You are on page 1of 182

Transport Modeling with

CTRAN/W
An Engineering Methodology

John Krahn

First Edition, May 2004

Copyright 2004 by GEO-SLOPE International, Ltd.


All rights reserved. No part of this work may be reproduced or transmitted in
any form or by any means, electronic or mechanical, including photocopying,
recording, or by any information storage or retrieval system, without the prior
written permission of GEO-SLOPE International, Ltd.
Printed in Canada.
Acknowledgements
To say that this book is by John Krahn grossly overstates the case. This book
is the result of a group effort by everybody at GEO-SLOPE. My name is listed
as author primarily for referencing convenience.
At the top of the list of contributors are Greg Newman and Leonard Lam.
All of us who participated in creating the content are grateful to Carola Preusser
and Patricia Stooke for their valuable assistance with editing and formatting this
book.

GEO-SLOPE International Ltd


1400, 633 6th Ave SW
Calgary, Alberta, Canada T2P 2Y5
E-mail: info@geo-slope.com
Web: http://www.geo-slope.com

CTRAN/W

Table of Contents

Table of Contents
1

Introduction ............................................................. 1
1.1

Typical applications ......................................................................1

1.2

Contaminant transport processes ................................................1


Transport processes ..............................................................2
Attenuation processes ...........................................................5

1.3

Advective contaminant transport ..................................................7

1.4

Advective-dispersive contaminant transport.................................8

1.5

Density-dependent contaminant transport ...................................9

1.6

About this book...........................................................................10

Material Properties................................................ 13
2.1

Dispersivity and diffusion............................................................13


Diffusion function .................................................................14

2.2

Adsorption function.....................................................................15
Dry density ...........................................................................16

2.3

Decay half life .............................................................................16

2.4

Function data application in solver .............................................17


Weighted splines..................................................................17
Best-fit splines......................................................................18

Boundary Conditions............................................. 21
3.1

Introduction.................................................................................21

3.2

Fundamentals.............................................................................22

3.3

Sources and sinks ......................................................................24

3.4

Advection and dispersion considerations ...................................24

3.5

Source concentration..................................................................25

3.6

Surface mass accumulation .......................................................26

Page i

Table of Contents

CTRAN/W

3.7

Exit review ..................................................................................27

3.8

Concentration vs. mass function ................................................30

3.9

Boundary functions.....................................................................32
General ................................................................................32

3.10

Time activated boundary conditions...........................................32

3.11

Null elements..............................................................................33

Analysis Types...................................................... 35
4.1

Seepage solution interpolation ...................................................35

4.2

Initial conditions ..........................................................................36


Using an initial conditions file...............................................36
Drawing the initial conditions ...............................................36

4.3

Particle tracking analysis ............................................................37

4.4

Advection-dispersion analysis ....................................................39

4.5

Density-dependent analysis (with SEEP/W only).......................40

4.6

Problem geometry orientation ....................................................41


Axisymmetric view ...............................................................41
Plan view..............................................................................42
2D view ................................................................................42

Numerical Issues .................................................. 43


5.1

Convergence ..............................................................................44
Vector norms........................................................................44

5.2

Numerical dispersion and oscillation ..........................................46

5.3

Peclet and Courant number criteria............................................46

5.4

Backward or central difference time integration .........................48

5.5

Mesh design ...............................................................................49

5.6

Time step design ........................................................................50

5.7

Use of the Peclet and Courant numbers criteria ........................50

Page ii

CTRAN/W

Table of Contents

5.8

Gauss integration order ..............................................................53

5.9

Equation solvers (direct or iterative)...........................................54

Visualization of Results......................................... 57
6.1

Node and element information ...................................................57

6.2

Particle information.....................................................................60

6.3

Mass accumulation.....................................................................62

6.4

Equipotential lines ......................................................................62


Projecting Gauss point values to nodes ..............................63

6.5

Contours .....................................................................................65

6.6

Water flow vectors and flow paths..............................................66

6.7

Flux sections...............................................................................66
Flux section theory...............................................................66
Flux section application........................................................67

6.8

Changes between selected times ..............................................68

6.9

Graphing.....................................................................................69

6.10

Reporting ....................................................................................72

Modeling Tips and Tricks ...................................... 73


7.1

Introduction.................................................................................73

7.2

Modeling progression .................................................................74

7.3

Problem engineering units..........................................................74

7.4

The use of infinite elements .......................................................76

7.5

Fracture flow simulation..............................................................77

7.6

Flux section location ...................................................................77

7.7

Unit flux versus total flux? ..........................................................78

7.8

Stopping and restarting an analysis ...........................................79

7.9

Element addition and removal....................................................80

Page iii

Table of Contents

CTRAN/W

Product Integration Illustrations ............................ 81


8.1

SEEP/W generated pore-water pressures in SLOPE/W stability


analysis.......................................................................................82

8.2

VADOSE/W generated pore pressures in SLOPE/W stability


analysis.......................................................................................84

8.3

SEEP/W dissipation of pore pressures generated in a QUAKE/W


earth quake analysis...................................................................89

8.4

SEEP/W velocity data in CTRAN/W contaminant transport


analysis.......................................................................................92

8.5

VADOSE/W velocity data in CTRAN/W contaminant transport


analysis.......................................................................................97

8.6

Density-dependent flow salt water intrusion............................98

8.7

Ground freezing and water flows (SEEP/W and TEMP/W)......101

8.8

Seepage-dependent embankment settlement (SEEP/W and


SIGMA/W) ................................................................................105

8.9

Uncoupled consolidation ..........................................................110

Illustrative Examples ........................................... 113


9.1

One-dimensional transport analysis .........................................113


Problem definition ..............................................................114

9.2

Transport analysis with a free exit boundary............................128


Problem definition ..............................................................129
Solution ..............................................................................129

9.3

Density-dependent transport analysis ......................................133


Static saltwater column ......................................................134
Problem definition ..............................................................134
Solution ..............................................................................134
One-dimensional horizontal saltwater flow ........................136
Problem definition ..............................................................136
Solution ..............................................................................136
Henry's problem for sea water intrusion ............................138
Page iv

CTRAN/W

Table of Contents

Problem definition ..............................................................138


Solution ..............................................................................139

10

Theory................................................................. 143
10.1

Flow velocity .............................................................................143

10.2

Governing equations ................................................................144

10.3

Finite element equations ..........................................................150

10.4

Temporal integration.................................................................154

10.5

Hydrodynamic dispersion matrix ..............................................155

10.6

Mass flux ..................................................................................157

10.7

Dispersive mass flux.................................................................157

10.8

Advective mass flux..................................................................159

10.9

Stored mass flux.......................................................................159

10.10

Decayed mass flux ...................................................................160

10.11

Mass quantity calculation .........................................................161

10.12

Exit boundaries.........................................................................162

10.13

Particle tracking ........................................................................164

10.14

Density-dependent flow ............................................................165

10.15

Seepage solutions from SEEP/W or VADOSE/W....................166

References................................................................... 169
Index ............................................................................ 171

Page v

Table of Contents

CTRAN/W

Page vi

CTRAN/W

Chapter 1: Introduction

Introduction

CTRAN/W is a finite element software product that can be used to model the
movement of contaminants through porous materials such as soil and rock. The
comprehensive formulation of CTRAN/W makes it possible to analyze problems
varying from simple particle tracking in response to the movement of water, to
complex processes involving diffusion, dispersion, adsorption, radioactive decay
and density dependencies.
CTRAN/W is integrated with SEEP/W and VADOSE/W, other GEO-SLOPE
software products that compute the water flow velocity for a problem.
CTRAN/W utilizes the seepage flow velocities to compute the movement of
dissolved constituents in the pore-water. CTRAN/W can only be used in
conjunction with either SEEP/W or VADOSE/W data. For density-dependent
analysis, CTRAN/W can only be coupled with SEEP/W.

1.1

Typical applications

CTRAN/W can be used to model many groundwater contaminant transport


problems. This section presents examples of the types of problems that can be
analyzed using CTRAN/W. It should be noted that CTRAN/W is designed to use
the seepage flow velocities computed due to flow in both the saturated and
unsaturated zones. Therefore CTRAN/W is formulated to model
saturated/unsaturated contaminant transport.
The next section provides a review of contaminant transport processes to
facilitate the later discussion of the applications for which CTRAN/W can be
used.

1.2

Contaminant transport processes

The factors which govern the migration of a contaminant can be considered in


terms of transport processes and attenuation processes. The transport processes
can be mathematically represented by equations based on flow laws. These
equations can be combined into a mass balance equation with those processes
causing the attenuation of the contaminant; this yields the general governing
differential equation for contaminant migration.

Page 1

Chapter 1: Introduction

CTRAN/W

Transport processes
The two basic transport processes are advection and dispersion. Advection is the
movement of the contaminant with the flowing water. Dispersion is the apparent
mixing and spreading of the contaminant within the flow system. The advection
and dispersion transport processes can be illustrated by considering a steady flow
of water in a long pipe filled with sand.
Consider the injection of a slug of contaminant mass into the pipe (Figure 1-1).
The mass flows along the pipe with a constant velocity v. This transport process
is called advection. As the mass moves along with the moving water, it also
spreads out (i.e., disperses). The contaminant mass occupies an increasingly
longer length of the pipe, thereby decreasing in concentration with time. The
spreading out of the contaminant is called dispersion.
Figure 1-2 illustrates the transport process when a continuous source of
contaminant mass is injected into the pipe. At some point in the pipe beyond the
injection location, the contaminant initially appears at a low concentration and
then gradually increases until the full concentration is reached. If only the
advection process is considered, the contaminant would arrive at some point in
the pipe as a plug with full concentration. Because of dispersion, however, the
full concentration arrives at a time later than the first appearance of the dispersed
contaminant, as shown in the figure.
Theoretically, the plug flow arrival time corresponds to the time when the fiftypercent concentration arrives. The time difference between the first arrival of the
dispersed contaminant and the arrival of the plug flow increases as the distance
from the injection point increases.
While the advection process is simply migration in response to the flowing water,
the dispersion process consists of two components. One is an apparent "mixing"
and the other is molecular diffusion.
The mixing component, often called mechanical dispersion, arises from velocity
variations in the porous media. Velocity variations may occur at the microscopic
level due to the friction between the soil particles and the fluid and also due to
the curvatures in the flow path, as illustrated in Figure 1-3. These velocity
variations result in concentration variations. When the concentration variations
are averaged over a given volume, the contaminant appears to have dispersed.

Page 2

CTRAN/W

Chapter 1: Introduction

Figure 1-1 The Migration and spreading of a contaminant slug in a


fluid flowing with velocity V
Page 3

Chapter 1: Introduction

CTRAN/W

Figure 1-2 Contaminant migration and spreading from a continuous


source in a fluid flowing with velocity V

Page 4

CTRAN/W

Chapter 1: Introduction

Figure 1-3 Factors causing mechanical dispersion


Molecular diffusion results in the spreading of contaminant due to concentration
gradients. This process occurs even when the seepage velocity is zero. Molecular
diffusion is dependent on the degree of saturation or volumetric water content of
the porous medium, an example of which is shown in Figure 1-4.
In equation form, the dispersion process is characterized as:

D = v + D*
where:
D

coefficient of hydrodynamic dispersion,

average linear velocity of the flow system,

dispersivity of the porous medium, and

D*

coefficient of molecular diffusion.

Attenuation processes
Contaminant migration in a porous medium is attenuated by chemical reactions
taking place during transport. These reactions can occur between the contaminant
mass and the soil particles or between the contaminant mass and the pore fluid.
Among these reactions, the process of adsorption is believed to be the most
important factor in attenuating the migration of contaminant.
Adsorption causes contaminant mass to be withdrawn from the moving water,
reducing the dissolved concentration and overall rate of contaminant movement.
Page 5

Chapter 1: Introduction

CTRAN/W

The amount of adsorption that occurs is a function of the contaminant


concentration within the porous medium. This relationship is described by an
adsorption function which relates the adsorption to the concentration. An
example of this relationship is shown in Figure 1-5.

Diffusion (log10)

0.001

0.0001

1e-005
0.0

0.2

0.4

0.6

0.8

1.0

Vol. W. C.

Figure 1-4 Example of molecular diffusion as a function of water


content

Page 6

CTRAN/W

Chapter 1: Introduction

1.0

Adsorption

0.8

0.6

0.4

0.2

0.0
0.0

0.2

0.4

0.6

0.8

1.0

Concentration

Figure 1-5 Example of adsorption as a function of concentration


In general, the adsorption characteristic of a contaminant in a soil is represented
by a function of S vs. C, where S is the mass of contaminant adsorbed per unit
mass of soil particles, and C is the concentration of the contaminant in the porous
medium. In the case of a linear function, the slope is called the distribution
coefficient, K d The slope represents the partitioning of the contaminant mass
between the solid (soil particles) and fluid phases of the porous medium. The
chemical reactions that cause the partitioning are assumed to be instantaneous
and reversible.
Another important attenuation process in the case of a radioactive contaminant
mass is radioactive decay. Radioactive decay causes a loss of contaminant mass
from the flow system. However, unlike adsorption, the decayed mass is
proportional to the travel time and is irreversible.

1.3

Advective contaminant transport

As described above, contaminant transport in soils involves the processes of both


advection and dispersion. Early in a contaminant transport analysis, it is often
useful to isolate the magnitude of purely advective transport without the extra
data input and computational requirements of including dispersion. It is
impossible to numerically solve the advection-dispersion equation when the
Page 7

Chapter 1: Introduction

CTRAN/W

dispersive component is small relative to the advective component, because the


numerical solution is unstable in these cases. To overcome this difficulty,
CTRAN/W has an option to simulate the purely advective contaminant transport
process using particle tracking.
In particle tracking, the dissolved solutes are represented by particles. Figure 1-6
presents an example of a particle tracking analysis. For each time step, the
particles are moved in space proportionally to the water flow velocity and the
time step size. The particle flow paths provide a graphical representation of the
contaminant plume movement caused by purely advective transport. The effects
of dispersion, adsorption, decay and density are not considered in a particle
tracking analysis.
12

Pond

11
10
9
8
7
6
5
4
3
2
1
0
0

10

11

12

13

14

15

16

17

18

19

20

21

22

23

metres

Figure 1-6 Example of a particle tracking analysis

1.4

Advective-dispersive contaminant transport

Quantification of the magnitude of advective flow is useful as a preliminary


analysis of contaminant transport. A more realistic analysis also includes the
effect of hydrodynamic dispersion. Hydrodynamic dispersion causes dilution of
contaminants both longitudinally, (in the direction of groundwater flow), and
transversely, (perpendicular to the direction of flow). Contaminant dilution
Page 8

CTRAN/W

Chapter 1: Introduction

caused by dispersion is a very significant component of contaminant transport


and therefore cannot usually be ignored. CTRAN/W provides the capability for
modeling contaminant transport with hydrodynamic dispersion.
The transport of certain contaminants, such as dissolved hydrocarbons, is
attenuated by reversible reactions with soil particles, such as adsorption. Other
contaminants, such as radioactive contaminants, undergo non-reversible decay
reactions that remove them from the groundwater during transport. CTRAN/W is
formulated to include the effects of absorption and decay type reactions during
contaminant transport.
Figure 1-7 shows the results of an advection-dispersion analysis of contaminant
migration from a surface pond.
Contaminant plume

Figure 1-7 Contaminant transport from a surface pond

1.5

Density-dependent contaminant transport

For problems where the dissolved solute density is significant, CTRAN/W has
the capability of performing density-dependent flow analyses. Density-dependent
problems include sea water intrusion into coastal aquifers, brine transport and
landfill leachate migration, to name just a few.
Figure 1-8 illustrates the CTRAN/W solution to the classic Henrys problem for
sea water intrusion. At the left boundary, freshwater enters at a constant rate
Page 9

Chapter 1: Introduction

CTRAN/W

while the right boundary of the aquifer is exposed to sea water constant head
conditions. The top and bottom boundaries have no flow. The contours show the
relative concentration of sea water, and the vectors show the relative magnitude
and direction of the water flow.
1.0
0.9
0.8
0.7

0.6

0 .1

0.5

0.

0.4

3
0 .5
0 .7

0.3

0 .9

0.2
0.1

0.0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Freshwater Inflow

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

Distance (m)
Sea Water Intrusion
Contours indicate relative concentration of sea water.

Figure 1-8 Sea water intrusion into a coastal aquifer

1.6

About this book

Modeling the movement of contaminants through soil with a numerical solution


can be very complex. Natural soil deposits are generally highly heterogeneous
and non-isotropic. In addition, boundary conditions often change with time and
cannot always be defined with certainty at the beginning of an analysis. In fact,
the correct boundary condition can sometimes be part of the solution as is the
case for an exit review boundary, where the direction of groundwater flow may
change between source and sink.
The movement of contaminants can not be modeled without a valid model for
groundwater flow in the system. That is why the MOST important aspect of this
type of model is to first be confident in the seepage solution. This book is NOT
about seepage modeling and it is assumed from this point onward, that the reader
Page 10

CTRAN/W

Chapter 1: Introduction

is familiar with and has read either the SEEP/W or VADOSE/W Engineering
Methodology books. This book is not a stand-alone reference.
While part of this document is about using CTRAN/W to do transport analyses, it
is also about general numerical modeling techniques. Numerical modeling, like
most things in life, is a skill that needs to be acquired. It is nearly impossible to
pick up a tool like CTRAN/W and immediately become an effective modeler.
Effective numerical modeling requires some careful thought and planning and it
requires a good understanding of the underlying physical fundamentals. Aspects
such as discretization of a finite element mesh and applying boundary conditions
to the problem are not entirely intuitive at first. Time and practice is required to
become comfortable with these aspects of numerical modeling.
Chapter 2 of the SEEP/W and VADOSE/W books is devoted exclusively to
discussions on the topic of How to Model. The general principles discussed in
that book apply to all numerical modeling situations, even though the discussion
there focuses on seepage analysis.
Broadly speaking, there are three main parts to a finite element analysis. The first
is discretization dividing the domain into small areas called elements. The
second part is specifying and assigning material properties. The third is
specifying and applying boundary conditions. Details of discretization are
provided in the SEEP/W or VADOSE/W book, while material properties and
boundary conditions as pertaining to transport analysis are discussed in detail in
their respective chapters here.
Transport modeling is numerically challenging because of the presence of a first
order transport term in the main differential equation. For this reason, it is
important to have an understanding of how that term affects the solution of the
equation and, in particular, how mesh size and time steps are critical to that
solution. The importance of the Peclet and Courant numbers will be introduced
and discussed, along with other numerical considerations in a chapter titled
Numerical Issues.
Two chapters have been dedicated to presenting and discussing illustrative
examples. One chapter deals with examples where geotechnical solutions are
obtained by integrating more than one type of analysis, and the other chapter
presents and describes how a series of different geotechnical problems can be
solved.

Page 11

Chapter 1: Introduction

CTRAN/W

A full chapter is dedicated to theoretical issues associated with transport and the
solution the finite element equations. Additional finite element numerical details
regarding interpolating functions and infinite elements are given in Appendix A
of the SEEP/W and VADOSE/W books.
The chapter entitled Modeling Tips and Tricks should be consulted to see if
there are simple techniques that can be used to improve your general modeling
method or to help gain confidence and develop a deeper understanding of finite
element methods, CTRAN/W conventions or data results.
In general, this book is not a HOW TO USE CTRAN/W manual. This is a book
about how to model. It is a book about how to solve transport problems using a
powerful calculator; CTRAN/W. Details of how to use various program
commands and features are given in the on line help inside the software.

Page 12

CTRAN/W

Chapter 2: Material Properties

Material Properties

This chapter describes the various soil transport properties that are required in the
solution of the CTRAN/W partial differential equation. It is important to have a
clear understanding of what the soil properties mean and what influence they
have on the type of results generated. This chapter is not meant to be an all
inclusive discussion of these issues. It is meant to highlight the importance of
various parameters and the implications associated with not defining them
adequately.
Well defined soil properties can be critical to obtaining an efficient solution of
the finite element equations. When is it acceptable to guess at a function and
when must you very carefully define one? This chapter will address these issues.

2.1

Dispersivity and diffusion

For one-dimensional flow, the hydrodynamic dispersion coefficient D is defined


above as:

D = v + D *
where:

dispersivity (material property),

DArcy velocity divided by volumetric water content (U/), and

D*

coefficient of molecular diffusion.

Dispersivity is the ratio of the hydrodynamic dispersion coefficient (d) divided


by the pore water velocity (v); thus a = d/v and has units of length. Typically, the
dispersivity varies from 0.1 to 100 m however field and laboratory tests have
indicated that dispersivity varies with the scale of the test. Large scale tests have
higher dispersivity than small lab column tests. An approximate value for
dispersivity is 0.1 times the scale of the system (Fetter, 1993). If you are
simulating contaminant transport in a 1 m long laboratory column, then
dispersivity ~ 0.1 m. However, if you are simulating transport in a large aquifer
greater than 1 km in extent, then use dispersivity ~ 100 m.
Dispersion in the direction of the water flow is usually higher than dispersion
perpendicular to the flow direction. Two dispersivity values are therefore
Page 13

Chapter 2: Material Properties

CTRAN/W

required to define the spreading process. Dispersivities in the flow directions are
designated as the longitudinal dispersivity L and the transverse dispersivity T .
In CTRAN/W if no coefficient of molecular diffusion (D*) is defined, then the
dispersion is equal to the diffusivity in the longitudinal and transverse directions
respectively.
Diffusion function
In general, the coefficient of diffusion D* is a function of the volumetric water
content, as shown in Figure 2-1. An empirical relationship between D* and
has been proposed by Kemper and Van Schaik (1966), however CTRAN/W
allows you to define any desired values of the coefficient of diffusion function as
a function of volumetric water content (i.e., the diffusion function).
The D* parameter and its dependence on water content is of significance only in
unsaturated flow and when the water flow rate is very low. The value of the
hydrodynamic dispersion coefficient is often governed by the water flow rate.
Therefore, it is often adequate to assume that D* is independent of , and to
define the relationship by a constant horizontal function.

Figure 2-1 Illustration of a diffusion versus water content function

Page 14

CTRAN/W

2.2

Chapter 2: Material Properties

Adsorption function

For the transport of a reactive substance, the movement of the mass is also
affected by the adsorption of the solute by the soil particles. As discussed above,
the amount of mass adsorbed can be defined in terms of the mass density of the
soil particles. From relationships developed in the Theory chapter, the adsorbed
mass Ms is:

M s = S d
The rate of change of the adsorbed mass is:

Ms
S
= d
t
t
The adsorption S is a function of concentration C with experimental results
usually plotted as S vs. C, as shown in Figure 2-2. The slope of the S vs. C
relationship is S / C and in the case of a linear relationship is usually referred
to as the distribution coefficient Kd.

Figure 2-2 Relationship between adsorption and concentration


Page 15

Chapter 2: Material Properties

CTRAN/W

For many dissolved contaminant and soil combinations, adsorption of


contaminant on the soil particles is linearly related to concentration (e.g. the Kd
term). CTRAN/W, however, allows a more general relation to be used to specify
the chemical partitioning by allowing the adsorption to be specified as a function
of concentration. The actual slope used in the solution of the equations will be
obtained from the function for any given concentration.
Dry density
The dry density term is the dry mass density of the porous medium. It is
multiplied by the adsorption quantity in the governing equation. The units of dry
density are (M/L3) must be consistent with the units of mass and length.

2.3

Decay half life

For the transport of a radioactive substance, mass may be lost during the
transport process due to radioactive decay of ions in the pore fluid and decay of
ions attached to the soil particles. The reduced concentration resulting from
radioactive decay, in terms of the initial concentration, is:
Equation 2-1 C = C0 e

where t is the elapsed time and is the decay coefficient. The decay coefficient
can be related to the half-life T of a decaying material. By definition, the halflife T is the elapsed time when the concentration of C/C0 = 1/2. Therefore:

C
1
=
= e - T
C0
2
which can also be written as:

ln 2
0.693
=
T
T

Differentiating Equation 2-1 with respect to time leads to:

C
= - C
t

Page 16

CTRAN/W

Chapter 2: Material Properties

The amount of radioactive mass in the pore-water Mw in an elemental unit


volume is C , (see above), or:

M w = C = C0 e t
The decay half-life must be specified in units of time that are consistent with the
units of diffusion. For example, if the diffusion coefficient is in meters per
second (m/sec), then the half-life must be specified in seconds.

