Monitoring above-zone temperature variations associated with CO2 and brine leakage from a storage...

transcript

ORIGINAL ARTICLE

Monitoring above-zone temperature variations associatedwith CO2 and brine leakage from a storage aquifer

Mehdi Zeidouni • Jean-Philippe Nicot •

Susan D. Hovorka

Received: 15 June 2013 / Accepted: 15 January 2014

� Springer-Verlag Berlin Heidelberg 2014

Abstract CO2 injection in saline aquifers induces tem-

perature changes owing to processes such as Joule–

Thomson cooling, endothermic water vaporization, exo-

thermic CO2 dissolution besides the temperature discrep-

ancy between injected and native fluids. CO2 leaking from

the injection zone, in addition to initial temperature con-

trast due to the geothermal gradient, undergoes similar

processes, causing temperature changes in the above zone.

Numerical simulation tools were used to evaluate temper-

ature changes associated with CO2 leakage from the stor-

age aquifer to an above-zone monitoring interval and to

assess the monitorability of CO2 leakage on the basis of

temperature data. The impact of both CO2 and brine

leakage on temperature response is considered for three

cases (1) a leaky well co-located with the injection well, (2)

a leaky well distant from the injector, and (3) a leaky fault.

A sensitivity analysis was performed to determine key

operational and reservoir parameters that control the tem-

perature signal in the above zone. Throughout the analysis

injection-zone parameters remain unchanged. Significant

pressure drop upon leakage causes expansion of CO2

associated with Joule–Thomson cooling. However, brine

may begin leaking before CO2 breakthrough at the leakage

pathway, causing heating in the above zone. Thus, unlike

the pressure which increases in response to both CO2 and

brine leakage, the temperature signal may differentiate

between the leaking fluids. In addition, the strength of the

temperature signal correlates with leakage velocity unlike

pressure signal whose strength depends on leakage rate.

Increasing leakage conduit cross-sectional area increases

leakage rate and thus increases pressure change in the

above zone. However, it decreases leakage velocity, and

therefore, reduces temperature cooling and signal. It is also

shown that the leakage-induced temperature change covers

a small area around the leakage pathway. Thus, tempera-

ture data will be most useful if collected along potential

leaky wells and/or wells intersecting potential leaky faults.

Keywords CO2 geological storage � Leakage �Temperature monitoring � Well � Fault

Introduction

Deep geological formations are used widely for fluid dis-

posal and are the main candidates for storage of CO2 as a

means of reducing greenhouse gas emissions to the atmo-

sphere. Safety of storage projects depends on the contain-

ment of CO2 within the target zone. However, the

confining system overlying the aquifer may include leak-

age pathways that permit injected and/or native fluids to

leak to subsurface formations and/or to the surface. Tem-

perature measurements at the injection zone and an over-

lying permeable zone can be acquired accurately at high

resolution at a low cost, from which information on CO2

leakage may be inferred. Temperature monitoring at the

subsurface has been used widely in the oil and gas industry

to track gas migration (Marinaro et al. 2011) and to mon-

itor hydraulic fracturing (Hoang et al. 2011). The thermal

response may have the potential to determine the fate of

injected CO2 during and after CO2 injection for seques-

tration purposes. Temperature has been monitored to assess

Electronic supplementary material The online version of thisarticle (doi:10.1007/s12665-014-3077-0) contains supplementarymaterial, which is available to authorized users.

M. Zeidouni (&) � J.-P. Nicot � S. D. Hovorka

Bureau of Economic Geology,

The University of Texas at Austin, Austin, TX, USA

e-mail: mehdi.zeidouni@beg.utexas.edu

Environ Earth Sci

DOI 10.1007/s12665-014-3077-0

CO2 storage performance in demonstration projects,

including Cranfield field, Mississippi, USA (Hovorka et al.

2013) and Ketzin, Germany (Liebscher et al. 2013). CO2

leakage from the injection zone will go through processes

that may induce a detectable temperature signal in the

above-zone monitoring interval (AZMI). These processes

are the same as those causing temperature changes in the

injection zone, namely the Joule–Thomson (JT) effect,

CO2-brine mutual dissolution, and heat exchange with the

surrounding rock-fluid system to which can be added the

geothermal gradient contributing to the cooling of leakage

fluids toward the surface.

Pressure propagates in the reservoir and decreases as

radial distance from the wellbore or wellfield increases.

Under reservoir adiabatic conditions and at the limit

DP ? 0, the process is isenthalpic, and temperature change

due to pressure reduction can be estimated by the JT

coefficient:

lJT ¼ limDP!0

� �H

Under the practical P–T conditions encountered for CO2

geological storage applications, the JT coefficient is a

positive number (Han et al. 2010). Therefore, pressure

reduction will be accompanied by thermal cooling. The

temperature reduction accompanied by fluid expansion is

referred to as JT cooling, even though the adiabatic condi-

tion assumption may not be fully valid. Thermal effects can

also be caused by mutual dissolution of CO2 and brine and

heat losses to underburden/overburden. At conditions suit-

able for CO2 storage (P 10–40 MPa and T 30–80 �C), the

enthalpy change (DH) associated with CO2 dissolution in

brine is negative, indicating an exothermic process. On the

other hand, vaporization of water into a CO2-rich phase is an

endothermic reaction and causes cooling (Han et al. 2010).

Assuming single-phase, steady-state flow and constant

thermophysical properties, Mathias et al. (2010) derived an

analytical solution to evaluate JT thermal cooling during

CO2 injection in a saline aquifer. They were able to closely

match numerical results obtained using the TOUGH2

numerical simulation package. Assuming constant ther-

mophysical properties of CO2 (density, viscosity, specific

heat capacity, and JT coefficient), their analytical solution

marginally overestimates temperature decline.

The thermal signature of CO2 injection in saline aquifers

caused by the above-mentioned processes is well studied in

the literature. Andre et al. (2010, 2011) investigated ther-

mal processes triggered by CO2 injection, including the JT

effect, heat of vaporization, and heat of dissolution. The JT

effect was identified mainly in the dry-out zone, whereas

CO2 dissolution was found to cause subtle large-scale

thermal effects. Andre et al. (2010) concluded that the

impact of thermal processes has to be considered mainly in

three types of scenarios: (1) for aquifers at PVT conditions

close to the critical point of CO2 at Pcrit = 7.38 MPa,

Tcrit = 31.04 �C; (2) for reservoirs at PVT conditions

favorable to a good expression of the JT coefficient, such as

in depleted gas fields (Oldenburg 2007); and (3) in the

presence of a large discrepancy between injection tem-

perature and reservoir temperature. Han et al. (2010)

investigated thermal effects within the injection well by

considering the effect of adiabatic compression (JT effect),

frictional energy loss, and heat exchange between injected

CO2 and the rock and fluids surrounding the injection-well

path from the surface. Adiabatic heating may be observed

under moderate injection rates, whereas high injection rates

maximize heat losses due to friction and reduce heat

exchange with the surroundings. Generally, the JT coeffi-

cient has an inverse relation with the density of CO2.

Tracking the CO2 plume using temperature measure-

ments in the injection zone has also been investigated. A

visible temperature signal may be detectable upon CO2

arrival at an observation well in the injection zone, as

shown by Hovorka et al. (2006) for the Frio CO2 injection

pilot test. Bielinski et al. (2008) investigated the possibility

of using temperature-measurement devices for CO2 moni-

toring in the injection zone. They studied the effect of

various parameters on the time-dependent temperature

signal measured at an observation well at some distance

from the injection point. They showed that a detectable

temperature effect at the observation well could be

obtained from which information on the shape of the CO2

plume (and, thus, about permeability distribution) could be

achieved. Hurter et al. (2007) examined the suitability of

thermal methods to monitor CO2 leakage. They discussed

the dynamic temperature signal associated with CO2

leakage and movement of CO2 in the porous rock. They

suggested that CO2 leakage can produce significant cooling

that may be easily detectable. Singh et al. (2012) have also

shown importance of cooling in the injection zone for CO2

injected in a gas field.

