You are on page 1of 27

Int J Geomath (2013) 4:27–53

DOI 10.1007/s13137-012-0044-3

ORIGINAL PAPER

A mathematical model of a passive scheme for acid


mine drainage remediation

M. Gouin · E. Saracusa · C. B. Clemons ·


J. Senko · K. L. Kreider · G. W. Young

Received: 6 January 2012 / Accepted: 3 October 2012 / Published online: 20 October 2012
© Springer-Verlag Berlin Heidelberg 2012

Abstract Acid mine drainage (AMD) is pollution that occurs when water trapped in
abandoned coal mines leaks out into the local environment. One of the most abundant
metals found in AMD is iron. Iron precipitates out of the system naturally in the form
of iron(III) hydroxide, leading to a decrease in fluid pH. Iron oxidizing bacteria can
promote this reaction process, assisting in the removal of Fe(II) ions from the AMD.
This paper develops a model of an AMD system that includes thin film sheet flow
and reactive transport of ions. In addition the model analyzes the precipitation of an
iron(III) hydroxide crust as a function of space and time. Several parameters are varied
to compare model predictions with field observations. Forcing low concentrations of
oxygen in the system slows the rate of iron(III) hydroxide crust growth. Increasing
the angle of inclination of the sheet flow while holding the flow rate fixed causes a
very small decrease in iron(III) hydroxide crust growth due to faster flow through a
thinner fluid film. Decreasing the flow rate of the system while maintaining a fixed
angle of inclination to simulate a thinning of the liquid layer produces a slower iron(III)
hydroxide crust growth rate. When the angle of inclination is increased with fixed fluid
film thickness, iron(III) hydroxide crust growth decreases due to downstream transport
of the ions. Oxidation and precipitation reaction rates had the greatest influence on
crust growth. An increase in iron oxidizing bacteria can increase the oxidation of
Fe(II). Therefore, an increase in the abundance of iron oxidizing bacteria can increase
the rate of iron(III) hydroxide crust growth.

M. Gouin · E. Saracusa · C. B. Clemons · K. L. Kreider · G. W. Young (B)


Department of Mathematics, The University of Akron, Akron, OH 44325-4002, USA
e-mail: gyoung1@uakron.edu
J. Senko
Department of Geology, The University of Akron, Akron, OH 44325-4101, USA

123
28 Int J Geomath (2013) 4:27–53

Keywords Acid mine drainage · Mathematical model · Passive remediation ·


Iron-oxidizing bacteria

Mathematics Subject Classification 92E20 Chemistry, classical flows, reactions ·


76A20 Thin fluid films
1 Introduction

Coal mine-derived acid mine drainage (AMD) occurs when iron sulfide phases
(notably pyrite; FeS2 ) that are exposed during mining operations come into contact
with O2 -rich fluids that infiltrate the abandoned mine works (Baker and Banfield 2003).
Aerobic oxidation of the sulfide moiety of iron sulfides leads to acidic conditions asso-
ciated with the production of sulfuric acid. The low pH enhances the dissolution of
adjacent metals (including Fe, Mn, and Al with Fe being the major metal of concern
in coal mine-derived AMD). The acidic and metal-rich fluids that ultimately emerge
from the abandoned mines infiltrate circumneutral surface waters, where the Fe ions
hydrolize and precipitate as Fe(III) (hydr)oxides, which coat stream beds and leave
the waterways largely devoid of plant and animal life (Herlihy et al. 1990). As such,
removal of dissolved Fe(II) from AMD is a major goal of AMD treatment approaches
(Senko et al. 2008).
Several AMD treatment strategies have been developed to remove Fe(II) from
AMD. These strategies require various levels of human intervention. Approaches
that require minimal human intervention (often referred to as passive treatment) are
preferred so that labor, energy, and maintenance costs can be minimized. A com-
monly implemented passive treatment approach is to divert the flow of AMD through
limestone-based systems (Johnson and Hallberg 2005). In such systems dissolution
of carbonate phases in the limestone neutralize the AMD, thus enhancing the kinetics
of Fe(II) oxidation (Eq. 1) and subsequent hydroloysis and precipitation of Fe(III) as
a variety of Fe(III) (hydr)oxide phases (Eq. 2).

4 Fe2+ + O2 + 4 H+ −→ 4 Fe3+ + 2 H2 O, (1)


Fe3+ + 3 H2 O −→ Fe(OH)3 + 3 H+ . (2)

However, the Fe(III) (hydr)oxides that are produced in such systems coat the lime-
stone (called armoring), limiting further dissolution of carbonate phases, and requiring
replacement or flushing of the systems.
A recently proposed strategy to remove dissolved Fe before limestone-based neu-
tralization is to facilitate the activities of aerobic Fe(II) oxidizing bacteria (FeOB),
which oxidatively precipitate Fe from the AMD (Eqs. 1, 2). We have observed such
activities in a variety of AMD-impacted systems in which the AMD flows as a sheet
over the terrestrial surface, see Fig. 1 (Senko et al. 2008; DeSa et al. 2010). The sheet-
flow facilitates aeration of the AMD and stimulates the activities of FeOB. Since these
systems that exhibit extensive capacity for Fe removal have developed with no human
intervention, they may represent an inexpensive and sustainable approach for AMD
treatment and minimize the armoring associated with traditional limestone-based
treatment systems.

123
Int J Geomath (2013) 4:27–53 29

Fig. 1 Overview of the sheet-flow region of the Mushroom farm system, showing overview of the region
studied (a), dissolved Fe(II) concentration (b), pH, and DO (c) in AMD from the emergence point at the
side of the pictured house over a distance of approximately 30 m

In order to implement engineered sheet-flow systems for the treatment of AMD,


robust predictive models are needed to guide the design of such systems. The
abundance of FeOB appears to exert control on the rates of Fe(II) oxidation
(Senko et al. 2011), and this is likely a reflection of the physiochemical setting of
the systems. Some parameters to predict the performance of AMD treatment systems
have been established by Kirby et al. (1999). They found that when pH is greater than
five, the oxidative precipitation of Fe is controlled predominantly by abiotic factors,
and not by the activities of FeOB. Kirby et al. (1999) also used a mathematical model
to evaluate the impacts of dissolved O2 (DO), Fe(II) concentration and pH on the kinet-
ics of Fe removal from AMD. Burke and Banwart (2002) created a model for abiotic
oxidation of iron using extrapolated literature values to determine abiotic empirical
rate laws for Fe(II) oxidation. They developed a steady state model that represents a
single, completely mixed reactor in the surface-catalyzed oxidation of Fe(II). Burke
and Banwart (2002) suggest Fe(II) removal can occur if the AMD is spread over a
sufficiently large surface area, similar to that observed in sheet-flow systems. Burgos et
al. (2002) developed a reaction-based biogeochemical model for the kinetics of Fe(III)
reduction by an Fe(III)-reducing bacterium, and found that independently-determined
reaction-based rate formulations in other experiments agreed. This finding implies that
simulation and prediction of complex biogeochemical systems may eventually be per-
formed using reaction-based models. However, reaction-based models are not always
easy to develop because reaction networks and mechanisms are sometimes difficult to
determine. For example, one must identify all microbial species active in a system, all
chemical reactions taking place both in the bulk fluid and at surfaces, the reaction rates
associated with these reactions, and how these rates vary with surrounding factors such
as temperature and ionic species concentrations.
In this paper we develop a two-dimensional model (see Fig. 2) of the biogeochem-
ical dynamics associated with a sheet-flow system in eastern Ohio, referred to as the
Mushroom farm (MF) system. At the MF, AMD emerges from the subsurface and
flows as a sheet. After flowing a distance of approximately 30 m, the AMD is aerated,

123
30 Int J Geomath (2013) 4:27–53

Flow

z
AMD
z = h( x , t )

Crust
z = s( x , t )

