the isentropic coordinate,
from the ERA-Interim reanalysis provided by the European Centre for Medium Range Weather
Forecast (ECMWF) . We describe the data and method used in this study in
Sects. and , respectively.
The reconstructed CO2 is compared with observations in Sect. .
The global monthly distribution of the zonal mean CO2 and its variability are
discussed in Sect. . Finally, Sect. provides
further discussions and conclusions.
Data
The reconstruction of the global distribution of CO2 from the upper troposphere
to the stratosphere through the use of back trajectories requires the value of the CO2
mixing ratio to be assigned at the lower tropospheric boundary. This is achieved by using
two different types of CO2 data: ground stations and
CarbonTracker .
In addition, in situ measurements of CO2 from balloons and aircraft were used
to validate the reliability of the model reconstructions .
Lower boundary condition of the backward trajectories
Two different observation-based data sets are used to assign CO2 to
air parcels transported along the backward trajectories. During the
1989–1999 time period, data from ground stations of the World Data Centre
for Greenhouse Gases (WDCGG, http://ds.data.jma.go.jp/gmd/wdcgg/) are
applied. The WDCGG is an international data centre participating in the WMO
Global Atmosphere Watch. It provides extensive data from ground stations and
aircraft measurements across the Earth that are non-homogeneously
distributed. The monthly CO2 data from ground stations (e.g. Mauna
Loa, South Pole, and others) located at different latitudes were used to
overcome the daily fluctuations of CO2 in the atmospheric boundary
layer. The criterion for selecting the ground stations is that they are
remotely located with respect to localized anthropogenic sources. In highly
industrialized regions, this criteria is performed by retaining only high-altitude stations to neglect the variability from localized sources at ground
level. The CO2 data are averaged from pole to pole in latitude
increments of 30∘ and over all longitudes to represent the global,
free tropospheric CO2 field. To better model the latitude dependence
of the seasonal cycle and to overcome discontinuities, the averaged
CO2 data obtained are then interpolated linearly in latitude. Since
the ground station locations are inadequate to define longitudinal
variability, we use a constant CO2 value at all longitudes for each
latitude bin. According to CarbonTracker residuals against the NOAA North
American aircraft network data at altitudes above the planetary boundary layer
(PBL)
(http://www.esrl.noaa.gov/gmd/ccgg/carbontracker/profiles.php), the
mean error of CarbonTracker is less than 1.25 ppmv at
500 hPa. This error should be added, in mean square, to the error
determined in this study as a result of model dispersion (Sect. 5.3).
For the 2000–2010 time period, we use CO2 output from the coupled
data assimilation system CarbonTracker with the TM5 transport model for the
lower boundary condition . CarbonTracker produces full
3-D dimensional output so that the backward trajectories are assigned to daily
CO2 mole fractions based on the latitude and longitude
(3∘ × 2∘ resolution) of the trajectory at the
5 km (500 hPa) level. The 5 km level was chosen to
represent the well-mixed free troposphere. The CarbonTracker system
assimilates CO2 observations from atmospheric stations and optimizes
underlying CO2 mixing ratio from the fluxes of the ocean, biosphere, biomass burning and
fossil fuel usage. These data are meant to achieve a complete and realistic
diagnostic of the lower-atmospheric CO2 and fluxes
(CarbonTracker-2013B, www.esrl.noaa.gov/gmd/ccgg/carbontracker/). The
version used in this study corrects an error in vertical mixing in the
previous versions.
Admittedly, the reduced sampling of the pre-2000 period and the lack of zonal variability
induces an increased uncertainty in our calculation. However, the zonal variability is
largely filtered out in the upper troposphere and lower stratosphere, especially during winter.
Moreover, we only provide CO2 reconstructions for 2000–2010. During this period,
the tropospheric conditions are mostly determined by CarbonTracker output due to the
fast transport time in the troposphere, and the zonal fluctuations are mostly washed out in
the stratosphere. The influence of pre-2000 data then decays exponentially and almost vanishes
after 2005.
In situ aircraft and balloon measurements
In the UTLS, there are strong horizontal and vertical gradients. These gradients
may occur on a small scale and exhibit high temporal and spatial variability. In this
study, we are interested in airborne measurements in order to validate our model as well as
to characterize the stratospheric variability and mixing process.
