International Journal of Greenhouse Gas Control 42 (2015) 188–199

Contents lists available at ScienceDirect

International Journal of Greenhouse Gas Control

journal homepage: www.elsevier.com/locate/ijggc

Satellite interferometry for high-precision detection of ground

deformation at a carbon dioxide storage site
Sergey Samsonov a,∗ , Magdalena Czarnogorska a , Don White b
a
Canada Centre for Mapping and Earth Observation, Natural Resources Canada, Ottawa, ON, Canada
b
Geological Survey of Canada, Natural Resources Canada, Ottawa, ON, Canada

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

Article history: At the Aquistore CCS site, located in the southeastern Saskatchewan, Canada, carbon dioxide (CO2 ) is
Received 24 December 2014 to be injected at variable rates of up to 1500 tonne/day. The storage reservoir consists of a dominantly
Received in revised form 29 July 2015 clastic, brine-filled interval (Deadwood and Winnipeg formations) which reside at 3150–3350 m depth.
Accepted 31 July 2015
Ground deformation at this site is being monitored to track pressure-induced uplift and potential upward
Available online 24 August 2015
migration of CO2 through faults and fractures. Deformation monitoring is conducted using space-borne
Differential Interferometric Synthetic Aperture Radar (DInSAR), capable of achieving millimeter precision
Keywords:
and meter spatial resolution over the entire monitored area. During June 2012–October 2014, prior to
Aquistore
Carbon capture and storage CO2 injection, two ascending and two descending high-resolution RADARSAT-2 data sets were acquired
Satellite interferometry and simultaneously processed with the advanced Multidimensional Small Baseline (MSBAS) DInSAR pro-
DInSAR ducing vertical and horizontal East-West deformation time series with six days temporal sampling, four
MSBAS times more frequent than the repeat cycle of each individual data set. Two years of monitoring prior to
RADARSAT-2 the onset of (CO2 ) injection allowed measurement of the deformation field of the background natural
Ground deformation and anthropogenic processes. Vertical and horizontal ground deformation was detected with the rates
of ±1.0 and ±0.5 cm/year and with precision of 0.3 and 0.2 cm/year (2) correspondingly. Background
motion (shape and magnitude) may resemble deformation signals due to potential upward migration of
CO2 through faults and fractures. Analytic elastic and poroelastic modeling was performed to estimate
the ground deformation that will be produced when injection begins. For this purpose, rock properties
determined from geophysical well logs and in situ temperature and pressure were used. For the elastic
model it was determined that a maximum of vertical deformation of 1.6 cm/year will be located around
the injection well, whereas the maximum of horizontal deformation of 0.6 cm/year will be located about
3 km away from the injection well. For a more realistic poroelastic model, it was determined that maxi-
mum vertical deformation will not exceed 1.6 cm and maximum horizontal deformation will not exceed
0.1 cm/year during the entire 25 year injection cycle. According to this model, the established monitoring
network cannot detect predicted horizontal motion since it is below its sensitivity, whereas for detection
of vertical motion ground-based monitoring sites need to be installed near and also at distance from the
injection well. Six day temporal sampling allows determination of the transient uplift phase. A proposed
MSBAS strategy overcomes limitations of the classical DInSAR, such as sparse temporal resolution and
the lack of ability to extract individual deformation components from the line-of-sight retrievals, and can
be implemented at other onshore CCS sites for operational monitoring, using readily available SAR data.
Crown Copyright © 2015 Published by Elsevier Ltd. All rights reserved.

1. Introduction
Carbon Capture and Storage (CCS) is one of a variety of means
that are currently being implemented (IPCC, 2005) to reduce the
amount of anthropogenic carbon dioxide that is released to the
∗ Corresponding author.
E-mail address: sergey.samsonov@nrcan-rncan.gc.ca (S. Samsonov).
atmosphere. CCS refers to a broad range of technologies for the cap-
ture (e.g., from fossil fuel power plants), transport, and long term
storage of CO2 in geological formations (IPCC, 2005; Gibbins and
Chalmers, 2008) including depleted oil and gas reservoirs, saline
aquifers and deep coal beds. CO2 storage also occurs in enhanced
oil recovery operations (Bachu et al., 2007).
Site monitoring is a key component of any CCS project
and is implemented based on the particular site characteris-
tics and requirements. The main objective of monitoring is to

http://dx.doi.org/10.1016/j.ijggc.2015.07.034
1750-5836/Crown Copyright © 2015 Published by Elsevier Ltd. All rights reserved.
 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199 189

