You are on page 1of 6

GEO-SLOPE International Ltd, Calgary, Alberta, Canada

www.geo-slope.com

1D Exit Review Example


1

Introduction

There are three basic boundary conditions for contaminant transport: 1) specified concentration
(Dirichlet); 2) specified dispersive flux (Neumann); and, 3) specified dispersive and advective flux
(Cauchy). In many cases, these boundary conditions are sufficient and easily defined. For example, a
contaminant source could be specified using either the Dirichlet (constant concentration) or Cauchy
(constant mass flux) type boundary condition. Unfortunately, knowledge of the far-field boundary
conditions is often not available a priori for many transport problems. In groundwater analyses, we often
deal with this uncertainty by placing the exit boundary sufficiently far from the area of the interest. The
same approach can be used in transport problems, and the boundary can either be specified as a constant
concentration or constant mass flux condition.
However, there are cases in which the plume unavoidably reaches the far-field boundary. In this case, it
is often convenient to use an exit review condition. Frind (1988) formulated a free exit boundary
condition for finite element analysis that allows mass to exit via both advective and dispersive flux (i.e.
Cauchy boundary condition). The objectives of this example are to: 1) demonstrate the use of the exit
review boundary condition; and, (2) compare the results from CTRAN/W to a closed-form solution of the
advection-dispersion equation.

Feature Highlights

GeoStudio features that are highlighted in this example include:


1. The use of a free-exit review boundary condition to handle the far-field transport boundary;
2. A comparison of the CTRAN/W solution to a closed-form solution of the advectiondispersion equation that includes a free-exit boundary condition; and,
3. A comparison of the results from two different exit review conditions in which mass is
allowed to exit by advective flux only (Neumann boundary where concentration gradient
normal to the boundary is zero) and with both advective and dispersive flux (Cauchy
boundary but with variable mass flux).

Closed-Form Solutions

Frind (1988) presents the following closed-form analytical solution to the advection-dispersion equation
for a one-dimensional semi-infinite medium with a Cauchy boundary condition:

C
x vt
C = o erfc
2
2 Dt

x + vt
vx
exp D erfc

2 Dt

( x vt ) 2
v ( x + vt ) v t
exp
+

1 +
D
4 Dt
D

[1]

where
C = concentration,
Co = specified concentration in the source boundary,
D = hydrodynamic dispersion coefficient,
v = average linear velocity,

CTRAN Example File: 1D Exit Review Example (pdf)(gsz)

Page 1 of 6

GEO-SLOPE International Ltd, Calgary, Alberta, Canada

www.geo-slope.com

t = elapsed time,
x = distance from the source boundary, and
erfc = complementary error function.

Boundary Conditions and Material Properties

The GeoStudio example file includes a steady-state seepage analysis (SEEP/W) and two contaminant
transport analyses (CTRAN/W). The SEEP/W analysis forms the Parent for each transport model
because it is the source of seepage velocities. The model domain is a one-dimensional column that is 40
m in length and 1 m in height.
The boundary conditions and material properties for the SEEP/W analysis were selected to produce a
constant seepage flux of 0.05 m/sec in the positive x-direction. The soil is assigned a hydraulic
conductivity of 0.5 m/sec and a saturated volumetric water content of 0.5. The left and right boundaries
are set to constant hydraulic heads of 5 m and 1 m, respectively, yielding a constant hydraulic gradient of
0.1 (i.e. 4 m/40 m). Accordingly, the average linear velocity (v) in the flow domain is 0.1 m/sec (i.e. v =
q/).
For each transport model, the left boundary is set to a constant concentration of 1.0 g/m3. The coefficient
of diffusion (D) is set to 0.0 m2/s and the longitudinal dispersivity (L) is 4 m, yielding a longitudinal
hydrodynamic dispersion (v + D) of 0.4 m2/sec. This example does not include adsorption or decay, so
an adsorption function has not been defined and the decay half-life and dry density are set to zero. The
activation concentration for the soil is set to 0 g/m3 to define the initial concentration in the model
domain. Each model was run for an elapsed time of 480 seconds using 48 time steps.
The two transport analyses in the file, labeled Exit 1 Qd > 0 and Exit 2 Qd = 0, make use of the exit
review conditions available in CTRAN/W. In both cases, the right boundary is set to a mass flux equal to
zero (Qm = 0) with the exit review condition selected. The first case allows mass to exit by both
dispersive and advective flux (Qd > 0), and is therefore consistent with the closed-form solution presented
above (i.e. Cauchy boundary). In contrast, the second case allows mass to exit only by advective flux
(Neumann boundary with concentration gradient set equal to zero); so there is zero dispersive flux (Qd =
0) and the concentration gradient at the exit boundary must be zero.