Aircraft observations have a vertical resolution of a few metres (during ascents
and descents) and a horizontal resolution of a few hundred metres, resulting from
the high sampling frequency of these instruments (0.5–2 Hz).
Therefore, the aircraft observations are able to sample the
small-scale variability of the tracers.
The SAGE III Ozone Loss and Validation Experiment (SOLVE) sought to establish
a comprehensive data set of UTLS trace gases and meteorological data in the northern
polar regions, including latitudinal gradients across the polar vortex.
Measurements were made in the Arctic high-latitude region during winter 1999–2000
using the NASA DC-8 and ER-2 aircraft, as well as balloon platforms and ground-based
instruments. CO2, CH4, and N2O were measured by several
instruments and used to calculate a composite mean age
as inusing earlier measurements.
In situ balloon-based CO2 profile measurements are also used as a
basis for comparisons with the reconstructed CO2 from the Lagrangian
transport model. The data sets used in this study are high-quality
observations with sufficient altitude coverage. They are measurements of
whole air samples collected cryogenically from balloons or in situ
measurements on-board a balloon gondola . Four
balloon flights were selected for which a full CO2 profile is
available (1) at Fort Sumner, New Mexico, USA (34.5∘ N), on
17 September 2004; (2) at Sanriku, Japan (39.33∘ N), on 30 May 2001;
(3 and 4) and at Aire-sur-l'Adour, France (43.75∘ N), on
24 September 2002 and 9 October 2001, respectively. Note that most profile
observations are from the May–October period, when stratospheric
variability in the Northern Hemisphere is expected to be lower than during
the winter period. The combined measurements of CO2 and SF6
were used by to derive the mean age of the air, but here we
focus on CO2.
A further data set, based on the CONTRAIL experiment was
used in the validation process of the reconstructed CO2 in the whole
troposphere (6–13 km) from 20∘ S to 60∘ N during
November 2005–2009. CO2 mixing ratios were measured during regular
flights by Japan Airlines from Japan to Australia, Europe, North America, and
Asia with continuous measuring equipment (CME) for in situ CO2
observations, as well as improved automatic air sampling equipment (ASE) for
flask sampling (for more details about the instrument see
). This data set provides significant spatial coverage,
particularly in the Northern Hemisphere .
Method of global CO2 reconstruction
To calculate the global CO2 distribution, air parcels are distributed from the
upper troposphere to the stratosphere and integrated backward in time.
Global backward trajectory
Backward deterministic trajectories were calculated using the Lagrangian
transport model, TRACZILLA , which is a modified version
of FLEXPART . TRACZILLA uses analysed winds to move air
parcels in the horizontal direction and performs direct interpolation from
data on hybrid levels. In the vertical direction, we used potential
temperature coordinates and total heating rates. We denote the trajectories
as diabatic following a convention established by .
At each level in the vertical, trajectories are initially distributed over
a longitude–latitude grid with 2∘ resolution in latitude and an
almost uniform spacing in longitude of 2∘/cos(ϕ), where ϕ
is the latitude, generating 10 255 air parcels on each level from pole to
pole. For the sake of simplicity, the vertical levels of the initial grid are
chosen to be the hybrid levels of the ECMWF model. In order to encompass the
whole stratosphere at any latitude and longitude, 30 levels from about
400 hPa (varying according to the surface pressure) to 2 hPa
were selected. Above 56 hPa, the hybrid levels are reduced to pure
pressure levels. Trajectories ending below the tropospheric boundary
condition at 500 hPa, at which we assign the free tropospheric
CO2 to each trajectory, were discarded during the initialization.
Trajectories ending above this boundary are integrated backward in time up to
10 years or until they cross the boundary condition. In practice,
stratospheric trajectories reach this boundary shortly (less than 2 months)
after crossing the tropopause. Ensembles of trajectories were launched at the
end of every month over the period 2000–2010 .
A schematic representation of the backward Lagrangian trajectories
of air parcels starting in a
(latitude × longitude × altitude) grid box. Here the
longitudinal extend of the box should be seen as the whole latitudinal
circle.