demonstrate that the storage site is safely and securely storing
CO2 and that it is not having detrimental effects on the sur-
rounding area. An important component of monitoring is the
measurement of geomechanical deformation due to increasing
reservoir pressure and potentially CO2 leakage through faults
and fractures (Rackley, 2009). Shallow and surface monitoring
used for ground deformation detection comprises tiltmeter and
GPS measurements, and more recently satellite interferometry
(Rutqvist et al., 2010; Vasco et al., 2010; Lubitz et al., 2013;
Rinaldi and Rutqvist, 2013; Rucci et al., 2013; Ramirez and Foxall,
2014).
In this study, we report on baseline Differential Interferomet-
ric Synthetic Aperture Radar (DInSAR) monitoring conducted at
the Aquistore CCS site in southeastern Saskatchewan, Canada,
near Estevan township (Fig. 1). Eighty satellite images have been
processed to image vertical and horizontal ground deformation
for the period of June 2012–October 2014, prior to the start of
CO2 injection. The purpose of this pre-injection study is to char-
acterize the vertical and horizontal ground deformation associated
with background natural and anthropogenic processes. Specific
objectives for the Aquistore site include: (1) establishing the
level of measurement precision and temporal resolution that can
be achieved, (2) constructing a baseline record of deformation
that is demonstrably non-injection related, and (3) determining
whether CO2 -injection related deformation will be discernable
in the presence of other sources (natural or anthropogenic) of
ground deformation. Toward the latter objective, simple analytic
and poroelastic modeling has been conducted to estimate the mag-
nitude and spatial extent of the expected deformation due to CO2
injection. Furthermore, we demonstrate how vertical and horizon-
tal deformation time series allow the detection of deformation
trend changes with minimal delay (e.g. Samsonov et al., 2013a),
which is not possible when only mean deformation (or deforma-
tion rate) is known (e.g. Rucci et al., 2013). These results are used
to evaluate the Aquistore ground network design and the ongoing
monitoring strategy.
2. The Aquistore CCS project
The Aquistore CCS project (Worth et al., 2014), started in 2012,
focuses on research, development, testing and adaptation of mon-
itoring methods for measurement and verification of CO2 storage.
A key objective is to integrate various data in order to deter-
mine subsurface fluid distribution, pressure change and associated
ground deformation. The Aquistore site serves as buffer storage for
the world’s first commercial post-combustion CO2 capture plant
at the Boundary Dam coal-fired power station. At the Aquistore
site CO2 injection will occur within brine-filled clastic strata of
the Deadwood and Winnipeg formations, which are the deepest
sedimentary units in the Williston Basin, and are below any oil
producing and potash-bearing formations (Norford et al., 1994).
The Deadwood and Winnipeg formations are located at more than
3000 m depth. Similar deep saline formations are found elsewhere
in western Canada, throughout North America and the world. The
rocks in these formations are well-suited for the storage of CO2
because they are porous and permeable, have huge volumes, and
are overlain by effective barriers and geologic seals that will impede
the upward migration of CO2 .
3. Satellite data and processing methodology
High-precision, high-resolution deformation maps are com-
puted using DInSAR processing methodology (Massonnet and Feigl,
1998; Rosen et al., 2000; Hooper et al., 2012) from the Synthetic
Aperture Radar (SAR) data repeatedly acquired by space-borne or
occasionally air-borne sensors. A number of advanced techniques
have been developed in order to mitigate temporal decorrelation
and to remove orbital and atmospheric noise affecting DInSAR
retrievals (Ferretti et al., 2001; Berardino et al., 2002; Hooper,
2008; Samsonov, 2010; Samsonov et al., 2011). In favorable con-
ditions DInSAR achieves millimeter precision in measuring ground
deformation at meter-scale areal resolution over a large area. Oper-
ational monitoring limitations of the early DInSAR included the

Fig. 1. Location of Aquistore CO2 storage site in Saskatchewan, Canada. (a) Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) 30 m resolution
Digital Elevation Model (DEM). RADARSAT-2 frames are outlined in black. Region of interest is outlined in brown. Extent in top-right corner shows location of study region
in North America. (b) LIDAR 1 m resolution DEM plotted over ASTER DEM. Reference region “R” assumed to be stable. Monitoring sites NE1, NE2, SE1, SE2, SE3, SITE, SW1,
NW1, NW2 are plotted in black. Points P1-P9 experiencing fast ground deformation are plotted in red. Location of injection site is plotted as red diamond. A – Souris river, B
– Boundary Dam Power Station, C – Rafferty Dam, D – Long Creek River, E – Boundary Dam, F – Boundary Dam Reservoir, GG’ – Rafferty-Boundary diversion canal.
190 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199

Table 1
SAR data sets used in this study: RADARSAT-2 Spotlight 18 (SLA18), Spotlight 12 (SLA12), Wide 2 Ultra-Fine 7 (U7W2); time span (in YYYYMMDD format), range × azimuth
SLC resolution, range × azimuth multilooking, azimuth  and incidence  angles, number of available SAR images N, and number of calculated interferograms M for each data
set.

InSAR set Span SLC res., m Multilook ◦ ◦ N M

R2 SLA18 (dsc) 20120605–20140923 1.6 × 0.8 3 × 12 351 44 25 184

R2 SLA12 (asc) 20120716–20140916 1.6 × 0.8 3 × 12 −170 40 20 156
R2 U7W2 (asc) 20120619–20140820 1.6 × 2.8 4×3 349 37 18 110
R2 U7W2 (dsc) 20120615–20141003 1.6 × 2.8 4×3 −170 37 17 81

Total 20120605–20141003 80 531

sparse temporal resolution due to the long (i.e. 24–41 days) satel-
lite orbital cycle and the lack of ability to distinguish vertical and
horizontal motion in the line-of-sight (LOS) retrievals. These lim-
itations are overcome in the advanced Multidimensional Small
Baseline (MSBAS) DInSAR that computes two dimensional (2D) ver-
tical and horizontal East-West deformation by combing SAR data
from various sensors and acquisition geometries, improving over-
all temporal resolution (Samsonov and d’Oreye, 2012; Samsonov
et al., 2013b, 2014a,b,c,d). While MSBAS is capable of measuring all
three components of deformation (3D), such processing requires
data from other than the near-polar orbiting sensors. Availability
of space-borne SAR data has been steadily improving making this
remote sensing technique a valuable, cost-effective alternative or
an addition to GPS and leveling.
In order to achieve high temporal and high spatial resolution
of deformation measurements at the Aquistore CCS site, high-
resolution ascending and descending RADARSAT-2 Spotlight and
Wide Ultra-Fine SAR data were acquired. The temporal and spa-
tial baselines of the RADARSAT-2 imagery are shown in Table 1
and Fig. 2. The Spotlight data consisted of ascending Spotlight 18
(SLA18, example is shown in Fig. 3) and descending Spotlight 12
(SLA12) images, acquired with the incidence angles of 44 and 40
degrees and the range-azimuth spatial resolution of 1.6 m × 0.8 m.
The Wide Ultra-Fine data consisted of ascending and descending
Wide 2 Ultra-Fine 7 (U7W2) images, acquired with an incidence
angle of 37 degrees and the range-azimuth spatial resolution of
1.6 m × 2.8 m. By using four independent data sets the overall tem-
poral resolution was increased by a factor of four from twenty four
to six days.
The initial DInSAR analysis was applied to each of four data
sets independently. The SLA18 and SLA12 data were “multilooked”
(i.e. spatially averaged) in range-azimuth by a factor of 3 × 12 and
the U7W2 data was multilooked by a factor of 4 × 3. Multilooking
increased the coherence of the interferograms, which is necessary
for accurate phase unwrapping and reduced their spatial resolution.
A single master for each data set was selected and remaining slave
images were co-registered and re-sampled to the master geome-
try. The topographic phase was computed from a 1 m resolution