θ
x
Soil

Fig. 2 A schematic of the proposed geometry of the AMD system including growth of an iron hydroxide
crust

and approximately 12 mM Fe(II) is removed from solution (Fig. 1). pH decreases con-
currently with Fe(II) removal, suggestive of the hydrolysis and precipitation of Fe(III).
The model also accounts for iron(III) hydroxide crust growth. Only the iron con-
taminant is taken into account for removal, ignoring removal of other ions present.
Ultimately, it may be possible to add a vegetated region to remove the other metal
ions, however, the main focus remains on the removal of dissolved iron. In order to
remove Fe(II), we utilize a sheet-flow system, which is simply a thin layer of liquid
AMD flowing downhill over relatively smooth land surface. A sheet-flow regime will
allow for efficient diffusion of O2 (Ardejani et al. 2010) from the surrounding air to
the liquid, which is necessary since aerobic iron-oxidizing bacteria consume it, and
oxygen is a reactant in the oxidation of Fe(II) (Eq. 1).
The Mushroom farm AMD sheet flow is 0.5–1 cm thick, covering an area of approx-
imately 45 m2 (Senko et al. 2011). We assume the bacteria are located only in the liquid
medium and not in the crust. A diagram of a cross-section of AMD flow is shown in
Fig. 2. In the figure, we define the location of the crust surface, ẑ = ŝ(x̂, tˆ ). The
location of the top of the air–liquid interface is defined as ẑ = ĥ(x̂, tˆ ). The distance
from the top of the crust to the air–liquid interface has a distance that is estimated to
be no larger than 2 cm. The solution of the thin liquid film problem and the reactive
transport problem (complete mixing is not assumed as is done in other models) are
uncoupled by holding liquid density constant and neglecting buoyancy driven flow
due to temperature gradients and concentration gradients.
Our model is similar to previous studies in that we model the change in concentration
of Fe(II) as a function of iron-oxidizing bacterial concentration through changes in
oxidation rates, see Senko et al. (2011). Our model differs from others since we
track the growth of the Fe(OH)3 crust as a function of oxidation rates, pH, and ion
concentrations. The assumptions of the proposed model are
– Bacteria are located only in the liquid.
– The model is two-dimensional in space.
– Flow is laminar.
– The model is meant to remove only iron, and no other metal ions.
– The density of the iron(III) hydroxide crust is constant throughout and the crust is
non-porous.
– The distance over which the AMD flows is far greater than the thickness of the crust.

123
Int J Geomath (2013) 4:27–53 31

– The thickness of the AMD fluid film is assumed to be constant in space, but allowed
to vary in time due to seasonal changes in precipitation.
– Temperature effects on bacteria activity are not considered.
– The capillary number relating surface tension forces to viscous forces is large.
– Bulk reactions are fast so that bulk rate laws approach equilibrium conditions and
are nearly zero in magnitude.
– The rate of iron(III) hydroxide crust growth is slow and so the surface reaction
rate law at the crust interface, which defines the crust growth rate in the model, is
nearly zero in magnitude.
– Mass transport from the liquid AMD to the non-porous iron(III) hydroxide crust
is small.

2 Hydrodynamic model

Figure 2 is a schematic of the proposed geometry with the bacteria treated as part of
the AMD fluid domain. We define all components with hats as dimensional variables.
Let the x̂-axis be the plane inclined at a small angle θ with respect to the surround-
ing terrain. The ẑ-axis is normal to the x̂-axis. The air–liquid interface is labeled as
ẑ = ĥ(x̂, tˆ), while the liquid–crust interface is assigned ẑ = ŝ(x̂, tˆ ).
There is no evidence that indicates the growth of iron hydroxide in deep water
(Senko et al. 2011). The sheet flow is at most 2 cm deep for the Mushroom Farm, thus,
the upper bound for the distance from ẑ = ŝ(x̂, tˆ) to ẑ = ĥ(x̂, tˆ ) is between 1 and 2 cm.
Temperature for this model is held constant and its value can be set according to the
time of year under consideration. Density is held constant, which allows us to ignore
buoyancy driven flow. The flow under consideration is two-dimensional. The liquid is
viscous and incompressible.
Conservation of mass is defined by the continuity equation

û x̂ + ŵẑ = 0, (3)
where û is the velocity component in the x̂-direction and ŵ is the velocity component
in the ẑ-direction. All subscripts represent the partial derivative with respect to the
subscripted variable. The Navier–Stokes equations that describe the conservation of
linear momentum are

−1
û tˆ + û û x̂ + ŵ û ẑ =
p̂x̂ + ν(û x̂ x̂ + û ẑ ẑ ) + g sin θ, (4)
ρ
−1  
ŵtˆ + û ŵx̂ + ŵ ŵẑ = p̂ẑ + ν ŵx̂ x̂ + ŵẑ ẑ − g cos θ, (5)
ρ
where ρ is the density of the AMD liquid, g is gravity, p̂ is fluid pressure, and ν is the
kinematic viscosity. The surface ẑ = ŝ(x̂, tˆ) is a solid. A no-slip condition is applied
at the crust–liquid interface,

û = ŵ = 0. (6)

123
32 Int J Geomath (2013) 4:27–53

The air–liquid interface, ẑ = ĥ(x̂, tˆ), is a free boundary. The conditions applied at
the boundary are: the kinematic condition, no shear stress in the tangential direction
and the normal force balance equation. The kinematic condition is

ĥ tˆ = ŵ − û ĥ x̂ . (7)

No shear stress in the direction tangent to the surface ẑ = ĥ(x̂, tˆ) is given by

2ĥ x̂ (−û x̂ + ŵẑ ) + (1 − ĥ 2x̂ )(û ẑ + ŵx̂ ) = 0. (8)

The force in the direction normal to the surface ẑ = ĥ(x̂, tˆ) is


 
μ û x̂ ĥ 2x̂ − ĥ x̂ (û ẑ + ŵx̂ ) + ŵẑ ĥ x̂ x̂
− p̂ + =σ 3
, (9)
1 + ĥ x̂ 2
(1 + ĥ 2x̂ ) 2

where σ is the surface tension of the AMD.

2.1 Reactive transport model

The reaction system is based on the oxidation of Fe(II) followed by its subsequent
precipitation in the form of Fe(OH)3 . In this model, the precipitation of Fe(II) to Fe(III)
is the rate determining step. The reaction system is based on four chemical species.
Reaction (1) is treated as a bulk reaction,

k1
4 Fe2+ + O2 + 4 H+  4 Fe3+ + 2 H2 O. (10)
k−1
Here, k1 is the forward reaction rate and k−1 is the backward reaction rate, where
the forward reaction rate is dependent on the concentration of iron-oxidizing bacteria.
Therefore, the rate law for this bulk reaction system is

Iˆ = −k1 ĈFe
4
2+ Ĉ O2 Ĉ H+ + k −1 Ĉ Fe3+ .
4 4
(11)

In the analysis that follows we assume the bulk reactions are at near equilibrium
conditions. Hence, the magnitude of Iˆ is small.
The following are governing equations for the concentrations of ions in the bulk,

∂ ĈH+ ∂ Ĉ + ∂ Ĉ +  
+ û H + ŵ H = ∇ · DH+ ∇ ĈH+ + Iˆ, (12)
∂ tˆ ∂ x̂ ∂ ẑ
∂ ĈO2 ∂ ĈO2 ∂ ĈO2  
+ û + ŵ = ∇ · DO2 ∇ ĈO2 + Iˆ, (13)
∂ tˆ ∂ x̂ ∂ ẑ
∂ ĈFe2+ ∂ Ĉ 2+ ∂ Ĉ 2+  
+ û Fe + ŵ Fe = ∇ · DFe2+ ∇ ĈFe2+ + Iˆ, (14)
∂ tˆ ∂ x̂ ∂ ẑ
∂ ĈFe3+ ∂ Ĉ 3+ ∂ Ĉ 3+  
+ û Fe + ŵ Fe = ∇ · DFe3+ ∇ ĈFe3+ − Iˆ. (15)
∂ tˆ ∂ x̂ ∂ ẑ

123
Int J Geomath (2013) 4:27–53 33

In these equations, Di is the mass diffusivity of the ith chemical species. Similarly,
Ĉi represents the concentration of the ith chemical species defined in Table 1.
At the air–liquid interface, ẑ = ĥ(x̂, tˆ), the concentration of oxygen, ĈO2 , is a
volatile gas. Oxygen can leave and enter the system at ẑ = ĥ(x̂, tˆ),
 