Calculation of global CO2
Once a parcel has reached the tropospheric boundary condition at
500 hPa at a given time and a given location, its CO2 mixing
ratio is assigned according to the mixing ratio at that time and that
location calculated from CarbonTracker and WDCGG. CarbonTracker was chosen
when a back trajectory reached the tropospheric boundary after 1 January 2000
and WDCGG was chosen when it was impacted before this date. Since WDCGG
provides surface data only, it was assumed that the vertical transport was
fast in the lower troposphere and induced only a negligible bias at
500 hPa in the CO2 mixing ratio, which was well verified (not
shown) in the inner tropical region where most parcels reach the boundary.
The assigned value was then used to reconstruct the CO2 mixing ratio
at the location and time of the trajectory initialization.
The monthly zonal mean CO2 for a given bin in latitude and altitude
was calculated as the average over all longitudes of the trajectories
initialized within this bin (Fig. ). The latitudinal
resolution of the bins is centred 2∘ equatorward of 68∘ and
decreases near the poles (69–73∘, 73–77∘, 77–81∘,
81–90∘). For each date, the average is made over 180 air parcels at
the equator and 67 air parcels at 68∘ N or S. Near the poles,
towards which the number of trajectories launched per degree of latitude
decreased to zero, larger intervals were chosen to maintain a sufficiently
large number of trajectories in the bins. This calculation uses the same
approach as Sect. 2.3 of . Further averaging over time is
performed to improve statistics and to reduce noise. These averaging
procedures are a simple way to account for mixing in the stratosphere and
gather a distribution of air parcels with different
histories within each bin.
As observed by , the number of backward trajectories
launched on a given date and remaining within the stratosphere after some
residence time, τ, decreases exponentially with τ.
showed that this relationship held for τ>3 year with an exponential decay parameter (b) equal to
0.2038yr-1 using ERA-Interim winds and heating rates. The
standard deviation of the mean (where each month was considered separately)
decayed at the same rate. After 10 years, 88 % of the particles
initialized in the stratosphere reached the troposphere. We followed
in using this property to correct the estimated
CO2 for the truncation of trajectory lengths at 10 years. If we
define G(J|t,τ) as the probability density of the residence time τ
at time t for parcels launched in the bin J, the monthly mean
stratospheric CO2 mixing ratio is
CO2(J,t)‾=∫0∞CO2T(t-τ)G(J|t,τ)dτ,
where CO2T is the tropospheric mixing ratio of CO2, which
is assumed here to be uniform for simplicity. The truncated version of this
integral, up to tf=10yr, can be calculated explicitly from the
backward trajectories as a mean for all parcels from bin J, which hit
the 500 hPa surface weighted by their proportion among all launched
parcels in bin J. Assuming that G(J|t,τ)=G(J|t,tf)exp(-b(τ-tf)) for t>tf, with CO2T governed by
an annual modulation added to a linear growth, CO2T(τ)=p0+p1×τ+a0×cos(2π(τ-φ)), the monthly mean
CO2 mixing ratio can be estimated as
CO2‾(J,t)=∫0tfCO2T(t-τ)G(J|t,τ)dτ+G(J|t,tf)bp0+p1(t-tf-1b)+ba0b2+4π2bcos2π(t-tf-φ)+2πsin2π(t-tf-φ),
where all times are in years. The contribution of the remaining air parcels
after 10 years of backward motion was thus accounted for by the integrated
term in Eq. (), where G(J|t,tf)/b is the proportion of parcels
in the bin that have not hit the 500 hPa surface at time tf. The
coefficients p0,p1,a0, and φ are estimated by fitting the Mauna Loa
CO2 data. The correction can also be applied below the tropopause
since
the only tropospheric parcels that live for 10 yr without hitting
the 500 hPa surface are among those that have been entrained in the
stratosphere.
Validations of the global reconstruction method
The reconstruction during SOLVE and in situ balloon campaigns are used to
validate the global reconstruction of CO2 and the ability of TRACZILLA
to reproduce the small-scale CO2 variations along the flight tracks.
