Sei sulla pagina 1di 10

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115, D19308, doi:10.

1029/2010JD014060, 2010

Physical atmospheric structure and tropospheric mixing


information in vertical profiles of atmospheric CO2 mixing ratios
A. Font,1,2 J.A. Morgu,1,2 and X. Rod1,2,3
Received 16 February 2010; revised 7 May 2010; accepted 12 May 2010; published 7 October 2010.
[1] Vertical distribution of atmospheric CO2 mixing ratios, as well as CO2 vertical
variance and gradient are related to the vertical stability at the time of measurement, to the
transport of coherent upstream plumes studied through changes in the upstream surface
influence or to the historic mixing processes and dispersive behavior. Three vertical
profiles of CO2 mixing ratios measured from 900 to 4200 m above sea level (masl) in 2006
at La Muela, Spain (LMU, 41.60N 1.10W, 570 masl) are examined. Changes in CO2
mixing ratio are associated with changes of the atmospheric physical parameters on the
day of the survey; and with the transport of coherent air masses. Its consistency is
examined through changes of the historical horizontal dispersion and chaotic mixing
dynamics of air masses during the four days prior to measurements. A climatology of
Lagrangian backward simulations run once a week at four altitudes (600, 1200, 2500 and
4000 masl) at LMU for 2006 shows that dispersion in these altitudes is superdiffusive
(exponent coefficient g > 1/2) and mixing follows a chaotic dynamics (power law
exponent l > 0) at all altitudes and in all seasons. Furthermore, a Horizontal Mixing
Discontinuity (HMD) at 2500 masl separates two layers with different constraints on
vertical mixing. Above the HMD, more coherent air masses and laminar transport
characterizes the dynamics of atmospheric horizontal mixing whereas below it,
filamentation and chaotic mixing dominate horizontal mixing. In the lower part of the
vertical profile, within the ABL, mixing takes place by convection. Chaotic mixing below
the HMD induces the boundary layer entrainment. Results highlight that there are two
main discontinuities in the air column which separates different atmospheric dynamics.
The ABL is driven by local meteorological conditions of the site at the sampling time;
the HMD is driven by the synopticscale historical mixing conditions of air masses. The
effect of horizontal transport in the free troposphere should be considered equally as
important as the local vertical mixing processes in the atmospheric boundary layer when
interpreting CO2 vertical gradients. The dispersive exponents can be used to identify the
transport of coherent plumes from anthropogenic emissions with different carbon
composition that the onesingle backtrajectories models do not detect.
Citation: Font, A., J.A. Morgu, and X. Rod (2010), Physical atmospheric structure and tropospheric mixing information in
vertical profiles of atmospheric CO2 mixing ratios, J. Geophys. Res., 115, D19308, doi:10.1029/2010JD014060.

1. Introduction motions [Sidorov et al., 2002]. In order to accurately interpret


observations of the CO2 mole fraction in terms of surface
[2] CO2 mixing ratios in the lower troposphere over con- sources and sinks, it is therefore vital that transport models
tinental areas at midlatitudes of the northern hemisphere are
not only portray the largescale circulation accurately, but
directly influenced by gas exchange with the surface vege-
also parameterize subgridscale mixing [Tans and Wallace,
tation below and also by anthropogenic emissions. Addi- 1999; Bakwin et al., 2003]. Measurements of CO2 above
tionally, CO2 mixing ratios are influenced by atmospheric
the boundary layer taken from aircraft samples are particu-
mixing processes driven by mesoscale circulation and large
larly useful in developing and testing transport parameteri-
scale advective transport as well as convective and turbulent
zation used in carbon cycle studies [Tans et al., 1990; Wang
et al., 2007]. Models used for globalscale inversions mis-
1
Institut Catal de Cincies del Clima, Barcelona, Spain. represent vertical mixing since they do not simultaneously
2
Also at Climate Research Laboratory, University of Barcelona, reproduce the annual average and the seasonal cycle of
Barcelona, Spain.
3
Also at Instituci Catalana de Recerca i Estudis Avanats, Barcelona,
measured vertical profiles [Stephens et al., 2007], probably
Spain. due to a misrepresentation of the depth of the boundary layer
[Gerbig et al., 2008] or the presence of filaments in tracer
Copyright 2010 by the American Geophysical Union. profiles not well reproduced [Lin et al., 2004]. Recent liter-
01480227/10/2010JD014060

