Chemical Geology 341 (2013) 102–109

Contents lists available at SciVerse ScienceDirect

Chemical Geology
journal homepage: www.elsevier.com/locate/chemgeo

An innovative method for continuous measurement of soil CO2 flux

Sofia De Gregorio ⁎, Marco Camarda, Santo Cappuzzo, Sergio Gurrieri
Istituto Nazionale di Geofisica e Vulcanologia, sezione di Palermo, via Ugo La Malfa 153, 90146 Palermo, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Herein, we present a method for continuous measurement of soil CO2 flux that is completely new and distinct
Received 5 October 2012 from existing instruments. The foremost difference is that instead of using an infrared gas analyser (IRGA),
Received in revised form 14 January 2013 the new device measures soil CO2 flux by means of a simple pressure sensor, measuring pressure transients
Accepted 22 January 2013
inside a closed polymeric tube inserted into the soil. This allows continuous measurements even in soil
Available online 29 January 2013
placed in environments that could potentially damage IRGA. In addition, due to the innovative operating
Editor: D.B. Dingwell principle, measurements of soil CO2 flux can be effortlessly performed also in strongly harsh weather condi-
tions. Theoretical equations were derived for calculating soil CO2 flux solely using measured transient values.
Keywords: The reliability of the equations was rigorously tested with a variety of experiments. Continuous measure-
Soil CO2 flux measurements ments over four months, acquired in a high-emission area on the Island of Vulcano, compared favourably
Continuous monitoring with the data obtained using an established method.
Polymeric membranes © 2013 Elsevier B.V. All rights reserved.
Methods of measurement

1. Introduction most used for extensive measurement is eddy covariance analysis

The emission of CO2 in natural environments is a phenomenon
that has always aroused great interest in the scientific community.
Over the years, many studies have investigated these emissions
from a variety of angles: evaluation of global CO2 emissions, ecologi-
cal and agriculture investigations (Raich and Schlesinger, 1992;
Batjes, 1996; Atkin et al., 2000; Shi et al., 2006; Fischer, 2008; Wang
et al., 2011); gas hazards (Beaubien et al., 2003; Hernández Perez et
al., 2003; Carapezza et al., 2011); volcano surveillance and geochem-
ical explorations (Chiodini et al., 1996; Gerlach et al., 2001;
Hernández et al., 2001; Faber et al., 2003; Granieri et al., 2006;
Bergfeld and Evans, 2011; Federico et al., 2011; Camarda et al.,
2012). In addition to CO2 emitted directly from volcanoes, CO2 of
natural origin is released into the atmosphere mainly through soil,
and as a result, most studies have focused on soil CO2 flux measure-
ments. These measurements have been performed with diverse
methods according to the aim of the research. Some methods perform
extensive measurements of CO2 emissions over a large area, while
others measure soil CO2 flux at a determinate point. Extensive
measurements are often used in research primarily to evaluate an
ecosystem's carbon exchange and/or total budget emission. Single
point measurements are used, inter alia, in agricultural studies; in
geochemical exploration with the aim to highlight an anomalous
zone of emission and to link soil degassing with tectonics; and in vol-
canic surveillance to monitor soil CO2 flux over time. The method
⁎ Corresponding author. Tel.: +39 0916809439; fax: +39 091680944.
E-mail address: s.degregorio@pa.ingv.it (S. De Gregorio).
(Law et al., 1999; Pilegaard et al., 2001; Lewicki and Hilley, 2009).
Single point measurements of soil CO2 flux can be performed directly
by measuring concentration gradients in the soil (Baubron et al.,
1990; Russell et al., 1998) or indirectly using tailor-made methods.
The two indirect methods mainly used are the accumulation chamber
method (ACM) (Norman et al., 1992; Bekku et al., 1995; Chiodini et
al., 1998) and the dynamic concentration method (DCM) (Camarda
et al., 2006). The two methods use very different operating principles
and have advantages and disadvantages related to the research for
which they are utilised (Carapezza and Granieri, 2004; Camarda et
al., 2009). However, both employ an infrared gas analyser (IRGA). In
both methods, a sample of soil gas is introduced into the instrument,
potentially exposing it to damage by corrosive gas species or liquid
water; consequently, neither method is widely applicable, especially
for continuous measurements, in environments with sulphurous or
acidic species, or in conditions that may lead to condensation phe-
nomena inside the instrument. In addition, these two methods cannot
be used to take continuous measurements in environments with
harsh weather conditions such as high wind speed, recurrent intense
rain, or thick snow, without great difficulty.
On these grounds, we have developed a new measurement
method for soil CO2 flux that is able to efficiently perform continuous
measurements in the above-mentioned extreme environmental
conditions and can also serve as a valid alternative to existing mea-
surement methods in any other environment. The method couples a
specially designed instrument, named CADEMASO, to a theoretical
model, based on the gas transport laws through porous media
(Sahimi, 1995). Experimental tests aimed to calibrate CADEMASO
and to test the reliability of the method are reported, and long-term

0009-2541/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.chemgeo.2013.01.011
 S. De Gregorio et al. / Chemical Geology 341 (2013) 102–109
continuous field measurements are compared with a canonical
method.
2. The instrument: CADEMASO
2.1. Operating principles
The operating principle of CADEMASO is based on the measure-
ment of pressure transients inside a closed polymeric tube that are
generated by the various rates at which gases permeate through it
in an equilibration process. The features of this transient depend on
the gas species involved as well as their partial pressures. The gaseous
phase usually present in soil is composed mainly of a binary system of
CO2 and air. In such a binary system (Fig. 1A), during an exchange
process between gas inside and outside the tube, there is a valid
linear relationship between the rate of increase of Pt inside the tube
at the beginning of the processes and CO2 partial pressures (De
Gregorio et al., 2009):
 