2.4

Function data application in solver

CTRAN/W uses spline interpolation techniques to create smooth, continuous


conductivity, and unfrozen water content functions. Once the spline fit is created
and adjusted to your satisfaction, the solver is capable of looking up any y
value on the splined function for any input x value. A typical example would
be for the solver to obtain a water content value from the spline data for any
value of soil water pressure. In all GEO-SLOPE products, the spline fit you
observe when you set up the function is identical to the spline data used in the
solver. What you see is what you get, so it is important to view the spline
function to ensure that it is smooth, defined over a full range of x values and
not discontinuous.
Weighted splines
Smooth curves were produced in the past using mechanical means. This involved
the use of a thin, flexible strip of wood or metal held in place with weights. The
flexible strip would bend in such a way that the internal energy due to bending
was at a minimum (see Lancaster and Salkauskas, 1986).
Such a curve can be described mathematically by defining the x-y coordinates of
the points (weights) and then computing the curvature at the points that
minimizes the internal energy term in the equations. Mathematically, this is
referred to as a natural spline (see Lancaster and Salkauskas, 1986).
A natural spline can have many undesirable humps and hollows when the data
points are not near the natural maximum curvature positions. Figure 2-3
illustrates this behavior.
Salkauskas, 1974, and Lancaster and Salkauskas, 1986, have developed a
procedure for controlling the undesirable humps and hollows. They called the
Page 17

Chapter 2: Material Properties

CTRAN/W

resulting curve a weighted spline. Figure 2-4 illustrates the effect on the shape of
the curve using the weighted spline interpolation technique.
CTRAN/W utilizes the weighted spline technique to create smooth conductivity,
unfrozen water content, boundary, and modifier functions. In addition to defining
a continuous and smooth function, spline interpolation also provides first
derivatives of the curve at any point. The first derivative of the unfrozen water
content function is the rate of release of phase change latent heat, and is needed
in the solution of the transient heat transfer equation.
Best-fit splines
Since all functions are approximations of real-world behavior, it is often
convenient to use measured data values for the definition of a function. These
values, however, usually do not lie along a smooth, continuous curve. A spline
function that is fit to these values will appear jagged and will not accurately
reflect the measured data. To overcome this problem, CTRAN/W allows you to
define a "best-fit" spline through the data points, as illustrated in the figures
below.
The input function commands provide a means of controlling how the spline
curve is fit to the data points. For each function, you can assign a "Fit Curve to
Data percentage" and a "Curve Segments percentage" between 0% and 100%.
When the curve is fit exactly (100%) to the data points, the spline passes through
each data point. As the curve fitting is reduced, the spline shape approaches a
straight line that passes close to each data point. This is useful when you want to
approximate a spline through laboratory-measured data points without moving
any of the data points.

Page 18

CTRAN/W

Chapter 2: Material Properties

Figure 2-3 Natural and weighted splines


When the curve segments are curved (100%) between data points, the curve is
defined as a natural spline. As the curve segments are made straighter, the curve
segments approach a straight line between data points. Straightening the curve
segments helps to prevent "spline overshoots" (extreme peaks or valleys in the
spline). It also allows you to define "step" functions that have straight line
segments between each data point.
When specifying a best-fit percentage, it is best to experiment with different
values until you obtain a smooth spline that still passes close to the data points.

Page 19

Chapter 2: Material Properties

CTRAN/W

185

180

Conductivity

175

170

165

160

155
-10

-8

-6

-4

-2

Temperature

Figure 2-4 Spline fit to data using a 30% "fit" value

185

180

Conductivity

175

170

165

160

155
-10

-8

-6

-4

-2

Temperature

Figure 2-5 Spline fit to data using an exact fit value


Page 20

CTRAN/W

Chapter 3: Boundary Conditions

Boundary Conditions

3.1

Introduction

Specifying conditions on the boundaries of a problem is one of the key


components of a numerical analysis. This is why these types of problems are
often referred to as boundary-valued problems. Being able to control the
conditions on the boundaries is also what makes numerical analyses so powerful.
Solutions to numerical problems are a direct response to the boundary conditions.
Without boundary conditions it is not possible to obtain a solution. The boundary
conditions are the driving force. What causes contaminant to transport? It is the
concentration difference between two points or some specified rate of
contaminant flux into or out of the system. The solution is the response inside the
problem domain to the specified conditions on the boundary.
Sometimes specifying conditions is fairly straightforward, such as defining the
concentration or contaminant flux conditions that exist on a year-round basis at
the leakage point beneath a waste collection pond. Many times, however,
specifying boundary conditions is complex and requires some careful thought
and planning. Sometimes the correct boundary conditions may even have to be
determined through an iterative process, since the boundary conditions
themselves are part of the solution, as for instance, the contaminant flux from a
seepage face where the seepage face is not active continually.
Due to the extreme importance of boundary conditions it is essential to have a
thorough understanding of this aspect of numerical modeling in order to obtain
meaningful results. Most importantly, it is essential to have a clear understanding
of the physical significance of the various boundary condition types. Without a
good understanding, it can sometimes be difficult to interpret the analysis results.
To assist you with this aspect of an analysis, CTRAN/W has tools which make it
possible verify that the results match the specified conditions. In other words, do
the results reflect the specified or anticipated conditions on the boundary?
Verifying that this is the case is fundamental to confidence in the solution.
This chapter is completely devoted to discussions on boundary conditions.
Included are explanations on some fundamentals, comments on techniques for
applying boundary conditions and illustrations of boundary condition types
applicable for various conditions.

Page 21

Chapter 3: Boundary Conditions

3.2

CTRAN/W

Fundamentals

All finite element equations just prior to solving for the unknowns untimely boil
down to:

[ K ]{ X } = { A}
where:
[K]

a matrix of coefficients related to geometry and materials properties,

{X}

a vector of unknowns which are often called the field variables, and

{A}

a vector of actions at the nodes.

For a transport analysis the equation is,


Equation 3-1

[ K ]{C} = {Q}

where:

{C}

a vector of the concentration at the nodes, and

{Q}

a vector of the contaminant flux quantities at the node.

The prime objective is to solve for the primary unknowns, which in a transport
analysis are the concentrations at each node. The unknowns will be computed
relative to the C values specified at some nodes and/or the specified Q values at
some other nodes. Without specifying either C or Q at some nodes, a solution
cannot be obtained for the finite element equation. In a steady-state analysis, at
least one node in the entire mesh must have a specified C condition. The
specified C or Q values are the boundary conditions.
A very important point to note here is that boundary conditions can only be one
of two options. We can only specify either the C or the Q at a node. It is very
useful to keep this in mind when specifying boundary condition. You should
always ask yourself the question: What do I know? Is it the C or the
contaminant flux, Q? Realizing that it can be only one or the other and how
these two variables fit into the basic finite element equation is a useful concept to
keep in mind when you specify boundary conditions.

Page 22

CTRAN/W

Chapter 3: Boundary Conditions

As we will see later in this chapter, flux across a boundary can also be specified
as a gradient or a rate per unit area. Such specified transport boundary conditions
are actually converted into nodal Q values. So, even when we specify a gradient,
the ultimate boundary condition options still are either C or Q.
Remember! When specifying transport boundary conditions, you only have one of two
fundamental options you can specify C or Q. These are the only options available but
they can be applied in various ways.

Another very important concept you need to fully understand is that when you
specify C, the solution to the finite element Equation 3-1 will provide Q.
Alternatively, when you specify Q, the solution will provide C. The equation
always needs to be in balance. So when a C is specified at a node, the computed
Q is the Q that is required to maintain the specified C. When Q is specified, the
computed C is the C that is required to maintain the specified flux Q.
Recognizing the relationship between a specified nodal value and the
corresponding computed value is useful when interpreting results. Assume you
know the specified flux across a surface boundary. Later when you check the
corresponding computed concentration at that node you may find that it is
unreasonably high or low. You would use your knowledge of the problem to
assess if the contaminant flux applied was reasonable. The C values are
computed based on Q and the soil properties, so it must be one of three things.
Knowing what to look for helps you to judge whether that is reasonable or not.
CTRAN/W always provides the corresponding alternative when conditions are
specified at a node. When C is specified, Q is provided, and when Q is specified,
C is provided. The computed Q values at nodes where a concentration is
specified are referred to as Boundary Flux values with units of mass of
contaminant per time (e.g. M/t). These Boundary Flux values are listed with all
the other information provided at nodes.
A third important fundamental behavior that you need to fully understand is that
when neither C nor Q is specified at a node, the computed Q is zero. Physically,
what it means is that the contaminant flux coming towards a node is the same as
the flux leaving the node. Another way to look at this is that no contaminant is
entering or leaving the system at these nodes. Contaminant leaves or enters the
system only at nodes where C or a non-zero Q has been specified. At all nodes
for no specified condition, Q is always zero. This, as we will see later in this
Page 23

Chapter 3: Boundary Conditions

CTRAN/W

chapter, has important implications when simulating features such as point


sources of contaminants at a single node.
The concentrations in a transport analysis are the primary unknowns or field
variables. A boundary condition that specifies the field variable (C) at a node is
sometimes referred to as a Type One or a Dirichlet boundary condition.
Transport gradient (flux) boundary conditions are often referred to as Type Two
or Neumann boundary conditions. You will encounter these alternate names in
the literature but they are not used here. This document on transport modeling
simply refers to boundary conditions as concentration (C) or contaminant flux
(Q) boundary conditions. Later we will differentiate between nodal flux Q and
specified gradients (rates of flux per unit area) across an element edge.

3.3

Sources and sinks

There is another type of boundary condition called a source or a sink. These


boundary conditions are sometimes referred to as a Type 3 boundary condition.
A typical sink might represent a drain at some point inside a mesh where
contaminant is added or removed. The important concept about sinks and sources
is that they represent mass flux into or out of the system.
In CTRAN/W flux boundary conditions can be applied along outside edges or
nodes of the mesh, or along inside edges or nodes. There is no difference in how
the equations are solved in either case. The only thing to watch for is that you
have an understanding of the area that the contaminant transports across.
There are a couple special types of boundary conditions that have been
formulated directly in CTRAN/W: the surface mass accumulation condition and
the exit review condition. These are discussed in more detail below.

3.4

Advection and dispersion considerations

When you specify the boundary conditions in a contaminant transport analysis, it


is important to recognize that there are two processes by which mass is carried
across a boundary: one is by advection and the other is by dispersion. The
advective component is due to the water movement across a boundary while the
dispersive process is due to the chemical (concentration) gradients between the
boundary nodes and the nodes immediately inside the boundary.

Page 24

CTRAN/W

Chapter 3: Boundary Conditions

The advective component of the boundary mass flux is related to the water flux
(Qw) across the boundary. This information is obtained from a SEEP/W analysis.
CTRAN/W uses the nodal Qw values from the SEEP/W head (H) file to compute
the advective boundary flux.
When specifying boundary conditions for a transport analysis, you will often find
it useful to first get a clear picture of the water flux across the boundaries. A clear
understanding of the boundary water flux is essential in the specification of
boundary conditions and the interpretation of computed results.

3.5

Source concentration

Consider the case illustrated in Figure 3-1 where the leakage from the lagoon is
the source of the contamination. If the concentration of the contaminated fluid in
the lagoon is known, you can specify the boundary condition type as Cs (the
concentration of the source).

Figure 3-1 Illustration of a lagoon with contaminated fluid


When you specify Cs as the boundary condition type, CTRAN/W uses the
concentration of the source to compute a nodal mass flux at the boundary. The
mass flux is computed as:

Qmass = Qw * Cs
where:
Page 25

Chapter 3: Boundary Conditions

CTRAN/W

Qmass =

nodal total mass flux,

Qw

nodal water flux from SEEP/W, and

Cs

user-specified concentration of the source.

Specifying a Cs boundary at the nodes is different than specifying a C boundary


with a value equal to Cs. By specifying Cs at a boundary, you are actually
defining a mass flux type boundary. When specifying C, you are defining a
specified concentration type boundary.
When you specify Cs at a node, the computed nodal concentration will be less
than Cs during the early stage of the transport process. After some time, the
computed concentration will become equal to Cs. However, if you specify a C
boundary with a value equal to Cs, the computed nodal concentration will be
equal to Cs immediately.
In general, using Cs as a boundary condition is a more realistic option than
simply specifying concentration at the nodes. In addition, it has the advantage of
not creating excessive initial concentration gradients, as is the case with
specified concentration boundaries. The gradual build-up of concentration at Cs
nodes tends to reduce numerical difficulties that may arise from excessively high
concentration gradients.
NOTE: Specifying a Cs boundary in a node with zero nodal flux is the same as
specifying the node as a zero mass flux boundary (i.e., Qm = 0.0 or qm = 0.0).

3.6

Surface mass accumulation

Slow, contaminated moisture flow to the ground surface (or evaporative water
flux) can result in an accumulation of the solute on the surface boundary. In such
cases, the water evaporates, but the solute remains and accumulates with time, as
illustrated in Figure 3-2.
The solute accumulation at the ground surface can be simulated with CTRAN/W
by specifying a zero mass flux at the surface (i.e., Qm = 0.0 or qm = 0.0). A zero
mass flux means that no mass gain or loss is allowed across the boundary. In
other words, contaminant mass carried by the water flow to the boundary is not
allowed to leave; consequently, the mass accumulates at the boundary.

Page 26

CTRAN/W

Chapter 3: Boundary Conditions

Figure 3-2 Illustration of mass accumulation on evaporative ground


surface
In technical terms, the advective flux carries the contaminant solute to the
boundary. Physically, the water flux will cause advective mass loss across the
boundary; however, because of the specified zero mass flux boundary condition,
the boundary develops a reverse dispersive mass flux equal in magnitude but
opposite in direction to the advective mass loss. The reverse dispersive mass flux
causes the increase in concentration.
For the mass to accumulate at the boundary there must be water flux loss across
the boundary. There will be no solute accumulation if there is no water flow
across the boundary (i.e., qw = 0.0).
The default boundary condition in CTRAN/W is a no-mass flux condition, (i.e.,
Qm = 0.0 or qm = 0.0). Specifying a boundary type as none is the same as
specifying Qm or qm equal to zero.

3.7

Exit review

At a boundary where neither the mass flux nor the concentration are known, or
where the nodal water flux may reverse in direction during the transport process,
you may specify the boundary as an Exit Review boundary. When a boundary
node is specified with Exit Review, the node is checked to see if an exit boundary
should be applied at each time step. If the water flux of the node is negative (i.e.,
Page 27

Chapter 3: Boundary Conditions

CTRAN/W

water flux is exiting at the boundary), the boundary condition of the node will be
changed to an exit boundary. If the water flux of the node is zero or positive, the
boundary condition of the node is not changed.
CTRAN/W offers two types of exit boundaries. The first and simplest option is
one that ignores the dispersive flux across the exit boundary (Qd = 0), and is
often referred to as a zero dispersive mass flux exit boundary. With this type of
exit boundary, contaminant mass is assumed to leave the exit boundary by
advection only. As a result, the concentration gradients at the boundary are
forced to be zero, which causes the concentration contours to be perpendicular to
the exit boundary. The second type of exit boundary condition accounts for both
advective and dispersive mass flux at the boundary (Qd > 0), and is referred to as
a free exit boundary. A free exit boundary accounts for both advective and
dispersive mass flux across the boundary, which generally gives more realistic
results. It is best to use the free exit boundary unless there is a specific reason for
using the zero dispersive mass flux exit boundary.
The Example problem in the Illustrative Examples chapter illustrates a typical
situation where an exit boundary is required. Contaminant mass is existing at the
downstream toe of dam. Since neither the mass flux nor the concentration are
known at the boundary, the boundary is specified as Qm =0 conditions with
review for free exit boundary (Qd > 0).
The Exit Review feature is particularly useful in a density-dependent flow
problem. Using Henrys sea water intrusion problem presented in the Illustrative
Examples chapter, sea water may enter the flow system along the bottom portion
of the sea water boundary and freshwater may exit along the upper portion. Since
the interface between the sea water and the freshwater is not known, it is best to
specify the entire vertical boundary as Cs = 1 and review for free exit boundary.
Figure 3-3 is the solution to Henrys sea water intrusion problem when the right
boundary is simulated as a C boundary condition with no exit review. Figure 3-4
is the solution to the same problem except that the right boundary is specified as
a Cs boundary type condition with exit review Qd > 0. In the latter case, the
nodes along the upper portion of right boundary were converted to an exit
boundary by the exit review feature.
Allowing the sea water boundary to be reviewed for free exit boundary
conditions provides a more physically realistic solution than when the
concentration is specified along the sea water boundary. The difference between
the two solutions is primarily in the concentration distribution along the upper
Page 28

CTRAN/W

Chapter 3: Boundary Conditions

portion of the sea water boundary. Without exit review, the concentration of the
nodes along the upper portion of the right boundary are equal to your specified
value at all times; whereas with exit review, the nodes may be converted to free
exit boundary, and the concentration is solved for at each time step.
1.0
0.9
0.8
0.7
0.6

2
0.

0.
4

0.5

0.3

0.8

0.
6

0.4

0.2
0.1
0.0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

Figure 3-3 contour of Henry's problem, no exit review


1.0
0.9
0.8
0.7
0.6

S
C

0.
4

0.4

0.
8

0.3

0.
6

0.
2

0.5

0.2
0.1
0.0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

Figure 3-4 Concentration contour of Henry's problem with exit


review
Page 29

2.0

Chapter 3: Boundary Conditions

CTRAN/W

Note that Exit Review specification is only allowed for mass flux type boundary
conditions, (i.e. Qm, qm and Cs), where the nodal concentration are computed.
Exit boundaries should only be applied to four-noded quadrilateral elements and
must be specified along only one edge, (defined by two nodes), of an element.
When each element matrix equation is calculated in SOLVE, a surface integral
term is formed for element edges that are exit boundaries. This surface integral
cannot be formed for higher order elements; and therefore, application of exit
boundaries to higher ordered elements is invalid.
The exit surface integral can be formed for three-noded triangular elements if two
of the three nodes are on an exit boundary. However, using an exit boundary on a
triangular element often is of limited value, since any adjacent triangular
elements usually have only one node on the exit boundary. In this case, the
surface integral is formed only for every second triangular element.
In summary, the best exit boundary results are obtained with four-noded
quadrilateral elements. Three-noded triangular elements can be used but do not
give the best results and higher-order elements cannot be used along an exit
boundary.

3.8

Concentration vs. mass function

A special type of boundary function can be used to simulate, for example, the
case where a contaminant flows into a body of fresh water. Figure 3-5 illustrates
this example. Initially, the boundary condition of the fresh water pond can be
specified as a zero concentration boundary, (i.e. C=0). As the contaminant flows
into the pond, the concentration of the water increases with time.
The concentration boundary condition of the pond therefore has to be modified
with time. The concentration of the pond can be computed at any time if you
know the accumulated mass discharged into the pond and the volume of the
pond. In equation form, the pond concentration is:

C pond =

accumulated mass
pond volume

This relationship can be used to develop a C vs. mass function, (Figure 3-6),
which can be specified as a boundary condition.

Page 30

CTRAN/W

Chapter 3: Boundary Conditions

When using this boundary type, CTRAN/W computes the accumulated mass that
flows into the pond at the end of each time step. This value, together with the
boundary function, is then used to compute the boundary concentration at the
start of a new time step.

Figure 3-5 Illustration of contaminant discharge into a fresh water


pond
20

Concentration

18

16

14

12

10
0

10

15

Mass

Figure 3-6 Illustration of a concentration versus mass boundary


function

Page 31

Chapter 3: Boundary Conditions

3.9

CTRAN/W

Boundary functions

CTRAN/W is formulated to accommodate a very wide range of boundary


conditions. In a steady state analysis, all of the boundary conditions are either
fixed concentrations or fixed flux values. In a transient analysis however, the
boundary conditions can also be functions of time or in response to transport
amounts exiting or entering the transport regime. CTRAN/W accommodates a
series of different boundary functions. Each one is discussed in this section.
General
A non-fixed boundary condition must be entered as a user-defined function. In
CTRAN/W all functions are defined using a combination of manually entered or
cut and paste data points, and all functions can be customized to suite your exact
needs. In certain cases, it may be desirable to have a stepped function. The
additional functionality of automatically fitting the data with a step function has
been added to the program. There is also an option to have a cyclic function
repeat itself over time, which saves you the task of defining it repeatedly.
In general, all functions are comprised of a series of x and y data points that are
fit by a spline curve. A spline curve is a mathematical trick to fit a curved shape
between a series of points. The simplest way to fit a series of data points is to
draw a straight line between the points. This is often a very poor way to represent
a non-linear function. The advantage of the spline is that is joins all data points
with a continuous smooth curve.
During the solution process, the solver uses the y value along the spline curve
for any required x value. It is therefore important to make sure the spline fit,
not your original data, portrays how you want the boundary condition applied.

3.10

Time activated boundary conditions

There are situations where the actual position of a boundary condition may
change with time. A typical case may be the placement of mining tailings. The
thickness of the tailings grows with time and yet there is always some water on
the surface. The position of the head equal elevation (zero pressure) changes with
time. Another case may be in the simulation of constructing an embankment in
lifts. The hydraulic boundary condition changes with time. This type of boundary
condition is only useful when elements are activated or deactivated with time for
the purpose of simulating fill placement or constructing an excavation.
Page 32

CTRAN/W

Chapter 3: Boundary Conditions

Time activated boundary conditions are currently under development for a future
version of GeoStudio.

3.11

Null elements

There are many situations where only a portion of the mesh is required. Parts of
the mesh that are not required in a CTRAN/W analysis can be flagged as null
elements. Consider a case where you want to use a similar mesh in different
GeoStudio analyses. In the case of a cutoff wall to prevent water flow, you could
set those elements as NULL in the seepage analysis. This would treat the
elements as if they were not a part of the mesh and water would not flow around
them. However, if you now wanted to model the contaminant transport in this
scenario, you would turn the elements back on in CTRAN/W and assign them
valid properties.
Elements present but not required in a particular analysis can be assigned a
material type (number) that does not have an assigned conductivity function.
Elements with this uncharacterized material are treated as null elements. As far
as the main solver is concerned, these elements do not exist.
Null elements are not boundary conditions in the traditional sense but they do
alter the actual solved domain and the locations where boundary conditions can
be applied. The boundary between active elements and null elements is just like
any other perimeter boundary,where conditions can be specified as shown below.

Page 33

Chapter 3: Boundary Conditions

CTRAN/W

Page 34

CTRAN/W

Chapter 4: Analysis Types

Analysis Types

There are two fundamental types of finite element contaminant transport


analysis: advection / dispersion, or density-dependent. A further option for
simple particle tracking is also available. In terms of the seepage solution used
with CTRAN/W, it can be steady state or transient. Full details of steady state
and transient seepage analysis are provided in the SEEP/W or VADOSE/W
engineering books. A description of each type of contaminant analysis and the
implications associated with each type are discussed in this chapter.

4.1

Seepage solution interpolation

The accuracy of the contaminant transport solution is directly dependent on the


accuracy of the seepage solution. In other words, you must be able to obtain a
reasonable seepage solution of a flow system first before conducting the transport
analysis.
CTRAN/W relies on a seepage solution generated by either SEEP/W or
VADOSE/W to perform the contaminant transport analysis. For a steady-state
seepage analysis, the seepage solution is assumed to be constant during the
transport process. For a density-dependent analysis, the time steps specified in
the CTRAN/W data file are used to compute the seepage solution in SEEP/W
(VADOSE/W does not support density-dependent analysis). For transient
seepage analysis, CTRAN/W interprets the seepage solution as a step function.
For example, if seepage solutions are only available for three elapsed time steps
at 100, 200 and 400 minutes, in a transport analysis, the seepage solution is
assumed to be constant between 0 to 100 minutes, 100 to 200 minutes and 400
minutes or more.
In situations where the boundary condition is not constant and a more precise
seepage solution is required, you may be required to do one of the following:

Use smaller time step increments, especially in a period in which


changes in boundary condition are anticipated. The approximation
error resulting from the step function decreases as the time step
increments become smaller.

Use the same time step increments in both the seepage model and
CTRAN/W, so that the exact seepage solution at a certain time is
computed rather than interpolated.
Page 35

Chapter 4: Analysis Types

4.2

CTRAN/W

Initial conditions

The initial concentration of all nodes must be defined in a transport analysis


regardless of whether it is an advection-dispersion or density-dependent problem.
CTRAN/W allows you to specify the initial conditions by either reading the data
from an initial conditions file, or by drawing the initial conditions. By default,
when initial conditions are not specified, CTRAN/W assumes the initial
concentration of all nodes to be zero. A zero initial concentration condition
represents a clean flow system at the beginning of a transport process.
NOTE: The initial concentration of a node is independent of the boundary
condition of the node. In other words, a particular node may have an initial
concentration of 100 units specified, with a C boundary condition of 0 units also
specified.
Using an initial conditions file
With this option, you specify an initial conditions file by using the Open Initial
Conditions command in CTRAN/W SOLVE. The initial conditions file must be
the result of a previous analysis on the same flow system at any time step.
NOTE: The file name of the initial conditions file may be different than the file
name of the definition data file, (e.g. EXAMPLE1.C10 and EXAMPLE2.CTR,
respectively), but it is critical that both files be based on the same finite element
mesh.
Drawing the initial conditions
With this option, you specify initial conditions directly by using the Draw Initial
Conditions command in CTRAN/W DEFINE. This is particularly useful when
the initial concentrations of the nodes are known. For example, consider the case
of leachate migration from a landfill, as illustrated in Figure 4-1. The shaded area
represents a volume of soil containing a known contaminant concentration. To
specify this as the initial condition for a transport analysis, you may set the
concentration at all nodes that are in contact with the shaded area to the known
initial concentration value.

Page 36

CTRAN/W

Chapter 4: Analysis Types

Surface Infiltration

C is known initially
at these nodes

Landfill

Figure 4-1 Establishing initial conditions by specifying


concentration in the contaminated region

4.3

Particle tracking analysis

In the initial stages of performing a contaminant transport analysis, it is


sometimes useful to isolate the advective component of contaminant transport to
get an idea of the contaminant travel distances and travel times. CTRAN/W
includes a particle tracking analysis capability for just this purpose.
The particle tracking feature analyzes purely advective transport problems. A
number of particles can be arbitrarily introduced to the flow system at any
position. Particles are assumed to be attached to the water and move in the
direction of the water flow with the same speed as the water flow.
You may track the movement of the particles forward in the direction of water
flow, or backward in the opposite direction, toward the entrance or source
boundaries. You may select the forward or backward tracking option using the
KeyIn Analysis Control command when defining the problem.
With the forward tracking option, particles are usually introduced at the source
boundaries. CTRAN/W computes the new positions of the particles according to
Page 37

Chapter 4: Analysis Types

CTRAN/W

the average linear velocity of the groundwater. Forward tracking is useful for
determining where a particle of contaminant may end up and approximately how
long a particle may take to arrive at a new position. It is also useful for
delineating the possible flow paths or contaminant plume from the source
boundaries. Figure 4-2 illustrates the migration of particles from the lagoon to the
right exit boundary using forward tracking. A total of seven particles are used in
this example.
With the backward tracking option, particles are usually introduced at the sink or
exit boundaries and are tracked backward to the source boundaries. For example,
Figure 4-3 illustrates particles that arrived at the right exit boundary not only
come from the lagoon but can also come from the upstream boundary. Therefore,
the backward tracking option is particularly useful in delineating the possible
sources of contamination to the sink or exit boundaries in transport problems
with multiple source boundaries.
NOTE: Backward tracking function is only valid in a steady-state seepage solution.