DO2 ∇ ĈO2 · n = K M T K H PO2 − ĈO2 . (16)

Here, K M T is the mass transfer coefficient for oxygen gas, K H is the Henry’s constant,
and PO2 is the partial

pressure

of oxygen. The unit normal vector at the ẑ = ĥ(x̂, tˆ)
−ĥ x̂ ,1
interface is n =  . All of the other species considered in this model are non-
1+ĥ 2x̂
volatile ions, satisfying no-flux conditions at this interface,

∇Ci · n = 0. (17)

Fe(III) in the liquid may precipitate and become a part of the iron-hydroxide crust
via Reaction (2),

κ−1
Fe3+ + 3 H2 O  Fe(OH)3 + 3 H+ , (18)
κ1

where κ1 and κ−1 are the forward and backward reaction rate coefficients respectively.
Therefore, the rate law is given by

J = −κ1 ĈFe3+ + κ−1 ĈH3 + . (19)

Field data (Senko et al. 2011) from the Mushroom Farm shows a slow crust growth
rate on the order of 1 cm/year. Hence, we assume the magnitude of J is small. If the
value of J is positive, by Eq. (19), κ−1 ĈH3 + must be greater in magnitude than κ1 ĈFe3+ .
By Le Châtelier’s Principle, equilibrium of Reaction (2) shifts to the left-hand side of
the reaction, causing the dissolution of Fe(OH)3 . Conversely, if J is negative, κ−1 ĈH3 +
must be less in magnitude than κ1 ĈFe3+ . Then the equilibrium in Reaction (2) shifts
to the right-hand side of the reaction, causing the precipitation of Fe(OH)3 . While
this reaction is occurring throughout the bulk, we model its effect through boundary
conditions at the iron hydroxide crust. This leads to

DH+ ∇ ĈH+ · ns = κ−1 ĈH3 + − κ1 ĈFe3+ , (20)


DFe3+ ∇ ĈFe3+ · ns = −κ−1 ĈH3 + + κ1 ĈFe3+ , (21)

where the normal vector to the crust liquid interface is ns = (√ x 2) .


−ŝ ,1
1+ŝx
Oxygen can leave and enter the surface ẑ = ŝ(x̂, tˆ) via absorption by the soil or the
roots of plants. For this reason, the net flux of oxygen traveling from the liquid into
the crust must be considered. We assume

123
34 Int J Geomath (2013) 4:27–53

DO2 ∇ ĈO2 · ns = Fc . (22)

Here, Fc is the net flux of oxygen into the iron-hydroxide crust which is assumed to
be known and small due to the non-porous structure of the crust. Since Fe(II) does not
participate in Reaction (2), there is no flux of Fe(II) ions into the crust at ẑ = ŝ(x̂, tˆ),

DFe2+ ∇ ĈFe2+ · ns = 0. (23)

The boundary condition defining the location of the crust–liquid interface is deter-
mined so that the normal velocity of the iron-hydroxide crust is driven by the rate
law (19)

ρ Fe(O H )3 ∂∂ŝtˆ  
 = −κ−1 ĈH3 + + κ1 ĈFe3+ , (24)
1 + ŝx̂2

where ρ Fe(O H )3 is the density of the iron hydroxide crust. This equation is a mass
balance equation which states that the rate of increase of iron hydroxide crust is equal
to the rate of loss of species from the liquid due to Reaction (2). Again, as mentioned
above, since the measured crust crust growth rate is slow, then the right hand side of
(24), which is −J , is assumed to be small.

3 Solution procedure

First, the solution to the dynamics of the fluid flow in the system is considered. This
is followed by analysis of the reactive transport equations. Once the solutions to these
equations are obtained, the iron hydroxide crust evolution, Eq. (24), can be determined.

3.1 Hydrodynamics solution procedure

The fluid-flow equations for the model are solved using asymptotic expansions defined
according to standard thin-film approximations (Oron et al. 1997). Equations (2), (3),
(4), (5), (6) and (8) are used to define solutions for p̂, û, and ŵ, which depend upon ŝ
and ĥ, the unknown free boundaries.
We assume that the length of the distance over which the fluid is flowing is much
larger in magnitude than the depth of the liquid. Let d be the characteristic depth of
the liquid and L be the characteristic distance over which the liquid travels. Since d is
on the order of a few centimeters, while L is on the order of meters at our location of
interest, this assumption is justified. We define the aspect ratio,  = Ld , where   1.
Table 1 summarizes the non-dimensional variables employed.
In Table 1 U , P, and T are the characteristic velocity, pressure, and time
2
scales respectively. The time scale, T =  Dd 3+ , is determined utilizing the non-
Fe
dimensionalized version of Eq. (24). For this reason, T  1. This is consistent with
field data which indicate that the growth of the iron-hydroxide crust is on the order of

123
Int J Geomath (2013) 4:27–53 35

Table 1 Non-dimensional
Dimensional variable Scaling Non-dimensional
variables
variable

Longitudinal dimension: x̂ L x̂
L

Transverse dimension: ẑ d ẑ
d
2
Longitudinal velocity: û U = ρgd
μ

U

Transverse velocity: ŵ W = ULd ŵL


Ud

Pressure: p̂ P = νU2L νU L

d
d2
Air–liquid interface: ĥ d ĥ
d

Crust–liquid interface: ŝ d ŝ
d

tˆ T =  Dd
2 tˆ
Fe3+
T

∗ ĈH+
Concentration hydrogen: ĈH+ CH + C∗ +
H

∗ ĈO2
Concentration oxygen: ĈO2 CO CO∗
2
2
ĈFe2+
Concentration Fe(II): ĈFe2+ C∗ C ∗ 2+
Fe2+
Fe
ĈFe3+
Concentration Fe(III): ĈFe3+ C∗ C ∗ 3+
Fe3+
Fe

about 1 cm per year. Since the unit of measurement for time is seconds, the magnitude
of T is large.
Applying this non-dimensional scheme to the hydrodynamic governing equations,
we find that the equation for conservation of mass is

u x + wz = 0. (25)

The Navier–Stokes equations become

 u t +  Re uu x + wu z = − px +  2 u x x + u zz + sin θ, (26)
Sc

and

3 wt +  3 Re uwx + wwz = − pz +  4 wx x +  2 wzz −  cos θ. (27)


Sc

The non-dimensional groups are defined in Table 2.


At the crust–liquid interface, z = s(x, t), the no-slip boundary condition becomes

u = w = 0. (28)

123
36 Int J Geomath (2013) 4:27–53

Table 2 Non-dimensional groups

Parameter Scaling Description

Reynolds number: Re Ud Inertial forces relative to viscous forces


ν
Schmidt number: Sc ν Mass transport by convection relative to transport by
DFe3+
diffusion
Capillary number: Ca σ Surface tension force of the liquid relative to the
μU
viscous force
Peclet number: Pe Ud Transport of concentration via convection relative to
Di
transport of concentration via diffusion
κ1 d
Damköhler number: Da(1) DFe3+ The forward chemical reaction rate relative to the
diffusive transport
 2
κ−1 d C ∗ +
Damköhler number: Da(−1) H The backward chemical reaction rate relative to the
DFe3+
diffusive transport

At the air–liquid interface, z = h(x, t), the non-dimensional representation of the


kinematic condition, Eq. (7), is

DFe3+
h t = w − uh x . (29)
Ud
As will be discussed shortly, we impose a small limit approximation to the non-
D
dimensional group UFed3+ . This will lead to a spatially constant AMD fluid film thick-
ness. Shear stress, Eq. (8), becomes

− 2 2 h x u x + 2 2 wz h x + u z +  2 wx −  2 h 2x u z −  4 h 2x wx = 0. (30)

The normal force balance is given by


 2  2 u x h 2x − u z h x −  2 wx h x + wz ¯ xx
 Cah
− p+   =  3 , (31)
1 +  2 h 2x 1 +  2 h 2x 2
σ
where Ca = μUis the capillary number. Consistent with thin-film flow approxima-
¯ ¯ is an O(1) constant. Hence, surface
tions, we take Ca to be large, Ca = Ca
2
, where Ca
tension forces are assumed to be large compared to viscous forces.
We assume the following asymptotic expansions

p = p0 + p1 +  2 p2 + · · · ,
u = u 0 + u 1 +  2 u 2 + · · · , (32)
w = w0 + w1 +  2 w2 + · · · .