In relatively static conditions, long after the end of the

CO2 injection period, thermal equilibrium prevails and

temperature gradient is controlled only by heat conduc-

tivity of the system. Heat conductivity of the porous

medium is a function of heat conductivity of the rock, as

well as its containing fluids. Because a rock filled with

brine has larger conductivity than a rock filled with CO2 or

both CO2 and brine, the presence of CO2 will make the

porous media less heat conductive. As a result, heat will be

more preserved, and temperature increase caused by

reduced heat conduction in the reservoir containing CO2

will be observed as an effect known as the blanketing effect

(Hurter et al. 2007; Wangen 1995; Zhang 1993). During

the injection period, dynamic-flow conditions prevail. Heat

will be transferred both by advection (owing to migration

Environ Earth Sci

of injected CO2 into the reservoir) and by conduction.

Given that CO2 migration in the reservoir is faster, the

thermal signature is controlled mainly by heat advection.

Heat conduction will affect the thermal signature only

slightly by reducing heat conductivity at the two-phase

(CO2-rich and aqueous phases) region. As a result, at

locations of increased CO2 saturation, heat transport may

be hindered.

Temperature can be monitored using point measure-

ments, wireline-deployed instruments that produce logs of

temperature versus depth, or fiber-optic distributed tem-

perature sensing (DTS) cables. In spite of its lower reso-

lution compared to conventional temperature sensors (e.g.

quartz gauges), DTS technology has the advantage of

providing continuous temperature measurements both in

time and in space along the cable. Upon careful calibration,

DTS can provide resolution and accuracy as small as 0.02

and 0.3 �C, respectively, as reported for a gas-hydrate

monitoring application at *1,200 m depth (Henninges

et al. 2005; Bielinski et al. 2008).

In this study, the usefulness of temperature data in

detecting leakage is investigated. First the numerical simu-

lation tool (CMG-GEM 2012) to be employed in modeling

the thermal response coupled with fluid flow is briefly dis-

cussed. Next, the capability of the CMG-GEM simulator in

modeling thermal response is assessed by comparing its

results with those reported using TOUGH2 (Pruess et al.

1999) for a benchmarking example problem. Then the

example problem is modified to account for leakage to the

overlying formation. Finally, the thermal response due to

leakage is investigated in three scenarios: (1) leakage

occurring through a leaky well co-located with the injection

well, (2) leakage through a leaky well located at a distance

from the injection well, and (3) leakage through a fault at a

distance from the injection well. Sensitivity of the temper-

ature signal to various parameters is investigated for the first

scenario.

Thermal simulation using CMG-GEM simulator

In modeling thermal processes in the reservoir, the CMG-

GEM (2012) compositional simulator is used. CMG-GEM

calculates temperature distribution by solving the energy

balance equation, coupled with flow equations. The energy

balance equation includes convection, conduction, and heat

loss to the surroundings. Enthalpy of the phases should be

calculated for the energy balance equation and is obtained

H ¼ DHE þ H� ð2Þ

where DHE is excess enthalpy of the fluid and H* is ideal

gas enthalpy at zero pressure and temperature. Excess

enthalpy can be calculated from the equation of state (EOS)

on the basis of the following:

DHE ¼ RTðZ � 1Þ þZV

� �� P

� �dV ð3Þ

where R is the universal gas constant and Z is the

compressibility factor; and P, V, and T are pressure, molar

volume, and temperature, respectively. Given EOS type

and parameters, DHE can be obtained. The ideal gas

enthalpy, H*, for each fluid is the sum of ideal gas

enthalpies of the fluid components weighted by their mole

fractions. The enthalpy of each component is obtained by

the following 5th-degree polynomial (API Technical Data

Book, Petroleum Refining 1983):

H� ¼X5

ajiTj ð4Þ

The coefficients aji are referred to as enthalpy coeffi-

cients that are listed in Table 1 for CO2 and brine.

As a result of changes in composition of the aqueous and

CO2-rich phases upon mutual dissolution, enthalpy of the

phases changes accordingly. For the range of injection-

zone temperatures for CO2 storage, enthalpy of the aqueous

phase decreases upon dissolution of CO2 (heat is gener-

ated), and enthalpy of the CO2-rich phase increases as

water vaporizes into the CO2-rich phase and heat is con-

sumed (Koschel et al. 2006). Therefore, temperature is

likely to increase at the dissolution front, and cooling may

occur at the vaporization front close to the wellbore. Heat

loss to the surrounding rock is incorporated using the

method of Vinsome and Westerveld (1980) and is con-

trolled by the cross-sectional area for heat loss, rock heat

capacity, rock mass density, and rock thermal conductivity.

Cubic EOS, including PR (Peng and Robinson 1976), is

not accurate in calculating solubility of CO2 in the aqueous

phase. For the CMG-GEM (2012) simulator, solubility is

obtained by applying the thermodynamic equilibrium.

Fugacity of CO2 in the CO2-rich phase is evaluated on the

basis of PR EOS, and fugacity of CO2 in the aqueous phase

is calculated based on Henry’s law:

fCO2¼ xCO2

HCO2ð5Þ

where xCO2and HCO2

are the mole fractions of CO2 in the

aqueous phase and CO2 Henry’s constant, respectively.

HCO2is a function of pressure, temperature, and salinity.

Variation of Henry’s constant with pressure and tempera-

ture is obtained on the basis of the Harvey (1996) corre-

lation. The effect of salinity on Henry’s constant is

obtained as described by Bakker (2003). Thermodynamic

equilibrium is applied to modeling of H2O vaporization in

the CO2-rich phase, for which the fugacity of H2O in the

Environ Earth Sci

CO2-rich phase is calculated based on PR EOS. Fugacity of

water in the aqueous phase is calculated on the basis of the

water mole fraction in the aqueous phase combined with

equations given by Canjar and Manning (1967) and Saul

and Wagner (1987).

Benchmarking

To validate results of the numerical simulation performed

by CMG-GEM, they are compared against those reported

in the literature by Han et al. (2012). They presented results

for CO2 injection in an idealized homogeneous, 2-D radial

zone using TOUGH2-ECO2N (Pruess 2005; Pruess et al.

1999). For their example problem (see Table 2 for detailed

specifications), the reservoir is bounded top and bottom by

no flow boundaries and extends 5 km radially. The depth to

the top of the injection zone is 900 m, its thickness is

182.9 m and is divided into 30 layers. Reservoir porosity is

10 %, and its lateral and vertical permeabilities are 10 and

1 mD, respectively. Aqueous- and CO2-rich-phase relative

permeability curves are obtained on the basis of models

given in Table 3. Reservoir initial temperature is 41.7 �C,

and its initial pressure varies over a 6.89- to 8.9-MPa range

from top to bottom, changing hydrostatically with depth.

CO2 is injected at a rate of 50 kton/year (69,185 sm3/day)

Table 1 Enthalpy coefficients for CO2 and brine used in CMG-GEM simulation

a0i a1i a2i a3i a4i a5i

CO2 0.097 0.158 -0.034e-03 0.148e-6 -0.966e-10 2.074e-14

Brine -1.930 0.447 -0.022e-3 0.030e-6 -0.057e-10 0.028e-14

Table 2 Descriptions of

example problems

* sm3, m3 at surface conditions

of 1 atm pressure and 25 �C

temperature; Empty cells, same

properties as left column; –,

property not applicable

Parameter Han et al.