∂P t ðt Þ A value of PinCO2 is kept negligible with respect PaCO2, a value of angular co-
¼ ðKpCO2 −Kpair ÞðP aCO2 −P inCO2 Þ
∂t t¼0 Vh
where [(∂Pt/∂t)t=0] is the first derivative of equation describing varia-
tion of total pressure inside the device at t = 0; A is the area of the mem-
brane surface, V is the internal volume of the tube, and h is membrane
thickness; KpCO2 and Kpair are the permeability coefficients of CO2 and
air, respectively; PaCO2 is the equilibrium partial pressure of CO2; and
PinCO2 is the initial partial pressure of CO2 inside the tube. The instru-
ment configuration (Fig. 1B) consists of a polytetrafluoroethylene
(PTFE) capillary tube connected at its ends to two-way electrovalves.
On the other end, one electrovalve is attached to a membrane pump.
The system is equipped with a pressure transducer interposed between
one end of the tube and the electrovalve. The use of PTFE as a polymeric
membrane ensures the generation of detectable pressure transients
A 1.8
Kpair<KpCO2
(PaCO -PinCO )
[(∂Pt/∂t)t=0]
Total pressure inside
1.6
PTFE tube (atm)
1.4
1.2
1.0
0 100
Time (minutes)
Fig. 1. Summary panels illustrating theoretical and practical characteristics of the CADEMASO device: (A) theoretical lines showing that in a matrix of CO2 and air, the value of
[(∂Pt/∂t)t = 0] is proportional to the difference between the partial pressure of CO2 inside (PinCO2) and outside (PaCO2) the PTFE tube. (B) Operative scheme and main components
of CADEMASO.
103
owing to the difference between KpCO2 and Kpair (De Gregorio et al.,
2005). The PTFE tube is inserted into the medium of interest, with the
rest of the components on the outside. The sampling sequence is divid-
ed into two steps: in the first step, the two electrovalves are kept open
with the pump active for 2 min. This operation has the aim to introduce
air into the tube and provides the disequilibrium condition necessary to
trigger the pressure transient. In the second step, the electrovalves are
closed, and the total pressure inside the device is measured every 30 s
for 10 min. The sampling sequence is completely automated using a
tailor-made electronic device, allowing continuous measurements.
2.2. Calibration coefficient
According to Eq. (1) for associating values of [(∂Pt/∂t)t=0] to PaCO2
values, we must determine the terms PinCO2 and [A / Vh(KpCO2 − Kpair)].
As shown in De Gregorio et al. (2009) the last term can be determined
experimentally by calculating [(∂Pt/∂t)t=0] for different (PaCO2 −
PinCO2) values and computing the angular coefficient of the straight
line fitting these values. Further, if in the calibration experiments the
ð1Þ
efficient that can be used as a calibration coefficient (C) for computing
PaCO2 directly from [(∂Pt/∂t)t=0] can be obtained. However, the obtained
C value is valid only for the specific features and conditions used in the
laboratory. In fact, as reported in formula (1), C depends on A/ Vh and
hence on PTFE tube geometry, volumes of fittings, and Kp values of
the gases. To use the device in the field, an understanding of the depen-
dence of C on the above-mentioned parameters was necessary.
2.3. Variability of calibration coefficient in response of different device
configurations
Knowledge of this variably geometric response is necessary be-
cause the configuration of the device used in the field will be unavoid-
ably dissimilar from the configuration adopted in the laboratory
gas air
B outlet inlet
electrovalve
2 2
pressure
transducer
0.8 atm
0.6 atm pump
electrovalve
0.4 atm
PTFE
0.2 atm tube
200
104 S. De Gregorio et al. / Chemical Geology 341 (2013) 102–109

during experimental calibration tests. Although the same PTFE tube
may be used, the fitting will always connect to the pressure transduc-
er and electrovalves in slightly different ways. According to Eq. (1),
provided the same PTFE tube is used, the C value will depend only
on the total volume. This means that we can compute the coefficients
for any configuration by knowing the volume and the volume of the
configuration used for experimental C coefficient determination. The
volume of each configuration can easily be computed by introducing
a known amount of gas and measuring the relative increase in inter-
nal pressure. A more accurate determination is recommended to per-
form measurements that use more than one amount of gas to ensure
that the eventual deformation phenomena induced by increases in
pressure are negligible.
2.4. Variations in temperature response
measurements. In Fig. 2, close to the series, the values of M (MTc)
are reported, computed using Eq. (2) and as Tr the calibration tem-
perature (Tc). The MTc values for the 3 series are exactly the same
showing as the influence of temperature on measurements is the
same for whichever concentrations. By using the coefficient MTc we
can compute the values of [(∂Pt/∂t)t = 0] reported at Tc according to
following equation:
Tg c ¼ Tg M =½ð1 þ MTc ðTm−TcÞ ð3Þ
where Tgc = [(∂Pt/∂t)t = 0] reported at Tc; TgM = [(∂Pt/∂t)t = 0] mea-
sured; and Tm is the measured temperature. We report [(∂Pt/∂t)t = 0]
as if it had been made at the calibration temperature to obtain the
real value of the CO2 concentration.

The C term includes the Kp values of the gaseous species involved.