In addition, we consider the boundary perturbations,

s(x, t) = s0 (x, t) + s1 (x, t) +  2 s2 (x, t) + · · · (33)


h(x, t) = h 0 (x, t) + h 1 (x, t) +  2 h 2 (x, t) + · · · (34)

123
Int J Geomath (2013) 4:27–53 37

Placing these expansions in Eqs. (25), (26), and (27), we obtain the leading order
problem
u 0x + w0z = 0,
− p0x + u 0zz = − sin θ, (35)
− p0z = 0,
subject to the conditions

u 0 = w0 = 0, (36)

at z = s0 (x, t), and

p0 = 0, u 0z = 0, (37)

at z = h 0 (x, t). The solution to this leading order system is

p0 = 0,

z2 s02
u 0 = sin θ − + h 0 z + − s0 h 0 , (38)
2 2

h 0x z 2 s02 h 0
w0 = sin θ − − s0 s0x z + s0x h 0 z + s0 h 0x z − + s0 s0x − s0 s0x h 0 .
2
2 2

To find the flow rate, Q, of the system we integrate u 0 from s0 to h 0 and find

h 0
sin θ
Q= u 0 dz = (h 0 − s0 )3 . (39)
3
s0

We use the leading order kinematic condition to define an evolution equation of the
liquid–air interface h 0 (x, t),

DFe3+
h 0t = w0 − u 0 h 0x . (40)
Ud

Table 3 lists the approximate values of mass diffusivity estimated from data in
Senko et al. (2011). Given the magnitude of DFe3+ , a film thickness on the order of
DFe3+
centimeters and flow velocities on the order of cm/s, we see that the ratio Ud has a
DFe3+
small magnitude. Because Udhas a small magnitude, we approximate the left-hand
side of Eq. (40) by zero. Substituting the velocity expressions, Eqs. (38) into (40), we
find

1  
− sin θ (h 0 − s0 )3 = 0. (41)
3 x

123
38 Int J Geomath (2013) 4:27–53

Table 3 Approximate values of 2


mass diffusivities Mass diffusivity Approximate value cm
sec

DH + 9 × 10−5
DO2 1.97 × 10−5
DFe2+ 7.26 × 10−5
DFe3+ 6.05 × 10−5

When considering Eqs. (39) and (41) together, we conclude that the derivative of the
flow rate Q with respect to x is zero. Hence the flow rate must be solely a function of
time. Furthermore, Eq. (39) provides a relationship between s0 and h 0 ,
 1
3Q 3
h0 = + s0 . (42)
sin θ
This indicates that the position of the liquid–air interface can be determined once
the position of the crust–liquid interface is determined. The location of h 0 is simply
a positive vertical shift of the location of s0 . Hence, the AMD fluid film thickness is
constant with respect to space, but can change in time if the flow rate Q changes in
time to reflect seasonal changes in precipitation. This constant thickness assumption
will significantly simplify the mathematical solution of the problem by uncoupling
the solution of the evolution equations for s0 and h 0 .

3.2 Reactive transport solution procedure

Applying the same non-dimensional scheme utilized in the hydrodynamic governing


equations, and the non-dimensional variables in Table 1, we find the reactive transport
governing equations become,
 DFe3+
 I¯d 2
CH+ t +  Pe uCH+ x + wCH+ z =  2 CH+ x x + CH+ zz + , (43)
DH + DH+ CH∗ +
 DFe3+
 I¯d 2
CO2 t +  Pe uCO2 x + wCO2 z =  2 CO2 x x + CO2 zz + , (44)
DO2 DO2 CO∗ 2

 DFe3+

CFe2+ t +  Pe uCFe2+ x + wCFe2+ z


DFe2+
 I¯d 2
=  2 CFe2+ x x + CFe2+ zz + ∗ , (45)
DFe2+ CFe 2+

CFe3+ t +  Pe uCFe3+ x + wCFe3+ z


 I¯d 2
=  2 CFe3+ x x + CFe3+ zz − ∗ . (46)
DFe3+ CFe 3+

123
Int J Geomath (2013) 4:27–53 39

Since the bulk reactions are approximately at equilibrium, the magnitude of Iˆ, Eq. (11),
is near zero. Hence we scale Iˆ to be small, Iˆ =  I¯.
The non-dimensional representation of Eq. (16) is

− 2 CO2 x h x + CO2 z

 =  K̄ M T K H PO2 − CO∗ 2 CO2 , (47)


1 +  hx
2 2

where DdO K M T =  K̄ M T . The scale for the mass transfer coefficient of oxygen
2
is chosen to be small to insure that the flux of oxygen at the air–liquid interface is
accounted for in the first correction for the system. Because the bulk reactions in the
system are small, this mass transfer coefficient must also be small. For Eq. (17) we
have

− 2 CH+ x h x + CH+ z
 = 0,
1 +  2 h 2x
− 2 CFe2+ x h x + CFe2+ z
 = 0, (48)
1 +  2 h 2x
− 2 CFe3+ x h x + CFe3+ z
 = 0.
1 +  2 h 2x

Non-dimensionalization of the flux conditions at the iron hydroxide crust leads to


 
− 2 CH+ x sx + CH+ z ¯ (−1) D 3+
 Da
 = Fe
C H+ 3
1 +  2 sx2 DH +
 
¯ (1) D 3+ C ∗ 3+
 Da Fe
− Fe
CFe3+ , (49)
CH∗ + DH+
 
− 2 CFe3+ x sx + CFe3+ z ¯ (−1) C ∗ +
 Da  
 =− H ¯ (1) C 3+ , (50)
CH+ 3 +  Da
∗ Fe
1 +  2 sx2 CFe 3+

where Da(1) =  Da ¯ (1) and Da(−1) =  Da


¯ (−1) . Da(1) and Da(−1) are Damköhler
numbers. The scalings for the Damköhler numbers are chosen to account for the small
rate of crust growth and the subsequent small magnitude of the surface rate law J as
defined in Eq. (19).
The oxygen transport condition across the surface z = s(x, t) is

− 2 CO2 x sx + CO2 z
 =  F̄, (51)
1 +  2 sx2
 
where  F̄ = C ∗ dDO Fc . The scale for the net flux of oxygen through the iron(III)
2
hydroxide crust is chosen to be small to insure that the flux of oxygen at the liquid–
solid interface is accounted for in the first correction for the system. In other words,

123
40 Int J Geomath (2013) 4:27–53

the net flux of oxygen through the crust is assumed to be limited by the nonporous
crust.
The no-flux condition of Fe(II) at the crust–liquid interface (z = s(x, t)) is

− 2 CFe2+ x sx + CFe2+ z
 = 0. (52)
1 +  2 sx2

We assume the following asymptotic expansions where  is small:

C H+ = CH+ + CH+ +  2 CH+ + · · ·


0 1 2

CFe2+ = CFe2+ + CFe2+ +  2 CFe2+ + · · ·


0 1 2
(53)
CFe3+ = CFe3+ + CFe3+ +  2 CFe3+ + · · ·
0 1 2

CO2 = CO20 + CO21 +  2 CO22 + · · ·

Substituting these expansions into Eqs. (43)–(46), we obtain the leading order reactive
transport governing equations

CH+ = 0, CO20zz = 0, CFe2+ = 0 and CFe3+ = 0. (54)


0zz 0zz 0zz

subject to

CH+ = 0, CO20z = 0, CFe2+ = 0, and CFe3+ = 0, (55)


0z 0z 0z

at the z = s0 (x, t) interface, and

CH+ = 0, CO20 = 0, CFe2+ = 0, and CFe3+ = 0, (56)