(2010)

Co-located

leaky well

Not-co-located

leaky well

Thermal gradient (�C/m) 0 0 0.03 0.03

Thermal conductivity [W/(m K)] 2.51

Rock density (kg/m3) 2,650

Rock heat capacity [J/(kg.K)] 1,000

Injected CO2 temperature (�C) 41.7

Injection-zone initial temperature (�C) 41.7

Injection-zone porosity (fraction) 0.1

Injection-zone lateral permeability (m2) 1e-14

Injection-zone vertical permeability (m2) 1e-15

Injection-zone thickness (m) 182.9 55

Reservoir radial extent (km) 5 24

Brine salinity (wt%) 15

Initial pressure at 1,000-m depth (MPa) 8.03

Injection-well radius (m) 0.3

Injection rate (kton/year, sm3*/day) 50, 69,185

Injection period (years) 2

Seal thickness (m) – – 5

AZMI thickness (m) – – 10

AZMI porosity (fraction) – – 0.1

AZMI lateral permeability (m2) – – 1e-14

AZMI vertical permeability (m2) – – 1e-15

Leaky-well radius (m) – – 0.3

Leaky pathway lateral permeability (m2) – – 1e-14

Leaky pathway vertical permeability (m2) – – 1e-14

Leaky pathway porosity (fraction) – – 0.1

Leak-injector distance (m) – – 0 100 100

Leaky fault width (m) – – – – 2

Environ Earth Sci

for 2 years in a 24.4-m-thick interval 12.2 m from the

bottom of the reservoir. Injected CO2 temperature is

identical to the initial reservoir temperature (41.7 �C), and

brine salinity is 15 wt%. Rock density and heat capacity

are 2,650 kg/m3 and 1,000 J/(kg K), respectively. Total

thermal conductivity is 2.51 W/(m K).

Compositional properties of aqueous and CO2-rich

phases used in CMG-GEM versus TOUGH2-ECO2N are

shown for a range of pressures and temperatures

encountered in simulating the Han et al. (2012) example

problem (Figure 1, ESM only). A good match is observed

for aqueous- and CO2-rich-phase viscosities and densi-

ties, as well as mutual solubilities. The temperature

change versus distance from the injection well at the

uppermost layer where CO2 is injected is shown in Fig-

ure 2 (ESM only) after 2 years of injection. In spite of

closely matching compositional properties, the tempera-

ture response given by TOUGH2 differs from that

obtained by CMG-GEM. TOUGH2 provides greater

heating, which corresponds to dissolution of CO2 and

brine and CMG-GEM provides greater JT cooling.

However, temperature responses have similar features,

displaying cooling close to the wellbore and heating

toward the CO2 front. Nevertheless, CMG-GEM is used

because of its more versatile compositional module

applicable to a broader range of temperature and pressure

conditions. Its use is also justified by a computed tem-

perature signal possibly smaller than the actual field

signal as approximated by TOUGH2 results. Conclusions

from this paper are therefore conservative in terms of

using the temperature signal that may be slightly stronger

than described here. Note that salt precipitation (Zeidouni

et al. 2009) is not considered here, which may affect the

thermal response near the wellbore. The pressure and

temperature range in this benchmarking example from

Han et al. are close to the critical point of CO2 where

small changes in pressure/temperature can cause large

changes in CO2 properties. So the thermal effects

reported tend to be the largest possible. The thermal

effects for the base case problem and leakage scenarios

studied in this paper will also tend to be the largest as

these cases are introduced based upon the Han et al.’s

example properties.

Base case

For the purpose of this work the parameters of the example

problem were modified to a situation in which CO2 is

injected over the whole thickness of the injection zone.

Also depth to the top of the injection zone was increased to

1,000 m and reduced injection-zone thickness to 55 m

(Table 2). Thus, the reservoir now extends from 1,000 to

1,055 m depth. While the reservoir depth was changed, the

pressure variation with depth is kept the same as that for

the benchmark model. Thus, at 1,000 m depth the pressure

is the same as that for the benchmark model at

*8.03 MPa. The reservoir radius was also increased from

5 to 24 km and the number of grids in radial direction from

60 to 400. The grid size in vertical direction remains

unchanged. This example is referred to as the base case, the

thermal response corresponding to which is shown in

Figure 3 (ESM only). After 2 years of injection, the max-

imum temperature change for the base case is 0.7 �C,

compared with 1.4 �C for the benchmark example. When

vertical gridding is neglected and the model is treated as a

single layer, more thermal heating and less cooling caused

by mutual dissolutions will be observed. The uppermost

layer of the base case will go through more vaporization

(and, thus, more cooling) than that of the single-layer case.

This is due to easier flow of the fresh CO2 caused by

increased permeability with respect to the CO2-rich phase

in the uppermost layer. Vaporization and dissolution fronts

for the uppermost layer of the base case will then be ahead

of those for the single-layer case. This implies more dis-

solution and greater heating of the uppermost layer could

be expected. However, the CO2-laden brine in the two-

phase region of the uppermost layer flows continuously by

gravity to the deeper layers, despite its slightly elevated

temperature, as it is being replaced by the buoyant CO2-

rich phase. As a result, less heating will be observed in the

uppermost layer.

The thermal response in Figure 3 (ESM only) can be

further compared with the analytical solution given by

Mathias et al. (2010), considering JT cooling as the only

thermal process. They derived the following equation for

the temperature variation near the wellbore:

Tðr; tÞ ¼

4phkln 1� cDqt

� �þ T0 for

phþ r2

aqlCO2

4phkln

� �þ Tw for

phþ r2

w� r2

8>><>>:

where T0 and Tw are initial temperature and injection-

well temperatures, respectively; q is volumetric injection

rate; rw is injection-well radius; lf is CO2-rich-phase

viscosity; h and k are injection-zone thickness and (lateral)

Table 3 Relative-permeability curves

Aqueous phase: van Genuchten function (1980)

kra ¼ffiffiffiffiffiS�p

1� 1� S�½ �1=k� �k

� �2

where: S� ¼ Sa�Swirr

1�Swirr, k = 0.457,

Swirr = 0.2

CO2-rich phase: Corey curve (1954)

krg ¼ 1� S n

1� S2

where: S ¼ Sa�Swirr

1�Swirr�Sgc, n = 2, Sgc = 0.05

Environ Earth Sci

permeability, respectively; and r and t are radial distance

and time. The constant cD is given by

cD ¼qf cf

/ð1� SwirrÞqf cf þ /Swirrqwcw þ ð1� /Þqscs

where q and c are density and specific heat capacity,

respectively; subscripts f, w, and s denote the CO2-rich

phase, aqueous phase, and rock, respectively; and Swirr

and / are irreducible water saturation and porosity,

respectively. For the analytical model constant JT coef-

ficient of 3 K/MPa and CO2 heat capacity of 1,016 J/

(kg K) were used, considering pressure and temperature

conditions at the injection well. The analytical model

displays less cooling (due to ignored vaporization) and no

heating (because of neglected dissolution) because it does

not consider the thermal effects corresponding to mutual

dissolution of CO2 and brine. The temperature minimum

will be closer to the CMG-GEM solution when the

vaporization of water in CO2 is neglected (Figure 3, ESM

only).