Fig. 2. Temporal and spatial baselines of RADARSAT-2 interferograms used in this study. Overall 80 SAR images spanning 20120615–20140923 (two years and three months)
produced 531 interferograms with perpendicular baselines less than 400 m.
 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199
Fig. 3. RADARSAT-2 Spotlight 18 (SLA18) Synthetic Aperture Radar intensity image
acquired on 20140823. Monitoring sites and reference region “R” are plotted in black.
Points P1–P9 experiencing fast ground deformation are plotted in red. RADARSAT-2
Data and Products © MacDONALD, DETTWILER AND ASSOCIATES LTD. (2014) – All
Rights Reserved. RADARSAT is an official mark of the Canadian Space Agency. (For
interpretation of the references to colour in this figure legend, the reader is referred
to the web version of this article.)
air-borne LIDAR Digital Elevation Model (DEM) and removed
from the interferograms. Adaptive filtering (Goldstein and Werner,
1998), phase unwrapping (Costantini, 1998) and geocoding pro-
cedures were applied. Residual orbital ramps were computed and
removed and the interferograms were re-sampled to a common
spatial grid with a resolution of 5 × 5 meters.
Advanced MSBAS processing was applied to compute vertical
and horizontal East-West deformation rates and time series. Highly
coherent interferograms acquired with small spatial and temporal
baselines were selected and deformation rates between sequen-
tial SAR acquisitions were computed by applying singular value
decomposition. The two dimensional deformation time series were
Fig. 4. Observed vertical (a) and horizontal (b) deformation rates computed by applying Multidimensional Small Baseline (MSBAS) technique to RADARSAT-2 data spanning
20120615–20140923. Monitoring sites and reference region “R” are plotted in black. Points P1–P9 experiencing fast ground deformation are plotted in red. SAR intensity
image is exposed in background in regions of low coherence.
191
reconstructed from the deformation rates and then the two dimen-
sional mean deformation rates were computed by fitting a line to
the time series. The MSBAS software is implemented in C++ and
is fast, in order to compute time series from 531 interferograms
created from 80 SAR images, only 20 min of processing time on an
engineering workstation was required.
A network of nine paired (i.e. for ascending and descending
acquisition) corner reflectors was installed around the injection
well in December of 2012 and January of 2014 at monitor-
ing sites NE1, NE2, SE1, SE2, SE3, SITE, SW1, NW1, NW2
(Fig. 1). The two-dimensional time series were computed for
these locations using interferograms coherent in the entire area.
This resulted in exclusion of winter interferograms that due to
snow cover were incoherent. Additional time series were com-
puted for regions P1–P9, where fast ground deformation was
observed.
4. Observed ground deformation prior to carbon dioxide
injection
The Aquistore site is located near Estevan township in south-
eastern Saskatchewan on the southern flank of the Souris river
(Fig. 1(a)). The Souris drainage basin is made of fertile silt and clay,
deposited by a former glacial lake. Part of the test area was strip-
mined for coal and has subsequently been reclaimed to a depth
of about 20–25 m. According to the LIDAR derived DEM (Fig. 1(b))
elevation ranges from 533 m in the Souris river channels to 593 m
in the mining area to the south, which was formed by both, mor-
phological processes and surface mining. Two earth-filled dams are
located around the Aquistore site: Boundary Dam, built in 1957 on
the Long Creek River and Rafferty Dam, built on the Souris river
in 1991. A 10-kilometre diversion canal connects Boundary and
Rafferty allowing water to be diverted from the Long Creek River
into Rafferty Reservoir. The area is strongly affected by glacioflu-
vial processes and anthropogenic changes due to mining activity
and infrastructure engineering.
The June 2012–October 2014 vertical and horizontal East-West
ground deformation rates for the Aquistore storage site are shown
in Fig. 4; note that these figures were saturated at ±0.8 and
192 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199
Fig. 5. Histograms of vertical (a) and horizontal (b) distributions of deformations rates from Fig. 4. Red arrows mark 2 or 95% confidence interval equal to 0.3 cm/year and
0.2 cm/year correspondingly.
0.4 cm/year, respectively, to improve visibility. Subsidence of up
to 1 cm/year is predominantly observed at former mining areas
and heave is visible in the north-east. The heave zone is elon-
gated parallel with the Souris river morphology, where the main
geological environment is an alluvial floodplain. A characteris-
tic East-West motion of up to ±0.5 cm/year is detected along the
Raferty-Boundary diversion canal. The 2 or 95% confidence inter-
val of the vertical and horizontal deformation rates is 0.3 and
0.2 cm/year correspondingly (Fig. 5).
Time series for selected regions P1–P9, shown in Fig. 6, describe
temporal evolution of deformation processes. Regions P1 and P2
are located on the western and eastern steep canal slopes, 4 and
3 m high respectively, which also act as the flood protective barri-
ers. These regions experience eastward and westward movements
of about 0.5 cm/year directed toward the canal. A seasonal heave
is observed in this time series during summer, while subsidence
occurs between late fall and early spring. This motion is likely
caused by erosion.
Regions P3–P5 are located near the canal outlet to Boundary
Dam reservoir. Geologically it is a contact between the moraine
eroded by the glacial melt stream sediments and moraine plain. The
vertical deformation map shows relatively fast subsidence reach-
ing 1 cm/year. Seasonal uplift is visible during the spring-summer
likely caused by sediments swelling.
Region P6 is located in the mining basin partially filled with
water and mine tailings. A high subsidence rate here can be caused
by compaction of mine tailings. A seasonal signal, such as uplift
during summer is also observed.
Region P7 is located in a small 7 m deep morphological basin.
The detected uplift of about 1 cm/year is likely produced by the
groundwater recharge after mining closure.
Region P8 is a 4 m high anthropogenic structure where sub-
sidence reaches 1 cm/year and minor horizontal motion is down
slope, toward the east. Deformation here could be caused by mass
movement, erosion, and compaction.
Region P9 is located in the intermittent water-body in the grass-
land that subsides with a rate of about 1 cm/year.
Overall MSBAS DInSAR results show ground deformation due
to the anthropogenic processes, such as the long-term open
mine exploitation, disturbances in groundwater regime, protec-
tive barriers contraction and mine tailings deposition. Natural
morphological changes, such as erosion and sediments swelling are
also contributing factors.
5. Simulated deformation due to carbon dioxide injection
Forward modeling was performed in order to simulate ground
deformation that would be produced by a maximum CO2 injection
rate of 1500 tonne/day at an average depth of 3250 m. Elastic and
poroelastic models were employed using representative rock prop-
erties and injection parameters for the Aquistore site. An elastic
point pressure source (i.e. Mogi model) was assumed to be located
at the injection depth (Mogi, 1958; Tiampo et al., 2000; Dzurisin,
2006; Samsonov et al., 2010). The increase in pressure produced by
CO2 injection resulted in the modeled surface deformation calcu-
lated according to (1) and shown in Fig. 7:
3V
ux (x, y, z) =x
4R3
3V
uy (x, y, z) =y
4R3 (1)
3V
uz (x, y, z) = −z0
4R3