Lagoon
Leakage

H1

H2

Figure 4-2 Illustration of particle flow path and plume using forward
tracking

Page 38

CTRAN/W

Chapter 4: Analysis Types

Lagoon
Leakage

H1

H2

Figure 4-3 Illustration of particle flow path and plume using


backward tracking
NOTE: Particle tracking simulates the transport of contaminant by the advection
process, (water movement), only. Other transport processes such as dispersion,
adsorption and decay are not considered. You may use the View Particle
information command in CTRAN/W CONTOUR to look at the computed results
of any particle at any location. The elapsed time of a particle at any location can
be interpreted as time required to move a dissolved contaminant molecule from
the starting position when dispersion, adsorption and decay are neglected.

4.4

Advection-dispersion analysis

Advection refers to the process by which solutes are transported by the bulk
motion of flowing groundwater. Dispersion refers to the phenomenon of
contaminant spreading from the path that it would be expected to follow
according to the advective hydraulics of the flow system. Virtually all
contaminant transport analyses require computation of advection and dispersion.
Adsorption refers to contaminant adsorption onto the solids of the porous
medium. Decay refers to removal of the contaminant by some form of decay
reaction, such as radioactive decay. Reactive, (adsorbing), or decaying
contaminants may have a significant effect on the contaminant concentration in
groundwater.

Page 39

Chapter 4: Analysis Types

4.5

CTRAN/W

Density-dependent analysis (with SEEP/W only)

For problems where the density of the contaminant is significantly different than
water, CTRAN/W has the capability of performing density-dependent transport
analyses. This feature is useful for solving problems such as sea water intrusion,
brine transport and landfill leachate migration, among others. Contaminant
density is modeled as varying linearly with concentration. Contaminant density
can be lower, higher or equal to the density of the native groundwater.
In density-dependent transport analyses, the flow velocities are dependent on the
contaminant concentration distribution and the concentrations are dependent on
the flow velocities. This circular dependence, or non-linearity, requires that the
seepage flow velocities and the contaminant concentrations be solved for
simultaneously by iterating at each time step. This type of non-linearity does not
exist for advection-dispersion transport analyses, where the flow velocities are
independent of the contaminant concentration distribution. For advectiondispersion analyses, the seepage velocities may be computed for all time steps
before calculating the contaminant transport.
To allow for the seepage velocities to be calculated at each time step during a
density-dependent analysis, CTRAN/W SOLVE starts and controls an instance of
SEEP/W SOLVE to perform the velocity calculations at each time step.
Therefore to perform a density-dependent transport analysis using CTRAN/W,
you must select the analysis type as Density-Dependent in both SEEP/W and
CTRAN/W.
In SEEP/W DEFINE you must specify the relative density of the contaminant at
a specified reference concentration. The relative density refers to the density of
contaminated water at the specified reference contaminant concentration. The
density of the contaminated water is assumed to vary linearly with increasing
contaminant concentration.
A relative density larger than 1.0 means the contaminant has a higher density
than water. Similarly, a relative density smaller than 1.0 means the contaminant
has a lower density than water. By default, the relative density of the contaminant
is 1.0, meaning that there is no density contrast between freshwater and
contaminated water as a function of concentration. Doing a density-dependent
analysis with relative density equal to 1.0 is essentially the same as doing an
advection-dispersion transport analysis, except that the seepage velocities will be
re-calculated within each iteration.

Page 40

CTRAN/W

Chapter 4: Analysis Types

The time step increments specified in CTRAN/W DEFINE are used in the
analysis for both SEEP/W and CTRAN/W. Time step information defined in
SEEP/W DEFINE is ignored in a density-dependent analysis. Similarly, the
maximum number of iterations allowed within a time step is specified in
CTRAN/W DEFINE. The maximum number of iterations specified in SEEP/W
is not used in density-dependent analyses.
It is important to recognize that the interpretation of head potentials in the
seepage solution of a density-dependent transport problem is somewhat different
than the usual interpretation in an advection-dispersion (density-independent),
transport problem. For density-dependent analyses, the total head at a node is
interpreted as the equivalent freshwater head. As the name implies, equivalent
freshwater head is an equivalent total head potential of freshwater. The density
contrast between contaminated water and freshwater adds a body force term on
the finite elements in the mesh. Therefore if density of the contaminated water
relative to freshwater increases with concentration (relative density greater than
1.0 at some concentration), then the equivalent freshwater head will increase with
contaminant concentration. Similarly, if the density of the contaminated water
relative to freshwater decreases with concentration, (relative density less than 1.0
at some concentration), then the equivalent freshwater head will decrease with
contaminant concentration.
A good example to illustrate the equivalent freshwater head is the static saltwater
column example in the chapter on Illustrative Examples. Although there is an
upward gradient in the equivalent freshwater head within the column, the upward
gradient is balanced exactly by the downward gradient caused by the body force.
As a result, there is no flow in the column.

4.6

Problem geometry orientation

The orientation of the defined problem in CTRAN/W will depend on what it was
specified as in the SEEP/W or VADOSE/W problem. More detailed discussion
of each type of view is provided in those other engineering books. A brief review
of the possible types of geometry views follows.
Axisymmetric view
An axisymmetric analysis can be used to simulate three-dimensional problems
with symmetry about a vertical axis of rotation. The problem is defined in two

Page 41

Chapter 4: Analysis Types

CTRAN/W

dimensions, but for analysis, it is as if the section is rotated about a vertical


central axis.
For an axisymmetric analysis, the computed mass flux is per unit radian if the
element thickness is specified as 1.0. If you want the computed flux value for the
entire circumferential area, you must either specify the element thickness as
6.2832 (i.e., 2 pi radians) before you do the analysis, or simply multiply the
above value by 2 pi after the solution is finished. You can change the element
thickness for the entire mesh with the Draw Element Properties command.
Plan view
A plan view analysis views the finite element mesh as lying on its side instead of
standing upright in a vertical plane. In CTRAN/W the only difference between a
plan view and a 2D view analysis is the way the area of a unit q contaminant
mass flux boundary is computed.
In a plan view, the area by which the q contaminant mass flux rate is multiplied
comes from the plan view areas of each element that contribute data to the q
specified nodes. The solver will compute the element areas and apply the flux, Q,
to the nodes.
2D view
In a 2D view, the area by which the q contaminant mass flux rate is multiplied
comes from the length of the element edges that the boundary condition is
specified. This length is then multiplied by the specified thickness of the element
(typically unit thickness into the screen) to obtain a total flux, Q, to apply at the
relevant nodes.

Page 42

CTRAN/W

Chapter 5: Numerical Issues

Numerical Issues

Entire textbooks can be written on numerical issues related to finite element


analysis. While modern computers and powerful graphics can make defining an
analysis quite fast and easy, they can not necessarily deal with some of the
intricate issues related to the concept of taking a natural process and breaking it
down into finite time and special domains (i.e. individual elements within a soil
geometry).
There are various ways to deal with many of the numerical issues, but the
unfortunate part is that there is no single method, approach, or technique that can
deal with all problems. Some numerical issues relate to restrictions in computer
hardware such as rounding off of non-integer variables during math operations;
some issues relate to non-linearity of soil properties; some issues relate to the fact
the physical equations being solved do not apply to all cases (for example,
CTRAN/W does not account for contaminants moving in both the air and liquid
phases); some issues relate to our inability to discretize a domain to small enough
element sizes; and other issues relate to the fact we have made the elements too
small!
There are numerical solvers that make use of adaptive meshing or adaptive time
stepping or both in attempts to be more suited to a wider range of problems. All
of these, however, have their limitations from a sound mathematical perspective
regardless of what the software developer will tell you. If you dont know what
the limitations are, it becomes somewhat risky to rely on a solver that claims to
handle it all.
Some finite element solutions attempt to march forward in time by considering
soil property values taken from the last, the current or the mid-time step average.
Some solvers simply make assumptions that limit their applicability to real life in
order to get a solution, such as moving the mesh to find the water table in a
seepage program instead of solving for the physics of flow above and below the
water table. Finally, some solvers may only work if you put in an initial guess of
the solution for the dependent variable being solved that is close to the desired
solution, in other words, start off the solution by pointing it in the right direction.
While it may appear all hope is lost, this is far from the case. Sound judgment
and common sense can usually overcome most of these challenges and result in
meaningful interpretations of how the soil will respond to changes in various
parameters.
Page 43

Chapter 5: Numerical Issues

CTRAN/W

It is not always possible to get an exact solution in many challenging cases, so


you should not necessarily be seeking an exact solution. If the problem is so
difficult that it is not solving reasonably, then it is very likely that either mistakes
have been made in the input, or, that you are pushing the envelope of the physical
theory applied in the model. This chapter looks at some of these issues as they
pertain to CTRAN/W.

5.1

Convergence

The objective of solving the finite element equations is to compute the


concentration at each node. For linear analyses when the material properties are
constant, the nodal concentrations can be computed directly. However, in the
cases of nonlinear analyses where there is adsorption as a function of
concentration, the correct material properties are not known at the start of the
analysis; consequently, an iterative scheme is required to solve the equations.
CTRAN/W uses a repeated substitution technique in the iterative process. For the
first iteration, the user-specified initial concentrations are used to define the
material properties. The material properties are updated in subsequent iterations
using the computed concentration from the previous iteration. The iterative
process continues until the iteration number reaches the maximum number
specified or until the results satisfy the convergence criteria.
Convergence means repeated solving of the nodal transport equations until the
computed solution does not change by more than a specified amount on
successive iterations. If the equations want to solve to the same results over and
over, then this means that the soil property values used in the solution agree with
the computed concentration values at each position in the mesh, and that the
influence of the boundary conditions (i.e., adding or removing mass) also are
reflected in the correct soil properties and concentrations. Everything must be
just right to balance both sides of the transport equation.
Vector norms
CTRAN/W does not use individual nodal concentrations to make the comparison
between successive iterations. Instead, it uses the percent change in the vector
norm a process that considers all nodal concentrations simultaneously. The
vector norm is computed as the square root of the sum of each nodal
concentration value squared, or in equation form:

Page 44

CTRAN/W

N=

Chapter 5: Numerical Issues

C
i =1

where:

the vector norm,

a counter,

the total number of nodes, and

Ci

the individual nodal concentrations.

The percent change in the vector norm is simply the change in the value N
between two successive iterations. If the change in N were exactly zero, then
there would be no change in solution on successive iterations. In general, there
are always very small differences between iterations so it is not reasonable to
expect a change in vector norm of zero.
You may wonder why we use the vector norm. It is a powerful way of
considering the entire problem in a single comparison, which prevents any
individual nodal concentration from potentially failing the convergence test in a
solution that really should be considered converged.
When the dependent variable (e.g. concentration in a transport analysis) gets
close to zero, very small changes in a single variable between successive
iterations can appear to be very big when converted to a percent difference
comparison. Consider if a single nodal concentration is changing from 0.01 to
0.02 between successive iterations. This is a change of 100% but likely
insignificant from a physical perspective. Its a failure in the logic of checking
convergence by individual nodal percent difference, because it lets small physical
changes become significant mathematical changes.
During the iterative process, CTRAN/W calculates the percent difference in
vector norm; which represents the percent difference in the concentration of all
nodes between two consecutive iterations. In an analysis where there is no
adsorption or in a particle tracking analysis, the problem is linear and is solved in
one iteration. Consequently, convergence parameters are ignored.

Page 45

Chapter 5: Numerical Issues

5.2

CTRAN/W

Numerical dispersion and oscillation

Numerical dispersion and oscillation are inherent in the finite element solution of
the advection-dispersion equation. The two phenomena are illustrated graphically
in Figure 5-1. Numerical dispersion tends to spread out the contaminant more
than what is predicted by analytical solutions. Numerical oscillation produces
concentrations higher or lower than the maximum and minimum specified
values.

Figure 5-1 The phenomena of numerical dispersion and oscillation


Numerical dispersion and oscillation cannot be eliminated; they can only be
controlled or minimized. The two widely used criteria are the Peclet Number and
the Courant number constraints. (See Frind and Pinder, 1982, and Daus, Frind
and Sudicky, 1983).

5.3

Peclet and Courant number criteria

The Peclet and Courant Number criteria are usually expressed in the onedimensional form as follows:

Page 46

CTRAN/W

Chapter 5: Numerical Issues

Peclet Number =

vx
2
D

Courant Number =

vt
1
x

where:

nodal spacing,

hydrodynamic dispersion coefficient,

average linear velocity, and

incremental time step.

In two-dimensional analyses, CTRAN/W computes the Peclet and Courant


Numbers at each Gauss point of an element using the following equations:

Px =

v x
2
D11

( Peclet Number in x-direction )

Py =

v y
2
D22

( Peclet Number in y-direction )

Cx =

vx t
1
x

( Courant Number in x-direction )

Cy =

v y t
y

( Courant Number in y -direction )

where:

maximum x-distance between the element corner nodes,

maximum y-distance between the element corner nodes,

D11 , D22

= hydrodynamic dispersion coefficient,


=

magnitude of linear average velocity,

Page 47

Chapter 5: Numerical Issues

CTRAN/W

vx

average linear velocity in x direction,

vy

average linear velocity in y direction, and

incremental time step.

The Peclet and Courant Number constraints provide the necessary conditions for
the finite element mesh design and the selection of time steps in transport
modeling. The Peclet Number constraint requires that the spatial discretization of
the flow regime is not larger than twice the dispersion potential of the porous
medium. The Courant Number constraint requires that the distance traveled by
advection during one time step is not larger than one spatial increment (i.e., one
element).
It is convenient to check for the Peclet and Courant numbers distribution within a flow
system by using the Draw Contours command. As a rule of thumb, when the Peclet
number is too big, the mesh size should be reduced, or alternatively, the material
dispersivity should be increased. When the Courant number is too big, the time step
increment should be reduced.

5.4

Backward or central difference time integration

Numerical dispersion and oscillation is also affected by the time integration


scheme. The Backward Difference approximation method (BDA) results in less
numerical oscillation but more numerical dispersion. On the other hand, the
Central Difference approximation method (CDA) results in less numerical
dispersion but more numerical oscillation.
A more stable answer can be obtained using the Backward Difference method.
However, if numerical stability can be controlled, a more accurate solution can
be obtained using the Central Difference method.
In summary, it is best to start with the more stable Backward Difference method
and then refine the solution with the Central Difference method once all other
factors in the analysis have been modeled correctly. In many cases, the
differences between the two methods are relatively small.

Page 48

CTRAN/W

5.5

Chapter 5: Numerical Issues

Mesh design

The Meshing chapter of the SEEP/W or VADOSE/W books provides general


modeling guidelines for the design of a finite element mesh for seepage analysis.
This section provides additional guidelines which should be followed in order to
extend the seepage analysis results for contaminant transport modeling.
In order to obtain stable solutions in a transport analysis, both the Peclet Number
and Courant number criteria must be considered in the design of the finite
element mesh. Rearranging the above Peclet and Courant number equations, the
spatial discretization requirements are:

x 2

D11
v

y 2

D22
v

x vx t
y v y t
The above equations indicate that the spatial discretization of the flow regime
should not be larger than twice the dispersion potential of the porous medium and
the distance traveled by advection during one time step. In most cases, the time
steps can be easily reduced to satisfy the Courant criterion, therefore, the design
of the mesh depending primarily on the seepage velocity and the dispersivity of
the materials.
Since the equations require knowledge of the average linear velocity, the Peclet
and Courant Numbers for each element will not be known at the time you create
the finite element mesh. Therefore, these criteria cannot be rigidly applied during
creation of the mesh. As a broad rule, the discretization in the major flow
direction should be about twice the longitudinal dispersivity L of the soil, and
the discretization in the minor flow direction should be about twice the transverse
dispersivity T of the soil.
This will provide a starting point for you to estimate the required spatial
discretization. Later, you can confirm the actual Peclet and Courant Numbers
Page 49

Chapter 5: Numerical Issues

CTRAN/W

computed by CTRAN/W and check that both the Peclet and Courant Number
criteria are satisfied in all time steps.

5.6

Time step design

Numerical dispersion and oscillation are directly affected by the time step
increments. To minimize numerical dispersion and oscillation, the Courant
Number constraint should be satisfied. The Courant Number constraint requires
that the distance traveled by the advective component of the transport process
during one time step ideally should not be larger than one element; that is, the
advective component should not jump across elements in one time step.
In order to satisfy the Courant Number constraint, the time increment should be:

x
vx

and,

y
vy

As a first approximation, you can estimate the time step increment based on the
average size of the elements and the average linear velocity. The Courant
numbers computed by CTRAN/W should be checked to confirm that the Courant
Number constraint is satisfied in all time steps.

5.7

Use of the Peclet and Courant numbers criteria

The Peclet and Courant Numbers must be viewed as a guide to minimizing


numerical dispersion and oscillation. Ideally, the Peclet Number should be less
than 2 and the Courant Number should be less than 1. Satisfying these criteria, in
general, would ensure a stable solution to the transport problem with minimal
numerical dispersion and oscillation. However, satisfying these criteria in all
elements for all time steps may be difficult to achieve in some practical
problems. It may require thousands of small elements and hundreds of small time
step increments depending on the material dispersivity and seepage velocity.

Page 50

CTRAN/W

Chapter 5: Numerical Issues

Generally, adherence to these criteria is most important where the concentration


gradients are steep, as is often the case during the initial time steps. Away from
regions with steep gradients, the criteria can be relaxed without creating major
difficulties. (Marsily, 1986) How rigidly you apply these criteria needs to be
judged in light of the objectives of a particular project.
Let us use Henry's Problem in the verification chapter as an example. Figure 5-2
shows the computed concentration contours at time step 35, and Figure 5-3
and Figure 5-4 illustrate the contour plots of the Peclet and Courant numbers,
respectively.

1.0
0.9
0.8
0.7
0.6
0.5

0.

0.4

0 .5

0.

0.3

0.7

0.2

0.0
0.0

0.9

0.1

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

Figure 5-2 Concentration contours in Henry's problem

Page 51

2.0

Chapter 5: Numerical Issues

CTRAN/W

1.0
0.9

1.
8

1.
4
0.8

1
0.7
0.6

0.6

0.5
0.4
0.3
0.2

0.2
0.1

0.4
0.0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

Figure 5-3 Peclet number contours of Henry's problem

1.0

0.9

0.8
0.7

0.6
0.5
0.4
0.3
0.2

0.1
0.0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

Figure 5-4 Courant number contours of Henry's problem

Page 52

2.0

CTRAN/W

Chapter 5: Numerical Issues

Near the upper right corner of the flow system, where the seepage velocity is the
highest, both the Peclet and Courant numbers criteria are not satisfied. However,
within most of the flow system, the Peclet number criterion is satisfied and the
Courant number is slightly higher than the suggested criterion. As a result, the
concentration contours indicate that a stable solution was obtained within most of
the flow system, and a small area with numerical instability is observed near the
upper right corner. The numerical instability causes small negative concentration
values to be computed at the upper nodes between the 0.1 and 0.2 concentration
contour lines. The concentration contours greater than 0.3 are not significantly
affected by the instability.
It is possible to improve the simulation such that both the Peclet and Courant
number criteria are satisfied and numerical instability is minimized. To do so will
require substantial refinement to the mesh especially in the upper right corner,
and a much finer time step increment. Since the area of interest of the Henrys
problem is in the lower right portion of the flow system, and only the 0.5
concentration contour line is used in the comparison with other models, we feel
that it is not necessary to further refine the analysis.

5.8

Gauss integration order

The details of numerical integration are provided in the appendices, along with a
discussion of how different integration orders can affect results for various types
of elements. Part of this discussion is repeated here as it pertains to improving
solution convergence.
The appropriate integration order is a function of the presence of secondary
nodes. When secondary nodes are present, the interpolating functions are
nonlinear and consequently a higher integration order is required. Table 5-1 gives
the acceptable integration orders.
Table 5-1 Acceptable element integration orders
Element Type

Secondary Nodes

Integration Order

Quadrilateral

no

Quadrilateral

yes

Triangular

no

Triangular

yes

Page 53

Chapter 5: Numerical Issues

CTRAN/W

It is also acceptable to use four-point integration for quadrilateral elements which


have secondary nodes. This is called a reduced integration order (see Bathe,
1982). Acceptable results can be obtained with reduced integration.
It is also possible to use three-point and nine-point integration with elements that
have no secondary nodes. However, the benefits of this are marginal, particularly
for quadrilateral elements. Nine-point integration for a quadrilateral element
involves substantially more computing than four-point integration, and there is
little to be gained from the additional computations. As a general rule,
quadrilateral elements should have secondary nodes to achieve significant
benefits from the nine-point integration.
The situation is slightly different for triangular elements. One-point integration
means the material properties and flow gradients are constant within the element.
This can lead to poor performance of the element, particularly if the element is a
zone of steep concentration gradient and there is active adsorption. Using threepoint integration, even without using secondary nodes, can improve the
performance, since material properties and gradients within the elements are
distributed in a more realistic manner. The use of three-point integration in
triangular elements with no secondary nodes is considered acceptable for
triangular elements in a mesh that has predominantly quadrilateral elements. This
approach is not recommended if the mesh consists primarily of triangular
elements with no secondary nodes.
In general, it is sufficient to use three-point integration for triangular elements
and four-point integration for quadrilateral elements. In situations where there is
adsorption and steep gradients within an element, it is best to use quadrilateral
elements with secondary nodes together with nine-point integration.

5.9

Equation solvers (direct or iterative)

CTRAN/W has two types of equation solvers built into it; a direct equation
solver and an iterative equation solver. Both offer certain advantages.
Select the direct equation solver option if you want the system equations to be
solved using a Gauss elimination skyline direct solver. The processing speed of
the direct solver is bandwidth (the maximum node number difference of all the
elements in a domain) dependent. In other words, the direct solver is very fast
when solving simple problems with small bandwidth, but it can be quite slow
when solving more complex problems with a large bandwidth. CTRAN/W
Page 54

CTRAN/W

Chapter 5: Numerical Issues

automatically sorts the nodes so that the bandwidth is the smallest possible value,
which helps the solution solve faster using the direct solver. By default, the direct
equation solver is selected.
Select the iterative equation solver option if you want the system equations to be
solved using a preconditioned Bi-Conjugate Gradient (BiCG) iterative solver.
The processing speed of the iterative solver is bandwidth (the maximum node
number difference of all the elements in a domain) independent. The iterative
solver can be slower than the direct solver when solving a simple problem with
small bandwidth, but it can be much faster than the direct solver when solving
more complex problems with a large bandwidth. If you choose the iterative
equation solver, the maximum number of iterations the solver can attempt is set
equal to the maximum number of degrees of freedom in the problem domain.
The iterative solver tolerance parameter is different from the finite element
solution tolerance, although they appear similar. The iterative solver tolerance is
the desired residual relative to the norm of the concentration vector ||C|| as a
solution for the global finite element matrices is obtained. The finite element
solution tolerance, on the other hand, is related to how the solution of all the
equations changes between two trials based on attempts to obtain the correct
material properties for the given boundary conditions.
If, for example, you choose an iterative solver tolerance of 10-6 it means that you
consider ||C|| to have errors in the range of 10-6. The iterative solver tolerances
have been defaulted to values that will work in most cases within a reasonable
computational effort. In the case where a high level of accuracy is required or
where the solution does not seem reasonable, you may need to tighten the solver
tolerance (e.g., to 10-8).

Page 55

Chapter 5: Numerical Issues

CTRAN/W

Page 56

CTRAN/W

Chapter 6: Visualization of Results

Visualization of Results

When you get to the visualization of results stage of a finite element analysis you
can congratulate yourself for having completed the hardest parts setting up the
geometry, defining meaningful soil property functions, and applying appropriate
boundary conditions to the mesh. If, at this point, you do not have the tools or the
understanding of how to interpret the massive amount of data that may have been
generated by the solver, then you have wasted your time.
This chapter describes the various types of output data that are computed by the
solver and it attempts to get you thinking about what the data is trying to tell you.
For example, did the solution solve properly? Did the boundary conditions you
applied get reflected in the actual solution? Did the soil respond how you thought
it would respond? If not, how to you methodically determine what to check next?
The chapter is structured to explain what type of data is available for
visualization. In the various sections, comments are provided that relate they type
of result data in question to how is should be used in the overall thought process.
Its a good idea to read this entire chapter.