Reconstruction of CO2 along aircraft flight track and balloon profiles
The procedure used here differs from that of the global reconstruction
described above in three main respects. First, the parcels were initialized
at locations distributed along the flight track or the balloon profile. In
the case of the ER-2 flights, parcels were released with the frequency of the
measurement, at 0.25 Hz , amounting to 900
locations per flight hour. In the case of the balloon flight
, the air parcels were distributed along the balloon
profile with a frequency higher than the tracer measurements. Namely, they
were released at 200 locations in the vertical, regularly distributed in
log pressure between 500 and 1 hPa at the same latitude–longitude
position as the balloon.
Second, we take into account that a single sample can be understood as a mixture of sub-parcels
arising from a large number of origins.
The simplest representation of this mixing is by a constant diffusion, which mainly acts
in the vertical direction, and it is well known that such a process can be represented by a
Wiener process.
Therefore, following , we released a large number of air parcels
(200 for the ER-2 flights, 5000 for the balloon profiles) from each measurement location.
The Lagrangian advection was modified such that on a time step δt the motion of a
given parcel located in X is
δX=u(X,t)δt+δηk,
where u is the wind fields, k is the vertical unit vector, and
δη≡w(t)δt is the product of the time step δt and a
Wiener process w approximated by 50 iterations of the white noise during a time step.
In the small δt limit, this is equivalent to a diffusion D=12<w2>δt.
The well-posedness in the backward time direction arises from the adjoint equation of the
Green function of advection–diffusion for more details see.
The value D in the lower stratosphere was estimated
by comparing the observed small-scale tracer fluctuations and their reconstructions.
The resulting value is D≈0.1m2s-1, which is applied to the
whole atmosphere in the present study.
Physically, this turbulent diffusion, which is about 4 orders of magnitude larger than the
molecular diffusion of CO2 in the air
1.6 10-5 m2s-1,
accounts for the small-scale motion missing in the ERA-Interim reanalysis winds.
It is noticeable that the diffusion is effective at dispersing the clouds of parcels emitted
from a single location only for a few days, after which dispersion by the resolved wind strain
dominates.
Third, the trajectories were integrated backward for 6 months, after which the CO2
mixing ratio was assigned according to the zonal mean CO2 value calculated from
the global
reconstruction at that time and at the locations of the parcels. The mean value and confidence
interval were calculated over all the initialized particles. The air parcels that reached
the 500 hPa level were assigned the CO2 mixing ratio on that surface.
Comparison of the reconstructed monthly mean CO2 from
backward trajectories with aircraft measurements from the SOLVE campaign
. Black shows reconstructed mean CO2 along the ER-2
flight track using the TRACZILLA Lagrangian transport model. Red shows observed
mean CO2. Green shows potential temperature of the ER-2 flight track.
The grey shaded area indicates the 95 % confidence interval calculated
from the reconstruction.
Comparison of observations and model reconstructions
In this section, we test the realism of CO2 reconstructions against several observation
data sets that span a large range of scales, geographical locations, and altitudes.
SOLVE campaign
Figure shows observed and reconstructed CO2 mixing
ratio time series from 16 flights during the SOLVE campaign.
Figure compares the observed versus reconstructed CO2
mixing ratios for each flight along with correlation coefficients and mean
distances (Δ in ppmv), defined as the sum of the absolute
difference between the observed and the reconstructed values divided by the
number of recorded values. The flight patterns include test flights at
subtropical (11 and 16 December 1999, 6 January 2000) and mid-latitudes
(11 January 2000), transit flights between mid- and high latitudes
(14 January, 16 March 2000), and flights inside the polar vortex or across
its edge (all other dates). In nearly all of the flights, the observed
CO2 falls within the 95 % confidence interval of the
reconstruction. It can be seen from Fig. that the correlation
is not a good indicator of the similarity between the observed and the
reconstructed curves because it can be high due to trends, even for cases that
exhibit large differences, such as 3 February 2000. The Δ value is a
much better metric of the agreement between the observed and reconstructed
CO2. In 6 cases out of 16 the agreement is excellent, with Δ≤0.36 ppmv, and the two curves agree fairly well even for the
magnitude of small-scale fluctuations. In four other cases with 0.49≤Δ≤0.61 ppmv, the two curves stay very close, with only a
few features missed by the reconstruction. In two other cases with 0.66≤Δ≤0.67 ppmv, the reconstruction shows some larger
deviations from the observations. On 11 December 1999, the reconstruction
missed the decrease in CO2 as the plane ascended at the beginning of
the flight and then stayed slightly too high for the subsequent horizontal
lag. The 27 January 2000 case is a flight from the inside of the polar vortex
to the outside, which was poorly reconstructed for the outside part between
10:30 UTC and 13:00 UTC. Using similar methods, showed
that the stratospheric tracers O3 and N2O could be
reconstructed for this flight, but they also found a large standard deviation for
the outside section where filaments of polar and extratropical air were
interleaved. The flights with the largest discrepancies (Δ>0.74 ppmv) on 14 January, 3 February, and 7 and 16 March 2000 can be
explained by flight tracks that followed the edge of the polar vortex. In
these flights, the reconstruction is very sensitive to any misplacement of
the vortex edge in the reanalysis and thus not useful as an evaluation of the
reconstructed CO2. In addition, it is important to emphasize that the
value of the applied diffusion (D≈0.1m2s-1) allows
the reconstruction to fit the observed small-scale variability
(∼ 1 km). See for a complete discussion on
this matter.