2.5. Geometry and placement
Kp is the product of the diffusion coefficient and the solubility
coefficient. Because both of these processes are influenced by temper-
CADEMASO was specially constructed to suit practical and theo-
ature, Kp, and hence C, will depend on temperature as well. Therefore,
retical requirements. The electrovalves, pump and pressure transduc-
to obtain concentrations that are corrected for temperatures that
er were housed in a watertight box. The PTFE tube was 500 cm long,
diverge from the calibration temperature, we must correct the
and it was connected to the measurement system with tight fitting.
recorded values of [(∂Pt/∂t)t = 0]. Essentially, we must be able to re-
Routine acquisition was implemented with a calculation programme
port the recorded value of [(∂Pt/∂t)t = 0] at the calibration tempera-
that automatically computed the value of [(∂Pt/∂t)t = 0] after
ture. We therefore performed three sets of experiments at different
performing pressure measurements, with the aim of completely auto-
concentrations and varying the temperature from 15 °C to 35 °C.
mating data collection. The PTFE tube was wrapped around a
We choose to perform three sets of experiments to ensure that
tailor-made holder 33 cm long and inserted vertically into the soil
according to Eq. (1), the influence of temperature change does not
starting from the air/soil interface (Fig. 3). The geometry and the
depend on concentration, but only on the change in Kp. The results
placement of the PTFE tube allow the concentration values corre-
are shown in Fig. 2; a linear relationship exists between the two pa-
sponding to the integral of CO2 concentration along the shallower
rameters for all concentrations and over the entire temperature
soil layers to be recorded. CADEMASO was equipped with an
range investigated. Consequently, from the equations of linear best
fit, we can easily calculate a coefficient M for reporting the measured
value of [(∂Pt/∂t)t = 0] to a fixed temperature. The coefficient M is
computed as:

M ¼ m=ðmTr þ cÞ ð2Þ

CADEMASO
where m and c are respectively the slope and the intercept of the
equation of best fit, and Tr is the temperature at which to report the

35 z=0
75%
30 56% −L
1
z axis

25% y =38.35x+1856.1 CCDMS = − ∫ [CO ](z )dz

Soil CO2 concentration gradient
2
R2 = 0.997 L
25 0
MTc= 0.014
[(∂Pt/∂t)t=0] . 105

20 y =29.56x+1408.5 z=-L
R2 =0.995
MTc= 0.014
15

10
y =13.80x+642.5
MTc= 0.014 R2 =0.992
5

0 CO2 flux
10 20 30 40
T (°C)

Fig. 2. Results of experiments to determine the change in C values in response to tem-
perature variation. The experiments performed at 3 different CO2 concentrations de-
fined a linear trend, and the generated equations for these near-straight lines were
used, according to Eq. (2) and using as Tr the calibration temperature (Tc), to compute
the correction factor (MTc).
Fig. 3. Geometry and placement. CADEMASO is placed at the soil–atmosphere interface
and the PTFE tube is vertically inserted into soil. With this peculiar arrangement, the
CCDMS value is the integral of soil CO2 concentrations from −L depth to the soil surface.
Horizontal colour lines represent a typical concentration gradient generate by a CO2
flux with advective and diffusive component.
 S. De Gregorio et al. / Chemical Geology 341 (2013) 102–109 105

additional temperature sensor for measuring this parameter inside
the soil.
3. Theoretical model
3.1. Conceptual model
Gas transport through porous media can occur via two different
processes: diffusion and advection. Diffusion is the movement of mat-
ter from a region of high concentration to a region of low concentra-
tion. Advection is the transport of matter induced by a pressure
gradient. Usually, both pressure and concentration gradients coexist
in natural soil, so gas transport occurs due to a combination of
advection and diffusion processes, and the flux Ji of a generic gas
species can be expressed as the sum of its diffusive and advective
components (Sahimi, 1995):
J i ¼ J d;i þ J ad;i ¼ −D∇C i þ C i v
shown in Fig. 4B, it will take into account the nonlinearity of the
soil concentration profile. For even higher soil CO2 flux, the advective
component will become dominant (Fig. 4C), and the deepest part of
the PTFE tube will be placed in a soil layer with a concentration of
nearly 100%. In the last two cases, an increase in soil CO2 flux will
not produce a linear rise in soil CO2 concentration along the vertical
profile (i.e., CADEMASO values), implying that a change in the
CADEMASO values in the flux function will not be linear. Neverthe-
less, the most important aspect is that CADEMASO values will change
in response to flux variations under the prevalent advective regimen,
although not linearly. The above examples demonstrate conceptually
that measurements made by CADEMASO depend on soil layer CO2
concentrations and their arrangement. Diverse relationships between
CADEMASO measurements and flux values must exist in response to
different prevalent transport modalities. Because CADEMASO values
inherently depend on ν values, they may be used to somehow define
the regimen in place during the measurement and hence the holding
ð4Þ relation.