D19308 1 of 10
D19308 FONT ET AL.: HORIZONTAL MIXING DISCONTINUITY D19308

ature stresses the importance of vertical profiles of CO2 or atmospheric boundary layer (ABL) [Seibert et al., 2000].
mixing ratio in evaluating surface to air fluxes in terrestrial In this paper, the height of the ABL is determined using
and oceanic environments [e.g., Gerbig et al., 2003]. How- radiosounding data from Zaragoza Airport, 15 km away
ever, the effect of horizontal transport should be also con- from LMU by the bulk Richardson number (BRN) method
sidered when interpreting vertical profiles. [Grimsdell and Angevine, 1998; Menut et al., 1999;
[3] Tracers are transported in the atmosphere by the Eresmaa et al., 2006; Sicard et al., 2006; Morille et al., 2007].
mixing of fluids with different properties in a multiscale The BRN as a function of altitude is defined as:
velocity field. Mixing processes stretch and fold the material
surfaces, reduce the length scales of the initial mass distri- g z  z0  z  z0 
Rib z 1
bution, and generate microstructures in the tracer distribu-  z u z2  z2
tion [Ottino, 1989; Shraiman and Siggia, 2000]. The
distortion of the shape of an air parcel occurs through the where  is the potential temperature, g is the acceleration due
action of synopticscale eddies and, because of wind shear, to the gravity, z is height, z0 is the height of the surface, and
produces filaments whose length scale is much larger than u and v are the zonal and meridian wind components,
their width [Pudykiewicz and Koziol, 1998; Stohl, 2001].The respectively. Beyond a critical value Ribc (critical Richardson
formation of streamers or filaments enhances the efficiency number) the atmosphere can be considered fully decoupled
of the microscale processes that mix air across air parcel from the ABL [Sicard et al., 2006]. Here, Ribc is set at 0.25,
boundaries changing the original chemical properties as in the works by Grimsdell and Angevine [1998] and
[Colette and Ancellet, 2006]. Gerbig et al. [2008].
[4] Previous papers interpret transport effects on observed 2.2. Dispersion Processes in the Atmosphere and
CO2 mixing ratio profiles by means of virtual potential Dispersion Indexes
temperature and backtrajectory analysis [Sidorov et al.,
2002] and/or chemical signal changes and backtrajectory [6] We do examine dispersion processes in the atmo-
analysis [Sturm et al., 2005; Lloyd et al., 2002]. Back sphere to evaluate the transport of coherent plumes which
trajectories simulates the mean path traveling of air masses are likely to induce a change in the measured CO2 con-
before arriving at the receptor or measurement site tracking centration through the use of the dispersive exponent (g),
one particle. However, backward modeling with Lagrangian the Lyapunov exponent (l) and the potential surface area
Particle Dispersion Models (LPDM) can replace traditional growth rate (b), as explained in the following subsections.
backward trajectory calculations as it provides more realistic 2.2.1. Dispersive Exponent (g)
simulations of air masses transport. Trajectories only account [7] The turbulent flow of a plume can be typified con-
the advection whereas backward modeling also includes sidering by the relative dispersion of Lagrangian particles
turbulence, convection and filamentation of the initial sam- [Maryon and Buckland, 1995; Huber et al., 2001; Colette
pling volume by the largescale advection [Stohl et al., 2002]. and Ancellet, 2006]. The rootmean square (rms) distance
In this paper, the presence of atmospheric layers which are in time, sr (t), is defined relative to the average position of
likely to preserve coherent information about upwind sources the plume as:
r
and sinks is highlighted through the analysis of profiles of 1 X 2
CO2 mixing ratios above the atmospheric boundary layer up r t di t 2
N
to 4000 m. Vertical distribution of CO2 mixing ratios profiles
in the low troposphere sampled over La Muela, Spain (LMU; where N is the number of particles, and di(t) is the distance
41.60N 1.10W; 570 masl) is analyzed and related to the of each particle from the center of the plume at time t
transport of coherent upwind plumes. The consistency of the (counted forward), considering both the meridional and
backplumes is evaluated through three indexes which inform zonal separations. The separation over time of neighboring
about the dispersion processes: the dispersive exponent (g), particles can be used to directly determine the intensity of
the Lyapunov exponent (l) and the potential surface area stirring. The stronger the stirring, the stronger is the strain
growth rate (b). First, a climatology made up of weekly on particles that are initially located close to each other, and
Lagrangian Particle Dispersion backward simulations started the faster these particles will separate [Stohl et al., 2004]. In
at 12 UT for 2006 is carried out to summarize the main- a turbulent plume, the increase of the rms distance over time
streamer processes in air masses at different altitudes before follows a power law [Huber et al., 2001], written as:
they arrive at the measurement site. Second, a sample of
three observed vertical profiles of atmospheric CO2 mixing r t  t  3
ratios is shown as an application of the study of the change
of the stretching and folding processes in the low tropo- where g is the dispersive exponent which characterizes the
sphere; and of the identification of plumes arriving at the dispersion of particles over time. Dispersion with g < 1/2 is
measurement site based on the dispersion indexes. termed subdiffusive, and occurs when particles spend long
stretches of time with nearzero velocity; and g > 1/2 is
2. Vertical and Horizontal Atmospheric Structure superdiffusive or turbulent, which arises from persistent
streaming motions due to the presence of rapid, long flights
2.1. Vertical Physical Atmospheric Structure and of particles [Solomon et al., 1994] such as those that may be
Atmospheric Boundary Layer Height produced by strong largescale shear. The particular case of
[5] Tracer gases are dispersed horizontally and vertically g = 1 is the ballistic one; and g = 1.5 is the Richardson
in the atmosphere by turbulence, and they eventually Obukhov (RO) law [Huber et al., 2001]. Here, whenever g is
become completely mixed above the socalled mixing layer 1.251.5, RO dispersion type is flagged.

