Marine and Petroleum Geology 131 (2021) 105162

Contents lists available at ScienceDirect

Marine and Petroleum Geology

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

Gravity effects of fluid storage and withdrawal in a reservoir from 3D

forward modelling
Paolo Mancinelli
Dipartimento di Ingegneria e Geologia, Università G. D’Annunzio di Chieti-Pescara, Via dei Vestini 31, Chieti, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: The gravity effects of the possible reservoir scenarios after primary exploitation are tested in this work. Starting
Time variable gravity from the exploitable volume after primary hydrocarbon production of the very small and deep Volve field in the
Numerical modelling North Sea, we model several scenarios and calculate 3D forward gravity signatures accordingly. Namely, we test
Gas and hydrate systems
water flooding by aquifer rise, carbon dioxide storage, hydrogen storage using different cushion gases, hydrogen
storage without cushion gas and hydrogen withdrawal. The differential gravity signature is calculated between
two consecutive steps and the results provide detectability thresholds for each scenario. To evaluate effects of
reservoir depth on the recovered gravity signatures, we repeat the calculations between 750 and 2750 m depth.
Results of the modelling provide reference values for gravity signatures related to fluid storage in the worst-case
scenario of a deep and thin (~100 m) reservoir and can provide valid constraints when mass loss estimation is
required in leaking reservoirs. When the denser carbon dioxide and water are tested, these always provide
detectable gravity signatures (>3 μGal) even at the maximum modelled depth, whilst storage or withdrawal of
hydrogen in the modelled depth range, often result in undetectable signatures.

1. Introduction Pudlo et al., 2013; Hagrey et al., 2014; Kim et al., 2015; Krahenbul et al.,
2015; Feldmann et al., 2016; Jacob et al., 2016; Pfeiffer et al., 2016;
The investigation and monitoring of reservoirs using time-lapse Kabuth et al., 2017; Wilkinson et al., 2017; Appriou et al., 2020; Goto
gravity techniques have significantly evolved in the last decades et al., 2020; Kabirzadeh et al., 2020 and references therein). Among the
thanks to improvements in gravimeters accuracy and data availability proposed techniques, 3D gravity modelling successfully located gravity
(e.g. Van Camp et al., 2017). New applications resulted in successful variations due to injection and withdrawal of hydrogen (Pfeiffer et al.,
monitoring of fluid production or injection sites at large scales such the 2016) or compressed air/gas (Hagrey et al., 2014) but differential
Prudhoe bay and North Sea sites (e.g. Hare et al., 1999; Eiken et al., gravity anomalies related to hydrogen storage and withdrawal was un­
2000; Ferguson et al., 2007; Alnes et al., 2008; Eiken et al., 2008; Fer­ detectable by modern gravimeters – i.e. time-lapse gravity anomalies
guson et al., 2008; Alnes et al., 2011) or at smaller scales (e.g. Jacob were <3 μGal (1 μGal = 1 × 10− 8 m s− 2).
et al., 2010; Elliott and Braun, 2016; Mancinelli, 2020). Compared to Despite several efforts aiming at the quantification of fluid storage
forward or inverse gravity modelling techniques, where large-scale in­ gravity effects and their variations even in the long period (e.g. Appriou
vestigations are allowed by wide gravity anomaly datasets (e.g. Man­ et al., 2020), the gravity signatures of different evolutive scenarios of a
cinelli et al., 2015; Dressel et al., 2018; Fedi et al., 2018; Mancinelli depleted reservoir were never addressed in a consequential modelling of
et al., 2019, 2020), time-lapse gravity requires acquisition of new each step considering all the possible fluids that can be temporarily or
high-precision data at each step of production/injection but allows permanently stored. In fact, all the literature focused on single steps,
detailed monitoring of fluid-related gravity changes. either of injection or production and often considered only one fluid to
In the frame of new environmental challenges driven by renewable be stored or withdrawn (mostly CO2, more recently H2). Similarly, the
energy exploitation and CO2 sequestration, a few research projects have gravity effect of rising aquifer in the reservoir and how it relates to the
been recently developed to investigate the feasibility of geophysical signature of the stored fluid was never addressed. Moreover, the gravity
monitoring of hydrogen and CO2 storage sites through simulations over signature of hydrogen storage or combined withdrawal with different
synthetic or real-case scenarios (Gasperikova and Hoversten, 2008; cushion gases (CO2 or N2) is tested and compared for the first time in this

E-mail address: [email protected].

https://doi.org/10.1016/j.marpetgeo.2021.105162
Received 28 January 2021; Received in revised form 17 May 2021; Accepted 18 May 2021
Available online 30 May 2021
0264-8172/© 2021 Elsevier Ltd. All rights reserved.
P. Mancinelli Marine and Petroleum Geology 131 (2021) 105162

work. ⎧ ⎫
To address these questions and contribute in the discussion about ∑M ⎪
⎨ ∫ z− z ⎪
⎬
(2)
0
differential gravity signatures related to fluid injection or production at gz (r0i ) = ρj γ
⎪ 3
dv
⎪
⎩ |r − r0i | ⎭
reservoirs, we present a series of 3D forward calculations produced in a
j=1
ΔVj