R = x2 + y2 + z02
where ux (x, y, z), uy (x, y, z), uz (x, y, z) are the annual deformation
rates in the x, y, and z directions, x and y are the distance in the x and
y direction from the center of the injection well along the earth’s
surface, −z0 is the source depth, R is the distance from the source to
the point with coordinates x and y at the surface, and V is the annual
volume change produced by injection. In Tiampo et al. (2000) it was
demonstrated that volume and pressure source representations
are identical. Parameters used for modeling are shown in Table 2.
The annual injection volume was estimated as the daily injection
rate multiplied by 365 days per year and divided by the density
of CO2 .
A number of limitations are inherent to this model. The point
source approximation of a finite-size source seems reasonable since
the actual source size is significantly smaller than its depth, partic-
ularly at the beginning of injection. The assumption of pure elastic,
instantaneous response and neglect of the effects of stratigraphy
are likely more significant limitations. Therefore, these modeling
 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199 193

Fig. 6. Vertical (in red) and horizontal (in green) time series of ground deformation for selected regions P1–P9 (Fig. 4) experiencing fast ground deformation and computed
by MSBAS method from RADARSAT-2 data.

results are recognized as a first-order approximation and are used
mainly to constrain the upper bound of possible deformation.
The actual deformation will likely be of a smaller magnitude and
broader extent. The vertical component shows uplift or heave
with a maximum magnitude of about 1.6 cm/year centered at the
injection well (Fig. 7(a)). The horizontal East-West component
shows zero motion above the injection well but increases with
the distance from the well and reaches its maximum of about
z0
0.6 cm/year at √ or about 2298 m away from the injection well
2
(Fig. 7(b)). The radial profile of deformation rates is shown in
Fig. 7(c). These deformation rates and locations of their maxima are
time-independent.
A poroelastic model was adopted to simulate ground deforma-
tion produced by CO2 injection at a deep brine aquifer of unlimited
spatial extent bounded from top and bottom by the impermeable
aquitards. The analytical solution for pressure buildup in a two-
phase (i.e. CO2 and brine) flow system with compressible brine and
rock formation was derived by Mathias et al. (2009a). Ground defor-
mation was computed for the estimated pressure source using the
analytical solution from Xu et al. (2012). Overall, this model com-
putes spatial and temporal pressure distribution in the aquifer as
a function of time since the injection, radial distance and depth
according to (2)–(3). The surface deformation is then computed
for the estimated pressure source according to (4)–(6)). A similar
approach was employed in Yang et al. (2015) for modeling ground
deformation at the enhanced oil recovery site in Texas (USA)

⎧  x     ⎫
⎪
⎪ 1 1 1 ˛ ⎪
⎪ 2
⎪ − ln − 1 + − ln + 0.5772 x ≤ 2 ⎪
⎪
⎪
⎪
⎪ 2  2 2 2 ⎪
⎪
1  ˛x    ˛x   ⎪
⎨     ⎪
⎬
1 x 1 1 ˛ 2
p(r, t) = P0 E1 + ln + 0.5772 + P0 − + − ln + 0.5772 2 < x < (2)
2 4 2 4 ⎪
⎪ 2  2 2 2 ⎪
⎪ 
⎪
⎪ ⎪
⎪
⎪
⎪     ⎪
2⎪
⎪
⎩ − 1 ln ˛x + 0.5772 ⎪
x≥ ⎭
2 4 
194 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199

Fig. 7. Modeled vertical (a) and horizontal (b) deformation rates computed for elastic point source (Mogi, 1958) due to injection of 1500 tonne/day of CO2 at depth of 3250 m.
Monitoring sites and reference region “R” are plotted in black. Spatial profile (c) of vertical and horizontal deformation rates from injection well to 40,000 m.