where Ci and Jd,i are the concentration and the diffusive flux of the gas
species, respectively, while Jad,i is the advective flux of the ith gas
species, D is the bulk diffusion coefficient and ν is the gas velocity.
From this theoretical basis, we investigated the expected relation-
ship between the values measured by CADEMASO and soil CO2 fluxes.
For low flux, ν values are rather low, and gas transport is almost total-
ly diffusive (Fig. 4A). In this condition, the concentration gradient is
practically linear, and the integral of concentration along a vertical
profile will nearly correspond to the mean of soil concentration
values at 33 cm depth and at soil air interface. Until the gradient of
soil CO2 concentrations is almost linear, the soil CO2 concentration
along the soil profile will increase linearly with increasing flux.
Hence, a linear relationship between CADEMASO measurements and
soil CO2 flux is expected. On the other hand, in the presence of higher
flux with a significant advective component, the change in concentra-
tion along the soil profile will not be more linear (Fig. 4B); the value
measured by CADEMASO will be higher than the mean of soil concen-
tration values at 33 cm depth and at soil air interface, because, as
3.2. Mathematical model
A rigorous theoretical expression linking the value of concentra-
tion measured by CADEMASO to the soil CO2 flux (JCO2) can be
deduced from the law governing gas transport through porous
media. CO2 transport through the soil is well described by the
so-called advective-diffusion model that takes into account CO2
transport due to pressure gradient as well as concentration gradient.
Hence, we began with the equation describing the variation of soil
CO2 concentration ([CO2]) under an advection diffusion regimen
along the z-axis from − L depth (negative downwards) to the soil
surface (Camarda et al., 2007). By means of this equation and the
appropriate calculations and substitutions, we obtain the relation be-
tween CO2 flux (JCO2) and CO2 concentration at − L depth ([CO2]−L):
ν½CO2 −L
J CO2 ¼ ð5Þ
1− expð−vL=DÞ

A CO2 (%) B CO2 (%) C CO2 (%)

0 50 100 0 50 100 0 50 100
0 0
CCD
C MS = 74%
C CD

CD
-10 M
Me

ea
M
S = M 10
MS

n ea
an

va 49 nv
=1

%
val

lu alu
e=
8%

-20 e= 20
ue

39 49
=1

% %
8%

-30 30

-40
Depth (cm)

40
CO2 (%)

-50 50

-60 60

-70 70

-80 80
CO2 flux = 800 g m-2 d-1 CO2 flux = 3400 g m-2 d-1 CO2 flux = 8500 g m-2 d-1
-90 90
v = 1×10-4 cm s-1 v = 2×10-3 cm s-1 v = 5×10-3 cm s-1
-100 100

Fig. 4. Soil CO2 concentration profiles. (A) Low flux condition: gas transport is mainly diffusive. (B) Intermediate flux condition: gas transport is diffusive and advective. (C) High soil
CO2 condition: gas transport is mainly advective. In (A), the soil CO2 concentration profile is linear as is the relationship between flux and CCDMS. For (B) and (C), owing to the
advective soil CO2 component, the concentration profile is not linear, nor is the relationship between flux and CCDMS.
106 S. De Gregorio et al. / Chemical Geology 341 (2013) 102–109

where D is the bulk diffusion coefficient, ν the gas velocity, and L the Under these conditions, Eq. (10) simplifies as:
length of soil section where the PTFE tube is inserted. The concentra-
tion value measured by CADEMASO (CCDMS) is the integral of the soil D 1
J CO2 ¼ : ð11Þ
CO2 concentration from −L to 0 depths: L 1−C CDMS

−L −L
Eq. (11) predicts that JCO2 does not linearly change with CCDMS
1 ½CO2 −L ½ expðvz=DÞ−1 when advection is dominant. As predicted by the conceptual model,
C CDMS ¼ − ∫ ½CO2 ðzÞdz ¼ ∫ dz
L 0 expð−vL=DÞ−1 under a different gas transport regimen, different relations between
0 ð6Þ
D½CO2 −L ð1− expð−vL=DÞ−νL=DÞ JCO2 and CCDMS are in effect. To compute soil CO2 flux values from
C CDMS ¼ :
νL½ expð−vL=DÞ−1 CCDMS, the definitions for the ranges of validity of Eqs. (8) and (11)
are of crucial importance. As shown later, because the modality of
gas transport is itself a function of soil CO2 flux and hence of soil
Eq. (6) mathematical demonstrates that, in agreement with the CO2 concentration along the soil profile i.e. CCDMS, the range of use
conceptual model, CCDMS, depends on both diffusive (D, [CO2]−L) of one equation with respect to other will be deduce by the CCDMS
and advective flux (v). Substituting the term [CO2]−L into Eq. (6) by values as well.
utilising Eq. (5), a theoretical equation for JCO2 as a function of CCDMS
is obtained: 4. Laboratory experimental tests