Defining a temperature change below which the thermal

response may be indistinguishable from instrument and

background noise is essential. On the basis of temperature

measurements at the Cranfield project (Nunez-Lopez and

Hovorka 2012) shown in Figure 4 (ESM only), it was

determined that temperature changes [0.15 �C (0.27 F)

may be considered a signal. Capability of acquiring high

resolution DTS data has also been demonstrated at Ketzin

site where the temperature has been extensively monitored

(Reinsch et al. 2013). The area of the thermal pulse is

therefore defined where the temperature change is

[0.15 �C in either direction from the initial temperature.

The modeled thermal pulse after 2 years of injection is

shown in Figure 5 (ESM only) and compared to the pres-

sure pulse and the location of the CO2 plume. The pressure-

change cutoff in defining the pressure pulse is 1 kPa as

suggested by the resolution of pressure sensors currently

used in petroleum industry and background noises such as

tidal effects (Zeidouni 2011 and references therein; Nogues

et al. 2011; Zeidouni and Pooladi-Darvish 2012). Whereas

the pressure pulse reaches 12.1 km from the injection well,

the temperature pulse does not extend beyond 42 m after

2 years. The CO2 plume at this time covers a 190-m radius

area. The thermal pulse covers a much smaller region than

the pressure signal or the CO2 plume (Figure 5, ESM only)

mainly because the large heat capacity of the rock and

brine weakens the temperature signal accompanied by CO2

flow into the reservoir.

Radius of the area covered by the temperature and

pressure pulses and the CO2 plume is plotted versus time in

Figure 6a (ESM only). The CO2 plume and pressure-pulse

radii are larger than the temperature pulse by one and four

orders of magnitudes, respectively.

Co-located leaky-well scenario

To investigate temperature response to leakage from the

storage formation, the base case is modified to allow for

leakage to an above-zone monitoring interval (AZMI)

above the injection horizon. The AZMI is separated from

the injection zone by an otherwise sealing interval. Unlike

the base case the geothermal gradient is accounted for by

considering 0.03 �C/m of temperature increase with depth

and 41.7 �C at the base of the reservoir at 1,053 m depth.

For this case (see Table 1 for detailed properties), thickness

of AZMI and seal is 10 and 5 m, respectively. All petro-

physical properties of AZMI are identical to those of the

injection zone. For convenience, a leaky well located at the

same location as the injection well is considered. This

maximizes the leakage rate to the AZMI while reducing

computational cost and it can be realistic as the injection

well itself may be leaky. The 10-m-thick AZMI is divided

into five layers. The leakage pathway has lateral and ver-

tical permeabilities of 10 mD with a 0.3-m radius.

Significant pressure drop occurs when CO2 flows from

the seal to the AZMI. After 2 years, the pressure drop from

the upper layer of the injection zone to the lower layer of

the AZMI is 2.40 MPa. As a result, the maximum JT

cooling effect occurs at the base of the AZMI, which is also

in closest thermal contact with the injection zone. Tem-

perature variation (after 2 years of injection) with distance

is shown for the base of the AZMI and compared with

those in the top layer of the injection zone and the seal

(only one layer) in Figure 7 (ESM only). The top layer of

the injection zone is initially at lower temperature than the

injected CO2. As a result, the thermal cooling caused by JT

effect appears as a local minimum followed by further

reduction in temperature with radial distance from the

injection well to reach the initial reservoir temperature.

Temperature values in the seal as well as in the AZMI are

characterized by an increase with radial distance from the

leak. As CO2 leaks toward the AZMI, it undergoes

expansion accompanied by significant JT cooling. There-

fore, what enters the AZMI (and is in communication with

the seal) is cooler than the rock and native fluids. Heat

transfer of the leaked CO2 with the rock and fluid system

leads to cooling of the AZMI and of the sealing interval. It

is visible that far from the leak the temperature in the seal

slightly increases which is due to heat transfer from the

injection-induced slightly elevated temperature in the

injection zone. Note that owing to the opposite effect of

pressure and temperature on CO2 viscosity, mobility of the

leaking CO2 changes only slightly. To present results in the

form resembling field DTS data, temperature change (T–Ti)

was plotted versus depth and time at the location of the

well (Fig. 1) for the 2-year injection period. The temper-

ature increases for all the layers of the injection zone

Environ Earth Sci

except for the bottom layer where the temperature change

is negligible. This is because the injected CO2 is at same

temperature as the bottom layer of the injection zone but at

higher temperature than the upper layers. The maximum

temperature increase of *1.5 �C is at the uppermost layer

of the injection zone. In the AZMI, significant cooling is

observed, especially at its base, and cooling becomes more

pronounced with time causing temperature reduction of

*5.7 �C. The temperature change in AZMI’s base already

exceeds the detection limit (0.15 �C) 7 h after start of CO2

leakage. The leakage rate after 2 years (that reaches a

maximum quickly after the start of injection) is 1.55 m3/

day which is 0.7 % of the injection rate at reservoir con-

dition (218.4 m3/day).

The area covered by the temperature pulse in the AZMI

is plotted versus time and is compared with that covered by

the CO2 plume and the pressure pulse in Figure 6b (ESM

only). Results are similar to those given for the injection

zone. As was already concluded in the base case, the area

covered by the temperature pulse is much smaller than that

covered by the CO2 plume and pressure pulse. After

2 years of leakage, the temperature pulse reaches only

13 m from the leak, as compared with 17 and 8,400 m for

CO2 plume and pressure pulse, respectively.

Sensitivity analysis

In the following, sensitivity of the temperature pulse to

different AZMI parameters is investigated. In analyzing

temperature response in the injection zone, Han et al.

(2012) identified the following as key parameters affecting

temperature change: porosity, vertical and horizontal per-

meabilities, injection rate, and injected CO2 temperature.

These parameters are varied for the AZMI and their effect

on the temperature pulse is investigated. Other important

parameters are leakage-pathway permeability, leakage

interval (length of leakage pathway) and rock heat capac-

ity. Throughout the analysis injection-zone parameters

remain unchanged.

Leakage-pathway permeability (kl)

Increasing leakage-pathway permeability increases leakage

rate, which enhances thermal advection in the AZMI.

Increasing leak permeability from 10 to 100 mD increases

the radius of the temperature pulse from 13 to 26 m

(Fig. 2a). Also, increasing thickness of the seal separating

the injection zone and the AZMI increases the leakage

interval. As a result, the leakage rate decreases, leading to a

reduced extent of the temperature pulse.

AZMI porosity

Varying AZMI porosity (assuming all pores contribute to

flow) may affect the extent of the thermal pulse. Higher

porosity limits CO2 migration and reduces heat advection

in the upper zone as a result of which temperature signal

should cover a smaller area. Simultaneously, higher

porosity allows for less heat adsorption by rock grains.

Heat adsorption by rock grains is a function of contact

surface area as well as grain volume. Based on Carman–

Kozeny equation (Carman 1956) increasing porosity at

constant permeability means increased contact surface

area per unit grain volume. The increased contact area

increases heat convection between the fluids and rock

grains. As a consequence, fluids and rock grains reach

equilibrium more quickly. It follows that the thermal

pulse covers a smaller area in response to increasing the

porosity. On the other hand, CMG-GEM already assumes

local temperature equilibrium and therefore, the local

temperature difference between the fluids and rock as

well as surface area does not appear in the governing

energy balance equation coupled with the flow equation

(see CMG-GEM 2012, Appendix F). The local tempera-

ture equilibrium assumption can be justified considering

large ([0.2) Biot number for other applications of heat

transfer coupled with flow in porous media (Satik and

Yortsos 1996; Javadpour et al. 2003). Thanks to smaller

heat adsorption by rock grains, temperature signal may

cover a larger area in response to increasing the porosity.

