Numerical Modeling for

Flow and Transport

Cive 7332
Lecture 7
Uses of Modeling
A model is designed to represent reality in such a way that
the modeler can do one of several things:

Quickly estimate certain aspects of a system (screening models,
analytical solutions, back of the envelope calculations)

Determine the causes of an observed condition (flow direction,
contamination, subsidence, flooding)

Predict the effects of changes to a system (pumping, remediation,
development, waste disposal)
Types of Ground Water Flow Models
Analytical Models (Exp and ERF functions)
1-D solution, Ogata and Banks (1961)
2-D solution, Wilson and Miller (1978)
3-D solutions, Domenico & Schwartz (1990)

Numerical Models (Solved over a grid - FDE)
Flow-only models in 3-D (MODFLOW)
MODPATH - allows tracking of particles in 2-D placed in flow
field produced from MODFLOW

Grid - Hydraulic Conductivity
Governing Equation for Flow
For two-dimensional transient flow conditions
Transient means that the water level changes with time
Steady state means it is constant in time.




S = storativity [unitless],
Q = recharge or withdrawal per unit area [L/T]
T = transmissivity [L
2
/T]


c
cx
T
x
ch
cx
|
\

|
.
|
+
c
cy
T
y
ch
cy
|
\

|
.
|
= S
ch
ct
+ Q
Poissons Equation
The basic flow equation in a homogeneous, confined, 2-D
aquifer at steady-state (S = 0), with sources and/or sinks




Can be solved analytically or numerically
Theis Analytical solution in cylindrical coordinates
Gauss-Seidel with Successive Over-Relaxation (SOR)

c
2
h
cx
2
+
c
2
h
cy
2
=
Q x, y
( )
T
Laplaces Equation
If there are no source/sink terms, Poissons equation
reduces to Laplaces Equation



Can also be solved analytically or numerically
Gauss-Seidel Iteration method
Successive Over Relaxation method
Solutions are generally smooth and well-behaved

c
2
h
cx
2
+
c
2
h
cy
2
= 0
Laplace Numerical Solution
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Numerical Solutions


For complex layered, heterogeneous aquifers, a numerical
approximation is required in non-uniform flow

Several types Finite Difference, Finite Element, and
Method of Characteristics

Finite difference approximations involve applying Taylors
expansions to the equations (flow and transport) and
approximating the derivatives in the equation

Other methods involve different approximations, but all
are based generally on the Taylor expansion
Taylors Expansion from Calculus
Taylors Series provides a means to predict a function f(x) value at
one point in terms of a function value and its derivatives at
another point.
Zero Order approximation might be f(x
i+1
) = f(x
i
)
Value at new point is same as at old point
First Order approximation is f(x
i+1
) = f(x
i
) + f(x
i
)(x
i+1
x
i
)
Straight line projection to next point
Second Order approximation captures curvature
f(x
i+1
) = f(x
i
) + f(x
i
)(x
i+1
x
i
) + f(x
i
) (x
i+1
x
i
)
2


2!
Approximation of f(x) by various orders
Zero Order
First Order
Second Order
True Fcn
X
i
X
i+1

f(x)
AX
Taylors Expansion from Calculus
Any function h(x) can be expressed as an infinite series:





h(x + Ax) = h(x) + Axh' (x) +
Ax
2
2
h' ' (x) +
h(x Ax) = h(x) Axh' (x) +
Ax
2
2
h' ' (x) +
Where h(x) is the first derivative and h(x) is the second derivative and so on.
Taylors Expansion for Second Derivative
Adding the above two eqns, and neglecting all the higher
terms, and rearranging terms gives a useful approximation
for h(x) :
h' ' x ( )=
d
2
h
dx
2
|
\

|
.
x
~
1
Ax
2
h x + Ax ( ) 2h x ( )+ h x Ax ( )
{ }
h(x) h(x + Ax) h(x - Ax)
True Function h(x)
Taylors Expansion for dh/dx
Neglecting second and all higher powers, and rearranging
terms gives a forward or backward difference approximation
for the first derivative or h(x) :

h' x
( )
=
dh
dx
|
\

|
.
|
x
~
1
Ax
h x + Ax
( )
h x
( ) { }
h(x) h(x + Ax) h(x - Ax)
True Function h(x)

h' x
( )
=
dh
dx
|
\

|
.
|
x
~
1
Ax
h x
( )
h x Ax
( ) { }
Forward Diff
Backward Diff
Numerical Soln to Laplaces Equation