2 of 10
D19308 FONT ET AL.: HORIZONTAL MIXING DISCONTINUITY D19308

2.2.2. Lyapunov Exponent (l) vals over 96h simulations, and also the particles residence
[8] The finitetime Lyapunov exponent describes the time in the 300 m above the ground level at the same time step
stretching processes in the atmosphere given the following (footprint layer).
expression [Maryon and Buckland, 1995; Cohen and [11] The dispersion exponents used to describe the
Kreitzberg, 1997; Stohl, 2001; Huber et al., 2001; Colette mixing processes (g, l and b) are calculated from the
and Ancellet, 2006] particles dispersion predicted by FLEXPART. The results
of the model may be affected by the assumptions made in
1 Drt
t; r0 ln  1 4 implementing the Langevin equations to account with the
t Dr0 stochastic fluctuations. However, as the aim of the paper is
not to propose a general scaling law for the quantification
where Dr(0) is an arbitrary horizontal distance between of the tropospheric mixing rather than discussing its vari-
particles at the initial point and Dr(t) is their distance after a ability, neither the implementation of the Langevin equa-
time t (counted forward) accounted as the standard deviation tions, nor the resolution of the meteorological data
distance of particles from the mean position of the plume; influence to the absolute dispersion indexes are considered
and t is the characteristic timescale of the dispersing eddies, in the analysis. Langevin equations are stochastic differ-
called typical eddy turnover time. If at any time l > 0, we ential equations describing Brownian motions in a potential
say that the system is chaotic [Pierrehumbert and Yang, field. Therefore, it is expected that the adjustment of the
1993]. Large l values indicate rapid filamentation and a dispersion of particles in time fits well a potential equation.
loss of coherency, and thus distorted and filamented retro-
plumes; whereas low values indicate high coherency and 3.2. Atmospheric CO2 Measurements
slow filamentation, corresponding to laminar mixing zones. [12] Atmospheric CO2 measurements were carried out
Moreover, the finitetime Lyapunov exponent is a tool for periodically in situ at LMU (41.60N 1.10W; ground level:
characterizing the chaotic properties of the flow [Joseph and 570 masl) within the ICARO Aircraft Program in Spain. LMU
Moustaoui, 2000]. is situated in the central part of the Ebro watershed, charac-
2.2.3. Increase in the Potential Surface Influence Area terized by a continental Mediterranean climate with semiarid
Over Time (b) features. The surrounding area is dominated by nonirrigated
[9] The increase in the potential surface influence area over arable lands, but along the River Ebro and its irrigation
time (b) describes the mixing processes in a layer adjacent channels, permanent irrigated lands are found. Moreover,
to the ground. The potential surface influence (PSI), defined occasional vineyards, sclerophyllous vegetation and conifer-
as the integrated time over which particles reside within the ous forests are also encountered. The city of Zaragoza
lowest 300 m above ground, increases its area over time (667,000 inhabitants) is situated 20 km east, downstream.
according to a power law: [13] Vertical profiles of CO2 mixing ratios were carried
t  t 5 out from 900 masl up to 4200 masl using a portable, non
dispersive, infrared, continuous measurement system developed
where a is the accumulated area of PSI at each time (t counted by AOS Inc. The mean precision of the system in laboratory
forward) and b is the exponent which characterizes the rate of conditions is s = 0.11 ppmv. Under flight conditions, the
increase over time. The lower b is, the slower the PSI area mean precision is estimated at s = 0.23 ppmv. For further
grows; atmospheric parcels spend more time in contact with details of the experimental procedure see Font et al. [2008].
the surface layer, and therefore are more influenced by sur-
face CO2 fluxes. Conversely, higher b are associated with 3.3. Study Setup
faster plumes that quickly escape from surface fluxes. [14] A climatology of weekly simulations for 2006 con-
sisting of 96h FLEXPART backward simulations centered
3. Experimental Design at LMU at 12 UTC at different altitudes (600, 1200, 2500
and 4000 masl) was computed. Briefly, a total of 204
3.1. Lagrangian Particle Dispersion Model Setup simulations (51 simulations for each altitude) were ana-
[10] The Lagrangian Particle Dispersion Model (LPDM) lyzed and they are discussed in section 4.1. Three vertical
FLEXPART v 6.2.4 [Stohl et al., 2005] was used to calculate CO2 profiles measured at LMU during 2006 are consid-
the position of infinitesimal air masses or particles which ered: 22nd February, 25th May, and 13th October.
arrive at LMU, Spain, at different altitudes within a vertical Atmospheric layers are identified by means of the shape of
profile. FLEXPART was driven by the global modellevel the vertical profile of the CO2 mixing ratios, and by the
data NOAANCEPGFS (National Oceanic and Atmospheric CO2 mixing ratios vertical variance and vertical gradient.
Administration National Centers for Environmental Predic- These layers matched both changes in the physical atmo-
tion Global Forecast System) with a horizontal resolution of spheric parameters of the local atmosphere on the day of the
1 1, 26 vertical layers, and a time resolution of 3 h (anal- flight and changes in the dispersive and mixing exponents
yses at 00, 06, 12, 18; forecasts at 03, 09, 15 and 21 UT). Ten obtained from 4day simulations (section 4.2).
thousand particles were released at defined points in a vertical
profile and transported 96 h back in time by the resolved 4. Results and Discussion
winds and by parameterized subgrid turbulence and convec-
tion motions. The release points are defined as small boxes 4.1. Annual and Seasonal ABL Height Together With
centered at LMU with dimensions 0.01 0.01 100 m. The Dispersive and Mixing Regimes
outputs were the position (latitude, longitude and altitude) of [15] Figure 1 shows the annual and seasonal ABL height
each of the 10000 particles tracked back in time, at 3h inter- distribution for 2006 at Zaragoza airport (LEZG) at 12 UT

3 of 10
D19308 FONT ET AL.: HORIZONTAL MIXING DISCONTINUITY D19308

than the one from radiosounding as in accordance with pre-


vious studies [Gerbig et al., 2008]. As vertical profiles were
sampled just after noon, the ABL height is not expected to be
significantly different from 12 UT radiosounding.
[16] With regard to the dispersive parameters, the growth
of the RMS distance in time (g) at all altitudes at LMU in
2006 is fitted by a power law function with correlation
coefficients ranging from r = 0.970 to r = 0.999 (a < 0.001)
as seen in Table 1. The g values were lower during MAM and
JJA (mean g of 1.15 and 1.10, respectively) and higher during
DJF and SON (1.26 and 1.25, respectively). As we move up
the vertical profile, larger g values were found, which indicate
enhanced dispersive regimes. At 600 and 1200 masl, the dis-
persive regime was ballistic during all seasons. Conversely, at
2500 and 4000 masl, ballistic regimes were only found in
MAM and JJA, while the RO law held in DJF and SON.
[17] Results agree with the study reported by Huber et al.
[2001] of worldwide climatology tropospheric turbulent dis-
persion. Power law behavior prevails at middle and high alti-
tudes during all seasons, yielding g values close to 1 (ballistic
dispersion) after 10 days for particle release along isentropic
surfaces (315K and 330K). Enhanced RO mixing regimes (g =
1.251.5) are found in winter in the northern hemisphere
when the wavethrough pattern is most pronounced, either
linked to the major mountain ranges or upstream of the storm
tracks. Maryon and Buckland [1995] remarked that stretching
and folding of plumes are enhanced in active cyclonic sys-
tems, and this leads to an increase in the mean separation of
particles and thus of g. The climatology study of tropospheric
ozone layers by Collete and Ancellet [2006] shows that
horizontal stretching is slightly more efficient in winter,
which is probably due to faster westerlies in the Atlantic
Figure 1. (a) ABL height boxplots at Zaragoza Airport for Ocean induced by a stronger latitudinal thermal gradient. The
2006 at 12 UT calculated using the bulk Richardson number larger dispersive exponents at LMU may be related to more
method, flagging the bulk Richardson critical number at frequent active cyclones, Atlantic in DJF and Mediterranean
0.25. (b) ABL height boxplots at LMU site for 2006 at in SON, in comparison to other seasons. Moreover, lower
12 UT from the NOAANCEPGFS meteorological archive. g values are found at low altitudes of the vertical profile as
Annual, DJF (December, January, and February), MAM topography (the Pyrenees mountain range) constrains the
(March, April and May), JJA (June, July and August) and mean separation of particles; or lighter winds prevent the
SON (September, October and November) boxplots are dispersion of air masses.
shown. The central and dotted line are the median (q2) and [18] For the increase in the PSI area over time (b; Table 1),
the mean values, respectively; the bottom and top edges of the power law fit has correlation coefficients ranging from r =
the box are the 25th (q1) and 75th (q3) percentiles respec- 0.987 to r = 0.999 (a < 0.001). DJF exhibited the largest
tively; the whiskers extend to the most extreme data points values (mean value of 1.7), followed by SON (1.5), MAM
not considered outliers; and outliers are plotted individually and JJA (both 1.4). b values were similar at 600, 1200 and
as dots whenever data is larger than q3 + 1.5(q3 q1) or smal- 2500 masl (mean value of 1.4) but increased considerably at
ler than q1 1.5(q3 q1). 4000 masl (mean value of 1.9). Results show that the growth
of the PSI area was faster in winter at high altitudes.
as calculated using equation 1 (Figure 1a); and the ABL [19] The finite Lyapunov exponents (l) calculated here by
height given by the NOAANCEPGFS meteorological averaging the 3h l for each 4day simulation, were all
model at LMU at 12 UT for 2006 (Figure 1b) (DJF: positive (Table 1). This indicates that trajectories separated
December, January and February; MAM: March, April and very fast and the system was sensitive to initial conditions
May; JJA: June, July and August; and SON: September, (chaotic advection). Seasonally, l values were slightly
October and November). The median ABL height calculated larger during the warm months (MAM and JJA, with l of
from radiosoundings (and NCEPGFS model) was 1136 masl 0.20 and 0.21 h1, respectively) with respect to the cooler
(1446 masl). The ABL was deeper during warm months and ones (DJF and SON, with l of 0.18 and 0.19 h1, respec-
shallower during cooler ones. MAM and JJA show median tively). Increasing height reduces filamentation and chaotic
values of 1300 masl (1600 and 1900 masl, respectively); in mixing, with more coherent and sheared air masses arriving
DJF 680 masl (770 masl) and 850 masl in SON (1200 masl). at the measurement site. The average l reduction in altitude
These results agree with the seasonality of sensible heat flux, (from the lowest level to the highest) was 9%, 13%, 10%
which is the main factor affecting ABL height. Globally, the and 12% for DJF, MAM, JJA and SON. These results agree
ABL height from the meteorological model is 30% larger with those reported by Colette and Ancellet [2006], where