0z z 0z 0z

at the z = h 0 (x, t) interface. Solution of these equations indicates that CH+ , CO2 ,
CFe2+ , and CFe3+ are functions of x. To determine these, we find and solve the O()
system of equations,

d2

C H+ + I¯ = Pe u 0 CH+ 0x , (57)
1zz DH∗ CH∗ +
d2  
CO21zz + I¯ = Pe u 0 CO2 , (58)
DO2 CO∗ 2 0x

d2  
CFe2+ + I¯ = Pe u 0 C 2+ , (59)
1zz ∗
DFe2+ CFe Fe0x
2+

d2  
CFe3+ − I¯ = Pe u 0 C 3+ , (60)
1zz ∗
DFe3+ CFe Fe0x
3+

123
Int J Geomath (2013) 4:27–53 41

 
where I¯ = −k1 CFe
4∗ C 4 ∗
2+ Fe2+ 0 C O2 C O2 0 C H+ C H+ 0 + k −1 C Fe3+ C Fe3+ 0 . The boundary
4∗ 4 4∗ 4

conditions are

   
¯ (−1) D 3+
Da ¯ (1) D 3+ C ∗ 3+
Da Fe
C H+ = Fe
CH3 + − Fe
CFe3+ , (61)
1z DH + 0 CH∗ + DH+ 1

CO21 = F̄, (62)


z
CFe2+ = 0, (63)
1z

¯ (−1) C ∗ +
Da
CFe3+ = − H ¯ (1) C 3+ ,
CH3 + + Da (64)
1z ∗
CFe 0
Fe 0
3+

at the z = s0 (x, t) interface.


At the z = h 0 (x, t) interface we have,

CH+ = 0, (65)
1z
 
K̄ M T K H+ PO2 − CO∗ 2 CO20
CO21 = , (66)
z CO∗ 2
CFe2+ = 0, (67)
1z

CFe3+ = 0. (68)
1z

To solve equations (54), each is integrated once with respect to z from z = s(x, t) to
z = h(x, t), followed by applying the respective boundary conditions at each interface.
This leads to a set of equations for the leading order concentration fields. Details can
be found in Gouin (2011).
For ease in later numerical analysis, the leading order equations are re-dimension-
alized. The leading order reactive transport equations are

Q̂ ĈH+ = −κ−1 ĈH3++ + κ1 ĈFe3+


0x̂ 0 0
 1/3  
3 Q̂μ
+ 4
k−1 ĈFe 3+ − k 1 Ĉ 4
Fe2+
Ĉ 4
+ Ĉ O20 , (69)
ρg sin θ 0 0
H0
 
Q̂ ĈO20 = K M T K H PO2 − CO20 − Fc

 1/3  
3 Q̂μ
+ k−1 ĈFe3+ − k1 ĈFe2+ ĈH+ ĈO20 ,
4 4 4
(70)
ρg sin θ 0 0 0
 1/3  
3 Q̂μ
Q̂ ĈFe2+ = k−1 ĈFe3+ − k1 ĈFe2+ ĈH+ ĈO20 ,
4 4 4
(71)
0x̂ ρg sin θ 0 0 0

123
42 Int J Geomath (2013) 4:27–53

Q̂ ĈFe3+ = κ−1 ĈH3 + − κ1 ĈFe3+


0x̂ 0 0
 1/3  
3 Q̂μ
+ −k−1 ĈFe
4
3+ + k 1 Ĉ 2+ Ĉ H+ Ĉ O20 .
4 4
(72)
ρg sin θ 0 Fe 0 0

From Eqs. (11) and (19), we may write Eqs. (69)–(72) as


 1/3
3 Q̂μ
Q̂ ĈH+ = −J + Iˆ, (73)
0x̂ ρg sin θ
 1/3
  3 Q̂μ
Q̂ ĈO20 = K M T K H PO2 − CO20 − Fc + Iˆ, (74)
x̂ ρg sin θ
 1/3
3 Q̂μ
Q̂ ĈFe2+ = Iˆ, (75)
0x̂ ρg sin θ
 1/3
3 Q̂μ
Q̂ ĈFe3+ =J− Iˆ. (76)
0x̂ ρg sin θ

Here, the left sides represent the downstream convective transport of each respective
ion, while the right sides represent reactions (1), through the Iˆ term and (2), through
the J term. Time dependent boundary conditions of the form

Ĉ(0, t) = Ĉ 0 (1 + 0.1 sin(4π tˆ/Tmax )) (77)

are applied for each species concentration at the mine source x = 0 to model seasonal
fluctuations in flow; Tmax = 31,536,000 s (1 year) and C 0 is an average measured
concentration at the source; the value for each species is listed in Table 4. In Eq. (74),
the flux through the gas–liquid interface is represented by the K M T term and the flux
through the solid–liquid interface is represented by Fc . Equations (11) and (19) indicate
that precipitation of iron(III) hydroxide crust occurs when J and Iˆ are negative.
Next, we non-dimensionalize Eq. (24), the iron-hydroxide crust growth equation,
and find

st D̄a(−1) CH∗ + D̄a(1) CFe 3+
 =− CH3 + + CFe3+ . (78)
1 +  2 sx2 ρ Fe(O H )3 0 ρ Fe(O H )3 0

Performing an asymptotic expansion, we have the leading order problem for the loca-
tion of the crust interface,
   ∗

D̄a(−1) CH∗ + D̄a(1) CFe 3+
s0t = − C H+ +
3
CFe3+ . (79)
ρ Fe(O H )3 0 ρ Fe(O H )3 0

We note that the crust growth rate in Eq. (79) is a function of CH+ and CFe3+ . We
0 0
re-dimensionalize Eq. (79) utilizing Table 1 and find

123
Int J Geomath (2013) 4:27–53 43

Table 4 Parameter values

Parameter Value

Forward reaction rate for reaction (2): κ1 0.17 cm s−1


Backward reaction rate for reaction (2): κ−1 1 cm L2 s−1 mg −2
Forward reaction rate for reaction (1): k1 2.5 × 10−7 s−1 mg−8 L8
Backward reaction rate for reaction (1): k−1 2.5 × 10−3 s−1 mg−3 L3
Dynamic viscosity: μ 0.01 g s−1 cm−1
Density of water: ρ 1 g cm−3
Gravity: g 980 cm s−2
Net flux of oxygen to the crust: Fc 1 × 10−7 cm mg s−1 L−1
Henry’s constant for oxygen: K H 41.6 mg L−1 atm−1
Partial pressure of oxygen: PO2 0.2 atm
Angle of inclination: θ 0.03π
180
Density of iron-hydroxide crust: ρ Fe(O H )3 3.3 × 106 mg L−1
Mass transfer coefficient for oxygen: K M T 0.37 cm s−1
3
Volumetric flow rate: Q̂·1 cm (y-direction) 3 cm
sec
0 at the mine source of AMD
Concentration ĈH + 0.063 mg L−1
Concentration ĈO0 at the mine source of AMD 3.0 mg L−1
2
Concentration Ĉ 0 2+ at the mine source of AMD 600 mg L−1
Fe
Concentration Ĉ 0 3+ at the mine source of AMD 5.6 mg L−1
Fe

   
κ−1 κ1
ŝ0tˆ = − ĈH3 + + ĈFe3+ . (80)
ρ Fe(O H )3 0 ρ Fe(O H )3 0

In summary, the system solved in the following numerical simulations consists of


the crust evolution Eq. (80) and the reactive transport system (69)–(72) subject to time
dependent boundary conditions (77).
A sensitivity analysis on the parameters of the system is conducted by making small
variations in the magnitude of Q̂, K M T , θ and all reaction rates. This will determine
whether small variations in the magnitude of each of these parameters causes drastic
differences in the output of the model.
Table 4 is a list of parameters used in the model, and their values. In Table 4, the
forward reaction rate k1 is derived from work done by Kirby and Cravotta (2005). The
remaining reaction rates: k−1 , κ−1 , and κ1 and the mass transfer coefficient, K M T , are
chosen to fit data obtained from the Mushroom farm (Senko et al. 2011).