Assume x = y [regular square grid]
Represent h(x
i
,y
j
) = h
i,j

h(x
i
+x, y
j
+y) = h
i+1,j+1
,

Thus replacing terms in Laplace, the F.D.E. becomes



Do the same for the j direction, and you have the following:
x
i
x
i+1
x
i-1
y
j
y
j-1
y
j+1
' ' h x
i
, y
j
( )
x
= ' ' h
i , j
~
1
Ax
2
h
i 1, j
2h
i , j
+ h
i +1, j
{ }

c
2
h
cx
2
+
c
2
h
cy
2
= 0 h
i,j

AX
AY
Approximation to Laplaces Eqn.
' ' h x
i
, y
j
( )
x
+ ' ' h x
i
, y
j
( )
y
~
1
Ax
2
h
i1, j
2h
i , j
+ h
i +1, j
{ }
+ h
i, j1
2h
i, j
+ h
i, j +1
{ }
| |
=
1
Ax
2
h
i1, j
+ h
i, j 1
4h
i, j
+h
i+1, j
+ h
i, j +1
| |
= 0
Summing terms and solving for h
i,j
gives:
h
i , j
=
1
4
h
i1, j
+ h
i, j 1
+h
i+1, j
+h
i , j +1
{ }
Application to a Simple 3x3 Grid
Start with boundary conditions and initial
estimate for h, assume h
1
= h
2
= h
3
= h
4
= 0

Get a new estimate for h
1
, call it h
1
m+1

h
1
m+1
= (top + right + bottom + left)/4
h
1
m+1
= {0 + h
2
+ h
3
+ 0}/4

Repeat four internal points - 2, 3, and 4
using same four-star average calculation
1 2
3 4
h = 0
h = 0
h = 0
h = 1
Application to a grid
h
1
m+1
= {0 + h
2
m
+ h
3
m
+ 0 } /4
h
2
m+1
= {0 + 0 + h
4
m
+ h
1
m
} /4
h
3
m+1
= {h
1
m
+ h
4
m
+ 1 + 0 } /4
h
4
m+1
= {h
2
m
+ 0 + 1 + h
3
m
} /4

Use the initial values and boundary
conditions for the first approximation
Use the updated values (m+1) for further
approximations (m+2)
Continue until the numbers dont change
much (convergence!)
1 2
3 4
h = 0
h = 0
h = 0
h = 1
FDE for Laplace - EXCEL
node # 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Iteration 0 0 0 0 0 0 0 0 0 0 1 1
1 0.0 0.0 0.0 0.0000 0.0000 0.0 0.0 0.2500 0.2500 0.0 1.0 1.0
2 0.0 0.0 0.0 0.0625 0.0625 0.0 0.0 0.3125 0.3125 0.0 1.0 1.0
3 0.0 0.0 0.0 0.0938 0.0938 0.0 0.0 0.3438 0.3438 0.0 1.0 1.0
4 0.0 0.0 0.0 0.1094 0.1094 0.0 0.0 0.3594 0.3594 0.0 1.0 1.0
5 0.0 0.0 0.0 0.1172 0.1172 0.0 0.0 0.3672 0.3672 0.0 1.0 1.0
6 0.0 0.0 0.0 0.1211 0.1211 0.0 0.0 0.3711 0.3711 0.0 1.0 1.0
7 0.0 0.0 0.0 0.1230 0.1230 0.0 0.0 0.3730 0.3730 0.0 1.0 1.0
8 0.0 0.0 0.0 0.1240 0.1240 0.0 0.0 0.3740 0.3740 0.0 1.0 1.0
9 0.0 0.0 0.0 0.1245 0.1245 0.0 0.0 0.3745 0.3745 0.0 1.0 1.0
10 0.0 0.0 0.0 0.1248 0.1248 0.0 0.0 0.3748 0.3748 0.0 1.0 1.0
11 0.0 0.0 0.0 0.1249 0.1249 0.0 0.0 0.3749 0.3749 0.0 1.0 1.0
12 0.0 0.0 0.0 0.1249 0.1249 0.0 0.0 0.3749 0.3749 0.0 1.0 1.0
13 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
14 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
Equations are programmed in Nodes 6, 7, 10, and 11.
All other nodes are boundary conditions, and do not change.
Cells 1,4,13, and 16 are the corners, and are not used in the calculations.
1 2 3 4 0 0 0 0
5 6 7 8 The grid has this 0 0.13 0.13 0
9 10 11 12 orientation in real life. 0 0.38 0.38 0
13 14 15 16 0 1 1 0
1
2
3
4
S1
S2
S3
S4
0
0. 1
0. 2
0. 3
0. 4
0. 5
0. 6
0. 7
0. 8
0. 9
1
Convergence Criteria
Convergence for a flow code requires that the change in
the solution at each point be less than a specified target,
called the convergence criterion, or sometimes epsilon, c
If c is too large, convergence will occur before a solution
is reached
If c is too small, convergence may take very long, or be
impossible due to oscillation
There is more than one way to define convergence,
including global measures, local measures, etc.
Gauss-Seidel Method
Uses partially completed iteration to estimate values for
the rest of the iteration.
h
1
m+1
= {0 + h
2
m
+ h
3
m
+ 0 }/4
h
2
m+1
= {0 + 0 + h
4
m
+ h
1
m
}/4
h
3
m+1
= {h
1
m+1
+ h
4
m
+ 1 + 0 }/4
h
4
m+1
= {h
2
m+1
+ 0 + 1 + h
3
m
}/4