real, deep and very small reservoir representing the worst-case scenario
for such tests. We model several possible scenarios after the production where ρj and ΔVj are the anomalous density and volume of the jth cell,
of the original hydrocarbons in place including the storage of CO2 and respectively.
storage and withdrawal of hydrogen. Furthermore, to evaluate contri­ We discretize the reservoir volume using 25 × 25 × 25 m cubic cells
butions of the depth of the reservoir on the retrieved differential gravity (Fig. 2) and assuming sealing conditions at the faults bounding the
signature, we repeat the same modelling with the reservoir depth Hugin sandstone. Finally, we homogenously distribute the mass varia­
ranging between 750 and 2750 m. tion due to gas injection or production within the entire reservoir vol­
ume that was left empty after the primary production.
Mass variations were computed considering the density of the
1.1. Data and methods modelled fluid (N2, H2O, CO2 or H2) at known or estimated pressure and
temperature of the reservoir at each modelled depth. Table 1 shows the
Modelling is performed in the Volve field (Fig. 1) due to availability densities used for the modelling steps. Pressure and temperature data
of production record and 3D geometrical model of the reservoir (Volve from the real reservoir at 2750 m depth were taken as a starting point to
data village webpage). Located in the Central North Sea, the Volve field estimate the pressure and temperature conditions at the other modelled
was discovered in 1993 and oil and gas production lasted between 2008 depths where measurements were not available. These estimates were
and 2016 with a total produced volume of 1.47 × 109 Sm3 (Sm3 rep­ produced assuming a ~1 × 107 Pa km− 1 hydrostatic pressure gradient
resents standard cubic meters at 15 ◦ C and 1010 × 102 Pa). This volume (Alnes et al., 2011) and a geothermal gradient of 26 ◦ C km− 1 (Volve
includes the oil, gas, and formation water that have been produced (1.5 documentation included in the dataset, see the reference list for a link to
× 109 Sm3) and the injected water (0.03 × 109 Sm3). Ranging between the web page hosting the dataset).
2750 and 3120 m depth, the reservoir is located in the Hugin Jurassic In the modelling procedure, we assume a homogeneous distribution
sandstones with an average reservoir thickness of ~100 m and porosity of the mass variation within the entire reservoir. Thus, we calculate a
of 20% ± 2.5 (Volve documentation included in the dataset). Given its mass variation induced by each operation on each cell according to the
size, depth and produced volumes (<1.7 × 109 Sm3), the Volve field modelled volume occupied by the injected fluid or freed by the with­
represents a very small and deep reservoir. Based on data included in the drawn fluid. Mass variation at each cell is provided by density contrast
Volve dataset, the density of the produced hydrocarbons at reservoir related to the inherited density of the cell at the end of the previous
conditions (3.28 × 107 Pa and 106 ◦ C) is 710 kg m− 3 (Volve docu­ modelling step and the density of the fluid at the current modelling step.
mentation included in the dataset, see the reference list for a link to the After 3D forward calculation, each injection or production step provides
web page hosting the dataset). The produced volumes over the eight a positive or negative gravity signature, respectively. Finally, the dif­
years of activity at the Volve field, resulted in a gravity signature that ferences between the maximum gravity observed after two consecutive
was recently estimated to be ~ -13 μGal (Mancinelli, 2020). steps, provide a differential gravity anomaly (Δgz) useful to evaluate the
Through 3D forward models, we simulate the gravity effects of detectability of the injection or withdrawal step. The detectability
different evolutionary scenarios of the reservoir after primary produc­ threshold for Δgz is set to 3 μGal for onshore scenarios (Pfeiffer et al.,
tion is completed. The possible cases are the following: (i) the reservoir 2016) and 6 μGal for offshore scenarios. However, these values repre­
can be used for CO2 storage; (ii) the reservoir is flooded by water if sent conservative thresholds soon to be overcame because ~3
strong aquifer push occurs; (iii) cushion gas is injected to eventually μGal-precision measures were achieved by Alnes et al. (2011) in the
stabilize the aquifer and store hydrogen to be withdrawn when needed; Sleipner offshore field, and ~2 μGal-precision onshore measures are
(iiii) hydrogen is stored without cushion gas. We do not consider the achievable using superconducting gravimeters (Kim et al., 2015).
usual scenario where hydrocarbons are stored in the reservoir because in To evaluate effects of reservoir depth on Δgz, after modelling at the
this case the amplitude of the gravity effect would be the same as that real depth of the Volve field (2750 m), we rigidly shift the reservoir
obtained during production if no over-pressure is introduced. Similarly, 1000 m upward in the second step and 2000 m in the third step to run
we assume no displacement of those fluids that remained in the porous the same forward models at 1750 and 750 m depth, respectively (Fig. 2).
media after primary production, such fluids like residual hydrocarbons
or irreducible water are assumed to not contribute in density changes of
2. Results
the reservoir volume. We also assume negligible surface deformation
and porosity changes related to eventual pressure build-up in the sur­
Differential gravity anomalies retrieved from the models are shown
roundings of the injection point.
in Figs. 3 and 4. In the following we report and discuss all the results.
The first model after hydrocarbon production (Fig. 3a–c) was per­
1.2. Modelling procedure formed assuming that the 1.47 × 109 Sm3 volume freed by hydrocarbon
production was used for carbon dioxide sequestration. In this case, we
Forward 3D calculations are performed using the algorithm proposed model only a complete filling of the reservoir volume and this results in
by Li and Oldenburg (1998), where the vertical component of the Δgz values of 12.1, 23.7 and 56 μGal with reservoir at 2750, 1750 or 750
gravity field due to density ρ(x,y,z) is given by: m depth. It is worth nothing that, carbon dioxide storage results in
∫ detectable gravity signatures (Fig. 3d–f).
z − z0
gz (r0 ) = γ ρ(r) 3
dv (1) Next, we model the possibility that after hydrocarbon production, in
V
|r − r0 | case of a strong aquifer push, the reservoir is entirely or half flooded by
water. In the case of a complete flooding, Δgz is calculated between the
where V is the anomalous mass volume, r0 = (x0, y0, z0) is the location of post-production gravity and the post-flooding gravity and we observe
the observation point, r = (x, y, z) locates the source and γ is the grav­ the highest positive differential gravity signature. In fact, Δgz ranges
itational constant. between 17.5 and 96.7 μGal with the reservoir at 2750 or 750 m depth,
If we assume a constant density contrast within each prismatic cell of respectively. If the reservoir is located at 1750 m depth, the retrieved Δgz
the 3D orthogonal mesh, the gravity field at the ith observation point is is 35.7 μGal (Fig. 3g–i). If the reservoir is only half flooded, the resulting
given by Δgz is 8.7, 17.9 and 48.3 μGal at 2750, 1750 and 750 m depth,