4 of 10
D19308 FONT ET AL.: HORIZONTAL MIXING DISCONTINUITY D19308

Table 1. Dispersive Exponent g and Type of Dispersion, Lyapunov Exponent l, and Dispersive Exponent b for
LMU for the 2006 Backward Lagrangian Dispersion Weekly Climatology Setup on a Seasonal Basis (DJF, MAM,
JJA, and SON) at Altitudes of 600, 1200, 2500, and 4000 masl
Altitude Season Dispersive Exponent (g) Dispersive Type Lyapunov Exponent (l) Dispersive Exponent (b)
610 masl DJF 1.12 Ballistic 0.199 1.39
MAM 1.15 Ballistic 0.216 1.40
JJA 1.09 Ballistic 0.224 1.33
SON 1.15 Ballistic 0.209 1.38

1210 masl DJF 1.17 Ballistic 0.197 1.41


MAM 1.14 Ballistic 0.217 1.41
JJA 1.01 Ballistic 0.225 1.34
SON 1.19 Ballistic 0.208 1.40

2500 masl DJF 1.36 Ballistic/RO 0.179 1.70


MAM 1.07 Ballistic 0.209 1.42
JJA 1.09 Ballistic 0.221 1.35
SON 1.27 Ballistic/RO 0.190 1.47

4000 masl DJF 1.37 Ballistic/RO 0.183 2.33


MAM 1.23 Ballistic 0.190 1.64
JJA 1.20 Ballistic 0.204 1.52
SON 1.40 Ballistic/RO 0.186 1.78

streamer formation is enhanced in air masses whose origin is [21] The existence of transport barriers related to different
in the ABL rather than in the free troposphere (FT). Lyapunov exponents has been pointed out in several studies
[20] As simulations were projected backward in time, [Pierrehumbert and Yang, 1993; Yang and Liu, 1997; Huber
l values became more equal in the vertical (see Figure S1a in et al., 2001] such as the tropicalextratropical one. This
the auxiliary material).1 At very early stages after plume barrier separates the slowly dispersing, weakly chaotic east-
release, there was quite rapid stretching and deformation of erly tropics from the highly dispersive, more strongly chaotic
the tracer cloud by atmospheric flow, with typical dispersive westerly extratropics [Huber et al., 2001]. Mixing regions
eddy turnover times of 0.9 h. Whereas, as t, l tended to indicate contiguous zones of large l, while transport barriers
a constant value as a consequence of the ergodicity of the to mixing are indicated by regions of small or vanishing l.
flow [Pudykiewicz and Koziol, 1998]. The eddy turnover Disjoint chaotic regions, such as might occur when some
times for particles 96 h from the receptor are 7 h. Fila- fluid is trapped within a coherent eddy, are characterized by
mentation occurred at the beginning of the simulations, when different l values. At LMU, a transport discontinuity appeared
eddy turnovers were fastest. In this study, the 96h mean to be set in the vertical, as l values were lower at 2500 masl
Lyapunov exponents were used for reasons of coherence than at 600 and 1200 masl, even though all l values were
with the other dispersive exponents (g and b). Nevertheless, positive, indicating that mixing followed chaotic dynamics at
the finite Lyapunov exponent values at the very beginning all the altitudes considered (Figure S1a in the auxiliary
of the simulationsup to 24 h or 18 h marked out the material). Simulations centered at 600 and 1200 masl experi-
96h mean l. The Lyapunov exponent is also interpreted as enced more chaotic mixing dynamics during the few hours
quantifying the loss of information from a system. At each previous to the arrival. At 2500 and 4000 masl, higher vertical
time step we lose some information, so that in the end there is stability and thus larger wind shear lead to enhanced dispersion
no reliable information and the system dissipates. Here, the of air masses above it.
loss of information relates to the fact that the consistency of
air masses diminishes with time and no coherent masses are 4.2. Vertical Profiles
found. The time the consistency of air masses is lost is cal- [22] For the three vertical profiles of CO2 mixing ratios
culated. Taken the mean and the variance values from the measured at 12 UT at LMU throughout 2006 (22nd Feb-
Lyapunov exponents at each 3h time step for all simulations ruary, 25th May and 13th October), changes in the CO2
for 2006 at each altitude, it is seen that the Lyapunov mixing ratios vertical variance (s2), expressed in ppmv2, and
exponent decayed over time following power law behavior in the CO2 vertical gradient (ppmv/m)2 are related to varia-
with a power exponent of 4/5 (Figure S1b in the auxiliary tions in the physical atmospheric structure the day of the
material; note that only 600 masl is shown). When the var- sampling (relative humidity (%), potential temperature (K),
iance is greater than the difference between two mean values wind direction (degrees) and speed (m/s), and the ABL height
from consecutive times is the time the air masses loss their (masl) computed by the bulk Richardson number method
consistency: 24 h at both 600 and 1200 masl, and 18 h at both from the WMO Zaragoza Airport 12UT radiosounding data;
2500 and 4000 masl. Therefore, the influence of consistent and extracted from the NCEPGFS meteorological model
air masses arriving to the measurement site is confined to the interpolated at the mean time when the profile was sampled),
day before sampling below 2500 m; whereas at higher alti- shown in light gray shading in Figures 2a2c. Changes
tudes, the influence is confined only to the previous 18 h. related to the transport of coherent backplumes are dealt with
changes in the mixing and dispersive regime exponents (g,
1
Auxiliary materials are available in the HTML. doi:10.1029/ l, b) obtained from the LPDM FLEXPART, and indicated
2010JD014060. with dark gray shading in Figures 2a2c.