Use m+1 update iteration values for h
1
and h
2
when
computing h
3
and h
4
Results in faster convergence over a large grid
Gauss-Seidel
node # 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Iteration 0 0 0 0 0 0 0 0 0 0 1 1
1 0.0 0.0 0.0 0.0000 0.0000 0.0 0.0 0.2500 0.2500 0.0 1.0 1.0
2 0.0 0.0 0.0 0.0625 0.0625 0.0 0.0 0.3281 0.3281 0.0 1.0 1.0
3 0.0 0.0 0.0 0.0977 0.0977 0.0 0.0 0.3564 0.3564 0.0 1.0 1.0
4 0.0 0.0 0.0 0.1135 0.1135 0.0 0.0 0.3675 0.3675 0.0 1.0 1.0
5 0.0 0.0 0.0 0.1203 0.1203 0.0 0.0 0.3719 0.3719 0.0 1.0 1.0
6 0.0 0.0 0.0 0.1230 0.1230 0.0 0.0 0.3737 0.3737 0.0 1.0 1.0
7 0.0 0.0 0.0 0.1242 0.1242 0.0 0.0 0.3745 0.3745 0.0 1.0 1.0
8 0.0 0.0 0.0 0.1247 0.1247 0.0 0.0 0.3748 0.3748 0.0 1.0 1.0
9 0.0 0.0 0.0 0.1249 0.1249 0.0 0.0 0.3749 0.3749 0.0 1.0 1.0
10 0.0 0.0 0.0 0.1249 0.1249 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
11 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
12 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
13 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
14 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
Equations are programmed in Nodes 6, 7, 10, and 11.
All other nodes are boundary conditions, and do not change.
Cells 1,4,13, and 16 are the corners, and are not used in the calculations.
1 2 3 4 0 0 0 0
5 6 7 8 The grid has this 0 0.13 0.13 0
9 10 11 12 orientation in real life. 0 0.38 0.38 0
13 14 15 16 0 1 1 0
1
2
3
4
S1
S2
S3
S4
0
0. 1
0. 2
0. 3
0. 4
0. 5
0. 6
0. 7
0. 8
0. 9
1
Successive Over-Relaxation - SOR
To speed up the convergence, overshoot the standard model.
Defining the Residual, c = h
i,j
m+1*
h
i,j
m
Using:

h
i,j
m+1
= h
i,j
m
+ ec = (1 e) h
i,j
m
+ eh
i,j
m+1*
,

where e is the relaxation factor and h
i,j
m+1*
is given by the
Gauss-Seidel approximation, we get:

h
i,j
m+1
=(1-e)h
i,j
m
+(e/4)(h
i-1,j
m+1
+ h
i,j-1
m+1
+ h
i+1,j
m
+ h
i,j+1
m
)