Results and Discussion

Figure 1 presents the results from CTRAN/W and the analytical solution for time steps of 80, 160, 240,
320, 400, and 480 seconds. In general, the results are in excellent agreement and suggest that the finitelength model behaves as an infinite-length model. The numerical solution drifts slightly from the
analytical solution near the exit boundary at times of 240, 320, and 400 seconds. This is due to numerical
dispersion that propagates into the model domain from the free exit boundary.
It should be noted that the CTRAN/W solution in Figure 1 was achieved with a Peclet # of 0.5 in the last
element. If the mesh size is changed such that the last element is 1 m in length (i.e. Peclet # = 0.25), the
numerical dispersion worsens and the results drift further from the analytical solution as shown in Figure
2. Conversely, increasing the last element size to a length of 3 m (not presented) causes the solution to
become unstable and incorrect.

CTRAN Example File: 1D Exit Review Example (pdf)(gsz)

Page 2 of 6

GEO-SLOPE International Ltd, Calgary, Alberta, Canada

www.geo-slope.com

1.0
0.9
0.8
0.7

C/Co

0.6
0.5
0.4
0.3
Closedform
Solution
CTRAN/W

0.2
0.1
0.0
0

10

15

20

25

30

35

40

Distance(m)

Figure 1 Comparison of CTRAN/W and the analytical solution with a free exit boundary with the
last element length equal to 2 m and a Peclet Number equal to 0.5
1.0
0.9
0.8
0.7

C/Co

0.6
0.5
0.4
0.3
Closedform
Solution
CTRAN/W

0.2
0.1
0.0
0

10

15

20

25

30

35

40

Distance(m)

Figure 2 Comparison of CTRAN/W and the analytical solution with a free exit boundary with the
last element length equal to 1 m and a Peclet Number equal to 0.25

Figure 3 compares the CTRAN/W results when both dispersive and advective fluxes are allowed at the
exit (Qd> 0) with the case when the dispersive flux at the exit is ignored (Qd = 0). Concentrations remain
lower at the exit boundary for a longer period of time when both types of mass fluxes are allowed.

CTRAN Example File: 1D Exit Review Example (pdf)(gsz)

Page 3 of 6

GEO-SLOPE International Ltd, Calgary, Alberta, Canada

www.geo-slope.com

Worth noting is that the two different boundary types have very little influence in the first 30 m of the
column. The differences occur primarily in the last 10 m and mostly at later times.
Also worth noting is that the concentration contours tend to be perpendicular to the exit boundary when
the dispersive flux is ignored. This is reflective of the zero concentration gradient at the boundary.
1.0
0.9
0.8
0.7

C/Co

0.6
0.5
0.4
0.3
Qd>0

0.2

Qd=0

0.1
0.0
0

10

15

20

25

30

35

40

Distance(m)

Figure 3 Concentration distributions with both dispersive and advective mass flux at the exit
boundary (Qd > 0; dashed line) and advective mass only (Qd = 0; solid lines)

Exit mesh elements

As discussed above, the element size at the exit boundary has an effect on the results. In the CTRAN/W
formulation, the type and orientation of the elements also has an effect. The best results are obtained
when rectangular elements are used adjacent to the exit boundary as illustrated in Figure 4. The size of
the elements needs to be governed by the flow velocity or Peclet Number
Exit elements