It follows that, in single-phase flow, these competing

processes may cancel each other out (Han et al. 2012;

Fig. 1 Temperature change (from baseline) at location of leaky well

(and injection well) versus time and depth. Temperature in the

injection zone does not vary with time as it is in immediate contact

with (and therefore determined by) the injection-well temperature

Environ Earth Sci

Fig. 2 Sensitivity of temperature-pulse extent in AZMI for various parameters of a leak permeabilities, b AZMI porosity, c horizontal and

vertical permeabilities, d injected CO2 temperature, e injection rate, f rock heat capacity, g geothermal gradient, and h AZMI thickness

Environ Earth Sci

Oldenburg and Pruess 1998). In the two-phase CO2-brine

system, increased porosity will further contribute to

reducing the thermal signature at the CO2 front by

reducing heat conductivity of the system. The area cov-

ered by the temperature pulse (Fig. 2b) for porosities of

0.1 and 0.3 suggests that the temperature pulse is almost

insensitive to porosity. However, the pulse covers a

slightly smaller area for a higher porosity, showing that

advective force and heat conductivity influence the extent

to which the temperature pulse can travel.

AZMI permeability (kv and kh)

Figure 2c shows the effect of increasing AZMI perme-

ability on the temperature-pulse extent. Both horizontal

and vertical permeabilities are increased by factor of 10.

The temperature-pulse area is decreased at early time but

increased at late-time period. In general, higher perme-

ability leads to a lower pressure gradient, which reduces

the JT cooling effect. Because pressure will diffuse to a

larger area, the temperature pulse travels a longer distance.

After a sufficiently long time, a larger area should be

covered by the temperature pulse. However, at earlier times

the temperature response will be too small to be detected.

The temperature pulse therefore covers a smaller area for

the higher permeability case at early time. Decreasing the

temperature-detection limit from 0.15 to 0.05 �C shortens

this early time period from 500 to 460 days (results not

shown for brevity).

Injected CO2 temperature

Injected CO2 may be at different temperature than the

brine-saturated injection zone. Non-isothermal modeling of

the wellbore is required to determine the temperature of

CO2 at the sandface. Lu and Connell (2008) and Han et al.

(2010) studied the non-isothermal behavior of CO2 injec-

tion wells by solving coupled heat, mass and momentum

equations using various fluid and thermodynamic proper-

ties. They have shown that CO2 generally undergoes

cooling followed by heating as the depth increases, a

process which has been also observed in the field as

reported for Ketzin project (Liebscher et al. 2013). As a

result the sandface temperature at the injection well may be

warmer or cooler than that of the injection zone. Injection

of cooler CO2 makes leaking CO2 even cooler, leading to a

stronger temperature signal, and, as a result, the area

covered by the temperature pulse will be larger. The effect

of changing injected CO2 temperature on the temperature-

pulse radius is shown in Fig. 2d. After 2 years of injection,

the temperature signal covers 30 m, for an injection tem-

perature of 38 �C, as compared with 13 m, for a 41.7 �C

injection temperature.

Injection rate

Increasing the (sandface) injection rate leads to an

increased leakage rate and faster advection of the thermal

signal. Therefore, the temperature pulse will cover a larger

area in response to injection at a higher rate. Figure 2e

shows that doubling the injection rate increases the radius

of the temperature pulse by 38 %.

Heat capacity of the rock

Increasing heat capacity of the rock implies that more heat

is required to increase its temperature, making the tem-

perature signal weaker and the area covered by the tem-

perature pulse smaller. Figure 2f shows the temperature-

pulse radius for various values of a rock’s heat capacity.

Increasing the rock capacity [from an initial value of

1,000 J/(kg K) given by Han et al. 2012] by 100 J/(kg K)

reduces the radius of the temperature pulse by only

*0.2 m. Decreasing the heat capacity by 100 J/(kg K)

increases the radius by 0.2 m as well.

Temperature gradient

The temperature-pulse extent is compared with that when

the thermal gradient is neglected in Fig. 2g. Based on this

figure, the thermal pulse weakens when the geothermal

gradient is considered. This is mainly because when geo-

thermal gradient is neglected, the AZMI is warmer and

leakage-related cooling is more pronounced. Thus, a case

in which the cooling due to JT effect just counterbalances

the lower temperature in the overlying aquifer due to the

geothermal gradient is possible. In such a case, leakage-

induced temperature signal will be negligible.

AZMI thickness

Thicker AZMI accommodates more vertical than lateral

flow of the leakage fluid while allowing for higher leakage

rate owing to its increased capacity. Thus, radial extent of

the temperature pulse will be reduced at the beginning due

to decreased lateral flow. But, after some time the increased

leakage rate will generate a larger thermal pulse. This is

shown in Fig. 2h where temperature pulse extent is com-

pared to a case with AZMI thickness of 20 m. At the end of

the injection period, the pulse extent is increased by 10 %

when doubling the thickness of AZMI.

Not-co-located leaky-well scenario

Because the leaky well is unlikely to be co-located with the

injection well, investigation of the temperature signal to

Environ Earth Sci

leakage through a well at a distance from the injection well

is important. Both CO2 and brine may be leaking in this

situation, making the thermal response more complex. To

investigate this problem, the co-located leaky-well exam-

ple above was modified by considering a 100-m distance

between the injection well and the leaky well (see Table 1

for case properties). Cartesian gridding is used for this

configuration, although it makes numerical calculations

more costly. To reduce the computational cost, the system

is made symmetric and only a quadrant of the system is

modeled. The system involves two identical leaky wells at

equal distance from the injection well. The quadrant model

then includes one leaky well (half the size of the leaky well

in the full system) and one injection well injecting

12.5 kton/year, or *17,300 sm3/day, which is � of the

central injection well of the full system. The same mesh

size is used in both x and y directions, covering the first

100 9 100 m2 around the well with a 2 9 2 m2 grid and

increasing it gradually beyond this area to ensure infinite

outer boundaries.

The HYBRYD refinement option is used to model the

cylindrical shape of the leaky well in Cartesian mesh. This

option allows refinement of a Cartesian grid block into a

cylindrical grid within the center plus an outer block,

whose outer boundary fits the shape of the parent grid.

Using 2 9 2 m2 for the parent grid of the leaky well, the

radius of the leaky well will be 0.37 m. The vertical

gridding and all other reservoir and leak properties are

identical to those of the co-located leaky-well case. The

total number of grid cells is 124,215 (91 9 91 9 15). A

thermal gradient of 0.03 �C/m is used.

The temperature variation at the base of AZMI is shown

in the Fig. 3. Note that temperature increases slightly at the

beginning owing to leakage of warmer brine from the

injection zone to the AZMI. After *6 months of injection,

CO2 reaches the leaky well, and the leakage fluid is no

longer a single-phase brine but a mix of CO2-rich and

aqueous phases (see Fig. 3 for aqueous- and CO2-rich-

phase leakage velocities). As a result of leakage and

expansion of the CO2-rich phase, *5 �C of cooling will be

observed in the AZMI. The pressure reduces from

10.26 MPa at the upper layer of injection zone to 8.36 MPa

at the lower layer of the AZMI at the end of the injection

period. The leakage rate at this time is 0.95 m3/day which

is 1.7 % of the injection rate at reservoir conditions

(55.8 m3/day). Temperature change (T–Ti) along the leaky-

well path is shown in Fig. 4 versus time and depth. In the

injection zone, slight cooling is observed at the base and

top of the reservoir, the latter of which is due to heat

exchange with the cooling seal, and the former corresponds

to leakage of warmer brine from the injection zone to the

AZMI. Although thermal cooling in the leaky well occurs

throughout the leakage pathway in the seal and the AZMI,

highest cooling within the AZMI occurs at its base of

owing to maximal CO2 expansion. The temperature change

in the AZMI (at the location of the leak) becomes detect-

able 1 day after start of CO2 leakage. The pressure

reduction from lower layer of the AZMI to its upper layer

is only *0.1 MPa compared to 1.9 MPa from the upper

layer of the injection zone to the lower layer of the AZMI

at the end of the injection period (Fig. 5, right panel). The

extent to which the CO2-rich phase may expand depends

on pressure decline and, thus, Darcy velocity of the leak.