2 To assess the reliability of the proposed model and the range of

ν LC CDMS
J CO2 ¼ : ð7Þ validity of Eqs. (8) and (11), several lab tests were performed using a
D½ expð−vL=DÞ−1 þ νL=D
gas flux simulator to imposed different CO2 flux values. In brief, this
apparatus is a cylindrical container filled with a soil sample (fine
Eq. (7) shows that the relationship between CCDMS and JCO2 de-
sand) with a diffuser placed at its base. To generate CO2 flux inside
pends on ν values and, hence, as predicted by conceptual model, on
the container, a tank of pure CO2 (99.99%) was connected to the diffus-
the prevailing gas transport modality. Furthermore, Eq. (7), provided
er, and CO2 was introduced at constant flux. The flux was measured
ν values are known, well describes the relation between JCO2 and
using an electronic mass flow metre with high accuracy (±1% of full
CCDMS for any flux values. Unfortunately, the Eq. (7) is impractical
scale) and good repeatability (±0.2% of full scale). At the beginning
for our aims because the term ν is hardly continuously measurable
of each test (before CO2 injection), the soil gas in each layer consisted
in the field. However, with realistic assumptions, the above equations
exclusively of air at atmospheric pressure. As CO2 entered the soil,
could be used to obtain direct relationships between JCO2 and CCDMS
CO2 concentration at different depths increased (at variable rates)
that are valid for different flux values and ν values. As previously
until a steady state was reached. Before reaching the steady state,
discussed, a realistic assumption for low flux value is that the gas
CO2 flux introduced into the base of the soil was higher than the CO2
transport is prevalent diffusive; hence, Eq. (7) can be simplified by
flux released by the system into the atmosphere. When the steady
expanding the term exp(− vL/D) in Taylor's series and neglecting
state was reached, the CO2 flux was the same at the inlet and outlet
terms that are higher than second order:
(see Camarda et al., 2006, for details). A wide range of CO2 fluxes
ranging from 2.4 × 10 2 up to 1.8× 104 g m−2 d−1 were investigated.
ν 2 LC CDMS 2DC CDMS We performed several experiments generating flux with diverse prev-
J CO2 ¼ ¼ : ð8Þ
D−vL þ v2 L2 =2D−D þ νL L alent gas transport modalities, starting from flux dominated by
diffusive transport, where CO2 flux is mainly driven by a concentration
gradient, to flux dominated by advective transport, where CO2 flux is
The above equation highlights that as in a case of a prevalent
mainly driven by pressure gradient. For each imposed soil CO2 flux
diffusive regimen, a linear relation exists between JCO2 and CCDMS.
under steady state conditions, we also measured soil pressure at a
Eq. (8) holds only until the assumption that diffusive gas transport
1-metre depth. The soil pressure measurements were used, in agree-
is prevalent is valid, i.e., for low soil CO2 flux with low ν values
ment with Darcy's law, to compute ν, whereas the value of the bulk
(Fig. 4A). For higher ν values, terms higher than second order are
diffusion coefficient (D) was calculated as described in Camarda et al.
no longer negligible. For high ν and CO2 flux values, when advection
(2007). During experiments, CADEMASO acquired data continuously,
becomes the main transport modality of CO2, Eq. (8) is no longer
with hourly frequency. Fig. 5 shows the relationship between CCDMS
valid, and another expression for JCO2 as a function of CCDMS must be
obtained. Thus, we substitute the term [exp(− vL/D) − 1] into
Eq. (6) by utilising Eq. (5) to give another theoretical equation for 100
JCO2 as a function of CCDMS:

75
D
CCDMS (%)

J CO2 ¼ vC CDMS þ ½CO2 −L : ð9Þ

L
50

Multiplying both sides of Eq. (9) by ½CO2 −L yields:

25
D
2
J CO2 ½CO2 −L ¼ v½CO2 −L C CDMS þ ½CO2 −L : ð10Þ
L
0
0 5000 10000 15000 20000
Bearing in mind the conceptual model, we can formulate realistic
Actual soil CO2 flux (g m-2 d-1)
assumptions to avoid the use of ν in computing flux. For elevated
v values, advective flux is dominant, and we can assume that Fig. 5. Relation between actual soil CO2 flux values and CCDMS. As expected, relatively
J CO2 ¼ ν½CO2 −L . Further, as shown by conceptual models, we can low flux values demonstrate a linear relationship, while for higher flux values, the re-
also realistically assume ½CO2 −L ¼ 1. lationship in not linear, and becomes almost exponential.