2
P. Mancinelli Marine and Petroleum Geology 131 (2021) 105162

Fig. 1. (a) Gravity anomalies over the study area, the free-air gravity anomaly at sea level and the Bouguer gravity anomaly on land (mod. from Olesen et al., 2010).
(b) Perspective view of a 3D model of the Volve field (mod. from Mancinelli, 2020). Light grey surfaces in (b) represent bounding faults. Coordinates in (b) are in
ED50 UTM 31 N.

3
P. Mancinelli Marine and Petroleum Geology 131 (2021) 105162

Fig. 2. Volve reservoir after discretization in 25 × 25 × 25 m cells. (a) perspective view, (b) northward view, (c) westward view, (d) downward view. Red volume
represents the original reservoir ranging between 2750 and 3120 m depth. Orange volume represents the same reservoir but shifted 1000 m upward and green
volume was shifted 2000 m upward. Coordinates are in ED50 UTM 31 N. (For interpretation of the references to colour in this figure legend, the reader is referred to
the Web version of this article.)

Table 1
Density and status of the modelled fluids according to pressure and temperature at each depth, see text for discussion (NIST, 2016). Due to their supercritical status at
these conditions, nitrogen, carbon dioxide and hydrogen show no distinct liquid or gas phases.
Depth (m) P (x107 Pa) T (◦ C) N2 H2O CO2 H2
3
2750 3.28 106 Density (kg/m ) 247 968 671 18
Status Supercritical Liquid Supercritical Supercritical
1750 2.28 80 Density (kg/m3) 197 981 651 14
Status Supercritical Liquid Supercritical Supercritical
750 1.28 54 Density (kg/m3) 128 991 574 9
Status Supercritical Liquid Supercritical Supercritical

respectively (Fig. 3j-l). It should be noted that, if a detectability cushion, its injection would provide detectable gravity signatures at all
threshold for gravimeters is assumed to be 3 μGal, all the above cases depths. In fact, we recover Δgz values of 7.5, 14.7 and 34.9 μGal with
would be easily observed by differential gravity measurements. reservoir at 2750, 1750 or 750 m depth (Fig. 4a–c). If N2 is used as
In the third modelling phase, we test the scenario where the reservoir cushion, we recover Δgz values of 2.7, 4.4 and 7.7 μGal with reservoir at
is used for hydrogen storage. In this case, prior to hydrogen injection we 2750, 1750 or 750 m depth. In this case, due to the lighter cushion gas,
evaluate the possibility of injecting cushion gas to prevent eventual the gravity signature would be detectable to a maximum reservoir depth
aquifer rise. Among the possible cushion gasses we test gravity effects of of ~2500 m.
CO2 and N2 (Oldenburg, 2003; Feldmann et al., 2016) excluding the After modelling the injection of cushion gas, we model the hydrogen
possibility that some original hydrocarbons may act as cushion. This last storage to occupy the residual 40% of the reservoir volume. In this case,
case, although is not uncommon, would be difficult to model due to the single hydrogen storage phase would not be detectable at any depth
uncertainties regarding the quantity of cushion gas required to stabilize because of the low density of the injected fluid at reservoir pressure and
the aquifer. For this reason, we test a scenario where 60% of the avail­ temperature (Table 1). In fact, even in the shallow reservoir case, we
able volume (0.88 × 109 Sm3) is used for cushion gas (CO2 or N2) and the recover maximum Δgz values of 0.4 μGal, significantly below the
remaining 40% is used for hydrogen storage. In the case CO2 is used as detectability threshold of 3 μGal (Fig. 4d–f).