Everything else being equal, the pressure drop (and phase

velocities) is larger if the leakage conduit is smaller. The

temperature variation along the leaky well is evaluated

considering various well radii at the end of injection period

(Fig. 5, left panel). As the size of the conduit increases, the

velocity of the CO2-rich phase decreases and CO2 goes

through less pressure reduction, resulting in less cooling.

Temperature response correlates well with CO2-rich-phase

velocity (Fig. 5). The higher the velocity, the more pro-

nounced the thermal cooling. The velocity of the fluids

decreases toward the top of the AZMI because of the upper

closed boundary. Note that, because of the pressure decline

along the leakage path, CO2 solubility also decreases,

leading to CO2 exsolution from the aqueous phase, which

may cause cooling effect if the pressure drop is significant

and/or the aqueous-phase leakage rate is high. Given the

conditions of this scenario, however, the change in aque-

ous-phase CO2 concentration upon leakage is negligible, as

is CO2 exsolution.

Leaky fault scenario

Running a temperature log or DTS system along wells

believed to be leaking is a conventional use of thermal

monitoring, and it has been shown to be effective in leak

0 0.5 1 1.5 2

Time (yr)

Temperature

Gaseous phase velocity

Aqueous phase velocity

Fig. 3 Temperature change at base of AZMI in response to mixed

leakage of brine and CO2. Leakage rates of aqueous and CO2-rich

phases through leaky well also shown

Environ Earth Sci

detection under certain conditions. However, in spite of

possibly large temperature perturbations at the location of

the leaky well (as shown earlier), it may be difficult to

detect well leakage based on temperature data at observa-

tion wells distant from the leaky well. Another setting to be

assessed is a fault providing a planar pathway for leakage.

Locations of faults are generally known and determined

based on a combination of geological, drilling, and seismic

data. Geometries in which potential observation wells

intersect faults include vertical wells crossing dipping faults

or non-vertical wells emplaced to intersect faults. Leakage

along the fault may cause a detectable temperature signal.

An observation well intersecting the fault may then provide

a temperature anomaly that can be translated to leakage.

In the following section, the temperature response caused

by fault leakage is investigated. The not-co-located leaky-

well scenario was modified to account for thermal effects of

fault leakage. All properties of the injection zone, AZMI,

and sealing interval, as well as injection rate and mesh size,

are identical to those of the not-co-located leaky-well case.

The system includes two identical and parallel leaky faults

at equal distances (100 m) from the injection well. Again,

symmetry allows for modeling using a quadrant. The 2-m-

wide fault zone, is laterally conductive, and is assumed to be

vertical. Fault permeability is 10 mD in all directions (see

Table 1 for detailed specifications).

Temperature variation versus depth at the base of the

AZMI at cell nx = 51 and ny = 1 is shown in Fig. 6 (left

panel). Injection well is located at cell nx = 1 and ny = 1.

Brine leakage before CO2 breakthrough causes a larger

thermal signal than it does in the not-co-located leaky-well

case. However, the temperature rise caused by brine leak-

age remains insignificant. In addition, unlike the not-co-

located leaky well, in which continuous temperature

Fig. 4 Variation of temperature change with time and depth along

leaky well. Note that unlike the co-located leaky well case, the

temperature in the injection zone slightly varies with time

Fig. 5 Variation of temperature (solid curves on left panel), CO2-rich-phase velocity (dashed curves on left panel), and pressure (dotted curves

on right panel) along leaky well at end of injection period (2 years) considering various leaky-well radii

Environ Earth Sci

decline was observed after CO2 breakthrough, the tem-

perature does not monotonically decrease. This is due to a

quick pressure buildup in the AZMI, which quickly reduces

the pressure drop over the leakage path (and therefore

reduces CO2-rich-phase velocity, as shown in Fig. 6). The

pressure drop from the upper layer of the injection zone to

lower layer of the AZMI is shown in Figure 10 (ESM only)

for both leaky fault and not-co-located leaky well scenar-

ios. The pressure drop for the leaky fault scenario reaches a

maximum after 5 month of injection (*1 month before

CO2 breakthrough at the fault) after which it declines

quickly toward the initial pressure difference between

AZMI and injection zone (*0.1 MPa). The pressure in the

AZMI starts declining after CO2 breakthrough but it

remains far from the initial difference for the not-co-

located leaky-well case. Note that in spite of the reduced

pressure drop for the leaky-well scenario, the CO2 leakage

velocity keep increasing (Fig. 3) due to increased relative

permeability of CO2-rich phase in consequence of its

increased saturation in AZMI.

The reduced pressure drop leads to less JT cooling and

non-monotonic temperature change after CO2 breakthrough.

This behavior is also shown in Fig. 7, in which a temperature

decline from *0.5 to 0.75 year followed by increase in

temperature is observed in the AZMI. Before CO2 break-

through, temperature in the AZMI increases as a result of

warmer brine leakage. In the uppermost layer of the AZMI,

temperature increased by 0.16 �C after 6 months, when CO2

leakage to the AZMI began. The thermal signal induced in

the AZMI by brine leakage exceeds the temperature

threshold (0.15 �C) 160 days after start of brine leakage,

while induced cooling due to CO2 leakage becomes

detectable 55 days after start of CO2 leakage. CO2 leakage

rate after 2 years of injection reaches 9.4 m3/day (*16.4 %

of injection rate at reservoir conditions).

The sensitivity of the temperature response to fault-zone

width is examined in Fig. 8. The fault-zone width is varied

from 1 to 3 m. JT cooling is more pronounced in the 1-m-

thick fault zone due mainly to larger pressure drops cor-

responding to higher CO2-rich-phase velocities.

Discussion

The thermal signature was modeled by considering some

idealizations and simplifications. It was assumed that both

leaky well and fault are homogeneous. However, assuming

that the fault is homogeneously permeable for its entire

length may be an oversimplification of the fault structure,

although it may be considered an end-member fault structure.

As shown earlier, such a structure may not generate a sig-

nificant temperature change. In the field, fault internal

structure is observed to be highly heterogeneous and is dif-

ficult to access. Spatial variations in porosity and perme-

ability and related non-uniform architecture are needed to

account for fault heterogeneity. One way to represent this

heterogeneity may be using multiple leaky wells contained in

the fault plane. Representing the fault zone using discrete

leaky wells will result in similar thermal behavior, and a large

cooling signal may be observed along the fault as a result.