Table 2 M0 0
Parameters used for modeling. Aquifer parameters Poisson’s ratio (), Young’s mod- P0 =
ulus (E), permeability (k), and porosity () are from well logs. CO2 parameters (for
2H0 k
T = 110 ◦ C, P = 35 MPa) density (0 ), viscosity ( 0 ), compressibility (c0 ); brine param- 0
 =
eters (for T = 110 ◦ C, P = 35 MPa) density (1 ), viscosity ( 1 ), compressibility (c1 ), and 1
formation compressibility (c2 ) are from NIST Chemistry WebBook (http://webbook. (3)
nist.gov) and Mathias et al. (2009b). 2H0 r 2
x =
M0 t
Property Value Unit
M0 0 (c1 + c2 )
Daily injection rate, M0 1500 tonne/day ˛ =
Annual injection volume, V 806,333 m3 /year
2H0 k
Injection depth (elastic), z0 −3250 m  ∞    z2
Injection depth range (poroelastic), z1 –z2 −3280 −3220 m (1 + )(1 − 2)
Aquifer thickness, H 60 m uz (r, z, t) = p(r  , t)
2E(1 − ) 0 0 z1
Poisson’s ratio,  0.27
Young’s modulus, E 25 GPa

z − z (1 − 4)z + (3 − 4)z  6z(z + z  )2  
Porosity,  0.07 × − − r  dz d dr
Permeability, k 2.25 × 10−14 m2 3 3 5
CO2 density, 0 679 kg/m3 (4)
CO2 viscosity, 0 5.648 × 10−5 Pa s  ∞    z2
CO2 compressibility, c0 6.94 × 10−9 Pa−1 (1 + )(1 − 2)
Brine density, 1 1186 kg/m3 ur (r, z, t) = p(r  , t)(r − r  cos  )
2E(1 − ) 0 0 z1
Brine viscosity, 1 56 × 10−5 Pa s
Brine compressibility, c1 0.27 × 10−9 Pa−1 
0.45 × 10−9 Pa−1
1 (3 − 4) 6z(z + z  )    
Formation compressibility, c2 × + − r dz d dr (5)
3 3 5
 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199 195

Fig. 8. Vertical deformation at different times computed for poroelastic model (Mathias et al., 2009a; Xu et al., 2012) due to injection of 1500 tonne/day of CO2 at depth of
3220–3280 m. Monitoring sites and reference region “R” are plotted in black.

Fig. 9. Horizontal deformation at different times computed for poroelastic model (Mathias et al., 2009a; Xu et al., 2012) due to injection of 1500 tonne/day of CO2 at depth
of 3220–3280 m. Monitoring sites and reference region “R” are plotted in black.
196 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199

 (for T=110◦ C, P=35 MPa) density (1 ), viscosity ( 1 ), compressibil-

 = (z − z  )2 + r 2 + r 2 − 2rr  cos 
 (6)
 = (z + z  )2 + r 2 + r 2 − 2rr  cos 
where ur (r, z, t) and uz (r, z, t) is the radial and vertical deformation
observed at distance r, depth z = 0 and time t since the beginning
of injection, M0 is the injection rate to depths from z1 to z2 that
produces pressure distribution p(r, t) in the aquifer of thickness H,
 is the Poisson’s ratio, E is the Young’s modulus, k is the perme-
ability,  is the porosity, r , z ,  are the variables of integration (
is the angle in the cylindrical coordinate system), P0 , x, ˛, ,  , and
 are the supplementary variables explicitly defined, 0 is the CO2
viscosity, 0 is the CO2 density, c0 is the CO2 compressibility, 1 is
the brine viscosity, 1 is the brine density, c1 is the brine compress-
ibility, c2 is the rock formation compressibility. Parameters used for
modeling are shown in Table 2. Aquifer parameters Poisson’s ratio
(), Young’s modulus (E), permeability (k), and porosity () were
estimated from well logs. CO2 parameters (for T=110◦ C, P=35 MPa)
density (0 ), viscosity ( 0 ), compressibility (c0 ); brine parameters
ity (c1 ) and formation compressibility (c2 ) were taken from NIST
Chemistry WebBook (http://webbook.nist.gov) and Mathias et al.
(2009b).
Deformation calculated for the poroelastic model are shown in
Figs. 8 and 9. Snapshots are presented for 1 month and 1, 2, 5, 10,
and 25 years after the beginning of injection. The vertical compo-
nent shows uplift or heave with a maximum magnitude of about
1.6 cm centered at the injection well. The horizontal component
shows maximum deformation of 0.1 cm after 25 years of injec-
tion located at the eastern and western boundaries of the mapped
region. The spatial and temporal profiles are shown in Fig. 10. It
can be seen that vertical deformation achieves a maximum value of
1.6 cm around the injection well immediately after the beginning of
injection. Uplift propagates laterally over time while its maximum
value around the injection well remains constant. After 25 years of
injection the spatial extent of the affected area is larger than 40 km
in radius. The horizontal deformation does not exceed 0.1 cm and
extends over time from the injection well in all directions.

Fig. 10. Modeled spatial vertical (a), spatial horizontal (b), temporal vertical (c), and temporal horizontal (d) deformation profiles computed for poroelastic model (Mathias
et al., 2009a; Xu et al., 2012) due to injection of 1500 tonne/day of CO2 at depth of 3220–3280 m.
 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199 197

Fig. 11. Observed and superimposed, modeled and observed, vertical time series of ground deformation for monitoring sites NE1, NE2, SE1, SE2, SE3, SITE, SW1, NW1, NW2
from Fig. 4 computed by MSBAS method from RADARSAT-2 data. Modeled time series were produced for elastic (Mogi, 1958) and poroelastic (Mathias et al., 2009a; Xu et al.,
2012) sources.