If e = 1, reduces to Gauss-Seidel
If 1 < e< 2, the method is over-relaxed - usually ~ 1.4 - 1.5
SOR - Marked Improvement
node # 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Iteration 0 0 0 0 0 0 0 0 0 0 1 1
1 0.0 0.0 0.0 0.0000 0.0000 0.0 0.0 0.3500 0.3500 0.0 1.0 1.0
2 0.0 0.0 0.0 0.1225 0.1225 0.0 0.0 0.3754 0.3754 0.0 1.0 1.0
3 0.0 0.0 0.0 0.1253 0.1253 0.0 0.0 0.3751 0.3751 0.0 1.0 1.0
4 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
5 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
6 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
7 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
8 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
9 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
10 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
11 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
12 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
13 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
14 0.0 0.0 0.0 0.1250 0.1250 0.0 0.0 0.3750 0.3750 0.0 1.0 1.0
Equations are programmed in Nodes 6, 7, 10, and 11.
All other nodes are boundary conditions, and do not change.
Cells 1,4,13, and 16 are the corners, and are not used in the calculations.
1 2 3 4 0 0 0 0
5 6 7 8 The grid has this 0 0.13 0.13 0
9 10 11 12 orientation in real life. 0 0.38 0.38 0
13 14 15 16 0 1 1 0
e = 1.4
1
2
3
4
S1
S2
S3
S4
0
0. 1
0. 2
0. 3
0. 4
0. 5
0. 6
0. 7
0. 8
0. 9
1
Numerical Simulation Models
Numerical models solve approximations of the governing
equations of flow and transport - over a grid system

Based on a grid or mesh laid over the study site
Helps organize large datasets into logical units
Data required includes boundary conditions, hydraulic
conductivity, thickness, pump rates, recharge, etc.)
Can be computationally intensive
Can be applied to actual field sites to help understand
complex hydrogeologic interrelationships.
Often-Used Numerical Models
MODFLOW (1988) - USGS flow model for 3-D aquifers
MODPATH - flow line model for depicting streamlines

MOC (1988) - USGS 2-D advection/dispersion code
MT3D (1990, 1998) - 3-D transport code works with
MODFLOW
RT3D (1998) - 3-D transport chlorinated - MODFLOW
BIOPLUME II, III (1987, 1998) - authored at Rice Univ 2-
D based on the MOC procedures.

Numerical Solution of Equations

Numerically -- H or C is approximated at each point of a
computational domain (may be a regular grid or irregular)
Solution is very general
May require intensive computational effort to get the desired
resolution
Subject to numerical difficulties such as convergence problems and
numerical dispersion
Generally, flow and transport are solved in separate independent
steps (except in density-dependent or multi-phase flow situations)
MODFLOW Introduction
Written in the 1984 and updated in 1988
Solves governing equations of flow for a full 3-D aquifer
system with variable K, b, recharge, drains, rivers, and
pumping wells.
Withdrawal and injection wells (rates may change with time)
Constant head boundaries or regions (ponds/rivers/fixed heads)
No-flow boundaries or regions (bedrock outcrops/water divides)
Regions of diffuse recharge or discharge (rainfall)
Observation wells
MODFLOW Features
MODFLOW
MODFLOW is a modular 3-D finite-difference flow code developed by
the U.S. Geological Survey to simulate saturated flow through a
layered porous media. The PDE solved is for h(x,y,z,t):






where K
xx
, K
yy
, and K
zz
are defined as the hydraulic conductivity
along the x, y, and z coordinate axis, h represents the
potentiometric head, W is the volumetric flux per unit volume
being pumped, S
s
is the specific storage of the porous material and
t is time.

t
h
S W
z
h
K
z y
h
K
y x
h
K
x
s zz yy xx
c
c
=
|
.
|

\
|
c
c
c
c
+
|
|
.
|

\
|
c
c
c
c
+
|
.
|

\
|
c
c
c
c
MODFLOW Features
MODLOW consists of a main program and a series of
independent subroutines grouped into packages.