Flow

Figure 4 Recommended shape and orientation of elements along the exit boundary

CTRAN Example File: 1D Exit Review Example (pdf)(gsz)

Page 4 of 6

GEO-SLOPE International Ltd, Calgary, Alberta, Canada

www.geo-slope.com

The use of triangular elements along the exit boundary as illustrated in Figure 5 should be avoided if at all
possible. The triangular elements with their apex on the boundary do not perform well for mass flux
calculations. One edge of the element must fall on the exit boundary for the element to have the proper
effect.
Exit elements

Flow

Figure 5 Triangular elements along the exit boundary should be avoid if at all possible

The use of a single transitional triangular element as shown at the slope toe in Figure 6 is not ideal, but in
the context of the whole system, one undesirable element may not have a significant effect on the results.

Figure 6 Illustration of a single triangular element at the slope toe

Such elements can be avoided by curving the corner with some additional Points as demonstrated in
Figure 7. If this does not seem to work, it may be necessary to use regions along the boundary with a
fixed quadrilateral mesh pattern. The extra effort involved, however, may not be warranted in the context
of how accurate all the material properties can be defined. This is something that needs to be decided in
the context of the project specific objectives.

Figure 7 Mesh with no triangular element at the slope toe

CTRAN Example File: 1D Exit Review Example (pdf)(gsz)

Page 5 of 6

GEO-SLOPE International Ltd, Calgary, Alberta, Canada

www.geo-slope.com

At the outset of an analysis it is good practice to start with an exit boundary with no dispersive flux and
use this boundary condition until reasonable and satisfactory results have been obtained. Once this has
been achieved the boundary condition can be modified to include dispersive mass flux to possibly refine
the analysis. As always, it is best to start simple and then move onto the more complex in steps.
A mass flux type boundary condition cannot be applied to a geometric Point. The boundary condition must be
applied to a Line so that indirectly the boundary condition is applied to the edge of an element. From the edge(s)
the proper nodal contributing areas are calculated.

Concentration at the entrance

It is important to note that the C/Co concentration at the entrance boundary starts below 1.0 and then
increases with time until it reaches the specified concentration of the source. This is consistent with the
physical condition at the entrance. The boundary has a specified constant concentration of 1 g/m3 with a
seepage flux (q) of 0.05 m/sec. Mass is therefore entering the domain via advection alone (q*C).
However, immediately inside the boundary, mass flux is occurring by both advection (q*C) and
dispersion (n*D*dC/dx) because a concentration gradient exists. As a result, the concentration at the
boundary must be less than 1.0 until the concentration gradient has dissipated (i.e. dC/dx = 0) and the
advective flux immediately outside the model domain is equal to the mass flux inside the domain. Note
again how the concentration contours become perpendicular at the entrance as the concentration gradient
dissipates.

Concluding Remarks

In this example, results from CTRAN/W are compared to the closed-form solution of the advectiondispersion equation with a free exit review boundary condition. A Cauchy-type boundary condition is
used in the analytical solution, allowing mass to exit the domain by both advective and dispersive flux.
The CTRAN/W results are in keeping with the analytical solution; however, the analysis demonstrates
that the results are sensitive to the dispersivity near the exit boundary. Generally, as the element length is
decreased near the exit (Peclet # decreases), numerical dispersion increases and the results drift from the
analytical solution. Exit elements that are too large (long), however, can lead to non-convergence.
The results confirm that the CTRAN/W implementation is consistent with the theoretical formulation.

Reference

Frind, E.O., 1988. Solution of the Advection-Dispersion Equation with Free Exit Boundary. Numerical
Methods for Partial Differential Equations, Vol. 4, pp. 301-313.

CTRAN Example File: 1D Exit Review Example (pdf)(gsz)

Page 6 of 6

You might also like