For nine monitoring sites NE1, NE2, SE1, SE2, SE3, SITE, SW1,
NW1, NW2 the currently observed and superimposed modeled and
observed vertical deformation time series are shown in Fig. 11. The
currently observed deformation time series show stable conditions
with moderate seasonal variability. Sites NE1, SE2, and SE3 show
moderate subsidence that was described previously in Figs. 4 and 6.
The simulated vertical time series show clear uplift with the maxi-
mum rates observed in the proximity of the injection well (i.e. SITE).
Simulated uplift is observed above the noise level at all monitoring
sites.
6. Discussion
Satellite interferometry or DInSAR can be used operationally
for detecting ground deformation at onshore carbon sequestra-
tion sites, prior, during and after CO2 injection. SAR data sets from
multiple satellites and acquisition geometries are combined to pro-
duce deformation time series with temporal sampling significantly
more dense than the temporal sampling of each individual data set.
The line-of-sight DInSAR retrievals are decomposed into the verti-
cal and horizontal deformation rates and time series using MSBAS
methodology. Detection of horizontal motion, triggered by injec-
tion, can be of particular value especially if such motion occurs on
previously unknown faults or fractures.
The six day sampling rate was achieved in this study by combin-
ing four RADARSAT-2 data sets with the individual repeat cycle of
24 days. Temporal sampling can be further improved by including
ascending and descending SAR data from different satellites, such
as X-band TerraSAR-X, Cosmo-SkyMed, C-band Sentinel-1, and L-
band ALOS-2. With this approach it seems feasible to achieve one
day sampling rate, similar to continuous GPS daily solutions, but
with denser spatial resolution and larger coverage. In this study five
meter spatial resolution over an 8 km × 8 km area was achieved.
While it was possible to increase the resolution to about 2 m it
seemed unnecessary.
In order to accurately detect deformation due to CO2 injection,
background deformation due to natural and anthropogenic pro-
cesses needs to be estimated first. It was determined that two years
of observations are sufficient for mapping background processes
with high precision of 0.2–0.3 cm/year. The deformation rate maps
for the Aquistore site showed active ground deformation prior to
CO2 injection. Vertical ground deformation with the maximum rate
of ±1 cm/year are caused by natural and anthropogenic processes,
such as, erosion, groundwater withdrawal and recharge, and post-
mining activities. Horizontal motion with the maximum rate of
±0.5 cm/year is caused by erosion of the man-made structures.
While it is impossible to guarantee that background deforma-
tion will remain the same over a long period of time, it can be
198 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199
assumed in this case. For the deformation process to change its state
external disturbance is needed. In this area most anthropogenic
deformation is caused by previous mining operations and con-
struction, which have already ended and future activities are not
expected. The magnitude of deformation due to natural processes
is expected to slowly reduce. Since more RADARSAT-2 data is being
acquired the background deformation maps can be recalculated at
reasonable intervals, for example, once or twice per year.
The large number of SAR images acquired during October
2010–June 2012 allowed computation of many highly coherent
small-baseline interferograms achieving nearly uniform tempo-
ral coverage throughout the year, except during the winter.
Nine paired (i.e. ascending–descending) corner reflectors were
installed at monitoring sites along with continuous GPS, tilt-
meters, and gravimeters. The time series for corner reflector sites
may achieve very high measurement precision (e.g. Rohmer and
Raucoules, 2012) but at the same time they may capture unde-
sirable motion due to monument settlement or miss localized
deformation due to potential upward migration of CO2 through
faults and fractures. Direct application of MSBAS DInSAR to corner
reflector processing is not obvious and needs to be further devel-
oped. The additional uncertainty, introduced during re-sampling
of interferograms to a common spatial grid, needs to be further
investigated.
Forward modeling based on a Mogi elastic pressurized point
source model due to injection of 1500 tonne/day of CO2 at depth of
3250 m determined that the extent of deformation due to simulated
injection is very broad. While the maximum of the vertical compo-
nent of 1.6 cm/year is located at the injection well, the maximum
horizontal component of about 0.6 cm/year is located 2298 m away
from the injection well. This means that the current location of the
monitoring sites is sub-optimal for the injection scenario chosen for
modeling. Whereas the vertical motion can be accurately mapped
at most sites as its magnitude exceeds the noise floor, the horizon-
tal motion at these sites is negligibly small. The sites NE2, SE2, SE3,
SW2, NW2 would capture the signal better if they were removed
from the injection well by 2298 m. In the case of the 2D MSBAS solu-
tion presented here, the maximum East-West deformation would
be observed precisely east and west of the injection well, whereas
in the case of 3D continuous GPS, North-South components would
also be well resolved, with the maximum precisely north and south
of the injection well.
Forward modeling using the poroelastic formulation of Mathias
et al. (2009a), Xu et al. (2012) due to injection of 1500 tonne/day of
CO2 at depth of 3220–3280 m produced a deformation field that is
likely more realistic. According to this model the maximum verti-