4
P. Mancinelli Marine and Petroleum Geology 131 (2021) 105162

Fig. 3. Differential gravity anomaly (Δgz) maps. From left to right: reservoir at 2750, 1750 and 750 m depth. (a–c) Δgz after primary hydrocarbon production. (d–f)
Δgz after CO2 storage in the entire exploitable volume (1.47 × 109 Sm3). (g–i) Δgz after water flooding of the entire reservoir. (j–l) Δgz after water flooding of half
reservoir. All values are in μGal. In all maps the anomaly is centred over the reservoir.

Finally, after injection of cushion gas we model the hydrogen with­ all the modelled depths even in the case of 100% recovery, because the
drawal phase assuming a 50% recovery factor of the hydrogen initially maximum Δgz related to a reservoir depth of 750 m is 0.9 μGal (Fig. 4l).
stored. We are also considering the realistic possibility that in the first
years of production the recovered hydrogen is impure due to mixing 3. Discussion
processes, and thus we model a 20% of the cushion gas being produced
together with hydrogen in the first withdrawal cycle. In Fig. 5, we plot the recovered Δgz and the modelled mass variation
In the case CO2 is used as cushion gas, the resulting Δgz after with­ causing it. To increase readability of the plot, we show absolute values of
drawal ranges between 1.6, 3.0 and 7.1 μGal with reservoir at 2750, maximum Δgz, these values should be considered negative for hydro­
1750 or 750 m depth (Fig. 4g–i). In this case, the gravity signature would carbon production and hydrogen withdrawal (mass loss), and positive
be detectable to a maximum reservoir depth of ~1750 m. On the other for injection phases (mass increase). These data represent an attempt in
hand, when N2 cushion is used, the withdrawal phase would be unde­ predicting detectable levels of gravity anomalies related to fluid injec­
tectable in the modelled depth range and likely only reservoirs within tion or production in a very small and deep reservoir with average
~500 m depth would provide detectable signatures because we retrieve porosity values of 20 ± 2.5%, relatively small thickness (~100 m) and
Δgz values of 0.5, 1.0 and 1.7 μGal at 2750, 1750 and 750 m depth, ranging between 750 and 2750 m depth.
respectively. There are several parameters that can affect these estimates, some of
Finally, we test hydrogen storage without cushion gas in the case of a them are related to the geometry and physics of the reservoir (e.g.
reservoir without aquifer or weak-to-null aquifer push (Fig. 4j-l). Also in thickness, porosity, permeability). Some others are related to chemical
this case, the hydrogen-related gravity change would be undetectable at and physical processes that can occur in the stored fluids and affect

5
P. Mancinelli Marine and Petroleum Geology 131 (2021) 105162

Fig. 4. Differential gravity anomaly (Δgz) maps. From left to right: reservoir at 2750, 1750 and 750 m depth. (a–c) Δgz after CO2 cushion gas injection in 60% of the
exploitable volume. (d–f) Δgz after hydrogen storage in 40% of the exploitable volume. (g–i) Δgz after withdrawal of 50% of stored hydrogen and 20% of CO2 cushion
gas. (j–l) Δgz after storage of 1.47 × 109 Sm3 of hydrogen without cushion gas. In this figure we show only maps related to CO2 cushion, Δgz values retrieved using N2
as cushion are given in the text. All values are in μGal. In all maps the anomaly is centred over the reservoir.

recovery efficiency. For example, during the hydrogen withdrawal Δgz by introducing a mass increase with strong vertical component.
models, we assume a 50% recovery factor for the stored hydrogen. This Conversely, enlarging the lateral extent of the reservoir will likely result
is mostly due to eventual losses due to bacterial degradation (e.g. Kabuth in smaller increase of Δgz due to the horizontal distribution of the mass
et al., 2017) and methanogenesis via hydrogen methanation if CO2 is increase.
used as cushion gas (Kabuth et al., 2017; Götz et al., 2016; Rönsch et al., The gravity forward models, once the differential gravity related to
2016) whose effects in term of hydrogen loss are difficult to quantify in each step are computed, result in detectability thresholds according to
the storage period (weeks to months). Moreover, the 50% recovery modern gravimeters capabilities. Unsurprisingly, when referred to the
factor for hydrogen is also accounting for eventual mixing between the same depth, the mass variation induced by injection/production activ­
hydrogen and the cushion gas that would likely affect the recoverability ities is linearly affecting the gravity response, while changes in depth
of the stored hydrogen. On the other hand, we model injection and provide non-linear effects. Noteworthy, the proportionality between the
production steps by only considering mass variations and this, as long as depth to the source and the observed Δgz is slightly less than quadratic,
the injected or produced volume is known and the density of the fluid is this is interpreted as a consequence of the geometry of the reservoir
well constrained, may result in reliable estimates of Δgz. whose horizontal extent is larger than its vertical thickness (Kabirzadeh
The exploitable volume of the reservoir plays a significant role in the et al., 2020).
whole process. The Volve reservoir represents an end-member in this Gravity signatures related to N2 being stored as cushion gas below
case because of its limited lateral extent, porosity and thickness. Thus, 2500 m depth are undetectable. Conversely, storage of CO2 is detectable
increasing the thickness of the reservoir will linearly affect the produced down to depth of 2750 m either if it is stored alone for sequestration or if