and imposed soil CO2 flux values. As expected, relatively low flux
values exhibit a linear relationship, while for higher flux values, the re-
lation becomes asymptotic. The CO2 flux values were computed from
CCDMS measurements using Eqs. (7), (8) and (11). Fig. 6 compares soil
CO2 flux values computed with these three equations with the values
measured by the electronic mass flow metre (i.e., the actual soil CO2
flux values). The graph plots a straight line with ratio 1:1, which is
representative of perfect agreement between measured and actual
CO2 flux values. The values calculated with Eq. (7) are very close to a
straight line with ratio 1:1 for the entire range of explored CO2 flux,
validating the proposed model. As previously mentioned, the values
computed by Eq. (8) are very close to a straight line with a 1:1 ratio
only for relatively low flux values and a mainly diffusive gas transport
regimen (JCO2 up to 1.6 × 103 g m−2 d−1 and CCDMS up to 40%). For
higher flux values, the use of Eq. (8) leads to underestimation of flux
values because the advective component is no longer negligible. Con-
versely, values calculated with Eq. (11) show a positive bias for low
soil CO2 flux, but show a good agreement with actual flux for higher
values of JCO2 (JCO2 > 6.0 × 103 g m −2 d −1, CCDMS > 58%) when advec-
tion is the dominant modality of CO2 transport. As stated, although
Eq. (7) is valid over the entire range of investigated fluxes, CCDMS re-
quires knowledge of ν values. Nonetheless, a relationship between
JCO2 and CCDMS not including this parameter and valid over the entire
range can be obtained by properly combing Eqs. (8) and (11)
because, as shown in Fig. 6, the first equation gives a very good esti-
mate for low flux values (JCO2 b 1.6 × 103 g m−2 d−1, CCDMS b 40%),
whereas the second equation estimates higher values well
(JCO2 > 6.0× 103 g m−2 d−1, CCDMS > 58%). Synthesising, we can write
that:
2DC CDMS D 1
J CO2 ¼ f þ ð1−f Þ
L L 1−C CDMS
where f is a function dependent on the gas transport modality; diffu-
sive transport f values approach 1, while mainly advective transport
values of f approach 0. As noted in the conceptual model, because the
value of CCDMS is itself a function of gas transport regimen, we can
equation (7)
15000
equation (8)
Computed soil CO2 flux (g m-2 d-1)
equation (11)
12000
equations (12+13)
9000
6000
3000
0 variations according to Eq. (3). The flux values for CADEMASO were
0 3000 6000 9000
Actual soil CO2 flux (g m-2 d-1)
Fig. 6. Experimental results. The straight line with ratio 1:1 is representative of perfect
agreement between computed CO2 flux values and those measured by electronic mass
flow metre (i.e., the actual soil CO2 flux values). Data computed using Eq. (7) with ν
measured values (open red squares) fit well to a straight line, confirming the reliability
of the proposed model. Data from Eqs. (8) (open blue triangles) and (11) (open blue
diamonds) fit for the specific flux values range where the assumption made for deduc-
ing the respective equations is valid. Finally, data obtained by combining Eqs. (12) and
(13) (open black circles) fit very well to a straight line over the entire range of fluxes.
S. De Gregorio et al. / Chemical Geology 341 (2013) 102–109 107
use values of CCDMS to define the f function. From experimental data,
we find that f must satisfy the following condition:
f ¼ 1 for C CDMS b40% and f ¼ 0 for C CDMS >58%
a function able to satisfy these conditions is a sigmoid:
1
f ¼ ð13Þ
1 þ exp½ðC CDMS −aÞ=b
where the values of the a and b constants can be determined by the
best fit of Eq. (12) to experimental values. The use of Eqs. (12) and
(13) provides the capability to calculate the soil CO2 flux for
CCDMS > 40% and CCDMS b 58% and therefore over the entire range of
explored flux. Fundamentally combining Eqs. (12) and (13), we
achieve a single mathematical equation relating soil CO2 flux to CCDMS
measurements. As shown in Fig. 6, the flux values computed by comb-
ing Eqs. (12) and (13) fit very well to a straight line with a 1:1 ratio,
confirming the reliability of the proposed equations in computing soil
CO2 flux values using only CCDMS measurements.
5. Results and discussion
5.1. Field site features
In January 2012, we installed a remote continuous monitoring
station on the Island of Vulcano. This island is an active volcano on
the Aeolian archipelago, located in the southern Tyrrhenian Sea
(southern Italy) off the north coast of Sicily (Italy). The station was
placed in an area well known for intense soil degassing, with great
ð12Þ spatial and temporal fluctuations (Diliberto et al., 2002; Carapezza
and Granieri, 2004). These characteristics made the location a sort
of natural laboratory where it was feasible to efficiently test the reli-
ability and the performance of CADEMASO. In addition to CADEMASO
we installed a probe for measuring soil CO2 flux with the DCM
(Camarda et al., 2006). To prevent perturbation in the measurements,
the DCM and CADEMASO measurement probes were placed in two
sites approximately 2 m apart. The two sites were characterised by
different soil CO2 fluxes owing to the wide soil CO2 flux spatial vari-
ability of the area, ranging up to an order of magnitude in the space
of a few metres (Carapezza and Granieri, 2004). The difference of
CO2 flux absolute values of the two sites could mainly ascribed to a
difference in permeability and/or distance from the CO2 source.
1
1:
tio
ra
5.2. Field data comparison
The station record hourly soil CO2 fluxes with both methods (DCM
and CADEMASO), including soil temperature and atmospheric param-
eters. The data were stored locally in a data-logger and automatically
transmitted daily to our laboratory via radio link. The four months of
records are shown in Fig. 7, and flux values (Fig. 7A) as well as soil
and atmospheric parameters (Fig. 7B and C) are reported. Soil tem-
perature varied from a minimum of 14 °C in February to a maximum
of 26 °C in May, and CO2 concentration values was corrected for these
12000 15000 computed with Eqs. (12) and (13). The two soil CO2 flux data sets
(CADEMASO and DCM) are in excellent agreement, displaying almost
perfect similarity across rather wide oscillations. A strong inverse cor-
relation between the temporal trend of soil CO2 flux and atmospheric
pressure can be found by inspecting Fig. 7. This feature is due to the
peculiar emission regime of the areas, which experience intense,
constant CO2 input from depth. Under this condition, a drop in
barometric pressure induces the movement of CO2 from depth layers
towards superficial portions of the soil, acting as a sort of pump. This
phenomenon, known as barometric pumping, is well described in the
108 S. De Gregorio et al. / Chemical Geology 341 (2013) 102–109

Soil CO2 flux (g m-2 d-1)

A

Soil CO2 flux (g m-2 d-1)

B

Fig. 7. Results of the field application at the Island of Vulcano and comparison with DCM. 24 point running average of: (A) soil CO2 flux measurements performed using DCM
(orange line) and the new method (blue line); and (B) atmospheric pressure, the red box indicates a peculiar long-lasting low atmospheric pressure period (see text for details).
(C) Soil temperature. The flux values reported for the new method were computed from CCDMS measurements, corrected for temperature changes, by applying the Eqs. (12) and
(13). The two soil CO2 flux data sets display wide oscillations that are temporally well correlated. The excellent agreement testifies that the new method works well in field con-
ditions. The wide fluctuations in soil CO2 flux are caused by changes in atmospheric pressure. Increases in atmospheric pressure induce a reduction in flux, whereas sharp decreases
promoted enhanced flux.

literature (Massmann and Farrier, 1992; Auer et al., 1996; Granieri et
al., 2003).
To further confirm the good concordance of flux values recorded
with the two methods, we plotted the correlation diagram between
the two datasets (Fig. 8). In the diagram, the points define a linear
trend with a high correlation coefficient, equal to 0.8. The slope of
the straight line of the best fit is lower than the slope of the straight
line with ratio 1:1, as consequence of different amounts of CO2 emis-
sion in the two sites. Further, the points with highest values of flux
recorded by CADEMASO (contained inside the grey ellipse in Fig. 8)
80000
Soil CO2 flux (g m-2 d-1)
slightly diverge from the major linear trend, displaying a higher
slope. A higher slope is indicative of the fact that values recorded by
CADEMASO, in the lower degassing site, approaches the values
recorded in the higher degassing site with CDM method. These values
were recorded at the beginning and during long-lasting low atmo-
spheric pressure period, that started at end of March 2012 (red box
in Fig. 7B). It can be supposed that, throughout this time, the regime
and the changes of atmospheric pressure were such, that the CO2
saturated soil was shifted upward, in the way as make the emission
of CO2 of the area more spatially homogeneous. However, leaving
aside the absolute value differences, the key result is excellent agree-
ment between the two data sets. On the whole, the result of field ap-
plication demonstrates the reliability of CADEMASO in continuously
recording soil CO2 flux over a wide range.