Comparison of the reconstructed monthly mean CO2 from the backward trajectories with
aircraft
measurements from the SOLVE campaign . Colours indicate different flights in
Fig. . R-squared and the Δ (in ppmv), defined as the mean of the
absolute value of model–observation differences, are shown in the legend. The dashed line
is the 1:1.
Balloon vertical profiles
In order to test the reconstruction over a larger vertical range of altitude,
Fig. shows a comparison of the vertical
profiles of the reconstructed mean CO2 by TRACZILLA with the
observations of four mid-latitude stratospheric balloon flights
. For three of the cases, most of the
measurements fall within the 95 % confidence interval of the reconstructed
profiles and the local maxima at 23 and 18 km in
Fig. c, d, respectively, are well reproduced.
These three profiles have relatively large CO2 mixing ratios in the
troposphere in common, which decrease with altitude. However, the
reconstructed profile in Fig. b is
1 ppmv smaller on average than the observed profile and misses the
large fluctuations above 20 km. This flight was performed from Aire-sur-l'Adour (France) when a cold front crossed the region, with strong local
tracer gradients in the lower stratosphere, as seen in the potential vorticity
map shown in the panel. In order to test the spread induced by this
meteorological structure, we have reconstructed eight vertical profiles at
1∘ distance around the initial profile. However, the observed spread
among this ensemble of profiles is too small to explain the discrepancy in
Fig. b. Notice that SF6-derived mean
ages are in good agreement with the reconstructed mean age of
. Therefore, we are left without any satisfactory
explanation for this case but to assume some undocumented instrument
malfunction.
Reconstructed vertical profiles of the monthly mean CO2
compared with each in situ stratospheric balloon observation
of CO2 .
Black curves show reconstructed vertical profiles of mean CO2.
Green squares show in situ balloon measurements of mean CO2.
Grey shading shows the 95 % confidence interval from the reconstruction.
The measurements were taken from Sanriku, Japan (39.33 ∘ N), on
30 May 2001 (a); Aire sur l'Adour, France
(43.75 ∘ N),
on 9 October 2001 (b) and on 24 September 2002 (c); and Fort Sumner, New Mexico, USA
(34.5 ∘ N), on 1 September 2005 (d), respectively. The different dashed
lines show the other eight reconstructed profiles surrounding the measurement on 9 October 2001.
The insert on the upper
right panel shows the potential vorticity (in PVU) on the 70 hPa surface
for 9 October 2001 at 12:00 UTC over France from ERA-Interim. The location of Aire sur
l'Adour is indicated by a diamond.
Temporal series
To obtain additional details about the upward propagation of the tropospheric CO2
seasonal cycle into the LMS and to evaluate the model near the lower boundary condition and the
tropical tropopause, we compare the time series of the reconstructed monthly mean CO2
(Sect. ) with the observations.
Figure a compares the time series of modelled monthly mean
CO2 with the measurements from CONTRAIL
in the tropical region 10∘ S–20∘ N and in the vertical
range 7–9 km between November 2006 and January 2010. The comparison
shows the ability of the model to capture the tropospheric CO2
seasonal variation and validates the tropospheric boundary condition.