5 of 10
D19308 FONT ET AL.: HORIZONTAL MIXING DISCONTINUITY D19308

Figure 2a. Measured vertical profiles of CO2 mixing ratios 1Hz data (dots) and mean CO2 concentration
for the altitudes where LPDM simulations where centered (crosses) (ppmv); Var: CO2 vertical variance
(ppmv2) within 200 m bins. The central altitude where the variance is calculated differs 100 m between
two consecutive calculations; Grad: 200mbin average of the squared CO2 vertical difference divided
by the altitude difference (pppmv2/m2) between two consecutive measurements. The central altitude where
the gradient is calculated differs 100 m between two consecutive calculations; potential temperature (, K)
and relative humidity (HR, %); wind speed (WS, m/s) and wind direction (WD, degrees) from Zaragoza
airport radiosoundings; mean Lyapunov exponent obtained from 96 h of simulation (l, h1); dispersive
exponent (g); bexponent of the power law fit of the PSI area growth in time. Horizontal dotted line shows
the ABL height (masl) calculated using the bulk Richardson number method at LMU for 22nd February
2006 from radiosounding at Zaragozas airport (ABL RS); long dashed line shows the ABL height (masl)
extracted from the GFSNCEP model, interpolated at the mean time when vertical profile was sampled
(ABL GFS). Light gray shading denotes changes either in the CO2 concentration, variance or gradient
related to the atmospheric physical properties; dark gray shading relates to changes of the mixing and dis-
persive regime exponents (g, l, b).

[23] Generally speaking, in the first meters of the profiles However, at lower altitudes is also possible to detect the
up to 2500 masl, changes on the CO2 concentration, vari- advection of coherent atmospheric layers as seen in Figure 2a,
ance and vertical gradient are related to the changes of either coincidental by a change of the dispersive exponents. A
ABL height or meteorological variables. It should be detailed description of the vertical profiles and the layers
noticed that there is a change of the mixing conditions of detected is done in the following sections.
transported air masses in two of the three vertical profiles 4.2.1. Profile on 22nd February 2006
shown (25th May and 13th October) at 2500 and 2200 masl, [24] The atmospheric CO2 concentration for this day indi-
respectively, in accordance with the presence of the HMD, cated different vertical layers ranging from 388 to 392 ppmv
discussed before in Section 4.1. CO2 changes recorded at (Figure 2a). The mean time when the profile was sampled was
higher altitudes (from 2500 up to 4000 masl) are mostly at 13 UT.
related to the transport of upwind air masses as indicated by [25] i. An increase in the absolute CO2 vertical gradient
changes in the mixing and dispersive regime exponents. value (0.05 ppmv2/m2) concurred at the ABL height

6 of 10
D19308 FONT ET AL.: HORIZONTAL MIXING DISCONTINUITY D19308

Figure 2b. Same as Figure 2a but for 25th May 2006.

derived from radiosounding at 12 UT (945.5 masl) and at centration between 2800 and 3000 masl of around 2 ppmv
1290 masl (NCEPGFS). and with the variance peak at 2800 masl of 0.1 ppmv2.
[26] ii. Between 1500 and 2100 masl, decreases in both [29] vi. The next layer (above 3000 and up to 3400 masl)
relative humidity (of 17%) and potential temperature (of 6) begins with a change of the relative humidity (decrease of
were matched by a continuous decrease in CO2 concentra- 50%) and b exponent (b2 at 3000 masl while b4 at
tion of roughly 2 ppmv that caused a peak in CO2 variance 3400 masl) and finishes with a peak of the CO2 mixing
of 0.17 ppmv2. ratio (393 ppmv at 3400 masl) linked to an increase in
[27] iii. Between 2200 and 2700 masl, the dispersive the CO2 variance (0.5 ppmv2) and vertical gradient
exponent changed from ROtype dispersion (g1.4) to a (0.06 ppmv2/m2). Such thin layer detected at 3400 masl is
ballistic dispersive regime (g1), and the Lyapunov expo- not expected to be reproduced by the LPDM with the coarse
nent increased from 0.22 h1 to 0.23 h1. These changes meteorological input used in the current study.
were accompanied by an increase in the CO2 mixing ratio 4.2.2. Profile on 25th May 2006
(from 390 to 393 ppmv), with low vertical variance but [30] The profile started at 13UT and finished 20 min later.
ended with a gradient peak at 2700 masl (0.05 ppmv2). How- The CO2 concentration ranged from 383 ppmv at 1000 masl
ever, the 96h PSI maps showed no changes in the origin of up to 385 ppmv at the highest level sampled, 4000 masl
air masses between 2200 and 2700 masl (see Figure S2 in (Figure 2b). Four discontinuities are described in this profile.
the auxiliary material): air masses accumulated the surface [31] i. A peak in both the CO2 variance (more than
influences from Western Europe and the northeast Atlantic 0.15 ppmv2) and CO2 gradient (0.2 ppmv2/m2) concurs
Ocean at all altitudes. The increase in CO2 concentration in with an increase of 53% in the wind speed between 900
this layer is due to the transport of air masses which kept the and 1400 masl.
signal of the Western Europe surface fluxes down to the [32] ii. The ABL height set between 1786 masl (radio-
measurement site as the g and l values indicate. sounding) and 1880 masl (NCEPGFS) is coincidental with
[28] v. A decrease of 50% of the relative humidity in the a modest variance value of 0.1 ppmv2.
atmosphere was associated with the decrease in CO2 con-