6
P. Mancinelli Marine and Petroleum Geology 131 (2021) 105162

detection becomes easier even for smaller masses, depending on the

depth of accumulation. Alternatively, an indirect leakage estimation can
be provided by repeated gravity measurements over the storage reser­
voir. Once the depth of the reservoir is known, the predicted gravity
signature of an injection/withdrawal period can be calculated if the
injected/withdrawn mass is known (Fig. 5). If a mass is lost during such
period and in between two measurement campaigns, it will affect the
latter gravity measurements proportionally to the leaked mass. If such
effects are above the detectability threshold, the leaked mass can be
estimated from the missing Δgz component.
In the modelling, we assumed no pressure build-up at injection
points and computed the models without over-pressuring the reservoir.
In other words, we used the gas volume produced in the primary phase
as the only available volume for storage – i.e. the exploitable volume.
Eventual over-pressures, if sealing conditions and the integrity of the cap
rock are preserved, will introduce an increase of the produced Δgz lin­
early proportional to the increase of the injected mass.
Density of the injected fluid plays a key role in the modelling pro­
cedure and accurate pressure and temperature values at the injection
point are thus fundamental to properly estimate these values. In fact, at
supercritical conditions the density of the fluid may result in rapid
changes even with small changes in pressure and temperature (e.g. Alnes
et al., 2011). However, it was demonstrated that diffusion and disper­
sion processes act similarly on normal fluids as for supercritical fluids
(Yu et al., 1999; Oldenburg, 2003) so the supercritical status of the
injected or withdrawn fluid will only affect its density. Considering the
pressure (3.28 × 107 Pa) and temperature (106 ◦ C) at reservoir, we used
fluid densities at these conditions for modelling at 2750 m depth. We
corrected the density values according to pressure and thermal gradients
from literature (Alnes et al., 2011; Volve documentation included in the
dataset) at the other modelled depths (Table 1) to showcase the effects of
depth, pressure and temperature on the fluid density and, in turn, on the
retrievable Δgz.
During the gas injection modelling, we assumed to be in the optimal
case of no aquifer push in order to allow uniform distribution of the
injected gas and avoid gravity override and viscous fingering (Feldmann
et al., 2016). However, despite this may be the case for some real res­
ervoirs, in some others the aquifer may partially or entirely flood the
reservoir during or after primary production operations. It follows that
aquifer push is another parameter that may conceal gravity effects
related to fluid injection and production. In the case of strong aquifer
Fig. 5. (a) Absolute values of Δgz recovered after 3D forward models of the push, the reservoir should always be filled in order to avoid water
reservoir. Values in red denote results with reservoir depth of 2750 m, orange flooding. In fact, if half of the reservoir is left empty, the gravity effect
values denote results with reservoir depth of 1750 m and green values represent produced by water flooding will completely conceal the gravity signa­
results with reservoir depth of 750 m. (b) Zoom of the plot in (a) to the area ture related to gas storage (Fig. 5). Moreover, such a scenario would also
with small mass variation values (<45 × 109 kg). Squares and triangles mark prevent any detection of possible leakage of the stored fluid. Never­
only cushion gas injection in (a) and hydrogen and cushion withdrawal in (b).
theless, reservoirs with strong aquifer push will likely represent a bad
(For interpretation of the references to colour in this figure legend, the reader is
scenario for gas storage in general, because the aquifer rise can lead to
referred to the Web version of this article.)
unpredictable pressures of the stored gas in the long period resulting in
possible leakage.
it is used as cushion gas. Withdrawal activities involving H2 and CO2
The masses injected in our simulations represent both short-term
cushion are undetectable if the reservoir is deeper than 1750 m (Fig. 5b).
periods of injection/production (Fig. 5b) or long-term injection plans
Within the modelled depth range, water flooding always represents a
(Fig. 5a) such those modelled by Appriou et al. (2020) where a total of
detectable phenomenon, even if it affects half of the reservoir.
150 × 109 kg of CO2 was injected at a 2.5 × 109 kg year− 1 rate. In this
In the modelling setup, we assumed that the faults surrounding the
latter case, the rate of injection may play a key role in the case of strong
reservoir are sealed, and no fluid migration is allowed outside of the
aquifer push. In fact, if the injection/sequestration of gas is slower than
reservoir. However, if potential migration paths are known after the
the aquifer rise, the available volume and in turn the injectable mass,
primary explorative phase, these can be included in the forward
will decay in time with obvious consequences on the retrievable Δgz
modelling. Otherwise, unknown migration paths can be identified and
signature both considering the contribution from the stored gas and the
monitored by proper gravity acquisitions provided that the leaked fluid
concealing effect of the aquifer. This implies that if a long-term storage is
accumulates in a monitored secondary lateral reservoir and produces a
planned over a reservoir with rising aquifer, the injection rate should
detectable gravity signature. In the case of a secondary reservoir at
consider the rising rate of the aquifer and how it will affect the storable
similar depth of the primary, the plots in Fig. 5 allow for a first-order
mass in the long-term.
estimation of the masses, depending on the observed gravity signa­
ture. In the case of a secondary reservoir that is shallower than the
primary – i.e. upward lateral migration of the leaked fluids, leakage