Temporal evolution of the monthly mean CO2 seasonal cycle from
TRACZILLA calculations (line) compared with CONTRAIL and ground measurements (circle).
(a) Comparison model with CONTRAIL in the tropospheric region above the tropospheric boundary
in the latitude range 10∘ S–20∘ N.
(b) In the upper tropospheric region close to the tropical tropopause and the latitude range
10∘ S–20∘ N, comparison with the average of surface station data
at Mauna Loa, Hawaii (19∘ N), and American Samoa (14∘ S) delayed by 15 days.
(c) Comparison model with CONTRAIL in the upper troposphere near the extratropical tropopause
at 50–60∘ N and at several heights from 7 to 15 km. The error estimated
from the reconstruction is indicated as vertical grey bars.
Figure b compares the modelled monthly mean CO2
time series in the altitude bin 16–17 km and between 10∘ S
and 20∘ N just below the tropical tropopause, where the tropospheric
air enters the stratosphere, with the average of ground-based CO2
data from Mauna Loa (19 ∘ N) and American Samoa (14 ∘ S)
delayed by 15 days. We find, consistent with and
, that the amplitude of the
two signals is the same, and we diagnose a delay of 2 months at a higher
altitude in the layer 18–19 km (not shown) in agreement with
. The shorter timescale below the tropopause is in
agreement with other studies .
Figure c shows the modelled monthly mean CO2 in the
latitude bin 50–60∘ N at different altitudes in the range
7–13 km between November 2005 and January 2010. These curves are
compared with CONTRAIL measurements in the same latitude band
. The modelled and measured CO2 differ by less than
1 ppmv, except for a few isolated months such as March 2006 and March
2009 and outliers such as April at 12–13 km. There is a shift on the
order of 4–6 months in the mean CO2 seasonal cycle above
11–12 km, in the lowermost extratropical stratosphere, with respect
to the tropospheric signal. This is due to the delay induced by the shallow
branch of the BDC also found by and .
The discrepancies are concentrated during the spring season, during which
large gradients of CO2 span the region, as discussed in
Sect. .
Global distribution of zonal mean CO2
The zonal mean distribution of CO2 illustrates the main features of
the BDC, such as mixing and transport variabilities through temporal and
spatial evolution. Figure illustrates the
typical seasonal variation of the monthly mean CO2 derived from the
Lagrangian reconstruction for 2010 as an example among the 11 years.
(a) Global distribution of the seasonal cycle of the reconstructed monthly mean
CO2 (in ppmv) in the upper troposphere and the lower stratosphere from 5 to 25 km for the odd months of 2010.
(b) Same as (a) but for the even months of 2010 and the altitude range from 5 to 45 km.
CO2 calculated on model levels is first interpolated to altitude levels using the latitude dependency of the zonally and monthly averaged geopotential.
(c) The standard error of the mean CO2 over the 2000–2010 period. The white contours show the isentropic surfaces.
Upper troposphere and lowermost stratosphere
The zonal mean distribution of CO2 in the free atmosphere, especially
above 5 km, is driven by the large-scale transport processes. Fast
quasi-isentropic mixing is combined with upwelling in the tropics and
downwelling in the extratropical lowermost stratosphere. Figure a shows the meridional and vertical
CO2 distribution during 6 different months in 2010. In the Northern
Hemisphere, the tropospheric monthly mean CO2 is dominated by a
strong seasonal cycle, reflecting the biospheric activity. The terrestrial
vegetation removes CO2 by photosynthesis during its growth phase and
returns CO2 to the atmosphere when it dies and decomposes.