Each package controls with a specific feature of the
hydrologic system, such as wells, drains, and recharge. The
division of the program into packages allows the user to
analyze the specific hydrologic feature of the model
independently.

MODFLOW Features
MODFLOW is one of the most versatile and widely
accepted groundwater models
It is particularly good in heterogeneous regions because it
allows for vertical interchange between layers, as well as
horizontal flow within the aquifers.
It also allows for variable grids to speed computation.
It has been applied to model thousands of field sites
containing a number of different contaminants and for a
number of different remediation applications.

Solution Methods
MODFLOW is an iterative numerical solver (SIP or SOR).
The initial head values are provided and the these heads
are gradually changed through a series of time steps, in the
case of a transient model run, until the governing equation
is satisfied. Time steps can be variable to speed output.
The primary output from the model is the head distribution
in x, y, and z, which can then be used by a transport model.
In addition, a volumetric water budget is provided as a
check on the numerical accuracy of the simulation.
MODFLOW - Input to Transport Models
Designed to create modern GUI to ease large data entry
and output graphical manipulation for applications to
complex field sites

GMS - 1995
Visual MODFLOW
Ground Water Vistas

PLUME
visualization
MODPATH - Pathlines
Designed to use heads from MODFLOW and linear
interpolation to compute velocity V
x
and or V
y
.
Particles can be placed in areas of known or suspected
source concentrations in order to create possible tracks of
contaminants in space and time - streaklines
Path results after two time steps
MODPATH
Designed to use heads from MODFLOW and linear
interpolation to compute velocity V
x
in the x direction:
V
x
= (1-f
x
)V
x(i -1/2)
+f
x
V
x(i +1/2)

Where f
x
= (x
p
- x
(i-1/2)
/ Ax
i,j
And x
p
is the x coordinate of the particle
i, j
(i + 1/2, j) (i - 1/2, j)
Particle Location
In Grid
V
x

Ax
Overlay of Grid on Map
Grid - Hydraulic Conductivity
Heads and Boundary Features
10 yr pathlines
Wei 03 2-layer final elev
(Weifinal)
Dec 2002 Original
MODPATH - EXAMPLE
Designed to create modern GUI to ease large data entry
and output graphical manipulation for applications to
complex field sites

Source
Alternative 2
Alternative 3
Alternative 2
Alternative 3 Alternative 4
Alternative 5
Contaminant Transport in 2-D
cCB
ct
=
c
cx
I
BD
IJ
cC
cx
J
|
\

|
.
|

c
cx
I
BCV
I
( )
' C W
E
C = Concentration of Solute [M/L
3
]
D
IJ
= Dispersion Coefficient [L
2
/T] - x and y
B = Thickness of Aquifer [L]
C = Concentration in Sink Well [M/L
3
]
W = Flow in Source or Sink [L
3
/T]
E = Porosity of Aquifer [unitless]
V
I
= Velocity in I Direction [L/T] - x and y
2-D CONTAMINANT TRANSPORT
Domenico and Schwartz (1990)
3-D solutions for several geometries (listed in Bedient et
al. 1999, Section 6.8) - spreadsheets exist
Generally a vertical plane, constant concentration source.



C x, y, z, t
( )
C
0
=
1
8
|
\

|
.
|
erfc
x vt
2 o
x
vt



(

(
erf
y +Y 2
2 o
y
x




(

(
(
- erf
y Y 2
2 o
y
x




(

(
(






`

)

erf
z + Z 2
2 o
z
x




(

(
(
- erf
z Z 2
2 o
z
x




(

(
(






`

)

Method of Characteristics USGS (MOC)
Written for USGS by Konikow and
Bredehoeft in 1978
Solves flow equations with Alternating
Direction Implicit (ADI) method
Solves transport equations via particle
tracking method and finite difference
Using velocities calculated from the flow
solution, v
x
= k
x
(h/x), particles are moved
Concentrations are based on the average
concentration of all particles in a cell at the
end of the time step
Velocity
MOC Concepts
Partial Differential Equation replaced by equivalent set of
ODEs called characteristic equations (approximated with
finite difference)
Particle in a cell is moved a distance proportional to the
seepage velocity within the cell
Accounts for concentration change due to advection
Remainder of governing equation solved by finite
difference methods
Accounts for concentration change due to dispersion, changes in
saturated thickness, and fluid sources
MOC/BIOPLUME II Capabilities
Can simulate effects of natural or enhanced
biodegradation:
Fast-equilibrium biodegradation
First-order biodegradation
Monod kinetics
Withdrawal and injection wells (pump and treat
systems)
Injection wells for oxygen enhancement
Observation wells within and outside plume area

