( ∫ ) ∫
encounters some number of existing wells. Each well is ∂ H H
(φFc(1 - Sres)h) + ∇ · Fc vc dz ) Qc dz (6)
characterized by a set of effective permeability values, with ∂t 0 0
∫ K(H - h)
H Nw
0
vb dz ) -
µb
∇ pB (9)
p(x, t) ) p0 + ∑ Q̃ G(x, x , h, t) + F(x, h, t)
i)1
i i (18)
0
Q̃ dz
any given well.
2.3.1. Fluid Injection into Confined Aquifers. In a previous
investigation (16), we considered the case of the constant-
(14)
rate injection into confined aquifers initially filled with brine.
The solution used the sharp interface assumption. It was
Here we have introduced the effective mobility given by (note
shown that the saturation distribution, given by h(x, t) has
that the relative permeability of brine kb ) 1)
a self-similar solution in time, such that the solution could
kch H - h be expressed in terms of a single space-time variable h(x, t)
λeff ) + (15) ) h(ri2/t), where ri ) ||x - xi||2 is the distance to the well. In
µcH µbH
self-similar coordinates, the equations governing the fluid
distribution and pressure are (see eqs 11 and 17 of ref 16)
Finally, within an aquifer, the source terms Qc and Qb appear
as point sources located at the wells, dh′ 2 d 1 - h′ dh′
Nw
) [
dχ χ dχ h′(λ - 1) + 1
2Γλhχ ′
dχ( +1 )] (19)
QR(x) ) ∑ δ(x - x )Q
i)1
i R,i (16) ∆p′ ) -
1
2Γ ∫ χ
ψ dχ
(h′(λ - 1) + 1)χ
+ F(h′) (20)
where δ(x - xi) is the Dirac delta function centered at the The dimensionless variables and parameters in these equa-
well location x ) xi. tions are defined as
Consider the formal solution p of eq 14 at some time t.
2πφH(1 - Sres)ri2
If the inverse of compressibility is large relative to the χ) (21)
perturbation from initial pressure in the system, which means Mc(t) ⁄ Fc
that pressure equilibrates fast relative to saturation changes,
2π(Fb - Fcw)gKλbH2 λc
we can approximate the pressure equations by superposition Γ) , λ) (22)
and introduce the quantities Q̃ λb
( )
fluid layering. This will be the case when the flow rates in
the well are small, or when the thickness of a CO2 plume at kR pBl+ - pTl-
a well is large. Using eqs 30 and 31 of ref 17, we see that this QRl ) -πrw2K w
l
FR + FRg (29)
µR Hl
situation occurs when the flow of the dense fluid (brine) in
the well is zero, Qd′ ) 0. Conditions for this are given by Equation 29 is used to describe the flux through a well
perforating caprock (or aquitard) layer l. We denote the well
k′ k′
(
ηouter2 < ηwell2 +
4Γλ)exp((h0)2Γλ) -
4Γλ
(26) by a subscript w, and the permeable formations immediately
above and below this layer are denoted by l+ and l-,
Here η ) ri/Kh is a dimensionless radial distance scaled by respectively. Note that we distinguish between pressures at
the horizontal permeability, and the dimensionless (inverse) the top and bottom of permeable formation through the use
vertical permeability is given by k′) H2/Kz. The outer radius of subscripts T and B, respectively.
of influence of the upconing is denoted by ηouter. For single- As the wells are treated as 1D objects intersecting the
phase flow, it is well-known (see, e.g., ref 2) that at steady aquifers, their internal volume is considered negligible.
state Therefore, saturation of each phase within the well segment
becomes a key parameter in the Darcy equation (eq 29) which
( )
Rη governs the value of the relative permeability kR and hence
Q̃
ηouter ) cη (27) controls the flow rate. Whenever a neighbor plume intersects
v̂H2
the well in the formation immediately below caprock section
where v̂ is the background flow rate (neglecting the effect of l (that is, formation l- in eq 29), the upconing solution of
the well), and the exponent Rη ) 1. For two-phase flow, we section 2.3.2 is used to determine flow rates for both fluids
obtain good results using Rη ) 2. in the well segment. Given the flow rates, the fluid saturation
When inequality 26 is violated, two-phase flow occurs in can be determined using the relative permeability function
the leaky well, as brine is pulled up into the well and flows (which is assumed to be knownshere we used linear
with the CO2. We describe this with the approach of ref 17, functions, but nonlinear forms can also be handled).