4 Solutions and results

The system (69)–(72) and (80) is solved numerically in MATLAB. The simulation
is run over a time period of one year to model seasonal fluctuations in flow. The
spatial domain spans 40 m because the Mushroom Farm data indicate that the Fe(II)

123
44 Int J Geomath (2013) 4:27–53

5
Mushroom Farm Data
Model Approximation
4

3
pH

0
0 1000 2000 3000 4000
^
Distance, x , from AMD Emergence Site (cm)
Fig. 3 The change in pH in the AMD as it flows away from the mine

concentration in the flow sheet is essentially depleted about 35m from the source. The
crust growth rate is slow, so (80) is solved by Euler’s method on a fixed time grid.
Because the species boundary conditions at the mine source are time dependent, the
reactive transport system (69)–(72) is solved on the spatial domain to obtain the species
concentrations at each time step. The system is moderately stiff, so the MATLAB
stiff solver ode15s is used for both accuracy and computational efficiency. In order
to achieve numerical convergence, the time domain is discretized with increment
dt = 52.56 min (10,000 time steps for 1 year) and the 40 m flow bed is discretized
with increment dx = 10 cm.

4.1 Analysis of chemical species

The simulation was run using the parameters listed in Table 4 to obtain the spatial
distributions of the pH and the chemical species concentrations through the AMD
flow bed, as well as the iron hydroxide crust growth. The spatial distributions vary
with time due to the time dependent boundary conditions at the mine source, but
Figs. 3, 4, 5, 6 show the pH and species concentration distributions only at the final
time step of the simulation. These distributions have the same shape at any given time;
they are simply scaled by the oscillating magnitude of the source.
Figure 3 is a plot of the predicted and measured pH profile for the AMD liquid. We
see that the predicted pH decreases from 4.2 to about 2.9 over a 35 m distance. The
predicted decrease in pH is consistent with the trend in field data (Senko et al. 2011),
although the field data show a much more rapid decrease to the limiting value than
do the predictions. From Eq. (69), the concentration of hydrogen ions is a function of
the source due to reaction (2) added to the sink due to reaction (1) multiplied by the
thickness of the liquid AMD. Because the pH of the system is decreasing, hydrogen
ion concentration is increasing. Hence, in Equation (69), we have J < 0 and Iˆ < 0.
An increase in hydrogen ion concentration indicates that reaction (2) is dominating

123
Int J Geomath (2013) 4:27–53 45

12
Model Approximation
Mushroom Farm Data
10

8
2
CO
6
^

0
0 1000 2000 3000 4000
^
Distance, x , from AMD Emergence Site (cm)
Fig. 4 The change in oxygen concentration in the AMD as it flows away from the mine

700
Mushroom Farm Data
600 Model Approximation

500

400
CFe2+
^

300

200

100

0
0 1000 2000 3000 4000
^
Distance, x , from AMD Emergence Site (cm)
Fig. 5 The predicted change in Fe(II) concentration in the AMD as it flows away from the mine

the control of pH. After 35 m, the pH becomes constant, and from Eq. (69) and the
discussions that follow, J = Iˆ = 0, because net changes in ion concentrations are
negligible past this point. This indicates that the minimum pH occurs about 35 m away
from the source.
Figure 4 is the concentration profile for dissolved oxygen in the AMD liquid. The
model approximation agrees well with data collected from the Mushroom Farm. In
Fig. 4 we see that dissolved oxygen concentration increases rapidly from 3 mg/L to
about 5.8 mg/L in the first few centimeters from the emergence site. This rapid rise
in oxygen concentration occurs because the AMD in the mine is oxygen-starved. The
mass transfer across the air–liquid interface in Eq. (70) controls the behavior of the
dissolved oxygen concentration for the first few centimeters from the emergence site.
After the initial jump in oxygen concentration, the dissolved oxygen concentration
remains a constant value of 5.8 mg/L until about 30 m from the emergence site.

123
46 Int J Geomath (2013) 4:27–53

5.4

C Fe3+ 5.2
^

4.8

0 1000 2000 3000 4000


^
Distance, x , from AMD Emergence Site (cm)
Fig. 6 The change in Fe(III) concentration in the AMD as it flows away from the mine

The constant concentration occurs because the net flux of oxygen from the air into
the liquid balances with the removal of oxygen due to bulk reactions and the loss of
oxygen to the iron(III) hydroxide crust. From 30 to 34 m from the AMD emergence
site there is a linear increase in oxygen concentration to 8.2 mg/L because Iˆ = 0
downstream. As shown in Fig. 5, all possible Fe(II) under the conditions imposed by
the model has been removed from the AMD 30 m from the emergence site. Equation
(71) indicates that having no net change in concentration of Fe(II) implies Iˆ = 0. Thus
mass transfer across the air–liquid interface controls the concentration of oxygen until
34 m from the AMD emergence site. At 35 m from the emergence site the net flux of
oxygen at the air–liquid interface equilibrates with the atmosphere.
Figure 5 is the predicted concentration profile for Fe(II) ions in the AMD liquid. We
see that concentration of Fe(II) decreases linearly from 600 mg/L to about 50 mg/L
over a 30 m distance. The linear behavior of the decrease in Fe(II) concentration is
also consistent with field data (Senko et al. 2011). From Eq. (71), the concentration
of Fe(II) is a function of the sink due to reaction (1) multiplied by the thickness of the
liquid AMD. The linear decreasing behavior in Fig. 5 indicates that the bulk reactions
in the system have a constant negative slope and so Iˆ, the sink due to bulk reactions, has
a negative value. Recall that Iˆ < 0 promotes iron(III) hydroxide crust growth. After
30 meters, the concentration levels of Fe(II) become constant, and from Eq. (71), the
bulk reactions have reduced to zero. This indicates that the maximum Fe(II) removal
occurs about 30 m away from the source.
Figure 6 is the concentration profile for Fe(III) ions in the AMD liquid. The con-
centration of Fe(III) decreases slowly from 5.6 mg/L to about 5.4 mg/L for the first
25 m from the mine. A rapid decrease in Fe(III) concentration occurs 25 to 35 m from
the emergence site, where the concentration of Fe(III) has the constant value of a little
over 4.7 mg/L. From Eq. (72), when Iˆ < 0 and J < 0, iron(III) hydroxide crust is
formed. The decrease in Fe(III) concentration implies that the magnitude of J is larger
than Iˆ multiplied by the thickness of the liquid from the emergence site until 35 m
from the emergence site, where both Iˆ and J approach zero. Because the thickness of
the AMD liquid is constant, the cause for the rapid decrease in Fe(III) concentration

123
Int J Geomath (2013) 4:27–53 47

20 2
^
Solid-liquid interface at 91.25 days Air-liquid interface, h

Interface Growth (cm)


Solid-liquid interface at 182.5 days 1.8 ^
Solid-liquid interface, s
Solid-liquid interface at 273.75 days
15 Solid-liquid interface at 365 days
1.6
10
s
^

1.4

5 1.2

1
0
0.0 7 7 7
0 1000 2000 3000 4000 1.0×10 2.0×10 3.0×10
^ ^
Distance, x , from AMD Emergence Site (cm) Time, t , (seconds)
(a) (b)
Fig. 7 a Iron-hydroxide crust growth at various times from the AMD emergence site to 40 m in distance
from the mine. b The growth in the ẑ = ŝ(x̂, tˆ) and ẑ = ĥ(x̂, tˆ) interfaces over 1 year at location 40 m from
the AMD emergence site

at 25 m from the emergence site must be due to a disparity in magnitude between


Iˆ and J .
Figure 7a shows the crust growth that occurs at 91.25, 182.5, 273.75, and 365 days
from the AMD emergence site to 40 m away from the mine. We see that there is fairly
constant growth of crust throughout the year from the emergence of the mine to about
30 m away. From Eq. (80), growth at the ẑ = ŝ(x̂, tˆ) interface is a function of hydrogen
ion and Fe(III) concentration. The concentrations of hydrogen ions and Fe(III) balance
in such a way that the rate of Fe(OH)3 growth is constant from the AMD emergence
site to about 25 m from the mine over the course of a year. Hence, the value of J
in Eq. (80) is constant from the AMD emergence site to about 25 m from the mine.
From Fig. 7a, at about 35 m from the mine, the iron(III)-hydroxide crust shows little
growth over the course of a year due to depletion of reactants downstream as shown
in Fig. 5. This is consistent with data from the Mushroom farm (Senko et al. 2011),
yet the growth rate of the model does exceed that observed from field data.
Figure 7b shows the seasonal changes for interface growth in a year, built into the
model. The flow rate, Q̂, is defined in Eq. (77) as a sine function where two periods
occur over the course of a year. This simulates the alternating dry and rainy seasons
that occur at the Mushroom farm. During wet seasons the sine wave is near its crest
and the crust is growing at a slightly greater rate. Seasons represented by the troughs
of the sine waves have a slightly slower crust growth rate.
Next we vary several parameters in the system to see how their changes affect
concentrations of chemical species and iron(III) hydroxide crust growth. The mass
transfer coefficient of oxygen is varied to simulate an oxygen starved and an oxygen
rich system. The flow rate, angle of inclination and fluid thickness are varied to simulate
slow and fast moving fluids. Finally, reaction rates are varied to analyze their impact
on the system.