cal deformation of 1.6 cm is observed around the well immediately
after injection. During the entire 25 year injection cycle the magni-
tude of uplift does not exceed its initial value but its spatial extent
grows to over 40,000 m in radius. The maximum horizontal defor-
mation does not exceed 0.1 cm, which is significantly below the
noise level of the monitoring network.
Assuming that measurement precision of vertical deformation
rates of 0.3 (2) does not change, the proposed measurement tech-
nique can be used for monitoring vertical deformation during the
entire 25 year period. Since the deformation signal due to injec-
tion is spatially correlated the advanced methodologies, such as
Principal Component Analysis (Jolliffe, 2002) or Singular Spectrum
Analysis (Golyandina and Zhigljavsky, 2013) can be employed to
extract signal from spatially and temporally correlated noisy data.
Comparison of modeling results from elastic and poroelastic for-
mulations may provide additional insights into interpretation of
the actual deformation.
The reference region R required by the MSBAS processing needs
to be positioned further from the injection well for it to remain
stable. Distinguishing the residual orbital ramps from deformation
would be easier if interferograms captured a larger area than what
has been shown in this study.
7. Conclusions
Satellite interferometry was used for detecting ground defor-
mation at the Aquistore CO2 storage site located in south-eastern
Saskatchewan, Canada. Very high precision of 0.2–0.3 cm/year
along with the six day temporal sampling, four times more frequent
than the sampling rate of each individual data set, was achieved
utilizing over two years (June 2012–October 2014) of RADARSAT-
2 Synthetic Aperture Radar data from two ascending and two
descending high resolution beams. Two dimensional deformation
maps of the background processes were produced that revealed
various vertical and horizontal motions. Numerical simulations
based on representative rock properties and injection parameters
for the Aquistore site showed deformation that would be produced
when injection begins. The spatial distribution of simulated defor-
mation suggests installation of ground monitoring sites including
DInSAR corner reflectors around the injection well and also at
a 2000–5000 m distance from the well in order to capture ver-
tical and horizontal deformation signals with the best possible
signal-to-noise ratio. The demonstrated MSBAS strategy over-
comes limitations of the classical DInSAR, such as sparse temporal
resolution and the lack of ability to extract individual deformation
components from the line-of-sight retrievals, and can be imple-
mented at other onshore CCS sites for operational monitoring,
especially now when SAR data is readily available.
Acknowledgments
We would like to thank the Canadian Space Agency for provid-
ing RADARSAT-2 data. Images were analyzed and plotted with GMT,
GNUPLOT and R software. The Petroleum Technology Research Cen-
tre provided logistical support for this work as well as project
information. SaskPower provided access to the Aquistore site. We
thank Drs Zhijie Xu and Jie Bao from the Pacific Northwest National
Laboratory for sharing their source code and Dr Simon A. Math-
ias from the Department of Civil and Environmental Engineering,
Imperial College London for his valuable comments.
References
Bachu, S., Bonijoly, D., Bradshaw, J., Burruss, R., Holloway, S., Christensen, N.,
Mathiassen, O., 2007. CO2 storage capacity estimation: methodology and gaps.
Int. J. Greenh. Gas Control 1 (4), 430–443.
Berardino, P., Fornaro, G., Lanari, R., 2002. A new algorithm for surface deformation
monitoring based on small baseline differential SAR interferograms. IEEE
Trans. Geosci. Remote Sens. 40, 2375–2383.
Costantini, M., 1998. A novel phase unwrapping method based on network
programming. IEEE Trans. Geosci. Remote Sens. 36, 813–821.
Dzurisin, D., 2006. Geodetic Monitoring Techniques. Springer.
Ferretti, A., Prati, C., Rocca, F., 2001. Permanent scatterers in SAR interferometry.
IEEE Trans. Geosci. Remote Sens. 39, 8–20.
Gibbins, J., Chalmers, H., 2008. Carbon capture and storage. Energy Policy 36 (12),
4317–4322.
Goldstein, R., Werner, C., 1998. Radar interferogram filtering for geophysical
applications. Geophys. Res. Lett. 25, 4035–4038.
Golyandina, N., Zhigljavsky, A., 2013. Singular Spectrum Analysis for Time Series.
Springer Briefs in Statistics. Springer, http://dx.doi.org/10.1007/978-3-642-
34913-3.
Hooper, A., 2008. A multi-temporal InSAR method incorporating both persistent
scatterer and small baseline approaches. Geophys. Res. Lett. 35, 16302.
Hooper, A., Bekaert, D., Spaans, K., Arkan, M., 2012. Recent advances in SAR
interferometry time series analysis for measuring crustal deformation.
Tectonophysics 514–517, 1–13, http://dx.doi.org/10.1016/j.tecto.2011.10.013.
IPCC, 2005. IPCC special report on carbon dioxide capture and storage. Prepared by
working group III of the Intergovernmental Panel on Climate Change.
Cambridge University Press, Cambridge, United Kingdom/New York, NY, USA,
pp. 442.
Jolliffe, I., 2002. Principal Component Analysis. Springer, http://dx.doi.org/10.1007/
b98835.
 S. Samsonov et al. / International Journal of Greenhouse Gas Control 42 (2015) 188–199
Lubitz, C., Motagh, M., Wetzel, H.-U., Kaufmann, H., 2013. Remarkable urban uplift
in Staufen im Breisgau, Germany: observations from TerraSAR-X InSAR and
leveling from 2008 to 2011. Remote Sensing 5 (6), 3082–3100.
Massonnet, D., Feigl, K., 1998. Radar interferometry and its application to changes
in the Earth’s surface. Rev. Geophys. 36, 441–500.
Mathias, S., Hardisty, P., Trudell, M., Zimmerman, R., 2009a. Approximate solutions
for pressure buildup during CO2 injection in brine aquifers. Transp. Porous