6.1

Node and element information

In order to understand what type of information can be viewed as results output,


it helps a bit to know how the data is obtained. To recap, you set up the problem
geometry, define material properties, and apply boundary conditions of either
known concentration or mass flux. The solver assembles the soil property and
geometry information for every Gauss point in every element and applies it to the
transport equation that is written for every node. Therefore, at each node we have
some applied boundary data, some interpolated soil property data and geometry
data. The solver then computes the unknown value in the equation for each node
the unknown value being either concentration or flux. It is the Gauss point data
that is used to set up the nodal equations so the Gauss point data written to the
output file is the actual data used in the solver.
Figure 6-1 is an illustration of the type of information that can be viewed for each
node in the finite element mesh. You can view information about geometry,
material properties, seepage results, concentrations and boundary fluxes as well
as gradients, water velocities etc. Of significance is the ability to view Courant
and Peclet information. This is discussed more in the contouring data section
below.
Page 57

Chapter 6: Visualization of Results

CTRAN/W

One key point to note in the figure below is that the nodal Boundary Mass Flux
quantity is exactly zero. This is an important point to understand because it can
help with your overall interpretation of results. This boundary flux is computed
by summing the contributing fluxes from each of the four Gauss points that
surround this node. So, if mass is flowing out of one Gauss region, it HAS TO be
flowing into an adjacent Gauss region. For all internal nodes with no user
boundary applied to them, the sum of all the mass fluxes at a node should equal
zero in a properly converged solution. In other words, there is mass balance at the
node.

Figure 6-1 Visualization of node information

Page 58

CTRAN/W

Chapter 6: Visualization of Results

If the node being viewed is a boundary condition node (not necessarily at the
edge of the geometry but with an allowed influx or out flux) then the summation
of all the mass fluxes at that node will not be zero because contaminant mass is
either gained or lost at that point.
Figure 6-2 is the corresponding Gauss point information for the Gauss point
located just above and to the right of the node illustrated in the previous figure.
The shaded region in the figure shows the contributing area of that Gauss point
and in this case, because the element is rectangular and has no secondary nodes,
this Gauss area is equal to one fourth of the total area of the element. The inset in
the figure below shows the type of data that can be viewed at each Gauss point. If
you consider the water content value of 39.8%, for example, you should realize
that this water content is assumed to exist throughout the Gauss point area
displayed; and you should next realize that if the element size is increased, the
estimate of the water content becomes less accurate as we are averaging it over a
larger area. The real trick to getting good finite element analysis results, is to
create a finite element mesh with just the right sized elements that are not too big
or too small, that can represent the highly non-linear soil properties within them,
and that can handle the potentially extreme boundary conditions you apply. Its
not always easy and there is no sure quick or automatic method to make that
happen.

Page 59

Chapter 6: Visualization of Results

CTRAN/W

Figure 6-2 Visualization of element Gauss point information

6.2

Particle information

This features displays information about a particle at a specified point in time.


When a particle time increment is selected to view, CTRAN/W reads all particle
files up to the selected time increment and displays the path over which the
particle traveled during these time increments. Using the View Particle
Information command, you can select any point along the particle path and view
the particle x-y coordinates, the time at which the particle arrived at the point, the
distance the particle traveled up to the point, and the average particle speed up to
the point. Figure 6-3 shows the type of data that can be accessed.

Page 60

CTRAN/W

Chapter 6: Visualization of Results

Figure 6-3 View particle information data


The position of the particles is a good indication of where the 50% concentration
value will be at any given time. Consider an advection dispersion problem
where some contaminant is moving due to water flow; and some is moving ahead
of the water due to molecular action, while some is retarded in progress due to
molecular action in the direction opposite to flow. The s shape of the
concentration plug illustrates this idea. Now, consider the mid-point of the s
shape. It is a point in concentration (the 50% point) and it is moving solely due to
advection. So, if we are only interested in the position of the 50% concentration
point, we can simply use a particle tracking analysis and not the full advection
dispersion solution.

Page 61

Chapter 6: Visualization of Results

6.3

CTRAN/W

Mass accumulation

This feature displays accumulated mass values within selected elements. In a


contaminant transport problem, it is often useful to find out the amount of mass
that has accumulated in any region of the mesh. The View Mass Accumulation
command allows you to select any group of elements and see the amount of mass
in the elements as illustrated in Figure 6-4.

Figure 6-4 View mass accumulation data

6.4

Equipotential lines

Equipotential lines are lines connecting all points of equal energy potential. The
energy in this case is represented by concentration and the equipotential lines are
called isochlors. In a one-dimensional analysis with mass flow in the vertical
direction, the equipotential lines are exactly horizontal. In a two-dimensional
analysis, the equipotential lines can exist in an infinite number of directions, as
shown in Figure 6-5.

Page 62

CTRAN/W

Chapter 6: Visualization of Results

C = 10

Free Exit

Figure 6-5 Illustration of isochlors generated in CTRAN/W program


Projecting Gauss point values to nodes
CTRAN/W performs contouring calculations based on parameter values at the
nodes. Since the primary parameter, concentration, is computed at the nodes,
these parameters can be contoured directly. However, secondary parameters
(velocity, gradient, Peclet / Courant number, and water content etc.), are
computed at the element Gauss points and must therefore be projected to the
nodes for contouring purposes.
In triangular elements, the Gauss point values are projected on the basis of a
plane that passes through the three Gauss points. For one-point integration, the
value at the Gauss point is also taken to be the value at the nodes (i.e., the Gauss
point value is constant within the element).
In quadrilateral elements, the Gauss point values are projected using the
interpolating functions. (For more information about interpolating functions, see
the appendix). In equation form,

x= N

{X }

where:

Page 63

Chapter 6: Visualization of Results

CTRAN/W

the projected value outside the Gauss points at a local coordinate


greater than 1.0,

<N>

a matrix of interpolating functions, and

{X}

the value of Gauss point variable.

The local coordinates at the element nodes are the reciprocal of the Gauss point
local coordinates when forming the element characteristic matrix. Figure 6-6 is
an example of the local coordinates at the element corner nodes when projecting
outwards from the four Gauss points in the element. The value of 1.7320 is the
reciprocal of the Gauss point coordinate 0.57735.
This projection technique can result in some over-shoot at the corner nodes when
variation in the parameter values at the Gauss points is large. For example,
consider that we wish to contour unfrozen water content and that in some
elements the water content at the Gauss points varies over the complete range of
the water content function. Projecting such a large variation to the nodes can
result in projected nodal water contents beyond the range of the water content
function.
Extreme changes in the parameter values at the Gauss points within an element
often indicate numerical difficulties (the over-shoot at the nodes being just a
symptom of the problem). This over-shoot can potentially be reduced by a finer
mesh discretization. Smaller elements within the same region will result in a
smaller variation of parameter values within each element, therefore lowering the
potential for encountering unrealistic projections.

Page 64

CTRAN/W

Chapter 6: Visualization of Results

Figure 6-6 Local coordinates at the corner nodes of an element with


four integration points

6.5

Contours

The power of using advanced graphical interfaces with finite element analysis is
that the computer can quickly convert thousands of pieces of data into
meaningful pictures. In the section above we introduced isotherms. We can use a
picture of the isochlors to tell us something about what is going on in the soil. In
particular, if we consider how close adjacent isochlors are to each other, we are
in effect, considering how steep the concentration gradient is. If we recall that the
amount of mass flow is equal to the concentration gradient multiplied by the
dispersion coefficient, we then have a fast and clear picture of where the areas of
high mass flow are in the domain we have modeled.
CTRAN/W is a powerful tool in that it will let you contour many different
parameters such as concentration, Courant number, Peclet number, pressure,
gradients, mass fluxes, water contents and more. Figure 6-7 is a contour of the X
direction Courant number values for the isochlors illustrated above. We know
that for an advection-dispersion type analysis, we want the Courant number to
remain low. This figure shows us that it ranges between o0.05 and perhaps 0.5,
which is acceptable. If this contour were to have shown us areas of high Courant
or Peclet value, we would have an idea of what to adjust in order to obtain a
better solution. Perhaps the mesh would need to be finer in one location or the
time steps smaller.

Page 65

Chapter 6: Visualization of Results

CTRAN/W

C = 10

0.05

0.2
0.1

5
5
Free Exit

0.1

0.2

0.05

0.3

Figure 6-7 CTRAN/W computed X direction Courant numbers

6.6

Water flow vectors and flow paths

Water flow vectors and flow paths are described in detail in the SEEP/W and
VADOSE/W engineering books.

6.7

Flux sections

CTRAN/W has the ability to compute the instantaneous flux across a userdefined section for either a steady-state or transient analysis. The view preference
command can be used to modify the flux type that is displayed on the flux
section. View Preferences allows you to display total, advective, dispersive,
stored, or decayed flux on the drawing. The stored flux is the rate at which mass
is stored in an element. A negative stored mass means that mass is accumulating
in an element (i.e., the total mass flow across the section is decreasing due to
adsorption). A positive stored flux means that mass is being released from the
element, thereby increasing the rate of mass flow across the section.
Flux section theory
Details of the flux section calculations are given in the Theory chapter.

Page 66

CTRAN/W

Chapter 6: Visualization of Results

Flux section application

1.6689

e-001

Flux sections can be used in many ways, because they can be drawn any place
across which you want to know the flux. You many want to check if an influx is
equal to an outflux such as illustrated in Figure 6-8. In this case, the values are
NOT the same which is to be expected because the problem is a transient analysis
and there is a change in the amount of mass storage over time.

12
11
10

C = 10

9
8
7

3.3035e-0

6
5
4

02

Free Exit

2
1
0
0

10

11

12

13

14

15

16

17

18

19

20

21

22

23

metres

Figure 6-8 Flux section used to check balance of mass inflow and
outflow
Flux sections do not have to be drawn as single straight lines. They can be made
of continuous attached segments as illustrated in the figure above. When a
multiple segmented flux section is drawn, the value of flux reported for the
section applies to the entire section, not any individual segment.
The key point to note when defining a flux section is to make the flux section
cross the sides of the elements and not the nodes of the elements. Also, if you
want to check the flux around a closed loop as illustrated for the SEEP/W
seepage drain nodes in Figure 6-9, make sure the end of the flux section crosses
over the tail of the first segment of the flux section.

Page 67

Chapter 6: Visualization of Results

CTRAN/W

Two words of caution: flux sections MUST be defined before you solve the
problem, because the program needs to calculate the values during the solution
sequence, not afterward. In addition, all flux values are reported as positive,
which means direction is not taken into account. This is required because the sign
of the flux value will depend on which way you draw the section. To avoid any
misinterpretation, all flux section values are reported as positive, and then you
can plot flux vectors in order to determine the direction of flow, if it is not
obvious based on your problem definition.

No flow boundary with


water table initially at surface

1.00

Sum of each nodal Q:


(3 x 0) = 0 plus
(2 x 0.15522) = 0.31044 plus
(2 x 0.3919) = 1.09424 plus
(1 x 0.40576) = 1.50000

0.75

0.50

Seepage
boundary

1.5000e+000
0.25

0.00

Q=0.5 m^3/time nodal flux boundary

Figure 6-9 Flux section used in SEEP/W around series of drain


nodes to check flow

6.8

Changes between selected times

It can often be useful to view how certain data changes with time and position.
CTRAN/W will let you simultaneously view the change in your defined
concentration value line for any or all of the time steps in your model as
Page 68

CTRAN/W

Chapter 6: Visualization of Results

illustrated in Figure 6-10. In this case, the concentration is set to show a 50%
value line at four different snapshots in time. The numerical value shown on the
actual line is an integer value representing the computed time step, not the actual
elapsed time.
12
11
10

C = 10

9
8
7
6

10

5
4

25

Free Exit

40

2
1
0
0

10

11

12

13

14

15

16

17

18

19

20

21

22

23

metres

Figure 6-10 Transient position of 50% concentration isochlor over


time
Most parameters cannot be contoured for more than a single time step at a time
because, unlike plotting the change in freeze-thaw line, most contour values vary
over the entire domain. In the event you want to view contours at different time
steps, you must select the time step in question and re-plot the contour as shown
in the next two figures. In the event you want more specific data output as a
function of time and position, you can use the graphing feature, which is
discussed next.

6.9

Graphing

The Draw Graph command allows you to plot a graph containing any of the
computed parameter values. If you are viewing a concentration time increment,
the following parameters can be plotted: concentration, total head, pressure,
pressure head, total mass, fluid mass (portion of the mass contained in fluids),
solid mass (portion of the mass on solids), adsorption, x-Peclet number, y-Peclet
Page 69

Chapter 6: Visualization of Results

CTRAN/W

number, x-Courant number, y-Courant number, x-velocity (log), y-velocity (log),


xy-velocity (log), volumetric water content, XX-dispersive coefficient (log), XYdispersive coefficient (log), and YY-dispersive coefficient (log).
If you are viewing a particle time increment, the following parameters can be
plotted: total head, pressure, pressure head, x-velocity (log), y-velocity (log), and
xy-velocity (log), volumetric water content.
The above listed parameters are the dependent variables of the graph. Any of the
dependent variables can be plotted versus the following independent variables:
nodal x coordinates, nodal y coordinates, and the distance between nodes
(starting at the first selected node).
The independent graph variable that you choose affects how the selected nodes
and time steps are used in the graph. If the graph independent variable is x
coordinate, y coordinate, or distance, then the parameter value at each selected
node is plotted versus the nodal coordinate or the distance between nodes. Each
selected time step is plotted as a separate line on the graph. If the graph
independent variable is time, then the parameter value at each selected node is
plotted versus the elapsed time for each of the selected time steps. Each selected
node is plotted as a separate line on the graph. These two cases are illustrated in
the Figure 6-11 and Figure 6-12 where the concentration beneath the lagoon is
compared versus position and time respectively. The nodes down the left edge of
the above illustrated examples are chosen for plotting data in the following
figures.
These illustrations show the default graph template used in the model. There is,
however, the ability to customize the scales, labels, and symbols so that the plots
can be used directly within your reports. In addition, the above figures were
simply copied from the screen and pasted into this document using the COPY
command inside CTRAN/W.

Page 70

CTRAN/W

Chapter 6: Visualization of Results

Concentration vs. Time


10
Node 4

Concentration

Node 5
Node 6

6
Node 7
4

Node 8
Node 9

2
Node 10
Node 11

0
0

500

1000

1500

2000

2500

3000

Time

Figure 6-11 Left side node concentration as a function of time


Concentration vs. Y
10
0.0000e+000
2.5000e+002
8

5.0000e+002
7.5000e+002

1.0000e+003
6

1.2500e+003
1.5000e+003
1.7500e+003
2.0000e+003

2.2500e+003
2.5000e+003
2.7500e+003

2
0

10

Concentration

Figure 6-12 Left side node concentration as a function of y


coordinate for each time

Page 71

Chapter 6: Visualization of Results

6.10

CTRAN/W

Reporting

In the event you need to further customize your output data, CTRAN/W will let
you view, then COPY data to many other programs, such as Microsoft Excel or
Word or other graphing packages. Once you decide what you want to graph, and
perhaps you have even viewed the graph on screen, you can select the option to
create a table of all of the data. The data in Table 6-1 shows the unfrozen water
content versus time for node 211. The data was copied directly from the
CTRAN/W generated graph and pasted as unformatted text in this Word
document.
This data can also be obtained directly from the actual output files created by the
solver. While this is not the general recommendation, if you are adventurous, you
can simply open the output data files at any given time step and manually extract
the necessary data. The very adventurous can even program macros within
Microsoft Excel to open files, get the data and close the files. This may be useful
if you want to automatically track the volume of water over a unique region of
the mesh at multiple time steps. Having said this, you should know that GEOSLOPE will not be helping you with any macros you want to write in Excel or
any other program. We just mention it here because it is do-able.
Table 6-1 Data copied from CTRAN/W and pasted directly into Word
Pt #

Time

Conc.9

Conc.10

Conc.11

+0.0000e+000

+0.0000e+000

+0.0000e+000

+0.0000e+000

+2.5000e+002

+7.7363e+000

+9.0826e+000

+1.0000e+001

+5.0000e+002

+9.1233e+000

+9.6720e+000

+1.0000e+001

+7.5000e+002

+9.5440e+000

+9.8327e+000

+1.0000e+001

+1.0000e+003

+9.7280e+000

+9.9011e+000

+1.0000e+001

+1.2500e+003

+9.8244e+000

+9.9364e+000

+1.0000e+001

+1.5000e+003

+9.8804e+000

+9.9568e+000

+1.0000e+001

+1.7500e+003

+9.9154e+000

+9.9695e+000

+1.0000e+001

+2.0000e+003

+9.9383e+000

+9.9778e+000

+1.0000e+001

10

+2.2500e+003

+9.9539e+000

+9.9834e+000

+1.0000e+001

11

+2.5000e+003

+9.9648e+000

+9.9873e+000

+1.0000e+001

12

+2.7500e+003

+9.9727e+000

+9.9902e+000

+1.0000e+001

Page 72

CTRAN/W

Chapter 7: Modeling Tips and Tricks

Modeling Tips and Tricks

7.1

Introduction

This chapter contains many useful hints about using the software and
understanding what it does. READ THIS CHAPTER!
CTRAN/W is a powerful analytical tool, but it will only provide valid solutions if
the boundary conditions, material properties, and time sequence are appropriately
defined. It is your responsibility to properly define the problem parameters and
ensure that the results produced are valid and reasonable.
This chapter presents some general modeling guidelines. The information
presented is not an exhaustive statement on the "how-to" of modeling a
contaminant transport problem. Instead, it is intended to provide suggestions on
how you might model various conditions, as well as to outline the implications of
certain modeling specifications.
There have been many occasions where GEO-SLOPE has been contacted by
clients with questions about how the model behaves in response to changes in
various parameters. If we do not know the answer, we conduct a numerical
experiment to test what will happen. The first few sections of this chapter
illustrate a few common examples of numerical experiments. You are strongly
encouraged to learn why these types of simple tests are so powerful in testing
how the program computes results BUT ALSO in enhancing your understanding
of how the physical mechanisms of flow through porous medium occurs.
A numerical experiment is carried out by making a very simple finite element
problem. It is useful to use a mesh that is one distance unit wide and one distance
unit high. This makes hand calculating flux values very simple and they can
easily be checked against the computed flux values. The following discussion
illustrates how some simple numerical experiments have been carried out to test
some simple, yet valid, questions.
When setting up these experiments, it is a good idea to input simple soil property
functions. In most cases, two data points are sufficient to define the conductivity
and storage function. Just as a reminder, give both functions some slope dont
make them horizontal!

Page 73

Chapter 7: Modeling Tips and Tricks

7.2

CTRAN/W

Modeling progression

Modeling the movement of contaminants through soil is a complex type of


analysis. As discussed in the Theory Chapter, the contaminant transport process
is not only governed by water movement, it is also influenced by dispersion,
molecular diffusion, adsorption, and radioactive decay. With so many factors
involved, it can be confusing and difficult to resolve the contribution of each
component in the transport process.
An important rule to follow in contaminant transport modeling is to progress
from the simple to the complex. It is good practice to initially define a simplified
version of the problem and then add complexity in stages. Moving from the
simple to the complex makes it easier to pinpoint difficulties with the model
when the results of the analysis are unrealistic.
For example, begin by analyzing a homogeneous problem with no adsorption and
no decay. Then add adsorption, decay, and other materials to the analysis in
separate stages. This incremental approach not only helps to evaluate the results
but also provides information on the effect of the various parameters.
Determining what causes unrealistic results can be difficult if all of the possible
complexities are included at the start of the problem analysis.
For complex transport problems that may involve multiple sources and exit
boundaries, it is useful to first do a particle tracking analysis before doing the
advection-dispersion transport modeling. Particle tracking has the advantage of
providing an approximate solution to the transport problem quickly with
minimum input requirements. Although only the advection process (water
movement) is considered in particle tracking, the solution does provide a clear
picture of the approximate contaminant plume.

7.3

Problem engineering units

Any system of units can be used for a contaminant transport analysis; the only
requirement is that you must be consistent with the units throughout both the
CTRAN/W and the corresponding seepage analysis. Fundamentally, you must
select the units for length (geometry), time, and force. Once you have selected
units for these parameters, all other units must be consistent. Table 7-1 and Table
7-2 present some typical sets of consistent units.

Page 74

CTRAN/W

Chapter 7: Modeling Tips and Tricks

Table 7-1 Consistent set of SI units


Parameter

Symbol

Units

Length

Time

Force

kN

Mass

g
2

kN/m2

kN/m3

Pressure

F/L

Unit Weight of Water

F/L

Hydraulic Conductivity

L/T

Concentration

M/L

g/m3

Diffusion Coefficient

2
L /T

m2/s

Decay Half-Life

sec

S (adsorption)

M/M

L and T (dispersivity)

m/s

g/g

Density

M/L

g/m3

Mass Flux

M/T

g/s

Table 7-2 Consistent set of imperial units


Parameter

Symbol

Units

Length

ft

Time

hr

Force

lbf

Mass

lb
2

psf

Pressure

F/L

Unit Weight of Water

F/L

pcf

Hydraulic Conductivity

L/T

ft/hr

M/L

lb/ft3

L /T

ft2/hr

L and T (dispersivity)

ft

Decay Half-Life

hr

Concentration
Diffusion Coefficient

S (adsorption)

M/M

lb/lb

Density

M/L

lb/ft3

Mass Flux

M/T

lb/hr

Page 75

Chapter 7: Modeling Tips and Tricks

CTRAN/W

The units of time are established once you select the units for hydraulic
conductivity. The units of pressure are established once you select the unit
weight of water. Generally, all units are defined by selecting the units of length
for the problem geometry, units for hydraulic conductivity, and the units for the
unit weight of water.
Concentration is often reported in units of milligrams per liter (One milligram per
liter (mg/l) is equivalent to one part per million (ppm) and is also equivalent to
one gram per cubic metre (g/m3). In equation form: mg/l = ppm = g/m3.
Therefore, concentrations reported in mg/l or ppm can be conveniently used in a
CTRAN/W analysis provided mass is defined in grams (g) and length is defined
in meters (m).
In summary, the key requirement is that the system of units must be consistent.

7.4

The use of infinite elements

The use of infinite elements in seepage analyses is discussed in the SEEP/W and
VADOSE/W books. To be fully compatible with the seepage solution, infinite
elements are also allowed in CTRAN/W. However, because of the sensitivity of
the transport solutions to the spatial discretization of the finite element mesh,
infinite elements must be used with caution in CTRAN/W.
A stable transport solution requires the satisfaction of the Peclet and Courant
number criteria. These criteria were developed for standard finite elements, and
the application of these criteria to infinite elements is unknown. Furthermore,
since the outer edge of an infinite element is deemed to be at infinity, the
computation of Peclet and Courant numbers within an infinite element is not
possible. CTRAN/W computes the Peclet and Courant numbers based on the
specified x- and y-coordinates of the infinite element. Therefore, satisfying the
Peclet and Courant number criteria in an infinite element may not necessarily
produce stable solutions in the infinite element.
Slightly negative concentrations can occur due to numerical oscillation. These
small negative concentrations will not likely affect the overall concentration
distribution throughout the system, but they can have a significant effect on the
system mass calculations. Infinite elements can be very big, and consequently,
small variations in concentrations can result in a large variation in the total
system mass calculations.

Page 76

CTRAN/W

Chapter 7: Modeling Tips and Tricks

Using infinite elements for density-dependent analyses is not recommended,


because element volume is not defined for infinite elements.
In summary, infinite elements must be used with caution in transport analyses.
Only simple boundary conditions, such as concentration or zero mass flux, are
recommended along the infinite boundary. You must be particularly careful in
interpreting the mass computations for infinite elements.

7.5

Fracture flow simulation

CTRAN/W can be used to analyze contaminant flow through fractures by


modifying the definitions for volumetric water content ( ) , adsorption (S), and
the dry bulk density ( d ) . For fracture flow, these parameters are defined as:

Volume of Fractures
Total Volume

S=

Mass of Solute Adsorbed on the Fractures


Unite Area of the Fractures

d =

Area of the Fracture Planes


Total Volume

The adsorption terms in the finite element equation involve ( d ) multiplied by


(S). Multiplying these two variables together results in units of mass adsorbed on
the fractures per unit volume. The area of the fracture planes vanishes as far as
the units are concerned.

7.6

Flux section location

Question: Does the location of a flux section within an element have any
influence on the computed flux value?
Answer: No. The flux section value will be the same regardless of whether the
section is drawn near the element edge or element middle. Figure 7-1 shows this
to be the case and it is true for a transient and steady state solution.

Page 77

Chapter 7: Modeling Tips and Tricks

CTRAN/W

2.6869e+002

2.6869e+002

Figure 7-1 Test to check flux section locations

7.7

Unit flux versus total flux?

There are many people who are unsure of the difference between a unit flux and
a total nodal flux. Do a simple test if you are unsure.
Question: How is a unit flux related to a total nodal flux in a 2D analysis?
Answer: The total nodal flux should be exactly equal to the unit flux multiplied
by the total length of the element edges that contribute to that node.
In the figure below, a unit flux of -500 g / time / meter edge length has been
applied to the top of the element. The top is a mass sink face which will let the
contaminant out. The flux sections drawn in the element confirm that the total
edge flux of -500 g / time has been converted by the solver into two equal total
nodal fluxes of -250 g / time each. For such a simple mesh, it is also possible to
use the View Node information option and click on each node to see the
computed total flux at each node. The sum of the individual total nodal fluxes is
the total flux across the element edge.

Page 78

CTRAN/W

Chapter 7: Modeling Tips and Tricks

Distance (m)

2.