4.2 Varying the oxygen mass transfer coefficient

We vary the value of the mass transfer coefficient, K M T . Low values for K M T simulate
an oxygen starved system, while larger values of K M T simulate an oxygen rich system.

123
48 Int J Geomath (2013) 4:27–53

1.2 11
Flow Rate = 3 Flow Rate = 3
Flow Rate = 5 10 Flow Rate = 5
Flow Rate = 7 Flow Rate = 7
1 9
0.8 8
7

2
CH+

CO
0.6

^
6
^

5
0.4
4
0.2 3
2
1000 2000 3000 0 1000 2000 3000 4000
^ ^
Distance, t , from AMD Emergence Site (cm) Distance, x , from AMD Emergence Site (cm)
(a) (b)
Fig. 8 a The change in hydrogen ion concentration in the AMD as it flows away from the mine, with
2
varying flow rate, Q̂ with units cm
sec . b The change in oxygen concentration in the AMD as it flows away
2
from the mine, with varying flow rate, Q̂ with units cm
sec

Greater concentrations of oxygen in the AMD lead to slight increases of the magnitude
of the negative value of Iˆ. This implies slightly lower concentrations for Fe(II), and
hydrogen ions, and a higher concentration for Fe(III). From Eq. (80), with lower
concentrations of hydrogen ions and greater concentrations of Fe(III), the rate of
iron(III) hydroxide crust growth increases slightly.

4.3 Increasing flow rate with a fixed angle of inclination (thicker fluid layer)

Varying the flow rate of the system while maintaining a fixed angle of inclination
 1/3
imposes changes in the thickness of the fluid film layer. In Eqs. (69)–(72), ρg3 Q̂μ
sin θ
2
is the thickness of the liquid layer. Increasing Q̂ (from our base value of 3 cms )
while maintaining a fixed angle of inclination simulates increasing the amount of bulk
solution. Increasing the fluid film thickness increases the impact of Iˆ on Eqs. (69)–
(72), causing a decrease in the concentrations of hydrogen ions and oxygen. On
the other hand, increasing Q̂ increases the concentration of Fe(II) and Fe(III). By
Eq. (80), lower concentrations of hydrogen ions and higher concentrations of Fe(III)
induce faster growth at the ŝ = (x̂, tˆ) interface. Hence, a large flow rate, Q̂, indi-
cates a greater rate of iron(III) hydroxide crust growth. These results are displayed in
Figs. 8, 9, 10.

4.4 Increasing angle of inclination with a fixed flow rate (faster fluid velocities)

We vary the angle of inclination (from 0.03 to 6 degrees) with fixed flow rate. By
Eqs. (69)–(72), changing values for θ with fixed Q̂ forces the fluid film thickness to
change. As the angle of inclination increases, the fluid film thickness decreases. In
turn, the fluid flow velocity increases. With increasing fluid flow velocity, there is less

123
Int J Geomath (2013) 4:27–53 49

600
Flow Rate = 3 Flow Rate = 3
Flow Rate = 5 Flow Rate = 5
500 Flow Rate = 7 5.8 Flow Rate = 7

400 5.6
CFe2+

CFe3+
5.4
300
^

^
5.2
200
5
100
4.8
0
0 1000 2000 3000 4000 0 1000 2000 3000 4000
^ ^
Distance, x , from AMD Emergence Site (cm) Distance, x , from AMD Emergence Site (cm)
(a) (b)
Fig. 9 a The change in Fe(II) concentration in the AMD as it flows away from the mine, with varying flow
2
rate, Q̂ with units cm
sec . b The change Fe(III) concentration in the AMD as it flows away from the mine,
2
with varying flow rate, Q̂ with units cm
sec

Flow Rate = 3 12 Flow Rate = 3


Flow Rate = 5 Flow Rate = 5
10 Flow Rate = 7 Flow Rate = 7
10
8
s
^
s
^

6
5
4
2
0 7 7 7 0
0.0 7 7 7
0.0 1.0×10 2.0×10 3.0×10 1.0×10 2.0×10 3.0×10
^ ^
Time, t , (seconds) Time, t , (seconds)
(a) (b)
Fig. 10 a The growth in the ẑ = ŝ(x̂, tˆ) interface over time at a location 13.5 m from the AMD emergence
2
site with varying flow rate, Q̂ in cm ˆ
sec with a fixed angle of inclination. b The growth in the ẑ = ŝ( x̂, t )
2
interface over time at a location 27 m from the AMD emergence site with varying flow rate, Q̂ in cm
sec with
a fixed angle of inclination

time for ions in the system to react and form Fe(OH)3 . Hence, with increasing θ , the
rate of formation of iron(III) hydroxide crust decreases.

4.5 Increasing angle of inclination with a fixed fluid film thickness


(increased flow rate)

Now we vary the angle of inclination with fixed values for the fluid film thickness.
By Eqs. (69)–(72), changing values for θ with fixed fluid film thickness forces Q̂ to
change. As the angle of inclination increases the flow rate must increase. Hence the
fluid velocity increases. With increasing fluid flow velocity, there is less time for ions
in the system to react and form Fe(OH)3 . Thus, with increasing θ , the rate of formation
of iron(III) hydroxide crust decreases upstream. However, more reactants are carried

123
50 Int J Geomath (2013) 4:27–53

downstream leading to an increase in crust growth in the downstream regions. This


provides a partial explanation for the differences in precipitation between the GumBoot
and Fridays-2 sites noted by Senko et al. (2008). Less precipitation at Fridays-2 may
be be due to the higher flow rates at this site.

4.6 Varying reaction (2) rates

Next, the backward reaction (2) rate, κ−1 and forward reaction (2) rate, κ1 are varied.
Increasing κ1 promotes the precipitation of Fe(OH)3 . By Reaction (2), as κ1 increases,
the concentration of Fe(OH)3 increases and the concentration of Fe(III) decreases.
Increasing κ1 leads to faster iron(III) hydroxide crust growth. Increasing κ−1 promotes
the dissolution of Fe(OH)3 and thus a slower iron(III) hydroxide crust growth.

4.7 Varying reaction (1) rates

Next, the backward reaction (1) rate, k−1 and forward reaction (1) rate, k1 are varied.
By Reaction (1), as k−1 increases, the concentration of Fe(III) decreases and the
concentrations of Fe(II), oxygen, and hydrogen ions increase. The lower concentration
of Fe(III) hinders Reaction (2), forcing a slower iron(III) hydroxide crust growth. By
Reaction (1), as k1 increases, the concentrations of Fe(II), oxygen, and hydrogen ions
decrease while the concentration of Fe(III) increases. The higher concentration of
Fe(III) promotes precipitation in Reaction (2), implying a faster iron(III) hydroxide
crust growth.
For larger values of k1 , the hydrogen ion concentration is smaller and Fe(III) con-
centration is larger. Equation (80) indicates that lower hydrogen ion concentrations
and higher concentrations of Fe(III) provide a faster iron(III) hydroxide crust growth
rate. The abundance of Fe(II) oxidizing bacteria is one of the factors that induce a
faster oxidation rate for Fe(II), or higher k1 value. This provides another possible
explanation for the differences in precipitation between the GumBoot and Fridays-2
sites, noted by Senko et al. (2008). Less precipitation at Fridays-2 may be due to a
smaller bacterial population.