7 of 10
D19308 FONT ET AL.: HORIZONTAL MIXING DISCONTINUITY D19308

Figure 2c. Same as Figure 2a but for 13th October 2006.

[33] iii. Between 2000 and 2300 masl, drier air (with a 2 to 2.8, associated with the advection of coherent air masses,
decrease of 64% in the relative humidity) coincides with which concurs with an increase in variance (0.3 ppmv2).
CO2 gradient of 0.15 ppmv2/m2 at 2300 masl. PSI maps for this day (see Figure S4 in the auxiliary material)
[34] iv. The changes in the three dispersive indexes present a change of surface influence above 2300 masl: while
(l from 0.22 h1 to more coherent air masses with l values of below this altitude PSI accumulates the northeast Atlantic
0.15 h1; b from 1.5 to 0, associated with a lack of surface and Mediterranean surface influence, above this altitude
influence at high altitudes (see Figure S3 in the auxiliary PSI accumulates the south of the Iberian Peninsula on the
material); and g from 1.5 (RO mixing) to a less dispersive trail of dense urban areas. The change of the PSI associated
value of 1, corresponding to a ballistictype dispersion) fits with the transport of air masses in a laminar flux leaded to
in the variance increase from 3000 up to 4000 masl with a an increase of the measured CO2 concentration between
peak value of 0.2 ppmv2 at 2800 masl. 2000 and 2500 masl.
4.2.3. Profile on 13th October 2006
[35] The profile was sampled around 11.20 UT. CO2 5. Discussion
vertical profile (Figure 2c) ranges from 381 ppmv to
382 ppmv. At 1000 masl CO2 mixing ratio is 381 ppmv, [38] In the three vertical profiles of atmospheric CO2
increasing up to 382 ppmv in the middle layers (2000 mixing ratios measured at LMU that had been shown (22nd
2500 masl) and decreasing again to 380 ppmv at high February, 25th May, and 13th October 2006), the vertical
altitudes (above 3500 masl). distribution of CO2 concentration is related to the physical
[36] An increase of the variance (0.15ppmv2) at 1200 masl structure of the atmosphere and to the transport of coherent
is related to the presence of the ABL height (1020 masl as air masses downstream to the measurement site. The trans-
seen by the NCEPGFS model). port of coherent plumes is described in terms of g, l and b.
[37] Between 2000 and 2500 masl, there is a change in the The CO2 mixing ratios result from the vertical atmospheric
mixing and dispersive exponents: g increases from 1.4 to 1.8; structure on the day of the measurements as well as mixing
l decreases from 0.23 h1 to 0.18 h1 and b increases from and historical dispersive processes of air masses transported
to the site.