0
50

0
+0
0e

2.
50
00
e+
00
2

5.0000e+002

0
5

Distance (m)

Figure 7-2 Test to compare unit flux and total flux

7.8

Stopping and restarting an analysis

CTRAN/W has a powerful Stop-Restart feature. The processing may be stopped


at any time step and then restarted after making changes in the problem. For
example, consider a transient problem with 10 time steps, and that you click the
Stop button after the 5th time step is complete. You can then make changes with
DEFINE and re-save the problem.
Typical changes you may want to consider when you stop and restart an analysis
include:

addition or removal of an element as discussed below,

changing adaptive time stepping parameters,

changing convergence parameters,

adjusting material properties to account for time dependent changes,


or
Page 79

Chapter 7: Modeling Tips and Tricks

7.9

CTRAN/W

adding or deleting a boundary condition

Element addition and removal

Any element can be considered to be nonexistent by assigning the conductivity


function number a value of zero. This feature makes it possible to simulate the
construction of embankments and excavations or the deposition of hot waste
material on existing frozen ground where the waste is placed in lifts.
It is necessary at the start of the problem to define the elements you anticipate
adding or removing. Defining the complete mesh at the beginning of an analysis
will maintain compatibility between files. The total number of elements and
number of nodes must not change as you progress from one stage of the analysis
to another. If the number of nodes or elements changes, it is no longer possible to
use previously computed results as initial conditions for a subsequent step
because the files will be incompatible.
The Draw Element Properties command in DEFINE is useful for applying a null
material type to specific elements as you are simulating the construction of
embankments or excavations. A null material type may be defined by using the
Key In Material Properties command to create a material type with a thermal KFunction of zero.
If you add an element using the stop-restart feature, you must consider that the
initial conditions of the element may not be realistic. By default, the program will
consider the newly added elements to have a concentration of zero, which may be
reasonable if the soil is clean but not if it is waste material being added. There is
a roundabout way to correct this problem in the short term and GEO-SLOPE
International Ltd. is working on a better solution for a future release.
In the short term, when the new elements are made active, a one second time step
can be solve where each node in the new elements is set at the correct
concentration using a C boundary condition. After the one second analysis, stop
the solver, remove the C boundary, and continue solving until the next desired
element addition elapsed time or until the completion of the simulation.

Page 80

CTRAN/W

Chapter 8: Product Integration Illustrations

Product Integration Illustrations

GeoStudio is a unique geotechnical engineering tool that has been designed to


allow seamless integration between various types of geotechnical engineering
analyses. It is now possible to define problem geometry one time and then apply
various types of boundary conditions in order to solve various partial differential
equations.
It is possible, for example, to use SEEP/W or VADOSE/W generated pore water
pressures in a SLOPE/W stability or QUAKE/W analysis. It is possible to account
for flowing water in a CTRAN/W contaminant transport process or TEMP/W heat
transfer processes related to natural or artificial ground freezing. It is possible to
specify different pore pressure conditions over time in order to carry out coupled
or uncoupled seepage consolidation and seepage influenced stressdeformation
analyses using SIGMA/W.
This chapter provides several examples that illustrate how GeoStudio facilitates the
smooth integration between various analytical tools. Specific example related
entirely to CTRAN/W are given in the following chapter.
The following examples are included as a sampling of what can be carried out
using GeoStudio.

SEEP/W generated pore pressures in SLOPE/w stability analysis

VADOSE/W generated pore pressures in SLOPE/W stability analysis

SEEP/W dissipation of pore pressures generated in a QUAKE/W earth


quake analysis

SEEP/W velocity data in a CTRAN/W contaminant transport analysis

VADOSE/W climate influenced velocity data in a CTRAN/W


contaminant transport analysis

SEEP/W with CTRAN/W density-dependent flow salt water


intrusion

TEMP/W ground freezing including water flow affects computed from


SEEP/W

SEEP/W seepage dependent embankment settlement coupling with


SIGMA/W

Page 81

Chapter 8 Product Integration Illustrations

8.1

CTRAN/W

Uncoupled consolidation using seepage and pressure data from


SEEP/W, VADOSE/W, QUAKE/W

SEEP/W generated pore-water pressures in SLOPE/W


stability analysis

The objective of this illustration is to observe how to use finite element pore-water
pressure results in a stability analysis. Including or deliberately ignoring negative
pore-water pressures can be critical to understanding and interpreting a slope
stability analysis.
In particular, the objectives of this illustration are to:

Set up and solve a steady-state finite element SEEP/W simulation. In


CONTOUR show the positive pressure heads that develop.

Set up a slope stability problem in SLOPE/W based on the SEEP/W


finite element mesh and computed pore-water pressure; determine the
critical slip surface and the factor of safety using the SEEP/W porewater pressure; and graph the pore-water pressure and strength along
the slip surface.

Repeat the analysis, but remove the advanced parameters from the soil
property information. Graph the pore-water pressure and strength
along the slip surface and note how the negative pore-water pressures
have been ignored).

Repeat the stability analysis, using the piezometric line option in


SLOPE/W to reflect a water table at an elevation of 10 m. In this case
we will not use advanced parameters.

The seepage portion of the analysis is illustrated in Figure 8-1. In general, the slope
is comprised of multiple layers with a finer, lower permeability layer located half
way up the slope face. Note that more coarse soil soils in region 1 and 3 have the
same hydraulic properties (Ksat 1x10-3 m/day). In addition, a steady-state
infiltration rate of q = 5.0x10-5 m/day is applied along the top and the slope with a
pressure equals zero condition on the downstream surface. A potential seepage
review has been applied along the face of the slope.

Page 82

CTRAN/W

Chapter 8: Product Integration Illustrations

26
24
22

Infiltration 5.0e-5 m/day

20

K(sat) = 1e-3 m/day 1

18

Seepage Face

16

K(sat) = 1e-5 m/day

14
12

K(sat) = 1e-3 m/day

Pressure = zero

10

8
6
4
0

10

20

30

40

50

60

Figure 8-1 Seepage problem definition


The soil property information for the SLOPE/W portion of the analysis is given in
Table 8-1.
Table 8-1 Slope soil information
Layer

Unit Weight

Phi

Cohesion

Unit Weight
above WT

Phi B

18

25

18

15

19

20

19

15

20

30

10

20

15

Figure 8-2 shows the location of grid and radius points applied in the stability
analysis, as well as the SEEP/W computed perched water table and the computed
factor of safety. SLOPE/W was able to read the seepage results directly from
SEEP/W in order to compute the actual pore-water pressures at the base of each
slice. Figure 8-3 shows the actual pore-water pressures applied on each slice.
Notice how the pore pressures on the slices change from negative to positive to
negative and back to positive as the slice number increases from left to right. This
type of pore-water pressure condition would not have been possible to accurately
establish without the use of a rigorous saturate-unsaturated seepage flow model.
Note in the SLOPE/W results that the finite element mesh used in the seepage
analysis is superimposed on the solution. The finite element mesh is not actually
used by SLOPE/W, but the data from the mesh at all locations can be used to
determine pore pressures on the base of any slice geometry.

Page 83

Chapter 8 Product Integration Illustrations

CTRAN/W

44
42
11

40
38
36
34
32

1.324

30

2
28

12

13

26
24

3
22
1

20

Infiltration

18
8

16

14

10

12
3

10

14

8
6

4
0

10

20

30

40

50

60

Figure 8-2 SLOPE/W solution showing perched water table from


SEEP/W
Pore-Water Pressure vs. Distance
30
25
20

Pore-Water Pressure

15
10
5

0
-5
-10
-15

-20
0

10

20

30

40

Distance

Figure 8-3 SEEP/W pore pressures at base of slices in SLOPE/W

8.2

VADOSE/W generated pore pressures in SLOPE/W


stability analysis

The objective of this illustration is to observe how to use climate coupled finite
element pore-water pressures results in a stability analysis. You will see how
important including negative pore pressure can be to a slopes stability.

Page 84

CTRAN/W

Chapter 8: Product Integration Illustrations

The VADOSE/W analysis mesh as well as the SLOPE/W grid and radius are
illustrated in Figure 8-1. In general, a large rainfall event was applied over the
undulating ground profile on the first day of a 30 day long transient climate
coupled VADOSE/W analysis. This heavy rainfall created ponded water
conditions in the low point near the center of the mesh. The rainfall day was
followed by 29 days of infiltration from the pond and evaporation through the
climate-ground boundary. Stability results are compared for the day just after the
rainfall (e.g., day 2 from VADOSE/W) as well as after a long drying period (e.g.
day 30). A comparison is also made between using and not using soil strength
parameters that account for increased strength with increased negative water
pressure (e.g. Phi-B).
The soil property information for the SLOPE/W portion of the analysis is given in
Table 8-1. Two stability simulations were carried out: one with a Phi B parameter
defined to account for the effect of soil water suction on strength and one without
Phi B.
Table 8-2 Slope soil information
Layer

Unit Weight

Phi

Cohesion

Unit Weight
above WT

Phi B

18

25

18

15

Figure 8-4 Stability problem definition (pwp from end of day 2 in


VADOSE/W)

Page 85

Chapter 8 Product Integration Illustrations

CTRAN/W

Figure 8-2 shows a minimum factor of safety of 1.16 in the stability analysis based
on advanced pore-water pressure parameters (e.g., soil suction strength effects) as
well as the VADOSE/W computed water. SLOPE/W was able to read the seepage
results directly from VADOSE/W in order to compute the actual pore-water
pressures at the base of each slice.
Figure 8-6 shows the computed factor of safety after the heavy rainfall day based
on soil property strength data that does not include the effect of suction on the
strength of the soil. In this analysis the factor of safety is reduced to less than 1.0.
One final analysis was carried out using advanced strength parameters to
determine the factor of safety after an extended drying period. Figure 8-7 shows a
factor of 1.74 after 29 days of infiltration from the pond (e.g. no rainfall) and
surface evaporation on the ground profile. It is clear that from this example that the
factor of safety is very dependent on ground-climate influences and on the
application of advanced soil strength parameters that take into account increased
strength due to negative water pressure in the soil (e.g. soil water suction).
Finally, Figure 8-8 shows the actual pore-water pressures applied on each slice in
the stability analysis for day 30 of the simulation. Notice how the pore pressures
on the slices change from positive to negative as the slice number increases from
left to right. This type of pore-water pressure condition would not have been
possible to accurately establish without the use of a rigorous saturate-unsaturated
climate coupled seepage flow model.
Note in the SLOPE/W results that the finite element mesh used in the seepage
analysis is superimposed on the solution. The finite element mesh is not actually
used by SLOPE/W, but the data from the mesh at all locations can be used to
determine pore-water pressures on the base of any slice geometry.

Page 86

CTRAN/W

Chapter 8: Product Integration Illustrations

1.159

Figure 8-5 SLOPE/W solution using water pressures from VADOSE/W


(solution including suction effect on strength)

0.999

Figure 8-6 SLOPE/W solution using water pressures from VADOSE/W


(solution NOT including suction effect on strength)

Page 87

Chapter 8 Product Integration Illustrations

CTRAN/W

1.739

Figure 8-7 Factor of safety after long drying period (e.g. day 30 porewater pressure data from VADOSE/W)

Pore-Water Pressure vs.


Distance
Pore-Water Pressure

10
0
-10
-20
-30
-40

Distance

Figure 8-8 Day 30 VADOSE/W pore pressures at base of slices in


SLOPE/W (for simulation with suction effects on strength)

Page 88

CTRAN/W

8.3

Chapter 8: Product Integration Illustrations

SEEP/W dissipation of pore pressures generated in a


QUAKE/W earth quake analysis

QUAKE/W can be used to analyze the transient build up of pore-water pressure


during an earthquake event. Consider the illustration in Figure 8-9 that shows a
phreatic surface in a dam with hydraulic fill on two sides of a core. The horizontal
acceleration earthquake record for this example is taken from the Pacoima dam
earthquake event and is illustrated in Figure 8-10. Note the duration of the
earthquake was only 40 seconds, yet during this time the pore pressures at point
A in Figure 8-11 increased significantly. The computed pore-water pressures at
this point during the entire earthquake are illustrated in Figure 8-12. In general, the
shaking almost doubled the pressure over pre-quake hydrostatic conditions.
1.14

(x 1000)

1.11
1.08
1.05
1.02
0.99
0.96
0.93
0.90
-500

-400

-300

-200

-100

100

200

300

400

500

Distance - feet
Figure 8-9 Dam pore-water pressure prior to earth quake

Page 89

600

Chapter 8 Product Integration Illustrations

CTRAN/W

Pacoima dam earthquake


0.6
0.5
0.4

Acceleration ( g )

0.3
0.2
0.1
0.0
-0.1
-0.2
-0.3
-0.4
-0.5
-0.6
0

10

20

30

40

50

Time (sec)

Figure 8-10 Horizontal acceleration record

Figure 8-11 Pore-water pressure contours after earthquake. Note


point A

Page 90

CTRAN/W

Chapter 8: Product Integration Illustrations

Pore-Water Pressure vs. Time

Pore-Water Pressure

9000

8000

7000

6000

5000
0

10

15

Time

Figure 8-12 Pore-water pressure (psf) build up at point A during quake


The QUAKE/W generated pore-water pressures can be used directly in SEEP/W as
the initial conditions in order to determine the time it takes for the quake generated
pressures to return to hydrostatic conditions. The same pore-water pressures
illustrated above can also be used directly in SLOPE/W in order to determine the
factor of safety versus time throughout the quake event. This is discussed in more
detail in the SLOPE/W engineering book.
Initial pressures to come from QUAKE/W

1.14

(x 1000)

1.11
1.08

Seepage face (Q=0)

Head BC = 1110 ft

1.05

P=0

1.02
0.99
0.96
0.93
0.90
-500

-400

-300

-200

-100

100

200

300

400

500

600

Distance - feet

Figure 8-13 Model set up for SEEP/W pressure dissipation analysis


Figure 8-13 shows the model set up for the SEEP/W analysis that will compute the
dissipation of the excess pore-water pressure generated during the earth quake. The

Page 91

Chapter 8 Product Integration Illustrations

CTRAN/W

boundary conditions for the seepage analysis are illustrated in the figure. In this
analysis it is assumed that the reservoir remains full with a total head of 1110 ft (or
102 ft pressure head). The seepage analysis was carried out to simulate 5 days and
the change in pore-water pressure at point A over time is illustrated in Figure
8-14 below.
Pressure vs. Time
9000

Pressure

8000

7000

6000

5000
0

100000

200000

300000

400000

500000

Time

Figure 8-14 Dissipation of excess pore pressure at point A over 5


days following quake

8.4

SEEP/W velocity data in CTRAN/W contaminant transport


analysis

Seepage data from a steady state or transient SEEP/W analysis can be used in the
GeoStudio CTRAN/W module to predict contaminant transport with or without
adsorption, diffusion and decay. This is a one-dimensional contaminant transport
problem verification example analysis with a free exit boundary. The CTRAN/W
results are compared with closed form analytical and published solutions.
To generate a one-dimensional steady-state flow system, a one-row finite element
mesh is created using SEEP/W DEFINE. The finite element mesh is 1 m high and
40 m long and consists of a total of 30 elements and 62 nodes. The SEEP/W files
containing this example are named EXIT.GSZ.

Page 92

CTRAN/W

Chapter 8: Product Integration Illustrations

The head differential and hydraulic conductivity are selected to produce a constant
seepage velocity U of 0.05 m/s in the positive x-direction. The volumetric water
content is defined as a constant value of 0.5. The average linear velocity is:

=U /
= 0.05 / 0.5
= 0.1 m/s
The dispersivity L is set to 4 m, and the molecular diffusion coefficient D* is set
to zero. The resulting hydrodynamic dispersion coefficient is:

D = Lv + D *
= 4 0.1 + 0.0
= 0.4 m 2 /s
The time step sequence consists of 48 steps. Results are presented for Time Steps
8, 16, 24, 32, 40 and 48 with total elapsed times of 80 s, 160 s, 240 s, 320 s, 400 s
and 480 s respectively. Computed results for these six time steps are included with
the CTRAN/W software.
The boundary condition at the left end of the problem is assumed to be a source
boundary with concentration of the source Cs specified as 1.0 unit/m3. The
boundary condition at the right end is specified as a free exit boundary (Qd > 0).
The initial concentration of the flow system is set to 0.0.
Frind, (1988) has presented an analytical solution to the transport equation with a
free exit boundary. The solution is approximated by the analytical solution for
transport in a semi-infinite medium. The concentration C as a function of x and t is
expressed as:

C =

Cs
2

x - vt
erfc
2 Dt

x - vt v( x +vt )
vx
exp D erfc
1 + D

2 Dt

( x vt ) 2
v t
+
exp

4 Dt
D

where:
C

concentration,

Page 93

Chapter 8 Product Integration Illustrations

Cs

specified concentration of the source,

hydrodynamic dispersion coefficient,

average linear velocity,

elapsed time,

distance from the source boundary, and

erfc

complementary error function.

CTRAN/W

In this verification example, no adsorption and no decay are considered. Figure


8-15 presents the CTRAN/W solutions using the central difference time integration
scheme (contained in the EXIT1 files). A comparison of the CTRAN/W solution
with the analytical solution using a free exit boundary is presented in Figure 8-16.
The Excel comparison indicates that the CTRAN/W solution is in excellent
agreement with the analytical solution. Almost identical solutions are obtained for
all three elapsed times in the first 35 m, while there are only slight differences
between 35 m and 40 m. The differences are likely due to the approximation of the
free exit boundary using a semi-infinite medium in developing the closed form
solution.
Figure 8-17 presents the CTRAN/W solution using the central difference time
integration scheme for the case with a zero dispersive mass flux exit boundary
condition (contained in the EXIT2 files). Figure 8-18 presents a comparison of the
CTRAN/W solution using a free exit boundary, with the CTRAN/W solution using
an exit boundary of zero dispersive mass flux (i.e., Type II boundary). As
expected, the concentration profile in the upstream portion is not sensitive to the
exit boundary conditions. However, there are significant differences in the
concentration profiles near the exit boundary. Since a zero dispersive mass flux
implies zero concentration gradient at the exit boundary, the nodal concentration at
the exit boundary has been forced to be the same as the nodes immediately to the
left of the exit boundary.
Both Figures below show that the concentrations at the entrance boundary are
increasing with time until the concentration is equal to the specified concentration
of the source. This is consistent with the physical relevancy of the entrance
boundary condition when the concentration of the source rather than the
concentration of the boundary nodes are specified.

Page 94

CTRAN/W

Chapter 8: Product Integration Illustrations

1.0
8.0000e+001

0.8
1.6000e+002

0.6

2.4000e+002

C
0.4

3.2000e+002

0.2

4.0000e+002

0.0

4.8000e+002

10

20

30

40

Distance

Figure 8-15 CTRAN/W solution of concentration versus distance at


various times using one free exit boundary

1.0
CTRAN/W Solution
(Exit Qd > 0)

0.9
0.8

Analytical Solution

Concentration

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

10

20

30

40

Distance

Figure 8-16 Comparison with analytical solutions for one exit


boundary condition

Page 95

Chapter 8 Product Integration Illustrations

CTRAN/W

1.0
8.0000e+001

0.8
1.6000e+002

0.6

2.4000e+002

C
0.4

3.2000e+002

0.2

4.0000e+002

0.0

4.8000e+002

10

20

30

40

Distance

Figure 8-17 CTRAN/W solution with two zero dispersive mass flux exit
boundary conditions
1.0
CTRAN/W Solution
(Exit Qd > 0)

0.9
0.8

CTRAN/W Solution
(Exit Qd = 0)

Concentration

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

10

20

30

40

Distance

Figure 8-18 Comparison with analytical solution for two exit boundary
conditions

Page 96

CTRAN/W

8.5

Chapter 8: Product Integration Illustrations

VADOSE/W velocity data in CTRAN/W contaminant


transport analysis

Seepage velocity and water content data from a steady state or transient
VADOSE/W analysis can be used in CTRAN/W to predict contaminant transport
with or without adsorption, diffusion and decay. The purpose of this example is to
show the influence of including climate (or vegetation) effects on the movement of
contaminants in soils.
In this example a simple advection dispersion analysis is carried out in
CTRAN/W after solving a transient VADOSE/W analysis. The CTRAN/W
program reads the transient seepage velocity and water content data at different
time steps and uses it in the solution of the advection dispersion equation.
The problem is kept quite simple for illustration purposes. In the seepage analysis,
a pressure equal to zero boundary condition is applied at the base of the low point
near the center of the mesh. This pressure boundary condition is saying there is a
small source of water at this location just enough to keep the ground saturated. It
is assumed that a source of contaminant is also present with the water at the ground
surface.
Figure 8-19 shows the position of the water table and the concentration contours
for the case where the model only allowed infiltration at the source of contaminant.
For this case, there is no evaporation demand along the rest of the ground surface
and therefore the contaminant is limited in its migration to a somewhat radial
pattern beneath the source.

Figure 8-19 Concentration contours without ground-climate coupling

Page 97

Chapter 8 Product Integration Illustrations

CTRAN/W

In Figure 8-20, the same concentration is applied in the pond location along with
the same source of water, however, the climate boundary condition is active at all
other ground surface nodes. The climate boundary in this case is removing about
5 mm per day of evaporation.
It is clear for the case with evaporation, that there is a lot more spreading of the
contaminant. The results show that the contaminant is being pulled up towards the
drying ground surface where it is either exiting with the vapor flow or being
deposited or both depending on the exit boundary condition specified in the
contaminant analysis.

Figure 8-20 Concentration contours with ground-climate coupling

8.6

Density-dependent flow salt water intrusion

Henry (1964), developed an analytic solution for a simplified sea water intrusion
problem. The Henry problem has since become a benchmark verification
example for many numerical models of density-dependent flow. However,
Croucher and OSullivan, (1995), noted that none of the published numerical
model comparisons with Henrys solution that they examined were able to match
Henrys solution to a great extent. In addition to outlining the possible reasons for
the discrepancies, Croucher and OSullivan, presented a new, highly accurate
numerical solution to the problem. Their numerical solution is used here for
comparison with the results from CTRAN/W.
The system being modeled is shown in Figure 8-21. It consists of a 2.0m long
section of a 1.0m thick aquifer where the right boundary is in direct contact with

Page 98

CTRAN/W

Chapter 8: Product Integration Illustrations

sea water and the left boundary has a constant influx of freshwater. The sea water
has a relative density, (specific gravity), of 1.025 at a reference concentration of
1.0. The concentration of sea water is fixed at 1.0 along the sea water boundary
and a fixed freshwater inflow rate of 6.6x10-5 m3/s is specified along the freshwater
boundary. The top and bottom boundaries are both impermeable. The aquifer is
homogeneous and isotropic and has a saturated hydraulic conductivity Ksat =
1x10-2 m/s, a porosity n = 0.35 and a velocity independent dispersion coefficient of
D = 1.89x10-5 m2/s. The aquifer is discretized using 0.05m square elements and the
solution is sought at steady state. The SEEP/W example file containing this
problem is called HENRY.GSZ, and the full definition of the problem may be
viewed using SEEP/W and CTRAN/W module DEFINE views.
Henry's Problem for Seawater Intrusion
Impermeable Top Boundary
1.0
0.9
0.8
0.7

Left Boundary
Freshwater
Qf = 6.6E-5 m2/s
C=0.0

Right Boundary
Seawater
Hs = 1.0m
C = 1.0

0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

Impermeable Bottom Boundary

Aquifer Properties
Ks = 1E-2 m/s
n = 0.35
D = 1.89E-5 m2/s

Seawater Density
SG=1.025 @ Cref=1.0

Axes lengths in metres.


Element thickness = 1.0m

Figure 8-21 Henry's problem definition


The above Henrys problem has been analyzed with CTRAN/W, and a steady state
solution is obtained at an elapsed time of more than 11,000 seconds (at time step
35). Figure 8-22 shows the computed seawater concentration contours at steady
state along with the water flow velocity vectors. At steady state, seawater enters
the aquifer across the lower portion of the sea water boundary via density induced
gradients and mixes with freshwater flowing in the opposite direction. The
constant influx of freshwater from the left freshwater boundary causes the diluted
seawater to exit the system across the upper portion of the sea water boundary. In
this way, a seawater flow cell is established in which the seawater toe migration

Page 99

Chapter 8 Product Integration Illustrations

CTRAN/W

towards the freshwater boundary is controlled by the rate of freshwater flow, the
density of the seawater, and the degree of mixing between the seawater and
freshwater. The degree of mixing is controlled by the dispersion coefficient used in
the modelling, which in this case is velocity independent.
It should be noted that in Figure 8-22 near the upper left of the aquifer, CTRAN/W
computed a few small negative concentration values. This slight numerical
oscillation is a direct result of the Peclet and Courant numbers being exceeded in
these areas because of the relatively high water velocity and relatively coarse mesh
and time step discretization. It is possible to eliminate the negative concentration
by reducing both the mesh size and the time step size, however, since we are more
interested in the solution in the lower portion of the flow system and we only use
the 0.5 concentration contours in the comparison, refinement to the finite element
mesh and time steps were deemed unnecessary in this case.
Freshwater
C=0

Seawater
C=1

1.00

0.75

0.50

0.
8

0.
6

0.
4

2
0.

0.25

0.00
0.00

0.25

0.50

0.75

1.00

1.25

1.50

1.75

2.00

Figure 8-22 Henry's problem verification - computed concentration


contours
Comparison of the CTRAN/W computed results with the highly accurate
numerical solution of Croucher and OSullivan is given in Figure 8-23. The figure
compares the 0.5 seawater concentration isochlors at steady state. It can be seen
that the results from CTRAN/W are almost identical to those of Croucher and
OSullivan.