60000 6. Conclusion
CADEMASO

We have presented a new, reliable, straightforward method for

40000 continuous measurement of soil CO2 flux that can serve as a valuable
alternative to existing methods. Unlike existing methods, CADEMASO
can perform soil flux CO2 measurements without introducing sample
gases inside the instrument and, in an almost completely passive way,
20000 induces no perturbation of shallow soil CO2 concentrations. Further
the PTFE, owing to its chemical composition and structure, is almost
totally inert at low and high temperatures (useful temperature up
0 to 260 °C).
0 20000 40000 60000 80000 These features confer to CADEMASO the capability to perform con-
DCM tinuous measurements in extreme weather and/or environmental
Soil CO2 flux (g m-2 d-1) conditions potentially damaging to existing instruments and/or
when usage of standard methods is challenging.
Fig. 8. Correlation diagram between soil CO2 fluxes measured with DCM method and The extreme variability of soil CO2 flux recorded (up to order of
CADEMASO. The points show a strong linear correlation with a correlation coefficient magnitude) on the Island of Vulcano highlights the importance of
of 0.8. Due to the wide soil CO2 flux spatial variability of the area, the slope of the
straight line of the best fit (grey line) is lower than the slope of the straight line with
continuous monitoring for investigations into volcano emissions
ratio 1:1. The grey ellipse contains data displaying a slope higher than best fit straight and/or budget estimates of CO2 emissions, especially in area with an
line (see text for details). emissions regime similar to the one investigated here.
 S. De Gregorio et al. / Chemical Geology 341 (2013) 102–109
References
Atkin, O., Edwards, E., Loveys, B., 2000. Response of root respiration to changes in tem-
perature and its relevance to global warming. The New Phytologist 147, 141–154.
Auer, L.H., Rosenberg, N.D., Birdsell, K.H., Whitney, E.M., 1996. The effects of barometric
pumping on contaminant transport. Journal of Contaminant Hydrology 24,
145–166.
Batjes, N.H., 1996. Total carbon and nitrogen in the soils of the world. European Journal
of Soil Science 47, 151–163.
Baubron, J.C., Allard, P., Toutain, J.P., 1990. Diffuse volcanic emissions of carbon dioxide
from Vulcano island, Italy. Nature 344, 51–53.
Beaubien, S.E., Ciotoli, G., Lombardi, S., 2003. Carbon dioxide and radon gas hazard in
the Alban Hills area (central Italy). Journal of Volcanology and Geothermal
Research 123, 63–80.
Bekku, Y., Koizumi, H., Nakadai, T., Iwaki, H., 1995. Measurement of soil respiration
using closed chamber method: an IRGA technique. Ecological Research 10,
369–373.
Bergfeld, D., Evans, W.C., 2011. Monitoring CO2 emissions in tree kill areas near the re-
surgent dome at Long Valley Caldera, California. U.S. Geological Survey Scientific
Investigations Report 2011-5038, p. 22.
Camarda, M., Gurrieri, S., Valenza, M., 2006. CO2 flux measurements in volcanic areas
using the dynamic concentration method: the influence of the soil permeability.
Journal of Geophysical Research 111. http://dx.doi.org/10.1029/2005JB003898.
Camarda, M., De Gregorio, S., Favara, R., Gurrieri, S., 2007. Evaluation of carbon isotope
fractionation of soil CO2 under an advective–diffusive regime: a tool for computing
the isotopic composition of unfractionated deep source. Geochimica et
Cosmochimica Acta 7, 3016–3027.
Camarda, M., Gurrieri, M., Valenza, M., 2009. Effects of soil gas permeability and
recirculation flux on soil CO2 flux measurements performed using a closed dynam-
ic accumulation chamber. Chemical Geology 265, 387–393.
Camarda, M., De Gregorio, S., Gurrieri, S., 2012. Magma-ascent processes during
2005–2009 at Mt Etna inferred by soil CO2 emissions. Chemical Geology
330–331, 218–227.
Carapezza, M.L., Granieri, D., 2004. CO2 soil flux at Vulcano (Italy): comparison
between active and passive methods. Applied Geochemistry 19, 73–88.
Carapezza, M.L., Barberi, F., Ranaldi, M., Ricci, T., Tarchini, L., Barrancos, J., Fischer, C.,
Perez, N., Weber, K., Di Piazza, A., Gattuso, A., 2011. Diffuse CO2 soil degassing
and CO2 and H2S concentrations in air and related hazards at Vulcano Island
(Aeolian arc, Italy). Journal of Volcanology and Geothermal Research 207,
130–134.
Chiodini, G., Frondini, F., Raco, B., 1996. Diffuse emission of CO2 from the Fossa crater,
Vulcano Island (Italy). Bulletin of Volcanology 58, 41–50.
Chiodini, G., Cioni, R., Guidi, M., Marini, L., Raco, B., 1998. Soil CO2 flux measurements in