8 of 10
D19308 FONT ET AL.: HORIZONTAL MIXING DISCONTINUITY D19308

[39] As seen in both the climatology and the individual [42] In the individual vertical profiles considered, the
vertical profiles studied, below 4000 masl there are at least various layers indicated are not necessarily static and might
two discontinuities with regard to mixing processes: the change in time and space in relation to enhanced or pre-
ABL height, related to vertical motions driven by local vented chaotic mixing conditions.
convection; and the horizontal mixing discontinuity (HMD),
related to enhanced or prevented horizontal mixing pro- 6. Conclusions
cesses. Therefore, in overall we can distinguish two main
discontinuities in the low troposphere: the ABL and FT, [43] At LMU, horizontal transport and mixing in the low
situated round 1500 masl (median ABL height at LMU from troposphere (below 4000 masl) is characterized by a super-
NCEPGFS model; see Section 4.1); and the transport and diffusive regime (g > 1/2) and chaotic mixing (l > 0)
stirring discontinuity, situated above 2500 masl, as seen in throughout the year, as calculated from simulations made by
the climatology study and in the individual profiles. Below LPDM FLEXPART. Different discontinuities between
the ABL, surface fluxes are vertically mixed by means of atmospheric layers and mixing processes are recorded in the
turbulence, whether free (convection) or mechanical (shear) vertical profiles of atmospheric CO2 mixing ratios, leading
generated; above it, in the FT, the surface fluxes need more to specific shapes (changes in the distribution of CO2) and
time to spread, typically, 1 or 2 months [Gerbig et al., 2003]. sizes (width of layers).
The horizontal mixing discontinuity separates two layers [44] The vertical propagation of CO2 surface fluxes within
with different horizontal mixing conditions. Below it, fila- the ABL and the transport of coherent air masses down to the
mentation and small eddy turnover lead to increased vertical measurement site determine the vertical CO2 distribution in
mixing. Above it, lower Lyapunov values (l) values the low troposphere (above 4000 masl). Vertical profiles
(<0.19h1) and high dispersive exponents (g) values (>1.25) contain information about the physical atmospheric structure
indicate that shear transport is more frequent. on the day of the survey and the historical mixing processes
[40] At LMU, the ABL and the HMD do not occur at the of transported air masses. In order to reproduce the high
same height (Figures 2a2c). Therefore, averaging the CO2 resolution of the vertical distribution of CO2 mixing ratios,
concentration over either the PBL or FT (i.e., used to vali- the low troposphere needs to be sliced into finer layers, rather
date the resulting effect of source and sink intensities and than considering only two (ABL and FT). Moreover, the
vertical mixing in atmospheric transport models; to estimate mean CO2 concentration above the ABL height still reflects
background concentration, etc.) does not reflect the struc- physical and chemical atmospheric structure that does not
tured horizontal mixing processes taking place. Further- correspond to a wellmixed atmospheric concentration.
more, the discretization of the atmosphere into fixed and [45] This paper highlights the importance of understand-
static layers does not capture the complexity of transport and ing the horizontal transport in evaluating the transport of
mixing of retroplumes, as seen in Section 4.2. In the vertical possible plumes derived from the Lagrangian dispersion
profiles shown (Figures 2a2c), the CO2 mixing ratio does models when analyzing the vertical distribution of CO2 at a
not become constant until 3500 masl. Above the ABL there single location and time. Gradients above the ABL should
is CO2 variability that should be taken into account when- be considered in any model or data comparison. This anal-
ever a data selection of profiles is done. ysis suggests that below 4000 masl there is significant
[41] The dispersive exponents have been used to detect variability which reflects horizontal transport of plumes that
consistent atmospheric layers which have been transported must be considered in any analysis. The advection of plumes
keeping its coherency and compared to CO2 vertical gra- to the measurement site can be assessed by changes in the
dients. The area where air masses reside before arriving at upstream surface influence with height, together with the
the measurement site at different altitudes of the vertical dispersive exponent (g), the chaotic dynamics in mixing (l)
profiles does not always provide enough information of the and the increase rate of the PSI area over time (b) obtained
transport of coherent plumes while the dispersive exponents from fully backward Lagrangian simulations.
do. Tracing the dominant backtrajectory for four or five days
as done as usually, mixes and erases the information [46] Acknowledgments. We would like to thank all the people who
contained in the structure of the lower troposphere. The static make the Aircraft Measurement Program ICARO in Spain possible, espe-
image of the region where air masses have reside before cially M. A. Rodrguez and I. Pouchet. We also thank Toni Fontanet for
arriving at the measurement site retrieved by LPDMs does his advice and all the crew of Top Fly from Sabadell Airport (Silver, Oriol,
Ral and many others) and lvaro Lapetra from Zaragoza airport. This
not always provide such kind of information as it could be research was funded by the Spanish Ministry of Education and Science
the same at different altitudes of the vertical profile (i.e., in (project ICAROREN200306089 and ICAROIIGL12398: Reduction
the 22nd March profile). Thus, the transport of a coherent of Uncertainties in the peninsular CARbon balance by means of Oscillating
tropospheric transects).
plume can be retrieved by LPDM simulations for the dif-
ferent heights and examining the mixing and dispersive
historical processes. However, the simultaneous sampling of References
other anthropogenic tracer gases such as carbon monoxide Bakwin, P. S., P. P. Tans, B. B. Stephens, S. C. Wofsy, C. Gerbig, and
A. Grainger (2003), Strategies for measurement of atmospheric column
[Paris et al., 2008], sulphur hexafluoride, ozone, reactive means of carbon dioxide from aircraft using discrete sampling, J. Geo-
nitrogen [Huntrieser et al., 2005; Stohl et al., 2007] could phys. Res., 108(D16), 4514, doi:10.1029/2002JD003306.
provide also this kind of information. Lack of measurements Cohen, R. A., and C. W. Kreitzeberg (1997), Airstreams boundaries in
of other tracer gases, as in this study, leads to a need of numerical weather simulations, Mon. Weather Rev., 125, 168183,
doi:10.1175/1520-0493(1997)125<0168:ABINWS>2.0.CO;2.
studying the variability of CO2 through the backprofiles of Colette, A., and G. Ancellet (2006), Variability of the tropospheric mixing and
transport characteristics to pick out the transport coherent of streamer formation and their impact on the lifetime of observed ozone
plumes which keep the signal of upwind fluxes. layers, Geophys. Res. Lett., 33, L09808, doi:10.1029/2006GL025793.