Page 100

CTRAN/W

Chapter 8: Product Integration Illustrations

1
0.9

Croucher and O'Sullivan, 1995

0.8

CTRAN/W

0.7

Elevation (m)

0.6
0.5
0.4
0.3
0.2
0.1
0
1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

X-Coordinate (m)

Figure 8-23 Comparison of 0.5 concentration isochlors

8.7

Ground freezing and water flows (SEEP/W and TEMP/W)

The GeoStudio module TEMP/W can be used to model natural and artificial
ground freezing without or with the addition of heat added by flowing water.
Consider the frozen ground that forms around a buried pipe with a pipe wall
boundary condition of -2 degrees Celsius. Figure 8-24 shows the ground
temperature profile after 2 years of freezing. It is clear from this figure that a
region of frozen ground has formed around the pipe and that the ground freezes to
a deeper depth beneath the pipe where there is less influence from the warmer
yearly average ground surface temperature. Details of this analysis and comparison
of the results for this case to other published data can be found in the TEMP/W
engineering book. The SEEP/W example file containing this problem is called
PIPE WITH FLOWING WATER.GSZ, and the full definition of the problem
may be viewed using SEEP/W and TEMP/W module DEFINE views.

Page 101

Chapter 8 Product Integration Illustrations

CTRAN/W

1.6
2.5

1.4

Pipe

1.2
1.0
0.8
0.6
0.4
0.2
0.0

-0.2
-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Figure 8-24 Temperature contours around freezing pipe without


flowing water
Figure 8-25 shows the temperature profile for same analysis except in this case the
influence of flowing water is considered. The SEEP/W module was set up such
that a hydraulic gradient of 0.27% was established across the region from right to
left with a saturated hydraulic conductivity of 0.1 m/day. It is clear from this figure
that the extents of the frozen ground are far less than for the static water condition
and that the shape of the frozen ground is more strongly influenced by the flowing
water than the warmer air temperature.
Figure 8-26 shows the actual computed water velocity vectors and total head
contours across the region after two years of analysis. This figure illustrates clearly
the diversion of water around the frozen ground region. Careful examination of the
figure shows that the velocity of water increases as it moves around the frozen
ground. This is because the cross sectional area available for flow is reduced due to
ice formation. For cases where several freezing pipes are installed to create a
frozen barrier wall, the increase in velocity of water between adjacent freeze pipes
can result in a situation where more heat is added by the water than can be
removed by the freeze pipes. When this occurs, closure of the frozen wall (due to
adjacent frozen regions growing together) will not happen. This effect can be seen
in Figure 8-27 for a typical shaft sinking artificial ground freezing project.

Page 102

CTRAN/W

Chapter 8: Product Integration Illustrations

1.6
2.5

1.4
1.5

1.2

Pipe

1.0
0.8
0.6
0.4
0.2
0.0
-0.2
-0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Figure 8-25 Temperature contours around freezing pipe with flowing


water
When carrying out this type of analysis, it is very important to have well-defined
material properties and to use a fine mesh discretization and small time steps. The
seepage velocity generated by SEEP/W and added into the TEMP/W finite element
partial differential equation is a linear term, and as such, the computed results are
quite sensitive to numerically appropriate element size and time steps. A first guess
at time step sizes can be made by computing the Courant number and ensuring that
it is less than a value of 1. The Courant number is given by:

C=

vt
x

where:

the Darcian velocity,

the time step, and

the element edge length.

Page 103

Chapter 8 Product Integration Illustrations

0.4

0.6

1.6045

0.2

1.6035

1.6025

5
1.601

1.6005

1.6
1.5
1.4
1.3
1.2
1.1 Pipe
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
-0.2 0.0

CTRAN/W

0.8

1.0

1.2

1.4

1.6

Figure 8-26 Water flow vectors and head contours around frozen pipe

Figure 8-27 Non closing "window" in shaft ground freezing


It is suggested when modeling convective heat transfer to use the adaptive time
stepping routine and set the maximum time step by first setting the Courant
number to 1 in the above equation and computing the permissible time step to
ensure numerical stability.

Page 104

CTRAN/W

Chapter 8: Product Integration Illustrations

Finally, while this example shows how water can influence freezing in ground, the
same combination of TEMP/W and SEEP/W can be used to study the influence of
freezing ground on the transient seepage of water when the parts of the ground are
subjected to cold temperatures. This might be the case for dam performance in
climates with both summer and winter seasons.
In both cases, you must set up a complete SEEP/W and TEMP/W analysis and
then start the solver process from the TEMP/W program. TEMP/W will launch
SEEP/W, pass it the ground temperatures and wait for SEEP/W to pass back water
content and computed seepage velocities. In the event that SEEP/W is passed a
ground temperature that is below the freezing point of water, it will compute a
reduced hydraulic conductivity corresponding to the magnitude of the temperature
below the phase change point. The calculation of the frozen ground conductivity is
discussed in the Material Properties chapter of this book.

8.8

Seepage-dependent embankment settlement (SEEP/W


and SIGMA/W)

The construction of an embankment on slightly over-consolidated clay is a


relatively complex modeling problem that requires a modeling approach
combining stress and seepage analyses.
In the following example, the subsurface clay stratum is 9 m deep and the water
table is 1 m below the ground surface. The 1m layer above the water table is highly
weathered, desiccated and fissured, making its behavior similar to a fine granular
soil. The embankment to be analyzed is shown in Figure 8-28 and is constructed of
relatively sandy soil. The height of the embankment is 5 m (El. 14) with 3:1 side
slopes and a 10 m crest width. Due to symmetry about the center, only half of the
cross section is used in the analysis.
The objective of this example is to determine the embankment settlement. Full
details of this example problem are given in the technical paper a351 available
on the GEO-SLOPE International Ltd. website.

Page 105

Chapter 8 Product Integration Illustrations

CTRAN/W

15
14

13

12

Elevation - metres

11
10
9
8
7
6
5
4
3
2
1
0
0

10

15

20

25

30

35

40

metres

Figure 8-28 Embankment problem configuration


The clay will be modeled as a Modified Cam-Clay (MCC) soil. It is highly plastic
with a liquid limit (LL) of 61%. A summary of all clay properties is given in Table
8-3. The embankment and upper meter of subsoil will be modeled as highly
permeable linear-elastic material, which is considered adequate since the major
settlement issue arises in the underlying compressible clay. The properties adopted
for the sand are:
E

2000 kPa

0.36

1.0 m/day

mv

3 x 10-4 (1/kPa)

The k and mv values are actually arbitrary for the sand, since steady-state porewater pressure conditions will be specified, meaning there is no change with time.
In a transient (consolidation) analysis, the software requires these values be
specified, but the actual values do not affect the results.

Page 106

CTRAN/W

Chapter 8: Product Integration Illustrations

Table 8-3 Summary of clay material properties


Property

Value

Cc (compression index)

0.46

(slope of normal consolidation line)

0.20

(slope of swelling line)

0.04

mv (coefficient of volume compressibility)

3 x 10-4 (1/kPa)

k (hydraulic conductivity)

5 x 10-3 m/day

(effective friction angle)

26 degrees

(slope of critical state line)

1.0

Ko (coefficient of earth pressure at rest)

0.56

(Poissons Ratio)

0.36

OCR (over consolidation ratio)

1.2

(specific volume)

2.2

The analysis will be run for 40 days. Figure 8-29 shows the embankment with the
numbers in each row of elements indicating at what time (day) the elements
become active. The clay will consolidate during the time between each lift
placement and for 15 days after the last lift is placed.
+25 +25 +25 +25 +25 +25
+25 +25
+19 +19 +19 +19 +19 +19 +19 +19 +19
+19 +19
+13 +13 +13 +13 +13 +13 +13 +13 +13 +13 +13 +13
+13 +13
+7

+7

+7

+7

+7

+7

+7

+7

+7

+7

+7

+7

+7

+7

+7

+7 +7

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1

+1 +1

Figure 8-29 Fill placement sequence specified in SIGMA/W


Important parameters and conditions to specify in SEEP/W include the coefficient
of volume compressibility mv , the hydraulic conductivity and the hydraulic
boundary conditions.
In SEEP/W the slope of the volumetric water content function is mv . In this
example, consolidation of the clay will occur under saturated conditions so the

Page 107

Chapter 8 Product Integration Illustrations

CTRAN/W

volumetric water content function can be defined as a straight line with two points
resulting in a line with a slope of 3 x 10-4 (1/kPa). The slope is (90 x 0.001)/300 =
0.0003.
The hydraulic conductivity of the clay will be constant and equal to the saturated
hydraulic conductivity, defined by a horizontal function at 5 x 10-3 m/day.
The pore-water pressure along the water table is zero and the total head is therefore
equal to the elevation. As a result, the boundary condition is defined as H (P=0).
Changes in pore-water pressure are not considered within the upper meter of clay
or the embankment fill. As a result, deformations in these zones due to changes in
pore-water pressure are not computed. It is important to note that pore-water
pressure changes within these zones are not allowed. Setting the boundary
conditions in these zones to be H (P=0) as illustrated in Figure 8-30, ensures this to
be the case. While the actual pore-water pressures may not be correct, the effect
has been removed, which is an objective of this analysis.
All the other boundary conditions can be left undefined, resulting in a zero flow
boundary at the vertical ends of the clay located below the water table and along
the bottom.
Figure 8-31 shows the pore-water pressure changes with time at a specified
location one meter to the right of the centerline and two meters below the water
table. The pore-water pressure immediately rises after the construction of a lift and
then falls until the next lift is placed. After the final lift has been placed, the porewater pressure at the specified location drops from a high of 90 kPa to about 55
kPa. Significant excess pore-water remains at the end of 40 days, indicating that
additional long-term settlement will occur.

Page 108

CTRAN/W

Chapter 8: Product Integration Illustrations

Figure 8-30 SEEP/W boundary conditions


Figure 8-32shows the settlement profiles along the original ground surface on days
0, 7, 13, 19, and 25. Of particular interest is the location of maximum settlement,
which does not occur along the centerline, but is located more toward the outer
edge of the fill in the early stages of loading. This result is due to the zone of lower
effective horizontal stresses as shown in Figure 8-33. In addition, the toe of the fill
has been lifted up slightly.

Pore-Water Pressure vs. Time

Pore-Water Pressure

110

90

70

50

30

10
0

10

20

30

40

Time

Figure 8-31 Pore-water pressure changes beneath embankment


versus time

Page 109

Chapter 8 Product Integration Illustrations

CTRAN/W

Y-Displacement vs. X
0.2
0.0000e+000

Y-Displacement

0.0
7.0000e+000

-0.2
1.3000e+001

-0.4
1.9000e+001

-0.6

2.5000e+001

-0.8
0

10

15

20

25

30

Figure 8-32 Ground surface settlement profiles at various times

25

30

25
30

Figure 8-33 Horizontal effective stress after second lift placed

8.9

Uncoupled consolidation

SIGMA/W can be used with SEEP/W, VADOSE/W, QUAKE/W and other


SIGMA/W projects in order to computed uncoupled consolidation. In this type of
analysis, the change in pore-water pressure at each time step is obtained from the

Page 110

CTRAN/W

Chapter 8: Product Integration Illustrations

external analysis and used in the solution of the consolidation equations inside
SIGMA/W.
More details of this type of analyses will be incorporated into this book in the next
edition. Specific questions can be addressed to support@geo-slope.com in the
interim.

Page 111

Chapter 8 Product Integration Illustrations

Page 112

CTRAN/W

CTRAN/W

Chapter 9: Illustrative Examples

Illustrative Examples

This chapter presents a variety of verification and typical real life example
problems analyzed with CTRAN/W. The real life examples are not case studies
as you would find them in a journal, but they can be used as a starting point for
setting up your own models.
The verification examples compare CTRAN /W solutions with benchmark
references that show that the software is functioning properly. The real life
examples help to illustrate how some of the unique features in CTRAN /W are
applied to practical situations.
A point to keep in mind while reviewing these examples is that they are set up
and solved using techniques and parameters that help illustrate the use of the
software in applying the theory. It is possible, likely, and recommended, that you
will need to adjust your unique parameters in order to fine tune your own
analysis and gain confidence in the results and your understanding of the
processes being modeled.

9.1

One-dimensional transport analysis

Included files

Verify.gsz

The first verification example problem is a one-dimensional contaminant


transport problem. This example problem can be used to test the various
fundamental aspects of the general transport equation implemented in
CTRAN/W. The following four cases are analyzed:

Case 1 : No adsorption and no decay

Case 2 : With adsorption

Case 3 : With decay

Case 4 : With adsorption and decay

Closed form analytical solutions are available in the literature for simple
problems involving steady-state seepage flow. Each transport case is first solved
numerically with CTRAN/W and then analytically with published closed form
solutions. Equations for the closed form solutions have been solved using
Page 113

Chapter 9: Illustrative Examples

CTRAN/W

Microsoft Excel spreadsheets and are presented in each case. Excel contains the
complementary error function, which greatly simplifies the computation process.
Comparing the CTRAN/W results with the analytical solutions is a means of
verifying the software.
Problem definition
The first step in the verification test is to run a steady-state seepage analysis. To
simulated the one-dimensional transport problem, a one-row finite element mesh
is created. The finite element mesh is 60 mm high and 3000 mm long and
consists of a total of 60 elements and 303 nodes.
The head differential and hydraulic conductivity are selected to produce a
constant seepage velocity U of 0.1 mm/s in the positive x-direction. The
volumetric water content is defined as a constant 0.5. The average linear
velocity is:

=U /
= 0.1/ 0.5
= 0.2 mm/s
The dispersivity is set to 100 mm, and the molecular diffusion coefficient D*
is set to zero. The resulting hydrodynamic dispersion coefficient is:

D = Lv + D *
= 100 0.2 + 0.0
= 20 mm/s
The time step sequence consists of 60 steps. Results are presented for Time Steps
20, 40, and 60 with total elapsed times of 2000 s, 4000 s, and 6000 s. Computed
results for these three time steps are included with the CTRAN/W software.
The boundary condition at the left end of the problem is specified as a constant
concentration boundary with C=1.0 unit/mm3. The boundary condition at the
right end is also set to a constant concentration boundary with C=0.0. The initial
concentration of the flow system is set to 0.0.

Page 114

CTRAN/W

Chapter 9: Illustrative Examples

Case 1 Solution: no adsorption and no decay


Case 1 is the analysis of the basic transport problem with no adsorption and no
decay. In other words, contaminant migration is only governed by the advection
and dispersion actions. The analytical solution to the basic transport equation
with no adsorption and no decay is provided by Ogata, 1970, as follows:

C =

C0
x - vt
x + vt
vx
+ exp erfc
erfc

2
D
2 Dt
2 Dt

where:
C

concentration,

Co

specified concentration at the source boundary,

hydrodynamic dispersion coefficient,

average linear velocity,

elapsed time,

distance from the source boundary, and

erfc

complementary error function.

Figure 9-1 presents the CTRAN/W solutions using the backward difference time
integration scheme. Figure 9-2 compares the CTRAN/W results with the
analytical results for the three elapsed times. The CTRAN/W solution compares
closely with the analytical solution. The analytical solution has a slightly steeper
curve than CTRAN/W solution. In other words, the CTRAN/W solution is
slightly smeared over the flow system. This phenomenon is related to numerical
dispersion, which is inherent in the finite element solution of the transport
equation.
Figure 9-3 presents the CTRAN/W solution using central difference time
integration scheme. A comparison of the CTRAN/W solutions with the analytical
results is presented in Figure 9-4. There is excellent agreement between the
CTRAN/W solution and the analytical solution. A detailed examination of the
CTRAN/W solution at early time steps using the CTRAN/W CONTOUR
function indicates that some of the computed nodal concentration is slightly
higher than the specified concentration at the boundary nodes. For example, the
maximum nodal concentration specified at the source boundary is 1.0 units.

Page 115

Chapter 9: Illustrative Examples

CTRAN/W

However, at Node 4, which is 5 mm away from the source boundary, the


computed nodal concentration at an elapsed time of 6000 s is 1.00096 units.
The phenomenon of having the computed concentration larger than the specified
maximum concentration or smaller than the specified minimum concentration is
related to numerical oscillation. This is also inherent in the finite element
solution of the transport equation.
Numerical dispersion and oscillation can only be minimized, not eliminated. In
general, numerical dispersion is more pronounced with the backward difference
time integration scheme, while numerical oscillation is more pronounced with the
central difference scheme. Techniques for minimizing numerical dispersion and
oscillation are presented in the chapter on Numerical Issues.

1.0

0.8

2.0000e+003

0.6
C
4.0000e+003

0.4

0.2

0.0

6.0000e+003

500

1000

1500

2000

2500

3000

Distance

Figure 9-1 CTRAN/W solution using backward difference time


integration
Page 116

CTRAN/W

Chapter 9: Illustrative Examples

Concentration

1.0
0.9

CTRAN/W Solution
(Backward Difference)

0.8

Analytical Solution

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

500 1000 1500 2000 2500 3000


Distance

Figure 9-2 Comparison of CTRAN/W solution with analytical solution


1.5

2.0000e+003

1.0
C
4.0000e+003

0.5

6.0000e+003

0.0

500

1000

1500

2000

2500

3000

Distance

Figure 9-3 CTRAN/W solution using central difference time


integration

Page 117

Chapter 9: Illustrative Examples

CTRAN/W

Concentration

1.0
0.9

CTRAN/W Solution
(Central Difference)

0.8

Analytical Solution

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

500 1000 1500 2000 2500 3000


Distance

Figure 9-4 Comparison of CTRAN/W solution and analytical solution


Case 2 Solution: with adsorption
Case 2 is the analysis of the transport problem in which contaminant may be
adsorbed to the soil particles. Adsorption in the transport process retards the
contaminant migration process. The effect of adsorption is expressed in a term
generally referred to as the retardation factor R, and is written as:

R = 1 + d K d
n

where:
Kd

distribution coefficient (slope of a linear adsorption relationship),

bulk (dry) mass density, and

porosity.

In the case of no adsorption, Kd is zero and R is equal to 1.

Page 118

CTRAN/W

Chapter 9: Illustrative Examples

In order to handle unsaturated soil and a non-linear adsorption function, the


equivalent retardation factor Re in the CTRAN/W formulation is derived as:

Re = + d
C

In other words, when there is no adsorption, the slope of the adsorption function
is zero and Re is equal to the volumetric water content of the soil. For
= 0.5 , the equivalent retardation factor Re is equivalent to 1/2 of the
retardation factor R. Therefore, to compare CTRAN/W results with an analytical
solution when R is equal to 2.0, Re must be equal to 1.0. This is accomplished by
setting d to 0.5 units/mm3 and defining the adsorption function as a straight line
with a slope of 1.0.
The analytical solution to the transport equation with adsorption is published by
Marsily (1986), as follows:

C =

x - (U/R ) t
x + (U/R ) t
Ux
C0
erfc
erfc
+ exp

2 D t/R
2 D t/R
2
Dd

d
d

where:
C

concentration,

Co

specified concentration at the source boundary,

Dd

hydrodynamic dispersion coefficient using Darcian velocity,

Darcian velocity (specific discharge),

volumetric water content,

retardation factor,

elapsed time,

distance from the source boundary, and

erfc

complementary error function.

Figure 9-5 presents the CTRAN/W solution using the central difference time
integration scheme with the retardation factor R equal to 2.0 (contained in the
verify2.gsz file). Figure 9-6 compares the CTRAN/W solution to the analytical
Page 119

Chapter 9: Illustrative Examples

CTRAN/W

solution. Figure 9-6 shows that the CTRAN/W solution is in excellent agreement
with the comparable analytical solution. Almost identical solutions are obtained
for all three elapsed times.
The reciprocal of R is known as the relative velocity of the mid-point
concentration. Described in equation form: (see Freeze and Cherry, 1979)

V
1
= r
R
V
where:
Vr

the velocity at the mid-point concentration with adsorption, and

the velocity at the mid-point concentration with no adsorption.

A retardation R value of 2.0 means that the relative velocity of the mid-point
concentration is 0.5. In other words, the case with retardation will migrate half
the distance of the case without retardation for the same elapsed time.
Figure 9-7 compares the CTRAN/W result with adsorption to the CTRAN/W
result without adsorption for three elapsed times. The solution indicates that the
transport process with adsorption is retarded by a factor of 2. For example, the
solution with adsorption at elapsed time of 4000 s is almost identical to the
solution with no adsorption at elapsed time of 2000 s. This observation is
consistent with the physical relevance of the retardation factor, R.

Page 120

CTRAN/W

Chapter 9: Illustrative Examples

1.0

0.8

2.0000e+003

0.6
C
4.0000e+003

0.4

0.2
6.0000e+003

0.0

500

1000

1500

2000

2500

3000

Distance

Figure 9-5 CTRAN/W solution with adsorption

Concentration

1.0
0.9

CTRAN/W Solution
(With Adsorption)

0.8

Analytical Solution

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

500 1000 1500 2000 2500 3000


Distance

Figure 9-6 Comparison of CTRAN/W and analytical solutions with


adsorption

Page 121

Chapter 9: Illustrative Examples

CTRAN/W

Concentration

1.0
0.9

CTRAN/W Solution
(With Adsorption)

0.8

CTRAN/W Solution
(Without Adsorption)

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

500 1000 1500 2000 2500 3000


Distance

Figure 9-7 Comparison of CTRAN/W solutions with and without


adsorption
Case 3 Solution: with decay
Case 3 is the analysis of the transport problem in which no adsorption takes
place, but mass may be lost due to radioactive decay. The analytical solution to
the transport equation with adsorption and no decay is published by Marsily
1986, as follows:

Ux
C0
C =
exp
exp (- x) erfc
2
2 Dd

Ux
C0
+
exp
exp ( x) erfc
2
2 Dd

x - t (U/) 2 + 4 D /
d

2 Dd /

x + t (U/) 2 + 4 D /
d

2 Dd /

where:

Page 122

CTRAN/W

Chapter 9: Illustrative Examples

U
=
+
2 Dd Dd
C

concentration,

Co

specified concentration at the source boundary,

Dd

hydrodynamic dispersion coefficient using Darcian velocity,

Darcian velocity (specific discharge),

volumetric water content,

coefficient of decay,

elapsed time,

distance from the source boundary, and

erfc

complementary error function.

In this verification example, a radioactive contaminant is assumed to have a


decay half-life T of 6931.5 s. This makes the coefficient of decay equal to
1.0e-04 s-1 (i.e., ln2/T). Figure 9-8 presents the CTRAN/W solutions using the
central difference time integration scheme (contained in the verify3.gsz). The
Excel comparison of CTRAN/W solution with the analytical solution for the case
with decay is presented in Figure 9-9. Figure 9-9 shows that the CTRAN/W
solution is in excellent agreement with the comparable analytical solution.
Almost identical solutions are obtained for all three elapsed times.
Figure 9-10 provides a comparison of CTRAN/W solutions between the case
with decay and the case with no decay. Since the mass lost due to decay is a
function of the mass present (concentration). The CTRAN/W results correctly
indicate that the largest effect of decay occurs near the source boundary where
the concentration is the highest. As the concentration approaches zero, the decay
component has no effect, since there is essentially no mass to decay.
Furthermore, the mass lost due to decay is also dependent on the elapsed time.
The longer the elapsed time, the greater the mass lost. This is correctly reflected
in Figure 9-10 by the difference in the curves with decay and the curves with no
decay.

Page 123

Chapter 9: Illustrative Examples

CTRAN/W

1.0

0.8

2.0000e+003

0.6
C
4.0000e+003

0.4

0.2

0.0

6.0000e+003

500

1000

1500

2000

2500

3000

Distance

Figure 9-8 CTRAN/W solution with decay

Concentration

1.0
0.9

CTRAN/W Solution
(With Decay)

0.8

Analytical Solution

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

500 1000 1500 2000 2500 3000


Distance

Figure 9-9 Comparison of CTRAN/W with analytical solutions


Page 124

CTRAN/W

Chapter 9: Illustrative Examples

Concentration

1.0
0.9

CTRAN/W Solution
(With Decay)

0.8

CTRAN/W Solution
(Without Decay)

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

500 1000 1500 2000 2500 3000


Distance

Figure 9-10 Comparison of solutions with and without decay


Case 4 Solution: with adsorption and decay
Case 4 includes the effect of both adsorption and radioactive decay in the
analysis of the transport problem. The analytical solution to the transport
equation with adsorption and decay is published by Marsily (1986) as follows:

C =

Ux
C0
exp
exp (- x) erfc
2
2 Dd

Ux
C0
exp
exp ( x) erfc
2
2 Dd

x - t (U/R) 2 + 4 D / R
d

2 Dd / R

x + t (U/R) 2 + 4 D / R
d

2 Dd / R

U R
=
+
Dd
2 Dd
Page 125

Chapter 9: Illustrative Examples

CTRAN/W

where:
C

concentration,

Co

specified concentration at the source boundary,

Dd

hydrodynamic dispersion coefficient using Darcian velocity,

Darcian velocity (specific discharge),

volumetric water content,

coefficient of decay,

retardation factor,

elapsed time,

distance from the source boundary, and

erfc

complementary error function.

Solutions for Case 4 are obtained using a retardation factor R of 2.0 and a decay
half-life T of 6931.5 s. Figure 9-11 presents the CTRAN/W solutions using the
central difference time integration scheme (contained in the verify4.gsz file). A
comparison of the CTRAN/W solution with the analytical solution for the case
with adsorption and decay is also presented in Figure 9-12. As is indicated in the
figure, excellent agreement is observed between the CTRAN/W solution and the
analytical solution. In fact, the two solutions are almost identical for all three
elapsed times.