7
P. Mancinelli Marine and Petroleum Geology 131 (2021) 105162

3.1. Noise sources affecting the detectability threshold small reservoirs due to the strong mass anomaly it produces. In fact,
water flooding may conceal gravity signatures related to other sources
Among the phenomena affecting the time-lapse gravity measure­ such gas storage or withdrawal even if only half of the reservoir is
ments there is a list of geophysical sources that can produce significant flooded. Similarly, operations involving CO2 result in differential gravity
effects or even conceal the monitored signal. In fact, our 3 μGal anomalies that are always detectable unless they provide small mass
detectability threshold is achievable and representative only if all the changes (<50 × 109 kg) at significant depths (>1750 m).
potential noise sources are addressed and eventually corrected. A Due to the small introduced mass changes, gravity changes related to
compelling discussion about all the noise sources is provided in Van hydrogen injection or withdrawal are undetectable at this reservoir size.
Camp et al. (2017) and references therein. In the following we briefly Only the production of hydrogen coupled with CO2 cushion gas would
discuss the most relevant for the application we tested in this work. be detectable from very small reservoirs at maximum depth of ~1750 m.
Local-scale and regional-scale sources of gravity noise can be The ideal conditions to detect differential gravity signatures during
distinguished. Among the local ones, we already modelled and discussed storage or withdrawal operations involving only hydrogen are given by
reservoir aquifer but did not mention the case of ground water mass shallow (<1000 m) and thick (≫ 100 m) reservoirs.
variations above the reservoir. The noise from this source can last de­ Finally, the data shown in Fig. 5 can provide valid support to esti­
cades and show maximum amplitudes of tens of μGals. Similarly, mate mass variations related to the observed Δgz. In fact, in the case of a
subsidence-related signals can potentially conceal gravity signatures suspected leakage of the reservoir, if the spilled fluid generates a
similar to those modelled in this work with periods spanning from detectable gravity signal after accumulation in a secondary monitored
months to decades. Moreover, also tides can provide similar noise with reservoir with known depth, the mass lost from the reservoir can be
even higher amplitudes. However, all these noise sources can be prop­ estimated from the observed Δgz. Alternatively, the monitoring of the
erly addressed by accurate piezometric monitoring of eventual ground primary reservoir may provide indirect estimates of the mass lost be­
water masses (e.g. Kim et al., 2015 and references therein), precise tween two surveys if the masses injected and/or withdrawn during the
levelling of gravity stations, and accurate tidal models. The first task cycle are known.
always represents a good practice in reservoir fields, while the tidal
modelling and station levelling are always required in gravity data Data availability
acquisition and processing and it all reduces to the accuracy of the in­
struments used to address these tasks. Among the regional-scale noise The data used in this work are available from sources in the public
sources listed by Van Camp et al. (2017) that are capable of generating domain: https://www.equinor.com/en/how-and-why/digitalisation-in-
noise amplitudes higher than those modelled in this work, mass dis­ our-dna/volve-field-data-village-download.html.
placements related to pre-seismic and post-seismic events can cause pore
pressure changes and deformation inside and around the reservoir.
Therefore, these parameters should be monitored over production or Declaration of competing interest
storage reservoirs as the accuracy of such monitoring will directly affect
the reliability of the time-lapse gravity measurements. Given the The authors declare that they have no known competing financial
porosity and thickness of the modelled reservoir, we assumed negligible interests or personal relationships that could have appeared to influence
ground deformation. Despite this might be the case, there are chances the work reported in this paper.
that surface deformation occurs following injection of large volumes in
confined reservoirs (Kabirzadeh et al., 2017a, 2017b, 2017b). In such Acknowledgements
cases, the magnitude of the free-air effect related to ground deformation
can be calculated (Kabirzadeh et al., 2017b, 2020) and removed from We warmly thank four anonymous reviewers for their constructive
the gravity signal. and insightful reviews. Constructive comments from the editor Luigi
Tosi are also warmly acknowledged. We warmly thanks Equinor for
4. Conclusions making available the Volve dataset. This work was supported by funds to
PM from the Department of Engineering and Geology of the Chieti-
The Volve field, given its exploitable volume and depth, represents Pescara University.
an end-member in the lower term of reservoir classification based on
size because it can be considered a very small and deep reservoir. Thus, References
the differential gravity signatures observed in this work, together with
the retrieved detectability thresholds, represent a minimum base of the Alnes, H., Eiken, O., Stenvold, T., 2008. Monitoring gas production and CO2 injection at
the Sleipner field using time-lapse gravimetry. Geophysics 73, WA155–WA161.
gravity effects induced by fluid storage and withdrawal in real reser­ https://doi.org/10.1190/1.2991119.
voirs. Nevertheless, some general considerations can be drawn from the Alnes, H., Eiken, O., Nooner, S., Sasagawa, G., Stenvold, T., Zumberge, M., 2011. Results
modelling above. from sleipner gravity monitoring: updated density and temperature distribution of
the CO2 plume. Energy Procedia 4, 5504–5511. https://doi.org/10.1016/j.
The most relevant parameters affecting differential gravity in­ egypro.2011.02.536.
vestigations over reservoirs are represented by the depth of the reser­ Appriou, D., Bonneville, A., Zhou, Q., Gasperikova, E., 2020. Time-lapse gravity
voir, aquifer push, exploitable volume, and densities of the fluids at monitoring of CO2 migration based on numerical modeling of a faulted storage
complex. International Journal of Greenhouse Gas Control 95, 102956. https://doi.
reservoir conditions. The combination of these parameters, together org/10.1016/j.ijggc.2020.102956.
with the accuracy of the monitoring techniques, drives the recover­ Dressel, I., Barckhausen, U., Heyde, I., 2018. A 3D gravity and magnetic model for the
ability of reliable differential gravity signatures. Moreover, the depth of Entenschnabel area (German North Sea). Int. J. Earth Sci. 107, 177–190. https://doi.
org/10.1007/s00531-017-1481-x.
the reservoir, the aquifer push and available volumes – i.e. the volume Eiken, O., Stenvold, T., Zumberge, M., Alnes, H., Sasagawa, G., 2008. Gravimetric
obtained during primary production, are well-known when a reservoir monitoring of gas production from the Troll feld. Geophysics 73, WA149–WA154.
has been discovered, parametrized and exploited for years using seismic, Eiken, O., Zumberge, M.A., Sasagawa, G.S., 2000. Gravity monitoring of offshore gas
reservoirs. In: Proceedings of the 70th Annual International Meeting, SEG, Expanded
borehole and laboratory data. Density of the injected fluid is a parameter
Abstracts (Tulsa, OK: SEG), pp. 431–434.
that needs careful attention, particularly when pressures and tempera­ Elliott, E.J., Braun, A., 2016. Gravity monitoring of 4D fluid migration in SAGD
tures at reservoir allows supercritical conditions that can lead to abrupt reservoirs – forward modelling. CSEG Rec 41, 16–21. https://csegrecorder.com/artic
changes in fluid density. les/view/gravity-monitoring-of-4d-fluid-migration-in-sagd-reservoirs.
Fedi, M., Cella, F., D’Antonio, M., Florio, G., Paoletti, V., Morra, V., 2018. Gravity
Water flooding of the reservoir results in differential gravity anom­ modelling finds a large magma body in the deep crust below the gulf of Naples. Italy.
alies that are always observable even at significant depths and in very Sci. Rep. 8, 8229. https://doi.org/10.1038/s41598-018-26346-z.