Media 79, 265–284.
Mathias, S., Hardisty, P., Trudell, M., Zimmerman, R., 2009b. Screening and
selection of sites for CO2 sequestration based on pressure buildup. Int. J.
Greenh. Gas Control 3, 577–585.
Mogi, K., 1958. Relations between the eruptions of various volcanoes and the
deformation of the ground surfaces above them: University of Tokyo. Earthq.
Res. Inst. Bull. 36, 99–134.
Norford, B., Haidl, R., Bezys, F.M., Cecile, M., McCabe, H., Paterson, D., 1994. Middle
Ordovician to Lower Devonian strata of the Western Canada Sedimentary
Basin. In: Mossop, G., I.S. (comp. (Eds.), Geological Atlas of the Western Canada
Sedimentary Basin. Canadian Society of Petroleum Geologists, Calgary, Alberta
and Alberta Research Council, Edmonton, Alberta, pp. 109–127.
Rackley, S., 2009. Carbon Capture and Storage. Elsevier.
Ramirez, A., Foxall, W., 2014. Stochastic inversion of InSAR data to assess the
probability of pressure penetration into the lower caprock at In Salah. Int. J.
Greenh. Gas Control 27, 42–58, http://dx.doi.org/10.1016/j.ijggc.2014.05.005.
Rinaldi, A., Rutqvist, J., 2013. Modeling of deep fracture zone opening and transient
ground surface uplift at KB-502 CO2 injection well, In Salah, Algeria. Int. J.
Greenh. Gas Control 12, 155–167, http://dx.doi.org/10.1016/j.ijggc.2012.10.
017.
Rohmer, J., Raucoules, D., 2012. On the applicability of Persistent Scatterers
Interferometry (PSI) analysis for long term CO2 storage monitoring. Eng. Geol.
147–148, 137–148, http://dx.doi.org/10.1016/j.enggeo.2012.07.010.
Rosen, P., Hensley, P., Joughin, I., Li, F., Madsen, S., Rodriguez, E., Goldstein, R., 2000.
Synthetic aperture radar interferometry. Proc. IEEE 88, 333–382.
Rucci, A., Vasco, D., Novali, F., 2013. Monitoring the geologic storage of carbon
dioxide using multicomponent SAR interferometry. Geophys. J. Int. 193,
197–208, http://dx.doi.org/10.1093/gji/ggs112.
Rutqvist, J., Vasco, D., Myer, L., 2010. Coupled reservoir geomechanical analysis of
CO2 injection and ground deformation at In Salah, Algeria. Int. J. Greenh. Gas
Control 4 (2), 225–230.
Samsonov, S., 2010. Topographic correction for ALOS PALSAR interferometry. IEEE
Trans. Geosci. Remote Sens. 48, 3020–3027.
Samsonov, S., d’Oreye, N., 2012. Multidimensional time series analysis of ground
deformation from multiple InSAR data sets applied to Virunga Volcanic
Province. Geophys. J. Int. 191, 1095–1108, http://dx.doi.org/10.1111/j.1365-
246X.2012.05669.x.
Samsonov, S., d’Oreye, N., González, P., Tiampo, K., Ertolahti, L., Clague, J., 2014a.
Rapidly accelerating subsidence in the Greater Vancouver region from two
199
decades of ERS-ENVISAT-RADARSAT-2 DInSAR measurements. Remote Sens.
Environ. 143, 180–191, http://dx.doi.org/10.1016/j.rse.2013.12.017.
Samsonov, S., d’Oreye, N., Smets, B., 2013a. Ground deformation associated with
post-mining activity at the French–German border revealed by novel InSAR
time series method. Int. J. Appl. Earth Observ. Geoinf. 23, 142–154.
Samsonov, S., Gonzalez, P., Tiampo, K., d’Oreye, N., 2013b. Methodology for
spatio-temporal analysis of ground deformation occurring near Rice Lake
(Saskatchewan) observed by RADARSAT-2 DInSAR during 2008–2011. Can. J.
Remote Sens. 39, 27–33.
Samsonov, S., Gonzalez, P., Tiampo, K., d’Oreye, N., 2014b. Modeling of fast ground
subsidence observed in southern Saskatchewan (Canada) during 2008–2011.
Nat. Hazards Earth Syst. Sci. 14, 247–257, http://dx.doi.org/10.5194/nhess-14-
247-2014.
Samsonov, S., van der Koij, M., Tiampo, K., 2011. A simultaneous inversion for
deformation rates and topographic errors of DInSAR data utilizing linear least
square inversion technique. Comput. Geosci. 37, 1083–1091.
Samsonov, S., Tiampo, K., Gonzalez, P., Manville, V., Jolly, G., 2010. Ground
deformation occurring in the city of Auckland, New Zealand, and observed by
Envisat interferometric synthetic aperture radar during 2003–2007. J.
Geophys. Res. Solid Earth 11, 5, http://dx.doi.org/10.1029/2009JB006806.
Samsonov, S.V., Tiampo, K.F., Camacho, A.G., Fernndez, J., Gonzlez, P.J., 2014c.
Spatiotemporal analysis and interpretation of 1993–2013 ground deformation
at Campi Flegrei, Italy, observed by advanced DInSAR. Geophys. Res. Lett. 41,
6101–6108, http://dx.doi.org/10.1002/2014GL060595.
Samsonov, S.V., Trishchenko, A.P., Tiampo, K., Gonzlez, P.J., Zhang, Y., Fernndez, J.,
2014d. Removal of systematic seasonal atmospheric signal from
interferometric synthetic aperture radar ground deformation time series.
Geophys. Res. Lett. 41, 6123–6130, http://dx.doi.org/10.1002/2014GL061307.
Tiampo, K., Rundle, J., Fernandez, J., Langbein, J., 2000. Spherical and ellipsoidal
volcanic sources at Long Valley Caldera, California, using a genetic algorithm
inversion technique. J. Volcanol. Geother. Res. 102, 189–206.
Vasco, D., Rucci, A., Ferretti, A., Novali, F., Bissell, R., Ringrose, P., Mathieson, A.,
Wright, I., 2010. Satellite-based measurements of surface deformation reveal
fluid flow associated with the geological storage of carbon dioxide. Geophys.
Res. Lett. 37 (3), L03303, 1–5.
Worth, K., White, D., Chalaturnyk, R., Sorensen, J., Hawkes, C., Rostron, B., Johnson,
J., Young, A., 2014. Aquistore project measurement, monitoring and
verification: from concept to CO2 injection. Energy Proc. 63, 3202–3208, http://
dx.doi.org/10.1016/j.egypro.2014.11.345.
Xu, Z., Fang, Y., Scheibe, T., Bonneville, A., 2012. A fluid pressure and deformation
analysis for geological sequestration of carbon dioxide. Comput. Geosci. 46,
31–37.
Yang, Q., Zhao, W., Dixon, T.H., Amelung, F., Han, W.S., Li, P., 2015. InSAR
monitoring of ground deformation due to CO2 injection at an enhanced oil
recovery site, West Texas. Int. J. Greenh. Gas Control 41, 20–28, http://dx.doi.
org/10.1016/j.ijggc.2015.06.016.