referring to the online appendix for the relevant equations. For wells that are not intersected by neighboring plumes
2.3.3. Well Flow Model. Flow in abandoned wells is in (so there is no upconing), we use Scl ) (Scl-1)ν where Scl-1 is the
general a complex process, which may include open well saturation in the segment below, and ν ≈ 0.5. This specifica-
total mass that may leave the geological system and should
be thought of as a proxy for the shallow groundwater and
atmosphere.
These results show the importance of including all layers
in an analysis of an injection scenario, with the concomi-
tant implications for site characterization and regulatory
requirements. We note that the rate of brine flow is also
captured in the model and shows a similar trend to CO2
leakage. However, due to lack of buoyancy, in some cases
we observe negative brine fluxes, associated with coun-
tercurrent flow of brine. A discussion of brine leakage
patterns is included in the online Supporting Information
FIGURE 4. Leakage into successive formations from a hypothetical for this publication.
injection at a location in Alberta. Aquifer numbering is from the Additional insights given by our simulator are illustrated
bottom up.
in Figure 5, which shows the distribution of secondary plume
sizes, sorted by layer. The largest plumes occur in the layer
values herein as illustrations of the variability we expect to immediately above the injection layer, with a distribution
see. Experimental methodologies for measuring the perme- clearly skewed toward higher values. Plumes become pro-
ability of abandoned wells and ongoing field campains are gressively smaller as we move upward in the stratigraphic
discussed elsewhere (7). Thus, each well segment was succession, and the distributions become more symmetric.
assigned an effective permeability value sampled randomly Again, the highest permeable layer is somewhat anomalous
from a log-normal probability distribution. Here, we assumed in its number of plumes, due to the boundary condition at
that log10 k ∼ N(µ ) log10(10 mD), σ2 ) 1), which corresponds the top of the domain. The atom of probability at 1 kg (labeled
to a most likely permeability value of 10 mD and a variance 100 in the figure) represents all smaller plumes, almost all of
of about an order of magnitude. In this example, we show which are insignificant. This information about plume sizes
a single realization of parameters from this distribution; a and their overall spatial distribution can provide practical
detailed investigation of the model behavior in response to guidance for activities such as the design of subsurface
the input parameter will be presented in a forthcoming monitoring and leakage detection strategies.
publication. Injection was modeled over a characteristic time
period of 30 years at 500 kg/s. Boundary conditions are Supporting Information Available
constant hydrostatic pressure. Note that while we only model Upconing equations used and the flow of brine predicted for
a single injection well in this run, this may also be considered the final example. This material is available free of charge via
as a proxy for a cluster of injection wells. the Internet at http://pubs.acs.org.
The accumulation of CO2 in the six permeable layers
above the injection layer, formed from a series of secondary Literature Cited
plumes, is shown in Figure 4. Similarly to the previous
results, we see a clear reduction in the amount of leakage (1) Bachu, S. and Celia., M. A. Assessing the Potential for CO2
Leakage, Particularly through Wells, from CO2 Storage Sites.
that occurs as additional layers are encountered moving The Science and Technology of Carbon Sequestration, AGU
upward in the stratigraphic sequence. This shows the Monograph; American Geophysical Union: Washington, DC,
importance of intervening permeable layers above the first 2008, in press.
impermeable layer and their ability to mitigate leakage (2) Bear, J. Dynamics of Fluids in Porous Media; Elsevier: New York,
rates. 1972.
(3) Celia, M. A.; Bachu, S.; Nordbotten, J. M.; Kavetski, D.; Gasda.,
The only layer where this pattern is not followed is the
S. Modeling critical leakage pathways in a risk assessment
upper-most permeable layer, where a no-flow boundary framework: Representation of abandoned wells. In Proceedings
above it leads to a somewhat artificial accumulation of mass. of the 4th Annual Conference on Carbon Capture and Seques-
In this sense, the upper layer in our model represents the tration, May 2-5, 2005, Alexandria, VA.