CO2 concentration increases during autumn and winter to reach a
maximum in April–May, followed by a rapid decay due to the spring biospheric
bloom, and reaches a minimum in July–August. The cycle is much weaker in the
Southern Hemisphere and is influenced by transport from the Northern
Hemisphere. The combined effect of fast isentropic mixing
and convection
propagates the cycle towards the tropics, creating both a horizontal and
vertical gradient . From
Fig. a, it is clear that during the Northern
Hemisphere winter, the concentration tends to follow the isentropes in the
extratropics for potential temperatures up to about 330 K. The
barrier effect of the subtropical jet generates a
strong meridional gradient near 30∘ N, which reaches a maximum near
350 K. Once it has reached the tropics, CO2 is then
transported upward by tropical convection and propagates into the
stratosphere through the BDC. Throughout the summer (June, July, and August),
while the tropospheric CO2 is removed from the atmosphere due to the
biosphere activity, a layer of high CO2 extends from the tropics to
the northern mid-latitudes into the lower stratosphere driven by the lower
branch of the BDC . This transport is promoted by the
Asian monsoon anticyclone, which traps young continental air lifted from the
surface and induces a flux to the extratropical stratosphere on its west
side as it is eroded across the jet
.
Due to the turnover time of this transport, the maximum concentration of
CO2 in the northern lower stratosphere lags behind that at the
surface by 4 to 6 months, and this concentration is essentially reached when the
surface concentration is at its minimum. The result is an inverted vertical
profile, which is at its maximum in July and persists over the summer. A
qualitative comparison between the reconstructed CO2 in
Fig. a and observations from
(see their Fig. 7) exhibits good agreement in the cycle of the tropospheric
and lower stratospheric CO2, and in particular in the cycle of the
inversion. There are, however, differences in the location and intensity of
the meridional gradient, which might be due to the specific sampling by
, and which gives a strong weight to the most intense region
of the Pacific jet stream.
(a) Reconstructed vertical profiles of the mean CO2
compared with CONTRAIL measurements for 2007 at 50–60∘ N. Dotted and dashed lines show vertical profiles of CO2 from TRACZILLA
(blue: May, orange: August). Symbols show in situ
aircraft measurements from the CONTRAIL campaign (magenta square: May, black triangle: August).
(b) Averaged monthly profiles of the reconstructed CO2 over the period 2000–2010 after
removal of the mean CO2 trend at each level and centred on 2007.
Red is January, black is March, blue is May, magenta is July, cyan is September, and green is November.
Middle and upper stratosphere
Figure b shows the CO2 global
distribution in the middle and upper stratosphere up to 42 km for
even months in 2010. As the tropospheric seasonal cycle is transported into
the middle and upper stratosphere through the tropical pipe, its amplitude
decreases upward because of the combined effect of the upwelling branch of
the BDC and mixing processes. The deep branch of the BDC is much slower than
the shallow branch and old air with low CO2 concentrations in the
middle and upper stratosphere. Younger air with high CO2 is isolated
in the tropical area, an effect that is at a maximum during northern
hemispheric winter, in agreement with the age of the air calculations
. The horizontal mixing homogenizes CO2 in
the mid- and high latitudes during summer. Because of this prior mixing, the
winter containment within the polar vortex generates only a weak polar
minimum (and no localized horizontal gradient averaged over the latitude
circle and it does not follow the CO2 or potential vorticity contours).
Uncertainty about CO2 global distribution
Figure c shows the monthly averaged
uncertainties about the reconstructed monthly zonal mean CO2 over the
2000–2010 period calculated from the trajectories. The uncertainty is
estimated as the standard error of the mean by assuming that the contributing
trajectories are independent samples. The standard error is performed for
each month over 2000–2010. As an illustration, the standard error is then
averaged over 11 years (Fig. c). The estimated
CO2 uncertainties reveal smaller values for the trajectories starting
in the troposphere than the trajectories starting in the stratosphere, which
have a longer transit time of several years to reach the lower boundary
condition where the CO2 value is assigned. As expected, the
uncertainty roughly scales with the transit time of the trajectories from the
upper troposphere to the stratosphere. The maximum uncertainty reaches
1 ppmv in the stratospheric polar regions where the mean age of the
air reaches a maximum during winter and sampling is lowest. Note that the
mean error of CarbonTracker on the initialization values should be added to
this uncertainty from the spread of the trajectories.
Spring–summer vertical profiles
In this section, monthly averaged CO2 profiles are investigated to better describe the
changes in the CO2 vertical structure within the upper troposphere and stratosphere.
The spring–summer reconstructed vertical profiles of CO2 are compared
with those from the CONTRAIL aircraft measurements for the year 2007 in the
50–60∘ N latitude range (Fig. a).