Page 126

CTRAN/W

Chapter 9: Illustrative Examples

1.0

0.8

2.0000e+003

0.6
C
4.0000e+003

0.4

0.2
6.0000e+003

0.0

500

1000

1500

2000

2500

3000

Distance

Figure 9-11 CTRAN/W solution with adsorption and decay


1.0
CTRAN/W Solution (With
Adsorption & Decay)

0.9

Analytical Solution

Concentration

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

500 1000 1500 2000 2500 3000


Distance

Figure 9-12 Comparison of CTRAN/W and analytical solutions with


adsorption and decay
Page 127

Chapter 9: Illustrative Examples

CTRAN/W

Figure 9-13 provides a comparison between the CTRAN/W solution for the case
with adsorption and decay and the solution for the basic case with no adsorption
and no decay. Figure 9-13 illustrates the correct combined effect of adsorption
and decay to the concentration profile. Not only is mass lost due to radioactive
decay, but the contaminant migration process is also retarded due to adsorption.
1.0
CTRAN/W Solution
(With Adsorption &
Decay)

0.9

Concentration

0.8

CTRAN/W Solution
(Without Adsorption &
Decay)

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

500 1000 1500 2000 2500 3000


Distance

Figure 9-13 Comparison of solutions with and without adsorption


and decay

9.2

Transport analysis with a free exit boundary

Included files

Exit1.gsz

Exit2.gsz

The second verification example problem is a one-dimensional contaminant


transport problem with a free exit boundary. The CTRAN/W results are
compared with closed form analytical and published solutions.

Page 128

CTRAN/W

Chapter 9: Illustrative Examples

Problem definition
To generate a one-dimensional steady-state flow system, a one-row finite
element mesh is created. The finite element mesh is 1 m high and 40 m long and
consists of a total of 30 elements and 62 nodes.
The head differential and hydraulic conductivity are selected to produce a
constant seepage velocity U of 0.05 m/s in the positive x-direction. The
volumetric water content is defined as a constant value of 0.5. The average
linear velocity is:

=U /
= 0.05 / 0.5
= 0.1 m/s
The dispersivity L is set to 4 m, and the molecular diffusion coefficient D* is
set to zero. The resulting hydrodynamic dispersion coefficient is:

D = Lv + D *
= 4 0.1 + 0.0
= 0.4 m 2 /s
The time step sequence consists of 48 steps. Results are presented for Time Steps
8, 16, 24, 32, 40 and 48 with total elapsed times of 80 s, 160 s, 240 s, 320 s, 400 s
and 480 s respectively. Computed results for these six time steps are included
with the CTRAN/W software.
The boundary condition at the left end of the problem is assumed to be a source
boundary with concentration of the source Cs specified as 1.0 unit/m3. The
boundary condition at the right end is specified as a free exit boundary (Qd > 0).
The initial concentration of the flow system is set to 0.0.
Solution
Frind (1988) has presented an analytical solution to the transport equation with a
free exit boundary. The solution is approximated by the analytical solution for
transport in a semi-infinite medium. The concentration C as a function of x and t
is expressed as:

Page 129

Chapter 9: Illustrative Examples

C =

Cs
2

x - vt
erfc
2 Dt

CTRAN/W

x - vt v( x +vt )
vx
exp D erfc
1 + D

2 Dt

( x vt ) 2
v t
exp

4 Dt
D

where:
C

concentration,

Cs

specified concentration of the source,

hydrodynamic dispersion coefficient,

average linear velocity,

elapsed time,

distance from the source boundary, and

erfc

complementary error function.

In this verification example, no adsorption and no decay are considered. Figure


9-14 presents the CTRAN/W solutions using the central difference time
integration scheme (contained in the exit1.gsz files). A comparison of the
CTRAN/W solution with the analytical solution using a free exit boundary is
presented in Figure 9-15. The Excel comparison indicates that the CTRAN/W
solution is in excellent agreement with the analytical solution. Almost identical
solutions are obtained for all three elapsed times in the first 35 m, while there are
only slight differences between 35 m and 40 m. The differences are likely due to
the approximation of the free exit boundary using a semi-infinite medium in
developing the closed form solution.
Figure 9-16 presents the CTRAN/W solution using the central difference time
integration scheme for the case with a zero dispersive mass flux exit boundary
condition (contained in the exit2.gsz files). Figure 9-17 presents a comparison of
the CTRAN/W solution using a free exit boundary with the CTRAN/W solution
using a exit boundary of zero dispersive mass flux (i.e., Type II boundary). As
expected, the concentration profile in the upstream portion is not sensitive to the
exit boundary conditions. However, there are significant differences in the
concentration profiles near the exit boundary. Since a zero dispersive mass flux
implies zero concentration gradient at the exit boundary, the nodal concentration
Page 130

CTRAN/W

Chapter 9: Illustrative Examples

at the exit boundary has been forced to be the same as the nodes immediately to
the left of the exit boundary.
Both sets of figures show that the concentrations at the entrance boundary are
increasing with time until the concentration is equal to the specified
concentration of the source. This is consistent with the physical relevancy of the
entrance boundary condition when the concentration of the source rather than the
concentration of the boundary nodes are specified.

1.0
8.0000e+001

0.8
1.6000e+002

0.6

2.4000e+002

C
0.4

3.2000e+002

0.2

4.0000e+002

0.0

4.8000e+002

10

20

30

40

Distance

Figure 9-14 CTRAN/W solution with free exit boundary

Page 131

Chapter 9: Illustrative Examples

CTRAN/W

1.0
CTRAN/W Solution
(Exit Qd > 0)

0.9
0.8

Analytical Solution

Concentration

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

10

20

30

40

Distance

Figure 9-15 Comparison of CTRAN/W and analytical solutions with


free exit boundary
1.0
8.0000e+001

0.8
1.6000e+002

0.6

2.4000e+002

C
0.4

3.2000e+002

0.2

4.0000e+002

0.0

4.8000e+002

10

20

30

40

Distance

Figure 9-16 CTRAN/W solution with zero dispersive mass flux exit
BC

Page 132

CTRAN/W

Chapter 9: Illustrative Examples

1.0
CTRAN/W Solution
(Exit Qd > 0)

0.9
0.8

CTRAN/W Solution
(Exit Qd = 0)

Concentration

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

10

20

30

40

Distance

Figure 9-17 Comparison of CTRAN/W and analytical solutions

9.3

Density-dependent transport analysis

To verify the density-dependent transport analysis computations, three example


problems are presented. The first example problem models a static, (no flow),
saltwater column, the second models one-dimensional horizontal saltwater flow,
and the third models Henrys problem for sea water intrusion into a coastal
aquifer. Although the first two examples are simplistic, they are included to show
that the computations do give intuitively correct results. Henrys problem for sea
water intrusion, while also somewhat simplistic, is the classic verification
example for numerical solutions of density-dependent flow problems.
Included files

Saltcol.gsz

Saltflow.gsz

Henry.gsz

Page 133

Chapter 9: Illustrative Examples

CTRAN/W

Static saltwater column


The static saltwater example verifies that SEEP/W computations of head and
fluid flux are correct. This example also illustrates how the equivalent freshwater
head varies with elevation when a constant concentration of saltwater is imposed
on the column.
Problem definition
Physically, the system consists of a 1.0m column of soil saturated with saltwater
having a relative density, (specific gravity), of 1.025. The column is discretized
using 20, 5cm square elements arranged in a column to a height of 1.0 m. The
reference concentration and relative density were specified as 1.0 and 1.025,
respectively. The boundary conditions at the bottom and top of the column were
specified as a total saltwater head of 1.0 m. The relative concentration of entire
column was specified to be 1.0 for every time step,
No initial conditions for head or concentration were specified. The analysis was
run using 5 time steps of 100 seconds each, and the results were viewed at the
final time step.
Solution
The variation of equivalent freshwater head along the saltwater column in this
example may be deduced intuitively. At the top of the column the elevation is
1.0 m and the saltwater pressure head is 0.0 m. At the bottom of the column the
elevation is 0.0 m and the saltwater pressure head is 1.0 m. Conversion of the
saltwater pressure head to the equivalent freshwater pressure head is performed
by multiplying the saltwater pressure head by the relative density of the
saltwater. Thus, the equivalent fresh water total head can be computed as:
Equation 9-1

H f = (H s Z )

s
+Z
f

where:
Hf

equivalent freshwater total head,

Hs

saltwater total head,

elevation, and

Page 134

CTRAN/W

s
f

Chapter 9: Illustrative Examples

relative density of saltwater (i.e., 1.025 in this example).

The interpretation of the seepage solution in a density-dependent flow problem is


somewhat different than the usual interpretation in a non-density-dependent flow
problem. In a density-dependent flow problem, the total head computed in
SEEP/W is actually the equivalent fresh water total head which includes the
body force due to the relative density of saltwater. Figure 9-18 shows the
equivalent freshwater head computed by SEEP/W along the column. The
equivalent freshwater head increase linearly from 1.0 m at the top to 1.025 m at
the bottom which is consistent with the prescribed boundary condition and the
equivalent freshwater total head represented by the Equation 9-1.

Total Head vs. Y


1.0

0.8

0.6

0.4

0.2

0.0
0.90

0.95

1.00

1.05

Total Head

Figure 9-18 Equivalent freshwater head along a static saltwater


column

Page 135

Chapter 9: Illustrative Examples

CTRAN/W

Although there is an upward gradient in the equivalent fresh water head within
the column, the upward gradient is counter balanced by the downward gradient
due to the body force. As a result, there is no net flow in the column.
One-dimensional horizontal saltwater flow
This problem verifies that heads and flow velocities are correctly computed by
SEEP/W for density-dependent problems. This problem also illustrates that the
flow direction may not be perpendicular to the equipotential lines of equivalent
freshwater head in a density-dependent flow problem.
Problem definition
The dimensions of the conduit are 1.0 m long by 0.1 m high by 1.0 m deep and
the conduit contains a porous medium with a saturated hydraulic conductivity of
1.0 m/s. The reference concentration and relative density at the reference
concentration are 1.0 and 1.025, respectively. Saltwater heads of 0.2 m and 0.1 m
are specified at the left and right boundaries. A relative concentration of 1.0 is
specified throughout the conduit at all time. The analysis was run using 3 time
steps of 100 seconds each, and the results were viewed at the final time step. No
initial conditions for head or concentration were specified.
Solution
As with the previous density-dependent verification example, the variation of the
equivalent freshwater head along the saltwater conduit may be deduced
intuitively. Figure 9-19 shows the saltwater head boundary conditions specified
at the end boundaries and the change in saltwater head along all four sides of the
conduit. Figure 9-20 shows the same boundary conditions in terms of equivalent
freshwater head. The values of equivalent freshwater head were calculated using
Equation 9-1. At points A through D, the equivalent freshwater heads are
HfA = 0.2025m, HfB = 0.2050, HfC = 0.1025 and HfD = 0.1000 with the head
changing linearly between any of these four points. Inspection of Figure 9-20
shows a vertical gradient of equivalent freshwater head of 0.025 m/m throughout
the conduit. This vertical gradient in equivalent freshwater head arises because
the body force of the saltwater.

Page 136

CTRAN/W

Chapter 9: Illustrative Examples

Figure 9-19 Saltwater head boundary conditions

Figure 9-20 Equivalent freshwater head boundary condition


The horizontal flow velocity, (Darcian velocity), may be calculated using:

vx = K s i fx =

K s H fx
Lx

where:

horizontal Darcian velocity,

Ks

saturated hydraulic conductivity,

ifx

gradient of equivalent freshwater head in horizontal direction,

H fx =

change in equivalent freshwater head in horizontal direction, and

Lx

horizontal length.

Page 137

Chapter 9: Illustrative Examples

CTRAN/W

From Figure 9-20 the change in equivalent freshwater head in the horizontal
direction across the conduit at any elevation is, H fx = 0.1025 m/m , and given
that Ks = 1.0 m/s and Lx = 1.0 m, then vx = 0.1025 m/s .
The results calculated by SEEP/W SOLVE are shown in Figure 9-21. As
expected the flow is completely horizontal with no vertical component and the
equivalent freshwater heads are identical to those calculated by hand. The fluid
flux is 0.01025 m3/s through a cross-sectional area of 0.1 m2, therefore the
Darcian velocity is 0.1025 m/s.

25
0.105

0.2

0.20

0
0.0

25

(x 0.001)

0.10

1.0250e-002
100

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Contours of equivalent freshwater head in metres.


Flux section indicates fluid flux in m3/s.
Axes scales are in metres.

Figure 9-21 One dimensional saltwater flow


Henry's problem for sea water intrusion
Henry, 1964, developed an analytic solution for a simplified sea water intrusion
problem. The Henry problem has since become a benchmark verification
example for many numerical models of density-dependent flow. However,
Croucher and OSullivan, 1995, noted that none of the published numerical
model comparisons with Henrys solution that they examined were able to match
Henrys solution to a great extent. In addition to outlining the possible reasons
for the discrepancies, Croucher and OSullivan, 1995, presented a new, highly
accurate numerical solution to the problem. Their numerical solution is used here
for comparison with the results from CTRAN/W.
Problem definition
The system being modeled is shown in Figure 9-22. It consists of a 2.0 m long
section of a 1.0 m thick aquifer where the right boundary is in direct contact with
sea water and the left boundary has a constant influx of freshwater. The sea water
has a relative density, (specific gravity), of 1.025 at a reference concentration of
1.0. The concentration of sea water is fixed at 1.0 along the sea water boundary
and a fixed freshwater inflow rate of 6.6X10-5 m3/s is specified along the
freshwater boundary. The top and bottom boundaries are both impermeable. The
Page 138

CTRAN/W

Chapter 9: Illustrative Examples

aquifer is homogeneous and isotropic and has a saturated hydraulic conductivity


Ks = 10-2 m/s, a porosity n = 0.35 and a velocity independent dispersion
coefficient of D = 1.89X10-5 m2/s. The aquifer is discretized using 0.05 m square
elements and the solution is sought at steady state.
Henry's Problem for Seawater Intrusion
Impermeable Top Boundary
1.0
0.9
0.8
0.7

Left Boundary
Freshwater
Qf = 6.6E-5 m2/s
C=0.0

Right Boundary
Seawater
Hs = 1.0m
C = 1.0

0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

Impermeable Bottom Boundary

Aquifer Properties
Ks = 1E-2 m/s
n = 0.35
D = 1.89E-5 m2/s

Seawater Density
SG=1.025 @ Cref=1.0

Axes lengths in metres.


Element thickness = 1.0m

Figure 9-22 Henry's problem definition


Solution
The above Henrys problem has been analyzed with CTRAN/W, and a steady
state solution is obtained at an elapsed time of more than 11,000 seconds (at time
step 35). Figure 9-23 shows the computed sea water concentration contours at
steady state along with the water flow velocity vectors. At steady state, sea water
enters the aquifer across the lower portion of the sea water boundary via density
induced gradients and mixes with freshwater flowing in the opposite direction.
The constant influx of freshwater from the left freshwater boundary causes the
diluted sea water to exit the system across the upper portion of the sea water
boundary. In this way, a sea water flow cell is established in which the sea
water toe migration towards the freshwater boundary is controlled by the rate of
freshwater flow, the density of the sea water, and the degree of mixing between
the sea water and freshwater. The degree of mixing is controlled by the
dispersion coefficient used in the modeling, which in this case is velocity
independent.
Page 139

Chapter 9: Illustrative Examples

CTRAN/W

It should be noted that in Figure 9-24 near the upper left of the aquifer,
CTRAN/W computed a few small negative concentration values. This slight
numerical oscillation is a direct result of the Peclet and Courant numbers being
exceeded in these areas because of the relatively high water velocity and
relatively coarse mesh and time step discretization. It is possible to eliminate the
negative concentration by reducing both the mesh size and the time step size,
however, since we are more interested in the solution in the lower portion of the
flow system and we only use the 0.5 concentration contours in the comparison,
refinement to the finite element mesh and time steps were deemed unnecessary in
this case.

1.0
0.9
0.8
0.7
0.6
0.5

0.

0.4

0 .5

0.

0.3

0.7

0.2

0.0
0.0

0.9

0.1

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2.0

Figure 9-23 Henry's problem computed concentration contours


Comparison of the CTRAN/W computed results with the highly accurate
numerical solution of Croucher and OSullivan (1995), is given in Figure 9-24.
The figure compares the 0.5 sea water concentration isochlors at steady state. It
can be seen that the results from CTRAN/W are almost identical to those of
Croucher and OSullivan.

Page 140

CTRAN/W

Chapter 9: Illustrative Examples

1.2
Croucher and O'Sullivan, 1995
1
CTRAN/W

Elevation (m)

0.8

0.6

0.4

0.2

0
1

1.2

1.4

1.6

1.8

2.2

X-Coordinate (m )

Figure 9-24 Comparison of Croucher and CTRAB/W concentration


solutions

Page 141

Chapter 9: Illustrative Examples

CTRAN/W

Page 142

CTRAN/W

10

Chapter 10: Theory

Theory

This chapter presents the methods, equations, procedures, and techniques used in
the CTRAN/W formulation. It is of value to be familiar with this information in
order to use the software. An understanding of these concepts will be of great
benefit in applying the software, resolving difficulties, and judging the
acceptability of the results.

10.1

Flow velocity

SEEP/W and VADOSE/W compute the specific discharge or DArcy velocity.


The specific discharge is the total flux Q divided by the full cross-sectional area
(voids and solids alike).
The actual cross-sectional area available for the water is less than the full area,
due to the presence of the solids. Consequently, the actual rate of movement of
the water is higher than the DArcy velocity. By definition, the porosity is the
volume of voids divided by the total volume. The cross-sectional area available
for the water flow is the porosity times the total cross-sectional area.
Therefore the average linear velocity of the pore fluid is: (see Freeze and Cherry,
1979)

v=

Q
nA

or,

v=

U
n

where:
U

specific discharge, or Darcian velocity, and

porosity.

Under saturated conditions, the volumetric water content ( ) is equal to the


porosity. The average linear velocity then is also equal to:

Page 143

Chapter 10: Theory

v=

CTRAN/W

CTRAN/W is formulated on the assumption that the average linear velocity can
be related to the volumetric water content for both saturated and unsaturated
conditions. SEEP/W, being a saturated/unsaturated flow model, computes the
volumetric water content. The SEEP/W specific discharge, (DArcy velocity),
divided by the SEEP/W volumetric water content is taken in the CTRAN/W
formulation as the average linear velocity.

10.2

Governing equations

The governing equation for contaminant transport is generally known as the


transport equation or the advection-dispersion equation.
For ease of presentation and discussion, the governing differential equation is
presented only in terms of one-dimensional transport. The actual CTRAN/W
formulation is based on the corresponding two-dimensional equation.
The solute transport equation can be derived by considering the mass flux q in an
elemental volume of porous material, as illustrated in Figure 10-1. The absolute
net mass flux across the element is:
net mass flux =

q
dx
x

Figure 10-1 Mass balance in a one dimensional element


To conserve mass, the time rate of change of the total mass M in the element
must be equal to the net mass flux. In equation form,

Page 144

CTRAN/W

Chapter 10: Theory

M
q
dx = dx
t
x

Equation 10-1

By definition, the concentration C is the mass M of dissolved solute in a unit


volume of water (solution). In equation form,

C=

M
Vw

or,

M = CVw
The volume of water per unit volume of the element is the volumetric water
content . The mass M per total unit volume then is:

M = C
Substituting for M in Equation 10-1 and dividing by dx leads to:

Equation 10-2

C
q
=t
x

The mass flux through the element arises from both advection and dispersion
processes. In equation form, these two mechanisms are:
advection = vC = UC
and,
dispersion = D

C
x

where:

average linear velocity,

volumetric water content,

concentration,

hydrodynamic dispersion coefficient, and

Page 145

Chapter 10: Theory

CTRAN/W

DArcy velocity (specific discharge).

The negative sign in the dispersion term indicates that the direction of the mass
flow is from a high concentration to a low concentration (that is, a negative
gradient). Substituting the previous two terms into Equation 10-2 leads to the
basic transport equation:

Equation 10-3

C

C

= + U C
- D
t
x
x

2
C
C
= D 2 U
x
x

This equation can be divided by , which leads to:

C
2C
C
= D
-v
2
t
x
x
where v is the average linear velocity. This is the form of the equation often seen
in publications. (Freeze and Cherry, 1979). CTRAN/W uses the more general
form as shown in Equation 10-3.
The hydrodynamic dispersion coefficient D is related to the dispersivity, average
linear velocity and diffusion coefficient D* by:

D = v + D *
For further discussion regarding this coefficient, refer to the section on the
Hydrodynamic Dispersion Matrix later in this chapter.
Equation 10-3 represents the basic transport of a non-reactive and nonradioactive substance; that is, there is no loss of mass due to adsorption or
radioactive decay. For general formulation, the loss of mass due to adsorption
and radioactive decay must be added to the transport equation.
For the transport of a radioactive substance, mass may be lost during the
transport process due to radioactive decay of ions in the pore fluid and decay of
ions attached to the soil particles. The reduced concentration resulting from
radioactive decay, in terms of the initial concentration, is:

Page 146

CTRAN/W

Equation 10-4

Chapter 10: Theory

C = C0 e t

where t is the elapsed time and is the decay coefficient. The decay coefficient
can be related to the half-life T of a decaying material. By definition, the halflife T is the elapsed time when the concentration of C/C0 = 1/2. Therefore,

C
1
=
= e - T
C0
2
which can also be written as:

ln 2
0.693
=
T
T

Differentiating Equation 2-1 with respect to time leads to:

C
= - C
t
The amount of radioactive mass in the pore-water Mw in an elemental unit
volume is C , (see above), or:

M w = C = C0 e t
The adsorption S is the amount of mass attached to the soil particles divided by
the mass of the solids. In equation form,

S=

mass of solute attached to the solids


mass of the solids

The mass per unit volume of the soil (solid) particles can be defined in terms of
the bulk (dry) mass density d of the soil. The parameter S is then defined as:

S =

Ms

or,

Page 147

Chapter 10: Theory

CTRAN/W

M = S d

Equation 10-5

where Ms is the amount of mass attached to a unit mass of soil particles.


In terms of radioactive decay,

M s = S d = d S0 e t
Therefore, the total radioactive mass M in both the fluid and solid phases is:

M = Mw + Ms
= C0 e t + d S0 e t
The rate of change of mass due to decay is then:

M
= -C -S d
t
The transport equation (Equation 10-3) can now be modified to include
radioactive decay. The result is:
Equation 10-6

C
2C
C
= D
-U
- C - S d
2
t
x
x

For the transport of a reactive substance, the movement of the mass is also
affected by the adsorption of the solute by the soil particles. As discussed above,
the amount of mass adsorbed can be defined in terms of the mass density of the
soil particles. From Equation 10-5, the adsorbed mass Ms is:

M s = S d
The rate of change of the adsorbed mass is:

Ms
S
= d
t
t
The adsorption S is a function of concentration C. Experimental results are
usually plotted as S vs. C, as shown in Figure 10-2. The slope of the S vs. C
relationship is S / C . In the case of a linear relationship, the slope is usually
Page 148

CTRAN/W

Chapter 10: Theory

referred to as the distribution coefficient Kd. The previous equation can then be
written as:

Ms
S C
= d
t
C t
Adding the adsorption term to Equation 10-6 gives the following governing
differential equation used in CTRAN/W:

C
S C
2C
C
+ d
= D
-U
- C - S d
2
t
C t
x
x

or,

Equation 10-7 + d

S C
2C
C
=
-U
- C - S d
D

2
C t
x
x

Figure 10-2 Relationship between adsorption and concentration

Page 149

Chapter 10: Theory

10.3

CTRAN/W

Finite element equations

CTRAN/W uses the same techniques as the seepage solution to develop the finite
element equation. Details regarding the interpolating function, the function
derivatives, the numerical integration scheme, and the implementation of the
infinite elements are all presented in the SEEP/W Engineering Methodology
book.
Using a Galerkin approach to deriving the finite element equation, the residual
function for the governing differential equation is:

R (c) = D

2C
C
S C

-U
- C - S d - + d
2
x
x
C t

The concentration anywhere in the element can be expressed in terms of the


nodal concentration as:
C = <N> {C}

where:
<N>

a matrix of interpolating (or shape) functions, and

{C}

the vector of nodal concentrations.

Substituting this into the prior equation leads to,

2 <N >{C}
<N >{C}
-U
- <N >{C} - S d
2
x
x
S < N >C

- + d
C
t

R (c) =D

The residual can now be multiplied by a weighting function <N>T, then


integrated over the volume, and then set to zero in order to minimize the residual.
The resulting equation is:

0 = D <N >T
v

2 <N >
dv {C}
x2

Page 150

CTRAN/W

Chapter 10: Theory

<N >
dv {C}
x

U <N >T
v

<N >
v

<N > dv {C}

S <N >T dv

T
+ d
<N > <N > dv {C},t
v
C

Applying Greens theorem, the first term can be integrated by parts to give:

2 <N>
T <N>
v D <N > x 2 dv = s D <N > n ds
<N>T <N>
dv
v D
x
x
T

where n is a direction unit vector normal to the surface.