As leaking CO2 continues to rise and expand toward

shallower depths, it may convert from supercritical fluid to

liquid phase (Pruess 2011). In this study, phase change was

not considered which occurs when there is a large interval

between injection formation and shallow zones. Such phase

change will be associated with further cooling similar to

that experienced from movement of natural gas in wells

and pipelines. As shown by Pruess (2011), liquid CO2

0 0.5 1 1.5 2

Time (yr)

Temperature

Gaseous phase velocity

Aqueous phase velocity

Fig. 6 Temperature variation at base of AZMI at the location of the

leak, cell nx = 51 and ny = 1. Injection well is located at cell nx = 1

and ny = 1

Fig. 7 Variation of temperature change with time and depth along

vertical well completed at fault center (nx = 51, ny = 1) throughout

all layers. Injection well is located at nx = 1, ny = 1

Environ Earth Sci

starts boiling as it goes through additional pressure decline,

leading to significant temperature reduction. Expansion of

the gas phase as it enters shallower depths causes further

cooling. Three-phase (aqueous, and liquid and gaseous

CO2-rich phases) flow may evolve consequently. Such a

three-phase flow is slower (owing to relative permeability

of the phases), allowing for stronger heat exchange with the

wall rocks. As a result, liquid CO2 heats up and may start

boiling, leading to increased CO2 flow rates again. Within

the three-phase zone, liquid and gas saturations, and

therefore CO2 leakage rates, may show strong fluctuations

(Pruess 2011). Also, creation of CO2 bubbles could sup-

press leakage rate because it increases viscosity of the brine

(Wang and Clarens 2012). The combined effects of these

processes should be studied to determine the temperature

signal along the leakage pathway toward the surface.

Impact of geochemical reactions was neglected in this

study. As discussed by Han et al. (2012), geochemical

interactions between CO2, rock, and native brine may

cause a visible thermal signature. Further research to

account for the thermal response associated with geo-

chemical interactions should be pursued. The analysis in

this study was limited to the injection period. Further

investigation is required to examine temperature variations

during the post-injection period, when CO2 and brine

backflow from the AZMI may occur.

Conclusions

The thermal response to CO2 injection was simulated using

the CMG-GEM simulator. Simulating a benchmark prob-

lem, the thermal results of which have been reported on the

basis of TOUGH2-ECO2N, it was found that temperature

responses of TOUGH2 and CMG-GEM have similar fea-

tures, providing cooling close to the well bore and heating

toward the CO2 front. However, the magnitude of the

temperature response is different in spite of closely

matching PVT properties used in CMG-GEM and

TOUGH2. This difference is rooted in differently calcu-

lating the enthalpy in CMG-GEM versus TOUGH2. As a

result, the temperature response corresponding to mutual

dissolutions given by TOUGH2 is larger than that given by

CMG-GEM.

A scenario with a leaky injection well (co-located leaky-

well case) was defined so that temperature behavior in the

injection zone and the AZMI can be examined. At a given

time, moving away from the injection well into the injec-

tion zone, temperature drops owing to the JT effect, before

increasing to equilibrate with the unperturbed temperature

of the injection zone. Leaking CO2 undergoes a significant

pressure drop and JT cooling upon leakage into the AZMI.

The CO2 entering the AZMI may therefore be cooler than

the AZMI brine. As a result, temperature in the AZMI may

Fig. 8 Variation of temperature (solid curves on left panel), CO2-rich-phase velocity (dashed curves on left panel) and pressure (dotted curves

on right panel) along cell nx = 51, ny = 1 at end of injection period (2 years) considering various widths of fault zone

Environ Earth Sci

increase monotonically with radial distance from the

injection well. Sensitivity analysis in the co-located leaky

well case revealed that the area covered by the temperature

pulse (defined by a temperature change of [0.15 �C)

depends more on boundary conditions than on static pe-

trophysical parameters of the injection zone or the AZMI.

The area is highly sensitive to injection-well parameters,

including injection rate and injection temperature, leakage-

pathway parameters, such as leakage conduit permeability,

and the geothermal gradient.

The temperature signal was further investigated for a

leaky well some distance from the injection well (not-co-

located leaky-well case). Slight heating was observed ini-

tially owing to leakage of warmer brine from the injection

zone to the AZMI. After CO2 breakthrough, JT cooling

dominated the thermal signature, causing monotonic tem-

perature decline at the leaky well. Temperature variation

along the leaky well was evaluated by considering various

well radii, indicating that as the size of the leakage pathway

increased, pressure drop decreased and CO2 experienced

less expansion, resulting in less cooling. Replacement of

the leaky well by a 2-m-wide leaky fault resulted in less

cooling, although more heating due to brine leakage was

observed prior to CO2 breakthrough at the fault.

In summary, it was shown that the temperature pulse

caused by CO2 leakage is local and restricted to the

immediate vicinity of the leakage pathway. However,

leakage can result in large temperature changes along the

leakage pathway. Running temperature logs along suspi-

cious wells and/or wells intersecting suspicious faults may

then provide useful information on the leakage potential

and amount.

Compared to temperature, the pressure pulse covers a

considerably larger area proving to be more useful with

sparsely spaced sensors. Unlike the temperature signal

which depends on the leaking fluid velocity, the strength of

the pressure signal is determined by the leakage rate.

Increasing the leakage conduit cross-sectional area

increases the leakage rate and thus increases the pressure

change in the AZMI. However, it decreases the flow

velocity, and therefore, reduces temperature cooling and

signal. It was shown that AZMI temperature may increase

for brine leakage and decrease for leakage of CO2. How-

ever, either CO2 or brine leakage induces pressure increase

in the AZMI. In addition, pressure changes in the AZMI

may be induced by geomechanical effects and not related

to any leakage (Kim and Hosseini 2013). However, tem-

perature changes in AZMI are only due to leakage.

Acknowledgments This project was funded in part by EPA STAR

grant R834384. The authors wish to thank three anonymous reviewers

as well as Axel Liebscher of GFZ German Research Centre for

Geosciences for their careful reviews of the manuscript and sugges-

tions for improvements. The authors thank Computer Modelling

Group Ltd. for providing the option of free access to the CMG sim-

ulation package. Amin Badamchizadeh and Vijay Shrivastava are

acknowledged for their help in applying CMG-GEM thermal mod-

eling of CO2 storage. Thank you also to Lana Dieterich who edited

the manuscript. Publication authorized by the Director, Bureau of

Economic Geology.

References

Andre L, Azaroual M, Peysson Y, Bazin B (2011) Impact of porous

medium desiccation during anhydrous CO2 injection in deep

saline aquifers: up scaling from experimental results at labora-

tory scale to near-well region. Energy Proced 4:4442–4449.

doi:10.1016/j.egypro.2011.02.398

Andre L, Azaroual M, Menjoz A (2010) Numerical simulations of the

thermal impact of supercritical CO2 injection on chemical

reactivity in a carbonate saline reservoir. Transp Porous Media

82(1):247–274. doi:10.1007/s11242-009-9474-2

Bakker RJ (2003) Package FLUIDS 1. Computer programs for

analysis of fluid inclusion data and for modelling bulk fluid

properties. Chem Geol 194(1–3):3–23

Bielinski A, Kopp A, Schutt H, Class H (2008) Monitoring of CO2

plumes during storage in geological formations using tempera-

ture signals: numerical investigation. Int J Greenhouse Gas Cont

2(3):319–328. doi:10.1016/j.ijggc.2008.02.008

API Technical Data Book, Petroleum Refining (1983) 4th edn.