Good agreement is obtained, including for the inversion of the CO2
vertical profile during August in the lower stratosphere. The monthly mean
CO2 vertical profiles, calculated by backward trajectories, exhibit a
complex vertical structure with gradient layers interspersed with no gradient
layers.
The annual structure of the profile is made apparent in
Fig. b, where we show averaged monthly
profiles over the period 2000–2010 after removing the mean CO2 trend
at each level. Starting from January, the increase in CO2 in the
troposphere penetrates upward in the stratosphere over the limited vertical
range of the extratropical transition layer
, which is over 2 to 3 km above
the tropopause, as is visible in the March profile. Between March and May,
another process occurs, which injects young air rich in CO2 above
13 km. This can only be due to a tropical intrusion promoted by the
weakening of the tropical barrier at the end of the winter. The profile
suggests (i) that the intrusion is deep from 13 to about 23 km,
(ii) that the well-mixed layer between 13 and 16 km is influenced by
the well-mixed tropical tropospheric profile at such altitudes, and (iii) that
the mixing layer between 16 and 23 km is also induced by the tropical
lower stratosphere vertical gradient. The mixing layer persists with the same
slope during the whole summer, and the bottom of the intrusion corresponds to
the maximum of CO2 when the inversion is at its maximum. During autumn,
when the subtropical barrier is re-established, the gradient weakens, the
residual well-mixed layer disappears, and the profile returns to the fairly
uniform slope of January.
Conclusions
Our study provides a monthly zonal mean distribution of CO2 spanning
the upper troposphere and the stratosphere over the time period 2000–2010,
established from observations and the state-of-the-art reanalysis
ERA-Interim. The zonal mean distribution of CO2 is a unique data set
of a critical trace gas that has a variety of uses for validating the
representation of upper tropospheric and stratospheric tracer distributions
in chemical transport models and chemical climate models, in particular
regarding the summer inversion of the CO2 profile in the Northern
Hemisphere. This CO2 product is also intended for satellite validation in
the upper troposphere and the stratosphere. It is used as a preliminary process
before (a prior) for inversion modelling and to analyse features of the
stratospheric–tropospheric exchange as well as the stratospheric circulation
and its variability. The reconstructed CO2 product contains zonal
mean, monthly mean mixing ratios in 77 latitude bins from 90 ∘ S to
90 ∘ N, and 36 vertical levels from 5 to 42 km.This
reconstructed monthly zonal mean CO2 exhibits a remarkable agreement
with CONTRAIL data, SOLVE, and in situ balloon measurements.
The comparison with SOLVE shows that a Lagrangian-diffusive model is able to reproduce the mean
value and the number of small-scale fluctuations that are recorded by in situ measurements along
flight tracks in the lower stratosphere. This reconstruction suggests that the distribution of
long-lived tracers, such as CO2, can be fully explained by the properties of
transport,
as resolved by meteorological analysis or reanalysis and a simple representation of sub-grid-scale effects as a diffusion.
In the northern hemispheric troposphere, the monthly mean CO2 is dominated by biospheric
activity and displays a strong seasonal cycle, which is vertically and horizontally propagated
to the tropopause and above in the lowermost extratropical stratosphere, on the one hand, and to
the tropics, on the other hand, where it reaches the tropopause and enters the stratospheric
Brewer–Dobson circulation. In regions of high horizontal mixing such as the mid-latitudes,
CO2 tends to be uniformly mixed at isentropic surfaces and its meridional gradients
are enhanced near transport barriers such as the subtropical jet during winter.
Transport of CO2 into the northern extratropical stratosphere above
the lowermost stratosphere is due to the export of tropical air. The long
circuit of CO2 from the extratropics to the tropics in the
troposphere and then back to the extratropics in the stratosphere induces a
time lag of 4–6 months such that the tropospheric and stratospheric
variability are almost opposite at mid-latitudes. The result is the
production of an inverted vertical CO2 profile during summer. In the
mid- and upper stratosphere, we found that as the tropospheric seasonal cycle
is transported into the stratosphere through the tropical pipe, its amplitude
is smoothed out because of the combined effect of the upwelling branch of the
BDC and quasi-horizontal mixing. A more confined tropical pipe is found in
the tropical band during winter and spring than during summer and autumn.