8
P. Mancinelli Marine and Petroleum Geology 131 (2021) 105162

Feldmann, F., Hagemann, B., Ganzer, L., Panfilov, M., 2016. Numerical simulation of Krahenbuhl, R.A., Martinez, C., Li, Y., Flanagan, G., 2015. Time-lapse monitoring of CO2
hydrodynamic and gas mixing processes in underground hydrogen storages. sequestration: a site investigation through integration of reservoir properties, seismic
Environmental Earth Science 75, 1165. https://doi.org/10.1007/s12665-016-5948- imaging, and borehole and surface gravity data. Geophysics 80 (2), WA15–WA24.
z. Mancinelli, P., Pauselli, C., Minelli, G., Federico, C., 2015. Magnetic and gravimetric
Ferguson, J., Klopping, F., Chen, T., Seibert, J., Hare, J., Brady, J., 2008. The 4D modelling of the central Adriatic region. J. Geodyn. 89, 60–70. https://doi.org/
microgravity method for waterflood surveillance: Part III – 4D absolute microgravity 10.1016/j.jog.2015.06.008.
surveys at Prudhoe Bay. Alaska. Geophysics 73, WA163–WA171. Mancinelli, P., Porreca, M., Pauselli, C., Minelli, G., Barchi, M.R., Speranza, F., 2019.
Ferguson, J.F., Chen, T., Brady, J.L., Aiken, C.L.V., Seibert, J.E., 2007. The 4D Gravity and magnetic modelling of central Italy: insights into the depth extent of the
microgravity method for waterflood surveillance II — gravity measurements for the seismogenic layer. G-cubed 20, 2157–2172. https://doi.org/10.1029/
Prudhoe Bay reservoir, Alaska. Geophysics 72, I33–I43. https://doi.org/10.1190/ 2018GC008002.
1.2435473. Mancinelli, P., 2020. Four dimensional gravity forward model in a deep reservoir. Front.
Gasperikova, E., Hoversten, G.M., 2008. Gravity monitoring of CO2 movement during Earth Sci. 8, 285. https://doi.org/10.3389/feart.2020.00285.
sequestration: model studies. Geophysics 73 (6), WA105–WA112. https://doi.org/ Mancinelli, P., Pauselli, C., Fournier, D., Fedi, M., Minelli, G., Barchi, M.R., 2020. Three
10.1190/1.2985823. dimensional gravity local inversion across the area struck by the 2016–2017 seismic
Goto, H., Ishido, T., Sorai, M., 2020. Numerical study of reservoir permeability effects on events in central Italy. J. Geophys. Res. 125, e2019JB018853 https://doi.org/
gravity changes associated with CO2 geological storage: implications for gravimetric 10.1029/2019JB018853.
monitoring feasibility. Greenhouse Gases-Science and Technology 10, 557–566. NIST, 2016. Thermophysical Properties of Fluid Systems. https://webbook.nist.
Götz, M., Lefebvre, J., Mörs, F., McDaniel Koch, A., Graf, F., Bajohr, S., Reimert, R., gov/chemistry/fluid/.
Kolb, T., 2016. Renewable power-to-gas: a technological and economic review. Oldenburg, C.M., 2003. Carbon dioxide as cushion gas for natural gas storage. Energy
Renew. Energy 85, 1371–1390. https://doi.org/10.1016/j.renene.2015.07.066. Fuels 17 (1), 240–246.
Hagrey, S.A., Kohn, D., Wiegers, C.E., Schafer, D., Rabbel, W., 2014. Feasibility study for Olesen, O., Ebbing, J., Gellein, J., Kihle, O., Myklebust, R., Sand, M., et al., 2010. Gravity
geophysical monitoring renewable gas energy compressed in pore storages. Journal anomaly map, Norway and adjacent areas. Scale 1:3 million. Trondheim: geological
of Geology and Geosciences 3, 5. https://doi.org/10.4172/2329-6755.1000169. survey of Norway. https://www.ngu.no/en/publikasjon/gravity-anomaly-map-n
Hare, J.L., Ferguson, J.F., Aiken, C.L.V., Brady, J.L., 1999. The 4-D microgravity method orway-and-adjacent-areas-scale-13-mill.