Washington

Canjar LN, Manning FS (1967) Thermodynamic properties and

reduced correlations for gases. Gulf Publishing Co., Houston

Carman PC (1956) Flow of gases through porous media. Butter-

worths, London 1

CMG-GEM (2012) CMG-GEM version 2012 user’s guide. Computer

modelling group Ltd. http://www.cmgl.ca/software/gem.htm

Han WS, Stillman GA, Lu M, Lu C, McPherson BJ, Park E (2010)

Evaluation of potential nonisothermal processes and heat

transport during CO2 sequestration. J Geophys Res 115(B7):

B07209. doi:10.1029/2009JB006745

Han W, Kim K-Y, Park E, McPherson B, Lee S-Y, Park M-H (2012)

Modeling of spatiotemporal thermal response to CO2 injection in

saline formations: interpretation for monitoring. Transp Porous

Media 93(3):381–399. doi:10.1007/s11242-012-9957-4

Harvey AH (1996) Semiempirical correlation for Henry’s constants

over large temperature ranges. AIChE J 42(5):1491–1494.

doi:10.1002/aic.690420531

Henninges J, Huenges E, Burkhardt H (2005) In situ thermal

conductivity of gas-hydrate-bearing sediments of the Mallik

5L-38 well. J Geophys Res-Sol Ea 110(B11)

Hoang H, Mahadevan J, Lopez H (2011) Interpretation of wellbore

temperatures measured using distributed temperature sensors during

hydraulic fracturing. In: SPE hydraulic fracturing technology

conference, SPE 140442, Woodlands, TX, 24–26 January 2011

Hovorka SD, Benson SM, Doughty C, Freifeld BM, Sakurai S, Daley

TM, Kharaka YK, Holtz MH, Trautz RC, Nance HS, Myer LR,

Knauss KG (2006) Measuring permanence of CO2 storage in

saline formations: the Frio experiment. Environ Geosci

13(2):105–121. doi:10.1306/eg.11210505011

Hovorka SD, Meckel TA, Trevino RH (2013) Monitoring a large-

volume injection at Cranfield, Mississippi—project design and

recommendations. Int J Greenhouse Gas Cont 18:345–360

Hurter S, Garnett A, Bielinski A, Kopp A (2007) Thermal signature of

free-phase CO2 in porous rocks: detectability of CO2 by

temperature logging. In: SPE Offshore Europe, SPE 109007,

Aberdeen, Scotland, UK, 4–7 September 2007

Environ Earth Sci

Javadpour F, Pooladi-Darvish M, Jeje A, Shen LC (2003) Effect of

transient temperature on MWD resistivity logs. Petrophysics

44:435–445

Kim S, Hosseini SA (2013) Above-zone pressure monitoring and

geomechanical analyses for a field-scale CO2 injection project in

Cranfield, MS. Greenhouse Gas Sci Technol. doi:10.1002/ghg.1388

Koschel D, Coxam JY, Rodier L, Majer V (2006) Enthalpy and

solubility data of CO2 in water and NaCl(aq) at conditions of

interest for geological sequestration. Fluid Phase Equilib

247(1–2):107–120. doi:10.1016/j.fluid.2006.06.006

Liebscher A, Moller F, Bannach A, Kohler S, Wiebach J, Schmidt-

Hattenberger C, Weiner M, Pretschner C, Ebert K, Zemke J

(2013) Injection operation and operational pressure–temperature

monitoring at the CO2 storage pilot site Ketzin, Germany—

design, results, recommendations. Int J Greenhouse Gas Control

15:163–173

Lu M, Connell LD (2008) Non-isothermal flow of carbon dioxide in

injection wells during geological storage. Int J Greenhouse Gas

Cont 2(2):248–258. doi:10.1016/S1750-5836(07)00114-4

Marinaro G, Etiope G, Gasparoni F, Furlan F, Bruni F (2011) Gas

seepage detection and monitoring at seafloor. In: 10th offshore

Mediterranean conference and exhibition, Doc. ID 2011-045,

Ravenna, Italy, 23–25 March 2011

Mathias SA, Gluyas JG, Oldenburg CM, Tsang CF (2010) Analytical

solution for Joule–Thomson cooling during CO2 geo-sequestra-

tion in depleted oil and gas reservoirs. Int J Greenhouse Gas

Cont 4(5):806–810. doi:10.1016/j.ijggc.2010.05.008

Nogues JP, Nordbotten JM, Celia MA (2011) Detecting leakage of

brine or CO2 through abandoned wells in a geological seques-

tration operation using pressure monitoring wells. Energy

Procedia 4:3620–3627

Nunez-Lopez V, Hovorka SD (2012) Subsurface monitoring of large-

scale CO2 injection at SECARBC’s phase III Cranfield Site. In:

carbon management technology conference, SPE 151504,

Orlando, Florida, USA, 7–9 February 2012

Oldenburg CM (2007) Joule–Thomson cooling due to CO2 injection

into natural gas reservoirs. Energy Convers Manage

48(6):1808–1815. doi:10.1016/j.enconman.2007.01.010

Oldenburg CM, Pruess K (1998) Layered thermohaline convection in

hypersaline geothermal systems. Transp Porous Media

33(1–2):29–63. doi:10.1023/A:1006579723284

Peng DY, Robinson DB (1976) A new two-constant equation of state.

Ind Eng Chem Fundam 15:59–64

Pruess K (2005) TOUGH2 fluid property module for mixtures of

water, NaCl, and CO2. Lawrence Berkeley National Laboratory,

Berkeley

Pruess K (2011) Integrated modeling of CO2 storage and leakage

scenarios including transitions between super- and subcritical

conditions, and phase change between liquid and gaseous CO2.

Greenhouse Gases Sci Tech 1(3):237–247. doi:10.1002/ghg.24

Pruess K, Oldenburg C, Moridis G (1999) TOUGH2 user’s guide,

version 2.0. Report LBNL-43134, Lawrence Berkeley National

Laboratory, Berkeley

Reinsch T, Henninges J, Asmundsson R (2013) Thermal, mechanical

and chemical influences on the performance of optical fibres for

distributed temperature sensing in a hot geothermal well.

Environ Earth Sci 70(8):3465–3480

Satik C, Yortsos YC (1996) A pore-network study of bubble growth

in porous media driven by heat transfer. J Heat Transfer

118(2):455–462. doi:10.1115/1.2825866

Saul A, Wagner W (1987) International equations for the saturated

properties of ordinary water substance. J Phys Chem Ref Data

16(4):893–901

Singh AK, Baumann G, Henninges J, Gorke UJ, Kolditz O (2012)

Numerical analysis of thermal effects during carbon dioxide

injection with enhanced gas recovery: a theoretical case study for

the Altmark gas field. Environ Earth Sci 67(2):497–509

Vinsome PKW, Westerveld J (1980) A simple method for predicting

cap and base rock heat-losses in thermal reservoir simulators.

J Can Petrol Technol 19(3):87–90

Wang S, Clarens AF (2012) The effects of CO2-brine rheology on

leakage processes in geologic carbon sequestration. Water

Resour Res 48(8):W08518. doi:10.1029/2011wr011220

Wangen M (1995) The blanketing effect in sedimentary basins. Basin

Res 7(4):283–298. doi:10.1111/j.1365-2117.1995.tb00118.x

Zeidouni M (2011) Analytical and inverse models for leakage

characterization of CO2 storage, Chapter 5, PhD dissertation,

University of Calgary, Calgary, Canada

Zeidouni M, Pooladi-Darvish M (2012) Leakage characterization

through above-zone pressure monitoring: 1—inversion

approach. J Petrol Sci Eng 98–99:95–106

Zeidouni M, Pooladi-Darvish M, Keith D (2009) Analytical solution

to evaluate salt precipitation during CO2 injection in saline

aquifers. Int J Greenhouse Gas Cont 3(5):600–611

Zhang YK (1993) The thermal blanketing effect of sediments on the

rate and amount of subsidence in sedimentary basins formed by

extension (vol 218, pg 297, 1993). Tectonophysics 225(4):551.

doi:10.1016/0040-1951(93)90317-D

Environ Earth Sci

Monitoring above-zone temperature variations associated with CO2 and brine leakage from a storage...

Documents