9 of 10
D19308 FONT ET AL.: HORIZONTAL MIXING DISCONTINUITY D19308

Eresmaa, N., A. Karppinen, S. M. Joffre, J. Rsnen, and H. Talvitie determination of the mixing height, Atmos. Environ., 34(7), 10011027,
(2006), Mixing height determination by ceilometer, Atmos. Chem. Phys., doi:10.1016/S1352-2310(99)00349-0.
6, 14851493, doi:10.5194/acp-6-1485-2006. Shraiman, B. I., and E. D. Siggia (2000), Scalar turbulence, Nature, 405,
Font, A., J. A. Morgu, and X. Rod (2008), Atmospheric CO2 in situ 639646, doi:10.1038/35015000.
measurements: Two examples of Crown Design flights in NE Spain, Sicard, M., C. Prez, F. Rocadenbosch, J. M. Baldasano, and D. Garca
J. Geophys. Res., 113, D12308, doi:10.1029/2007JD009111. Vizcaino (2006), Mixedlayer depth determination in the Barcelona
Gerbig, C., J. C. Lin, S. C. Wofsy, B. C. Daube, A. E. Andrews, B. B. coastal area from regular lidar measurements: Methods, results and lim-
Stephens, P. S. Bakwin, and C. A. Grainger (2003), Toward constraining itations, Boundary Layer Meteorol., 119, 135157, doi:10.1007/s10546-
regionalscale fluxes of CO2 with atmospheric observations over a conti- 005-9005-9.
nent: 1. Observed spatial variability from airborne platforms, J. Geophys. Sidorov, K., A. Sogachev, U. Langendrfer, J. Lloyd, I. L. Nepomnjashiy,
Res., 108(D24), 4756, doi:10.1029/2002JD003018. N. Vygodskaya, M. Schmidt, and I. Levin (2002), Seasonal variability of
Gerbig, C., S. Krner, and J. C. Lin (2008), Vertical mixing in atmospheric greenhouse gases in the lower troposphere above the eastern European
tracer transport models: Error characterization and propagation, Atmos. Taga (Syktyvkar, Russia), Tellus, Ser. B, 54, 735748.
Chem. Phys., 8, 591602, doi:10.5194/acp-8-591-2008. Solomon, T. H., E. R. Weeks, and H. L. Swinney (1994), Chaotic advec-
Grimsdell, A. W., and W. M. Angevine (1998), Convective boundary layer tion in a twodimensional flow: Levy flights and anomalous diffusion,
height measurement with wind profilers and comparison to cloud base, Physica D, 76, 7084.
J. Atmos. Oceanic Technol., 15, 13311338, doi:10.1175/1520-0426(1998) Stephens, B. B., et al. (2007), Weak northern and strong tropical land car-
015<1331:CBLHMW>2.0.CO;2. bon uptake from vertical profiles of atmospheric CO2, Science, 316,
Huber, M., J. C. Mcwilliams, and M. Ghil (2001), A climatology of turbu- 17321735, doi:10.1126/science.1137004.
lent dispersion in the troposphere, J. Atmos. Sci., 58, 23772394, Stohl, A. (2001), A 1year Lagrangian climatology of airstreams in the
doi:10.1175/1520-0469(2001)058<2377:ACOTDI>2.0.CO;2. Northern Hemisphere troposphere and lowermost stratosphere, J. Geo-
Huntrieser, H., et al. (2005), Intercontinental air pollution transport from phys. Res., 106(D7), 72637279, doi:10.1029/2000JD900570.
North America to Europe: Experimental evidence from airborne mea- Stohl, A., S. Eckhardt, C. Forster, P. James, and N. Spichtinger (2002), On
surements and surface observations, J. Geophys. Res., 110, D01305, the pathways and timescales of intercontinental air pollution transport,
doi:10.1029/2004JD005045. J. Geophys. Res., 107(D23), 4684, doi:10.1029/2001JD001396.
Joseph, B., and M. Moustaoui (2000), Transport, moisture and rain in a Stohl, A., O. R. Cooper, and P. James (2004), A cautionary note on the
simple monsoonlike flow, J. Atmos. Sci., 57, 18171838, doi:10.1175/ use of meteorological analysis fields for quantifying atmospheric mixing,
1520-0469(2000)057<1817:TMARIA>2.0.CO;2. J. Atmos. Sci., 61, 14461453, doi:10.1175/1520-0469(2004)061<1446:
Lin, J. C., C. Gerbig, B. C. Daube, S. C. Wofsy, A. E. Andrews, S. A. Vay, ACNOTU>2.0.CO;2.
and B. E. Anderson (2004), An empirical analysis of the spatial variability Stohl, A., C. Forster, A. Frank, P. Seibert, and G. Wotawa (2005), Technical
of atmospheric CO2: Implications for inverse analyses and spaceborne note: The Lagrangian particle dispersion model FLEXPART version 6.2,
sensors, Geophys. Res. Lett., 31, L23104, doi:10.1029/2004GL020957. Atmos. Chem. Phys., 5(9), 24612474, doi:10.5194/acp-5-2461-2005.
Lloyd, J., et al. (2002), A trace gas climatology above Zotino, central Stohl, A., C. Forster, H. Huntrieser, H. Mannstein, W. W. McMillan,
Siberia, Tellus, Ser. B, 54, 590610. A. Petzold, H. Schlager, and B. Weinzierl (2007), Aircraft measurements
Maryon, R. H., and A. T. Buckland (1995), Tropospheric dispersion: The first over Europe of an air pollution plume from Southeast Asiaaerosol and
ten days after a puff release, Q. J. R. Meteorol. Soc., 121, 17991833, chemical characterization, Atmos. Chem. Phys., 7, 913937, doi:10.5194/
doi:10.1002/qj.49712152802. acp-7-913-2007.
Menut, L., C. Flamant, J. Pelon, and P. H. Flamant (1999), Urban boundary Sturm, P., M. Leuenberger, J. Moncrieff, and M. Ramonet (2005), Atmo-
layer height determination from lidar measurements over the Paris area, spheric O2, CO2, and d 13C measurements from aircraft sampling over
Appl. Opt., 38, 945954, doi:10.1364/AO.38.000945. Griffin Forest, Perthshire, UK, Rapid Commun. Mass Spectrom., 19,
Morille, Y., M. Haeffelin, P. Drobinski, and J. Pelon (2007), STRAT: An auto- 23992406, doi:10.1002/rcm.2071.
mated algorithm to retrieve the vertical structure of the atmosphere from Tans, P. P., and D. W. R. Wallace (1999), Carbon cycle research after
singlechannel lidar data, J. Atmos. Oceanic Technol., 24, 761775, Kyoto, Tellus, Ser. B, 51, 562571.
doi:10.1175/JTECH2008.1. Tans, P. P., I. Y. Fung, and T. Takahashi (1990), Observational constraints
Ottino, J. (1989), The Kinematic of Mixing: Stretching, Chaos and Trans- on the global atmospheric CO 2 budget, Science, 247, 14311438,
port, Cambridge Univ. Press, New York. doi:10.1126/science.247.4949.1431.
Paris, J.D., et al. (2008), The YAKAEROSIB transcontinental aircraft Wang, J.W., A. S. Denning, L. Lu, I. T. Baker, K. D. Corbin, and K. J.
campaigns: New insights on the transport of CO2, CO and O3 across Davis (2007), Observations and simulations of synoptic, regional and
Siberia, Tellus, Ser. B, 60, 551568. local variations in atmospheric CO2, J. Geophys. Res., 112, D04108,
Pierrehumbert, R. T., and H. Yang (1993), Global chaotic mixing on isen- doi:10.1029/2006JD007410.
tropic surfaces, J. Atmos. Sci., 50, 24622480, doi:10.1175/1520-0469 Yang, H., and Z. Liu (1997), The threedimensional chaotic transport and
(1993)050<2462:GCMOIS>2.0.CO;2. the great ocean barrier, J. Phys. Oceanogr., 27, 12581273.
Pudykiewicz, J. A., and A. S. Koziol (1998), An application of the theory
of kinematics of mixing to the study of tropospheric dispersion, Atmos. A. Font, J.A. Morgu, and X. Rod, Institut Catal de Cincies del Clima,
Environ., 32(24), 42274244, doi:10.1016/S1352-2310(98)00181-2. Dr Trueta, 203, 3a planta, E08005 Barcelona, Spain. (afont@ic3.cat)
Seibert, P., F. Beyrich, S.E. Gryning, S. Joffre, A. Rasmussen, and P. Tercier
(2000), Review and intercomparison of operational methods for the

10 of 10

Potrebbero piacerti anche