volcanic and geothermal areas. Applied Geochemistry 13, 543–552.
De Gregorio, S., Gurrieri, S., Valenza, M., 2005. A PTFE membrane for the in situ extrac-
tion of dissolved gases in natural waters: theory and applications. Geochemistry,
Geophysics, Geosystems 6. http://dx.doi.org/10.1029/2005GC000947.
De Gregorio, S., Camarda, M., Cappuzzo, S., Gurrieri, S., 2009. Using pressure transients
within a polymeric membrane for gas composition measurements. Geochemistry,
Geophysics, Geosystems 10. http://dx.doi.org/10.1029/2009GC002683.
Diliberto, I.S., Gurrieri, S., Valenza, M., 2002. Relationships between diffuse CO2 emis-
sions and volcanic activity on the island of Vulcano (Aeolian Island, Italy) during
the period 1984–1994. Bulletin of Volcanology 64, 219–228.
Faber, E., Morà, N.C., Poggenburg, J., Garzà, N.G., Teschner, M., 2003. Continuous gas
monitoring at Galeras volcano, Colombia: first evidence. Journal of Volcanology
and Geothermal Research 12, 13–23.
109
Federico, C., Camarda, M., De Gregorio, S., Gurrieri, S., 2011. Long-term record of CO2
degassing along Mt. Etna's flanks and its relationship with magma dynamics and
eastern flank instability. Geochemistry, Geophysics, Geosystems 12. http://
dx.doi.org/10.1029/2011GC003601.
Fischer, T.P., 2008. Fluxes of volatiles (H2O, CO2, N2, Cl, F) from arc volcanoes.
Geochemical Journal 42, 21–38.
Gerlach, T.M., Doukas, M.P., McGee, K.A., Kessler, R., 2001. Soil efflux and total emission
rates of magmatic CO2 at the Horseshoe Lake tree kill, Mammoth Mountain,
California, 1995–1999. Chemical Geology 177, 101–116.
Granieri, D., Chiodini, G., Marzocchi, W., Avino, R., 2003. Continuous monitoring of CO2
soil diffuse degassing at Phlegraean Fields (Italy): influence of environmental and
volcanic parameters. Earth and Planetary Science Letters 212, 167–179.
Granieri, D., Carapezza, M.L., Chiodini, G., Avino, R., Caliro, S., Ranaldi, M., Ricci, T.,
Tarchini, L., 2006. Correlated increase in CO2 fumarolic content and diffuse emis-
sion from La Fossa crater (Vulcano, Italy): evidence of volcanic unrest or increasing
gas release from a stationary deep magma body? Geophysical Research Letters 33.
http://dx.doi.org/10.1029/2006GL026460.
Hernández Perez, P., Notsu, K., Tsurumi, M., Mori, T., Ohno, M., Shimoike, Y., Salazar, J.,
Pérez, N., 2003. Carbon dioxide emissions from soils at Hakkoda, north Japan.
Journal of Geophysical Research 108, 2210. http://dx.doi.org/10.1029/
2002JB001847.
Hernández, P.A., Notsu, K., Salazar, J.M., Mori, T., Natale, G., Okada, H., Virgili, G.,
Shimoike, Y., Sato, M., Perez, N.M., 2001. Carbon dioxide degassing by advective
flow from Usu volcano, Japan. Science 292, 83–86.
Law, B.E., Baldocchi, D.D., Anthoni, P.M., 1999. Below-canopy and soil CO2 efflux in
ponderosa pine forest. Agricultural and Forest Meteorology 94, 171–188.
Lewicki, J.L., Hilley, G.E., 2009. Eddy covariance mapping and quantification of surface
CO2 leakage fluxes. Geophysical Research Letters 36. http://dx.doi.org/10.1029/
2009GL040775.
Massmann, J., Farrier, D.F., 1992. Effects of atmospheric pressures on gas transport in
the vadose zone. Water Resources Research 28, 777–791.
Norman, J.M., Garcia, R., Verma, S.B., 1992. Soil surface CO2 fluxes and the carbon
budget of a grassland. Journal of Geophysical Research 97, 18845–18853.
Pilegaard, K., Hummelshøj, P., Jensen, N.O., Chenc, Z., 2001. Two years of continuous
CO2 eddy-flux measurements over a Danish beech forest. Agricultural and Forest
Meteorology 107, 29–41.
Raich, J.W., Schlesinger, W.H., 1992. The global carbon dioxide flux in soil respiration and
its relationship to vegetation and climate. Tellus Series B: Chemical and Physical
Meteorology 44B, 81–99.
Russell, C.A., Voroney, R.P., Black, T.A., Blanken, P.D., Yang, P.C., 1998. Carbon dioxide
efflux from the floor of a boreal aspen forest. II. Evaluation of methods — verification
by infrared analysis of a dynamic closed chamber. Canadian Journal of Soil Science
78, 311–316.
Sahimi, M., 1995. Flow and Transport in Porous Media and Fractured Rock: From
Classical Methods to Modern Approaches.VCH Publisher, New York.
Shi, P.L., Zhang, X.Z., Zhong, Z.M., Ouyang, H., 2006. Diurnal and seasonal variability of
soil CO2 efflux in a cropland ecosystem on the Tibetan Plateau. Agricultural and
Forest Meteorology 137, 220–233.
Wang, R., Willibald, G., Feng, Q., Zheng, X., Liao, T., Brüggemann, N., Butterbach-Bahl, K.,
2011. Measurement of N2, N2O, NO, and CO2 emissions from soil with the gas-flow-
soil-core technique. Environmental Science and Technology 45 (14), 6066–6072.