for waterflood surveillance: a model study for the Prudhoe Bay reservoir, Alaska. Pfeiffer, W.T., al Hagrey, S.A., Kohn, D., Rabbel, W., Bauer, S., 2016. Porous media
Geophysics 64, 78–87. https://doi.org/10.1190/1.1444533. hydrogen storage at a synthetic, heterogeneous field site: numerical simulation of
Jacob, T., Bayer, R., Chery, J., Le Moigne, N., 2010. Time-lapse microgravity surveys storage operation and geophysical monitoring. Environmental Earth Science 75,
reveal water storage heterogeneity of a karst aquifer. J. Geophys. Res. 115, 1–18. 1177. https://doi.org/10.1007/s12665-016-5958-x.
https://doi.org/10.1029/2009JB006616. Pudlo, D., Ganzer, L., Henkel, S., Kühn, M., Liebscher, A., et al., 2013. The H2STORE
Jacob, T., Rohmer, J., Manceau, J.-C., 2016. Using surface and borehole time-lapse project: hydrogen underground storage – a feasible way in storing electrical power in
gravity to monitor CO2 in saline aquifers: a numerical feasibility study. Greenh. Gas. geological media? In: Hou, M.Z., Xie, H., Were, P. (Eds.), Clean Energy Systems in
Sci. Technol. 6, 34–54. https://doi.org/10.1002/ghg.1532. the Underground: Production, Storage and Conversion. Springer Series in
Kabuth, A., Dahmke, A., Beyer, C., Bilke, L., Dethlefsen, F., et al., 2017. Energy storage in Geomechanics and Geoengineering. Springer, Berlin, Heidelberg. https://doi.org/
the geological subsurface: dimensioning, risk analysis and spatial planning: the 10.1007/978-3-642-37849-2_31.
ANGUS+ project. Environmental Earth Science 76, 23. https://doi.org/10.1007/ Rönsch, S., Schneider, J., Matthischke, S., Schlüter, M., Götz, M., Lefebvre, J.,
s12665-016-6319-5. Prabhakaran, P., Bajohr, S., 2016. Review on methanation from fundamentals to
Kabirzadeh, H., Kim, J.W., Sideris, M.G., 2017a. Micro-gravimetric monitoring of current projects. Fuel 166, 276–296. https://doi.org/10.1016/j.fuel.2015.10.111.
geological CO2 reservoirs. Int. J. Greenh. Gas Control 56, 187–219. Van Camp, M., de Viron, O., Watlet, A., Meurers, B., Francis, O., Caudron, C., 2017.
Kabirzadeh, H., Sideris, M.G., Shin, Y.J., Kim, J.W., 2017b. Gravimetric monitoring of Geophysics from terrestrial time-variable gravity measurements. Rev. Geophys. 55,
confined and unconfined geological CO2 reservoirs. Energy Procedia 114, 938–992. https://doi.org/10.1002/2017RG000566.
3961–3968. Volve data village webpage. https://www.equinor.com/en/how-and-wh
Kabirzadeh, H., Kim, J.W., Sideris, M.G., Vatankhah, S., 2020. Analysis of surface gravity y/digitalisation-in-our-dna/volve-field-data-village-download.html last visited.
and ground deformation responses of geological CO2 reservoirs to variations in CO2 (Accessed 27 January 2021).
mass and density and reservoir depth and size. Environ Earth Sci 79, 163. https:// Wilkinson, M., Mouli-Castillo, J., Morgan, P., Eid, R., 2017. Time-lapse gravity surveying
doi.org/10.1007/s12665-020-08902-x. as a monitoring tool for CO2 storage. Int. J. Greenh. Gas Control 60, 93–99.
Kim, J.W., Neumeyer, J., Kao, R., Kabirzadeh, H., 2015. Mass balance monitoring of Yu, D., Jackson, K., Harmon, T.C., 1999. Dispersion and diffusion in porous media under
geological CO2 storage with a superconducting gravimeter—a case study. J. Appl. supercritical conditions. Chem. Eng. Sci. 54, 357–367.
Geophys. https://doi.org/10.1016/j.jappgeo.2015.01.003.