5 Conclusions

We develop a mathematical model for an AMD system that includes hydrodynamic


and reactive transport. A thin-film fluid flow approximation is applied to the equa-
tions of fluid mechanics. Chemical kinetics are used to describe the reactive transport
of chemical species in the bulk solution. The hydrodynamic and reactive transport
components are coupled via the free boundaries defining the crust–liquid and liquid–
air interfaces. We make the following assumptions to simplify the analysis of the
governing equations:
– Bacteria are located only in the liquid.
– The model is two-dimensional in space.
– Flow is laminar.

123
Int J Geomath (2013) 4:27–53 51

– The model is meant to only remove iron, and no other metal ions.
– The density of the iron(III) hydroxide crust is constant throughout and the crust is
non-porous.
– The distance over which the AMD flows is far greater than the thickness of the
crust.
– The thickness of the AMD fluid film is assumed to be constant in space, but allowed
to vary in time due to seasonal changes in precipitation.
– Temperature effects on bacterial activity are not considered.
– The capillary number relating surface tension forces to viscous forces is large.
– Bulk reactions are fast so that bulk rate laws approach equilibrium conditions and
are nearly zero in magnitude.
– The rate of iron(III) hydroxide crust growth is slow and so the surface reaction
rate law at the crust interface, which defines the crust growth rate in the model, is
nearly zero in magnitude.
– Mass transport from the liquid AMD to the non-porous iron(III) hydroxide crust
is small.

The major components that control the formation of iron(III) hydroxide crust are
concentrations of Fe(II) and Fe(III), dissolved oxygen, pH, flow rate, angle of incli-
nation, and reaction rates. Model predictions of spatial and temporal concentration
profiles, as well as iron hydroxide crust growth are consistent with field data.
Model predictions are plotted in comparison with data collected from the Mushroom
Farm. These concentration profiles include Fe(II) concentration, amount of dissolved
oxygen, and pH. The pH predicted by the model followed the same general trend
as measured data. The predicted amount of dissolved oxygen closely fits the data
collected from the Mushroom Farm because the mass transfer coefficient for oxygen,
K M T , is chosen to fit the data. The linear decreasing profile for the change in Fe(II)
concentration in the x̂-direction closely fits the field data. Because the maximum
amount of Fe(II) removal occurs approximately 30 m from the AMD emergence site,
this would be the optimal location for the placement of a limestone bed to prevent
armoring and raise pH levels.
There is no available measurement of Fe(III) concentration from the Mushroom
Farm. We predict a decrease in Fe(III) consistent with growth of the iron(III) hydrox-
ide crust. Fe(III) concentration is reduced to nearly zero in the bulk because it is being
removed via precipitation. Model predictions indicate a crust growth rate of approx-
imately 10 cm/year. This is a slow rate of growth, similar to that of the Mushroom
Farm, yet larger than the growth rate from field data which appears to be approximately
1 cm/year. We are not yet able to explain this discrepancy, but it may be attributable
to the reductive dissolution of Fe(III) (hydr)oxide phases by Fe(III) reducing bacte-
rial activities (Senko et al. 2011) and subsequent leaching of soluble Fe(II) into the
underlying soil or to physical removal of Fe(III) (hydr)oxide particles by fluid flow.
Several parameters are varied to observe predictions made by the model. Forcing
low concentrations of dissolved oxygen in the model hindered the oxidation of Fe(II),
subsequently slowing the rate of iron(III) hydroxide crust growth. Increasing the angle
of inclination while holding the flow rate fixed caused a slight decrease in the oxidation
of Fe(II) due to faster flow through a thinner fluid film. It also caused a very small

123
52 Int J Geomath (2013) 4:27–53

decrease in iron(III) hydroxide crust growth. Increasing the flow rate of the system
while maintaining a fixed angle of inclination simulated an increase in the depth of
liquid layer of bulk solution. When the flow rate was increased, the model predicts a
faster iron(III) hydroxide crust growth rate. Because more bulk solution is available,
more iron is oxidized, providing higher concentrations of Fe(III), one of the reac-
tants for precipitation of Fe(OH)3 . When the angle of inclination was increased with
fixed fluid film thickness, the flow rate increased. This caused a decrease in iron(III)
hydroxide crust growth due to downstream transport of the ions.
Reaction rates for the oxidation of Fe(II) as well as those for the precipitation of
Fe(OH)3 were varied to observe the sensitivity of model predictions as a function of
uncertainty in the values of these rates. When the backward rate for reaction (2), κ−1 ,
and the reverse rate for reaction (1), k−1 , were increased, the quantity of Fe(II) oxi-
dized decreased and in turn the rate of iron(III) hydroxide crust production decreased.
When the forward rate for reaction (2), κ1 , and the forward rate for reaction (1) k1 ,
were increased, both promoted an increase in Fe(OH)3 crust growth rate. The system
response was most sensitive to changes in these two rates. An increase in iron oxi-
dizing bacteria can increase the oxidation of Fe(II), due to an increased value of k1 .
Therefore, an increase in the abundance of iron oxidizing bacteria can increase the
rate of iron(III) hydroxide crust growth.
Future work for the model may include:
– Collection of field data from other site locations to provide insight about the use-
fulness of the model for systems other than the Mushroom Farm.
– Development of higher dimensional spatial models.
– Addition of a limestone bed at the end of the sheet flow where maximum removal
of Fe(II) is predicted.
– Addition of vegetation to the system and subsequent removal of metal ions other
than just Fe(II).
– Inclusion of the thermal effects on bacterial activity.
– Generalizing the flow rate and the angle of inclination to functions of x̂.

References

Ardejani, F.D., Shokri, B.J., Bagheri, M., Soleimani, E.: Investigation of pyrite oxidation and acid mine
drainage characterization associated with Razi active coal mine and coal washing waste dumps in the
Azad shahr-Ramain region, northeast Iran. Environ. Earth Sci. 61, 1547–1560 (2010)
Baker, B.J., Banfield, J.F.: Microbial commuities in acid mine drainage. FEMS Microbiol. Ecol. 44, 139–152
(2003)
Burgos, W.D., Royer, R.A., Fang, Y.: Theoretical experimental considerations related to reaction-based
modeling: a case study using Iron(III) oxide bioreduction. Geomicrobiol. J. 19, 253–287 (2002)
Burke, S.P., Banwart, S.A.: A geological model for removal of Iron(II)(aq) from mine water discharges.
Appl. Geochem. 17, 431–443 (2002)
DeSa, T.C., Brown, J.F., Burgos, W.D.: Laboratory and field-scale evaluation of low-ph Fe(ii) oxidation at
Hughes Borehole, Portage Pennsylvania. Mine Water Environ. 29, 239–247 (2010)
Gouin, M.: Acid mine drainage remediation utilizing iron-oxidization bacteria. Master’s thesis, The
University of Akron (2011)
Herlihy, A.T., Kaufmann, P.R., Mitch, M.E.: Regional estimates of acid mine drainage impact on streams
in the mid-Atlantic and southeastern United States. Water Air Soil Pollut. 50, 91–107 (1990)
Johnson, D.B., Hallberg, K.B.: Acid mine drainage remediation options: a review. Sci. Total Environ. 338,
3–14 (2005)

123
Int J Geomath (2013) 4:27–53 53

Kirby, C.S., Cravotta III, C.A.: Net alkalinity and net acidity 1: theoretical considerations. Appl. Geochem.
20, 1920–1940 (2005)
Kirby, C.S., Thomas, H.M., Southam, G., Donald, R.: Relative contributions of abiotic and biological factors
in Fe(II) oxidation in mine drainage. Appl. Geochem. 14, 511–530 (1999)
Oron, A., Davis, S., Bankoff, S.: Long-scale evolution of thin liquid films. Rev. Mod. Phys. 69(3), 931–980
(1997)
Senko, J.M., Bertel, D., Quick, T.J., Burgos, W.D.: The influence of phototrophic biomass on Fe and S
redox cycling in an acid mine drainage-impacted system. Mine Water Environ. 30, 38–46 (2011)
Senko, J.M., Wanjugi, P., Lucas, M., Bruns, M.A., Burgos, W.D.: Characterization of Fe(II) oxidiz-
ing bacterial activities and communities at two acidic Appalachian coalmine drainage-impacted sites.
ISME J. 2, 1134–1145 (2008)

123

You might also like