Substituting this into the first term in the previous equation results in:

0 = D <N >T
s

<N >
<N >T <N >
ds {C} - D
dv {C}
v
n
x
x

U <N >T
v

<N >
dv {C}
x

<N >T <N > dv {C}

S <N >T dv

T
+ d
<N > <N > dv {C},t
v
C

Page 151

Chapter 10: Theory

CTRAN/W

The derivatives of the interpolating functions are designated as the gradient


matrix [B]. Substituting [B] for the gradient terms and rearranging the terms
leads to:
Equation 10-8

[B ]T [D ][B ] dv {C} (Term 1)

+ <N >T [U ][B ] dv {C}

(Term 2)

<N >T <N > dv {C}

(Term 3)

T
+ + d
<N > <N > dv {C},t
v
C

= D <N >T
s

<N >
ds {C}
n

S <N >T dv

(Term 4)

(Term 5)
(Term 6)

The surface integral on the right hand side of Term 5 represents the dispersive
mass flux across the boundary, and is the natural boundary condition. However,
there may also be advective mass flux across the boundary which will be equal to
the specific discharge at the boundary Ub times the concentration.
A more practical boundary condition is the specification of the total mass flux
across the boundary (i.e., the Cauchy type or the Third type boundary condition).
This total mass flux q will be the sum of the dispersive flux and the advective
flux across the boundary. In equation form,

q = D

C
+ U bC
n

Expressing the element concentration C in terms of the nodal concentration {C}


and the shape function <N>: (See above)
Equation 10-9 q = D

<N >{C}
+ U b <N >{C}
n

Page 152

CTRAN/W

The term,

Chapter 10: Theory

<N >{C}
signifies that the flow is normal (n) to the boundary.
n

Equation 10-7 can be rearranged as:

<N >{C}
= q - U b <N >{C}
n

Therefore, Term 5 of Equation 10-6 can be expressed as:

D <N >T

<N >
ds {C} =
n

q <N >T ds

U b <N >T <N > ds {C}

Substituting this in Equation 10-6 leads to the general finite element equation
used in the CTRAN/W formulation:
Equation 10-10

[B ]T [D ][B ] dv {C}

+ <N >T [U ][B ] dv {C}

(Term 2)

(Term 1)

<N >T <N > dv {C}

(Term 3)

T
+ + d
<N > <N > dv {C},t
v
C

(Term 4)

+ U b <N >T <N > ds {C}

(Term 5)

= q <N >T ds

(Term 6)

S <N >T dv

(Term 7)

In abbreviated form, the finite element equation can be written as:

Page 153

Chapter 10: Theory

CTRAN/W

[ K1 ]{C} + [ K 2 ]{C} , t = {Q}{K1}{C} + {K 2 }{C} , t = Q


where:
{K1}

the element characteristic matrix,

the Dispersive Mass Flux (Term 1),


+ the Advective Mass Flux (Term 2),
+ the fluid phase Decayed Mass Flux (Term 3),
+ the boundary Advective Mass Flux (Term 5),

{K2}

{Q}

the element capacitance (storage) matrix,

the Stored Mass Flux (Term 4),

the mass flux entering or leaving the element,

the specified Nodal Total Mass Flux (Term 6), and


+ the solid phase Decayed Mass Flux (Term 7).

10.4

Temporal integration

The finite element solution for the transport equation is a function of time as
indicated by the {C},t term in the finite element equation. The time integration
can be performed by a finite difference approximation scheme. CTRAN/W uses
the following general equation for the time integration:

(t [ K ] + [ K ]) {C } = (t {Q } + (1 ) t {Q })
1

+ ([ K 2 ] (1 ) t [ K1 ]) {C0 }
where:

time increment,

a ratio between 0 and 1,

{C1}

nodal concentration at end of time increment,

{C0}

nodal concentration at start of time increment,

{Q1}

nodal mass flux at end of time increment,


Page 154

CTRAN/W

Chapter 10: Theory

{Q0}

nodal mass flux at start of time increment,

{K1}

element characteristic matrix, and

{K2}

element capacitance (storage) matrix.

CTRAN/W allows you to select between the Backward Difference


Approximation Method (BDA) with being set to 1, and the Central Difference
Approximation Method (CDA) with being set to 0.5. For more information
about which time integration method to use, see Backward or Central Difference
Time in the chapter on Numerical Issues.
As indicated by the above equation, in order to solve for the nodal concentration
at the end of the time increment, it is necessary to know the nodal concentration
at the start of the time increment. Stated in general, the initial conditions must be
known for solving the transport equation.
When the initial conditions are not specified, CTRAN/W assumes the initial
nodal concentration as zero.

10.5

Hydrodynamic dispersion matrix

For one-dimensional flow, the hydrodynamic dispersion coefficient D is defined


above as:

D = v + D *
where:

dispersivity (material property),

Darcian velocity divided by volumetric water content (U/), and

D*

coefficient of molecular diffusion.

Dispersion in the direction of the water flow is usually higher than dispersion
perpendicular to the flow direction. Two dispersivity values are therefore
required to define the spreading process. Dispersivities in the flow directions are
designated as the longitudinal dispersivity L and the transverse dispersivity T .

Page 155

Chapter 10: Theory

CTRAN/W

For two-dimensional flow, as used in the CTRAN/W formulation, the


hydrodynamic dispersion coefficient D can be defined in matrix form: (See Bear,
1979)

D11
D
21

D12
D22

where:
2

v
v2
D11 = L x + T y + D*
v
v
2
v
v2
D22 = T x + L y + D *
v
v
vx v y
D12 = D21 = ( L T )
v
v = vx2 + v y2
In general, the coefficient of diffusion D* is a function of the volumetric water
content, as shown in Figure 10-3. An empirical relationship between D* and
has been proposed by Kemper and Van Schaik, (1966) CTRAN/W allows you to
define the desired values of the coefficient of diffusion function as a function of
volumetric water content (i.e., the diffusion function).
The D* parameter and its dependence on water content is of significance only in
unsaturated flow and when the water flow rate is very low. The value of the
hydrodynamic dispersion coefficient is often governed by the water flow rate.
Therefore, it is often adequate to assume that D* as independent of , and to
define the relationship by a constant horizontal function for a CTRAN/W
analysis.

Page 156

CTRAN/W

Chapter 10: Theory

Figure 10-3 Relationship between D* and water content

10.6

Mass flux

CTRAN/W can compute the total mass flux across a user-specified section. The
mass flux across a section is composed of four components; the dispersive mass
flux Qdis, the advective mass flux Qadv, the stored mass flux Qsto, and the
decayed mass flux Qdec. The total mass flux Q across a section is the sum of all
four components, as represented by the following equation:

Qtotal = Qdis + Qadv + Qsto + Qdec


10.7

Dispersive mass flux

The dispersive mass flux is computed in the same way as SEEP/W computes the
water flux across a section. For full details see the Theory chapter of the SEEP/W
Engineering Methodology book.
As described in the SEEP/W book, consider a mesh with only one element

Page 157

Chapter 10: Theory

CTRAN/W

Figure 10-4 Illustration of mass flux section


The finite element equations of the dispersive mass flux term for one element can
be expressed as follows:

d11
d
21
d31

d 41

d12
d 22
d32

d13
d 23
d33

d 42

d 43

d14 C1 Qdis 1

d 24 C2 Qdis 2
=

d34 C3 Qdis 3

d 44 C4 Qdis 4

The d coefficients in the above equation are a representation of the dispersion


properties of the element. Therefore, the dispersive mass flux from Node i to
Node j is:

Qdis ij = dij Ci d ji C j
The total dispersive mass flux through the flux section shown in Figure 10-4 is:

Qdis = Qdis 21 + Qdis 24 + Qdis 31 + Qdis 34


The imaginary flow lines from one side of the section to the other side are known
as subsections. CTRAN/W identifies all subsections across a user-defined flux
Page 158

CTRAN/W

Chapter 10: Theory

section, computes the mass flux along each subsection, and then sums the
subsection values to obtain the total across the mass flux section.

10.8

Advective mass flux

Both the advective mass flux term and the boundary advective mass flux term are
used in the advective mass flux calculations.
The finite element equations of the sum of the advective mass flux term and the
boundary advective mass flux term for one element can be expressed as follows:

a11
a
21
a31

a41

a12

a13

a22
a32
a42

a23
a33
a43

a14 C1 Qadv 1

a24 C2 Qadv 2
=

a34 C3 Qadv 3

a44 C4 Qadv 4

The a coefficients in the above equation are a representation of the advective


properties of the element. Therefore, the advective mass flux from Node i to
Node j is:

Qadv ij = aij Ci a ji C j
The total advective mass flux through the flux section shown in Figure 10-4 is:

Qadv = Qadv 21 + Qadv 24 + Qadv 31 + Qadv 34


10.9

Stored mass flux

The stored mass flux is computed in the same way as SEEP/W computes the
stored flux across a section.
The finite element equations of the stored mass flux term for one element can be
expressed as follows:

Page 159

Chapter 10: Theory

s11
s
1 21
t s31

s41

s12
s22
s32
s42

s13
s23
s33
s43

CTRAN/W

s14 C1 Qsto 1

s24 C2 Qsto 2

s34 C3 Qsto 3

s44 C4 Qsto 4

where C1 , C2, , C3 and C4 are the changes of concentration at the various


nodes between the start and the end of a time step. In general, the average change
of concentration from Node i to Node j can be expressed as:

Cij =

Ci + C j

Therefore, the stored mass flux from Node i to Node j due to a change in storage
is:

Qsto ij = sij

Cij
t

The total stored mass flux through the flux section shown in Figure 10-4 is:

Qsto = Qsto 21 + Qsto 24 + Qsto 31 + Qsto 34


10.10 Decayed mass flux
Both the fluid-phase decayed mass flux term and the solid-phase decayed mass
flux term are used in the decayed mass flux calculations.
For the fluid phase, the finite element equations of the decayed mass flux term
for one element can be expressed as follows:

df11

df 21
df 31

df 41

df12

df13

df 22
df32
df 42

df 23
df33
df 43

df14 C1 Q fdec 1

df 24 C2 Q fdec 2
=

df 34 C3 Q fdec 3

df 44 C4 Q fdec 4

Page 160

CTRAN/W

Chapter 10: Theory

The df coefficients in the above equation are a representation of the fluid phase
decay properties of the element. Therefore, the fluid phase decay mass flux
calculated at the mid-point of Node i and Node j is:

Q fdec ij = dfij

(Ci + C j )
2

As presented in the governing equation, the solid-phase decayed mass flux term
is expressed in the form of nodal mass flux to be subtracted from the specified
nodal total mass flux term. The nodal mass flux due to decay in the solid phase,
Qsdec can be evaluated by the numerical integration of the solid phase decayed
mass flux term.
Therefore, the solid-phase decayed mass flux calculated at the mid-point of
Node i and Node j is:

Qsdec ij =

Qsdec i + Qsdec

and, the total decayed mass flux at the mid-point of Node i and Node j is:

Qdec ij = Q fdec ij + Qsdec ij


The total decayed mass flux through the flux section shown in Figure 10-4 is:

Qdec = Qdec 21 + Qdec 24 + Qdec 31 + Qdec 34


10.11 Mass quantity calculation
CTRAN/W computes the mass quantity in both the fluid phase, Mf, and the solid
phase, Ms of an element by numerical integration of the following equations:

M f = <N >T <N > dv {C} and


v

M s = d <N >T <N > dv {S }


v

In performing the numerical integration of the above equations, CTRAN/W


computes the concentration at the Gauss point based on the nodal concentration
and the interpolation function. The Gauss point concentration is then used to
Page 161

Chapter 10: Theory

CTRAN/W

compute the Gauss point adsorption S from the adsorption function. The
volumetric water content at each Gauss point is obtained from the material
file generated by SEEP/W or VADOSE/W.
CTRAN/W computes the mass quantity at each Gauss point within an element.
The Gauss point mass quantity is calculated based on a weighted distribution of
the element mass quantity to the Gauss points. For the fluid phase, the
distribution is weighted according to the product of *C at each Gauss point.
Similarly, for the solid phase, the distribution is weighted according to the
adsorption S at each Gauss point.
The mass quantity of each element is then summed to give the total mass quantity
in the flow problem.

10.12 Exit boundaries


A unique solution to the transport equation requires the specification of boundary
conditions around the entire flow boundaries. In general, a boundary may be
specified as a known concentration boundary or a known mass flux boundary. In
certain cases, however, neither the concentration nor the mass flux is known, and,
consequently, special considerations are required. This is the case at a boundary
where mass leaves the flow system (i.e., an exit boundary).
Two approaches are available for dealing with exit boundaries. One is to ignore
the dispersive portion of the mass flux and account only for the advective part of
the flux (Qd=0). A second, more advanced alternative is to account for both
dispersive and advective mass flux at the exit boundary (Qd>0).
Equation 10-10 includes the term,

D <N >T

<N >
ds {C}
n

This term represents the dispersive mass flux across a flow boundary and is the
natural boundary term derived as a result of the finite element formulation of the
transport equation. The simplest option is to set this natural boundary term to
zero at the exit boundary (Qd=0). This approach is sometimes referred to as a
second type boundary condition. Setting the dispersive mass flux to zero has the
effect of forcing the concentration gradient to zero at the exit boundary.

Page 162

CTRAN/W

Chapter 10: Theory

Frind (1988), has proposed a method whereby the natural boundary term is
solved and both the dispersive and advective mass flux are accounted for at the
exit boundary (Qd>0). This procedure results in what is referred to as a free exit
boundary. CTRAN/W uses this method in solving for the natural boundary term.
With a free exit boundary, the nodes along the exit boundary do not behave like
ordinary boundary nodes. Rather, they behave like all other internal nodes,
making it possible to compute the concentration at these nodes without assuming
a certain concentration of mass flux to the nodes.
The free exit approach does not force the concentration gradient to zero, as with
the second type (i.e. zero dispersive mass flux) alternative. The effect of the two
methods is illustrated in Figure 10-5. Note that the solutions from the two
methods are the same except near the exit boundary that is specified on the right
hand side of the flow problem.
1.0
CTRAN/W Solution
(Exit Qd > 0)

0.9
0.8

CTRAN/W Solution
(Exit Qd = 0)

Concentration

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0

10

20

30

40

Distance

Figure 10-5 Comparison of concentration profiles computed with


CTRAN/W for two exit boundary conditions

Page 163

Chapter 10: Theory

CTRAN/W

10.13 Particle tracking


It is difficult to solve the advection-dispersion finite element equation when the
flow velocity is high and the dispersivity is low. In the limit when the diffusion
coefficient D* and the dispersivity parameters L and T are set to zero, it is
impossible to solve the advection-dispersion equation with the CTRAN/W
general finite element formulation.
That said, CTRAN/W includes a separate particle tracking capability to analyze
purely advective transport problems. Particles are introduced at the entrance or
source boundaries. Particles are assumed to be attached to the water and to move
in the direction of the water flow with the same speed as the water flow. The
particle tracking feature can be used to determine where a particle of contaminant
may end up and approximately how long it may take for a particle to arrive at a
new position. A locus of the particle positions at all time steps depicts the
migration path of the particle. It is also a means of producing a graphical
representation of the migration of contaminant in a flow system with no
dispersion, no adsorption and no decay.
The particle tracking process is conceptually simple and straightforward. At the
start of the process, CTRAN/W locates the host element of each particle, then
computes the average linear velocities vx and vy of each particle using the
interpolating (or shape) functions evaluated at the coordinates of each particle.
The size of the host element is also determined. The projection time step
increment is determined based on the courant number criterion
For each projection time step increment, the change in position of each particle,
x and y is then calculated according to the following equation:

x = t vx
y = t v y
where:

projection time step increment,

average linear velocity in the x direction, and

average linear velocity in the y direction.

Page 164

CTRAN/W

Chapter 10: Theory

The particle tracking process is repeated until your specified elapsed time has
been reached or the particle has reached the boundary of the flow system.
CTRAN/W provides the option of tracking particles forward in the direction of
water flow, or backward in the opposite direction toward the entrance or source
boundaries. Backward tracking is accomplished by multiplying all linear
velocities by -1.0. All computations are carried out in exactly the same way as for
forward tracking. In forward tracking, particles stop at the exit boundaries,
whereas in backward tracking, particles stop at the entrance or source boundaries.
Therefore, forward tracking is useful in delineating the possible flow paths or
contaminant plume from source boundaries, whereas the backward tracking
option is useful in delineating the possible sources of contamination flowing to
exit boundaries.
NOTE: Backward tracking is only possible in a steady-state flow system.

10.14 Density-dependent flow


Density-dependence refers to contaminant transport problems where the density
of the contaminated water is significantly different than the density of the native
groundwater. In these cases, the density contrast will affect the flow dynamics of
the system. CTRAN/W allows the simulation of density-dependent transport
problems by coupling with SEEP/W using an iterative procedure. VADOSE/W
does not support density-dependent analysis. The density effect is accommodated
by the addition of a body force term to the groundwater flow governing equation
in SEEP/W. For more information about the body force term for densitydependent flow, see the Density-Dependent Flow section in the Theory chapter
of the SEEP/W book.
Since the body force term is added to the seepage governing equation in
SEEP/W, there is no special treatment to the finite element formulation in
CTRAN/W for density-dependent flow. However, the solution procedure is
different for density-dependent flow problems compared to advection-dispersion
problems. For density-dependent problems, the groundwater velocities and
concentrations must be solved for simultaneously at each time step because the
groundwater velocities are dependent on contaminant density and the
contaminant density is in turn dependent on concentration. At each time step,
SEEP/W uses the contaminant concentrations to calculate the density body force
term for the groundwater flow governing equation and then solves for equivalent
freshwater heads and groundwater velocities. CTRAN/W then reads the
Page 165

Chapter 10: Theory

CTRAN/W

groundwater velocities and solves for concentrations. Before proceeding to the


next time step, the solution continues iteratively until the groundwater velocities
and concentrations are compatible. The iterations are complete when either the
percentage change in both the vector norm of nodal pressure head and nodal
concentration are smaller than your specified convergence tolerances, or if the
maximum number of your specified iterations are reached.
CTRAN/W assumes that density varies linearly with concentration. Therefore
density-dependent problems require specification of the contaminant relative
density, (density relative to freshwater), at some reference concentration. For
example, the density of seawater is 1.025 times that of freshwater. Therefore the
relative density value is specified as 1.025 at a reference concentration of 1.0. In
this particular case the reference concentration is a relative concentration
representing 100% seawater. It should be noted that specification of the relative
density at the reference concentration defines a linear variation of relative density
with concentration. In the example given here, the relative density varies from
1.0 at a seawater concentration of zero, (freshwater), to 1.025 at a concentration
of 1.0.
In order to have any density effect, there must be a density contrast between the
contaminated water and freshwater. This means that the relative density of the
contaminated fluid must be different (larger or smaller) than the density of
freshwater. If there is no density contrast between the contaminated water and the
freshwater, then the density body force term in the groundwater flow governing
equation will be zero and the problem is essentially an advection-dispersion
transport problem. The relative density and reference concentration values are
specified in SEEP/W DEFINE under Key In Analysis Control.

10.15 Seepage solutions from SEEP/W or VADOSE/W


CTRAN/W relies on SEEP/W or VADOSE/W to provide the seepage solution in
a transport analysis. You must have access to both programs to undertake a
transport analysis. CTRAN/W requires the knowledge of the following quantities
from the seepage model output files:

Nodal total head H and Nodal flux Qw from the head file

DArcy velocity U from the velocity file

Volumetric water content from the material properties file

Page 166

CTRAN/W

Chapter 10: Theory

For practical purposes, CTRAN/W and the seepage model are allowed to have
independent time step increments. However, depending on the type of analysis
and the specified time step increments, CTRAN/W selects the appropriate
seepage solution to be used in the transport analysis.
For a density-dependent flow problem, the seepage solution is computed by an
iterative procedure controlled by CTRAN/W; therefore, the same time step
increments must be used in both SEEP/W and CTRAN/W. To avoid redundancy,
CTRAN/W makes use of the user-specified time step sequences in the
CTRAN/W analysis settings.
For non-density-dependent flow problem (i.e., advection-dispersion transport and
particle tracking), the seepage solutions must be computed using SEEP/W or
VADOSE/W before launching CTRAN/W. In a steady-state seepage problem,
CTRAN/W uses the steady-state seepage solutions for all time steps in the
analysis. In a transient seepage problem with time step increments the same in
both the seepage model and CTRAN/W, the seepage solution at the same time
step is used in the transport analysis. However, in a transient seepage problem
with different time step increments set up in analysis settings, CTRAN/W
assumes that the seepage solution is constant within a time step increment. For
example, assume that the seeepage elapsed times for Step 1, Step 2 and Step 3
are 50, 100 and 200 respectively. CTRAN/W uses the seepage solution of Step 1
for elapsed time between 0 and 50, the solution of Step 3 for elapsed time
between 100 and 200 in the transport analysis.
CTRAN/W uses the seepage output file(s) with the highest elapsed time for all
time steps that have an elapsed time greater than the highest seepage elapsed
time. This allows a CTRAN/W analysis to be continued past the steady-state
point in a seepage analysis.

Page 167

Chapter 10: Theory

CTRAN/W

Page 168

CTRAN/W

References

References
Bear, J., 1979. Hydraulics of Groundwater. McGraw-Hill, London, England.
Daus, A.D., Frind, E.O., and Sudicky, E.A., 1983. Comparative Error Analysis in
Finite Element Formulations of the Advection-Dispersion Equation.
University of Waterloo.
Freeze, R.A., and Cherry, J.A., 1979. Groundwater. Prentice-Hall.
Frind, E.O., 1988. Solution of the Advection-Dispersion Equation with Free Exit
Boundary. Numerical Methods for Partial Differential Equations. 4, John
Wiley & Sons, Inc.
Frind, E.O., and Pinder, G.F., 1982. The Principal Direction Technique for
Solution of the Advection-Dispersion Equation. Proceedings of 10th
World Congress of the IMACS, Concordia University, Montreal,
Canada.
Kemper, W.D., and Van Shaik, J.C., 1966. Diffusion of Salts in Clay-Water
Systems. Soil Science of America Proceedings, Vol.30.
Lancaster, P., and Salkauskas, K., 1986. Curve and Surface Fitting: An
Introduction. Academic Press.
Marsily, Ghislain de., 1986. Quantitative Hydrogeology: Groundwater
Hydrology for Engineers. Academic Press.
Ogata, A., 1970. Theory of Dispersion in a Granular Medium. Fluid Movement
in Earth Materials. Geological Survey Professional Paper 411-I. U.S.
Government Printing Office, Washington.
Salkauskas, K., 1974. C' Splines For Interpolation of Rapidly Varying Data.
Rocky Mountain Journal of Mathematics, Vol. 14, No. 1.
Segerlind, L.J., 1984. Applied Finite Element Analysis. John Wiley and Sons.

Page 169

References

CTRAN/W

Page 170

CTRAN/W

Index

Index
adaptive time stepping............43, 79

Decay half life ............................. 16

Adsorption function......................15

Density-dependent
contaminant
transport ..................................... 9

Advection and dispersion .............24


Advective contaminant transport....7

Density-dependent flow............. 165


Diffusion function........................ 14

Advective-dispersive contaminant
transport......................................8

Dispersivity and diffusion............ 13

Attenuation .....................................5

Dry density .................................. 16

axisymmetric ................................41

equipotential line ......................... 62

boundary conditions10, 11, 21, 22,


23, 24, 32, 33, 44, 55, 57, 59

Exit boundaries .......................... 162

Boundary conditions

Exit review................................... 27
Finite element equations ............ 150

concentration vs. mass function30


exit boundary............................28
exit conditions ..........................28
free exit boundary.....................28
mass accumulation ...................26
source concentration.................25

Flow velocity ............................. 143


flux section ............................ 66, 67
Fracture flow simulation.............. 77
Free exit boundary ..................... 128
Governing equations.................. 144
Graph

Boundary Conditions....................21
boundary functions .......................32
conductivity ....17, 18, 33, 63, 73, 80
convergence..................................44
Darcian velocity .........................143

parameters.......................... 69, 70
Hydrodynamic dispersion.... 13, 155
Infinite elements .......................... 76
Mass accumulation ...................... 26
Mass flux

Page 171

Index

CTRAN/W

advective.................................159

Surface mass accumulation.......... 26

decayed...................................160

Time integration ........................ 154

dispersive................................157

Time Integration .......................... 48

stored ......................................159

time step................43, 69, 70, 72, 79

Mass quantity calculation ...........161

Time step design.......................... 50

material properties11, 44, 54, 55,


57, 79

total flux................................. 78, 79

Mesh Design.................................49

total head.......................... 44, 45, 62


transient ....................................... 69

Numerical
Dispersion
and
Oscillation ................................46

transport processes......................... 1

numerical experiments .................73

units ............................................. 23

Peclet and Courant Number Criteria


..................................................46

Units

phreatic line ..................................68

View

consistent set............................ 74

restart ......................................79, 80

mass accumulation................... 62

Source concentration ....................25

particle information ................. 60

Sources and sinks .........................24

volumetric water content17, 18, 63,


64

splines...............................17, 18, 19

Weighted splines.......................... 17

Page 172

Consistent sets of units


Parameter

Symbol

Units (SI)

Length

ft

Time

hr

Force

kN

lbf

Mass

Pressure

F/L

Units
(Imperial
)

lb
2

kN/m

psf

Unit Weight of Water

F/L

kN/m

pcf

Hydraulic Conductivity

L/T

m/s

ft/hr

Concentration

M/L

g/m3

lb/ft3

Diffusion Coefficient

2
L /T

m2/s

ft2/hr

L and T (dispersivity)

ft

Decay Half-Life

sec

hr

S (adsorption)

M/M

g/g

lb/lb

Density

M/L

g/m

Mass Flux

M/T

g/s

lb/ft3
lb/hr

Consistent sets of units


Parameter

Symbol

Units (SI)

Length

ft

Time

hr

Force

kN

lbf

Mass

Units
(Imperial
)

lb

kN/m2

psf

kN/m3

pcf

Pressure

F/L

Unit Weight of Water

F/L

Hydraulic Conductivity

L/T

m/s

ft/hr

Concentration

M/L

g/m3

lb/ft3

Diffusion Coefficient

2
L /T

m2/s

ft2/hr

ft

Decay Half-Life

sec

hr

S (adsorption)

M/M

g/g

lb/lb

Density

M/L

g/m3

lb/ft3

Mass Flux

M/T

g/s

lb/hr

L and T (dispersivity)