BIOPLUME II Concepts
Can simulate effects of natural or enhanced biodegradation
Adjustment is made at each time step in numerical grid

Background D.O.
Initial Hydrocarbon
Concentration
Reduced Oxygen
Concentration
Oxygen
Depletion
Reduced Hydrocarbon
Concentration
+
=
BIOPLUME II
Introduction to Solution Methods and
Model Mechanics
What does it do?
Two dimensional finite difference model for simulating
natural attenuation due to:
advection
dispersion
sorption
biodegradation

How Does BPIII Solve Equations?
Contaminant transport solved using the Method of
Characteristics
Particles travel along Characteristic lines determined by
flow solution.
Particles carry mass
Advection solved via particle movement
Dispersion solved explicitly
Reaction solved explicitly
First order decay
Instantaneous Biodegradation
Initial location of particle
New location of particle
Flow line and direction of flow
Computed path of particle
Particle Movement
Limitations/Assumptions
Darcys Law is valid
Porosity and hydraulic conductivity constant in time,
porosity constant in space
Fluid density, viscosity and temperature have no effect on
flow velocity
Reactions do not affect fluid or aquifer properties
Ionic and molecular diffusion negligible
Vertical variations in head/concentration negligible
Homogeneous, isotropic longitudinal and transverse
dispersivity
Limitations of Biodegradation
No selective or competitive biodegradation of
hydrocarbons (lumped hydrocarbons)

Conceptual model of biodegradation is a simplification of
the complex biologically mediated redox reactions that
occur in the subsurface
End of
Simulation?
End of
Pumping
Period?
End of
Time Step?
Start
Read Geologic, Hydrogeologic
& Chemical Input Data
Generate Uniformly Distributed Particles
Make New/Remove Old Particles @ Edges
Compute Explicitly Conc. at Nodes
Compute Dispersion Coefficients
Determine t for Explicit Calculations
Move Particles
Calc. Avg. Conc. in Each Cell
Compute Hydraulic Gradients
Adjust Concentration of Each Particle
Compute Ground Water Velocities
Compute Mass Balance
Stop
Summarize and Print Results
Y
N
Y
N
N Y
BIOPLUME II Flowchart
HOW TO SET UP A MODEL
1. Data Collection & Analysis
2. Modeling Scale
3. Discretization
4. Boundary Conditions
5. Parameter Estimation
6. Calibration
7. Sensitivity Analysis
8. Error Estimation
9. Prediction

SOURCE DATA
Mass of contaminant
Q, C
0

Discrete vs. Continuous

Nature of contaminant

Chemical stability
Biological stability
Adsorption
PARAMETER ESTIMATION
1. Porosity
2. Dispersivity
3. Storage coefficients
4. Hydraulic conductivity
5. Thickness of unit
6. Recharge rates

REGIONAL SCALE -
QUANTITATIVE
Aquifer characteristics
Background gradients
Geology
Recharge sources
LOCAL SCALE - WATER QUALITY
Site history
Site characterization
Source definition
Nature of contamination
Plume delineation
MOC TIMING PARAMETERS
Total Simulation Time

1st pumping period 2nd
NPMP = 2

For Each Pumping Period

PINT = pumping period in yrs


NTIM = # of time steps in pumping period
MOC BOUNDARY CONDITIONS
Two types

Constant Head
Water Table = constant

Constant Flux
Flow rate Q
Concentration C
0

MOC BOUNDARY CONDITIONS
Specifications of NCODES
For Each Code in NOEID map

LEAKANCE (s
-1
)
vertical hyd. conduct. / thickness

CONCENTRATION OF CONTAMINANT

RECHARGE RATE (ft/s)

NOTE
For constant head cells set LEAKANCE to 1.0
MOC SOURCE DEFINITION
Injection well
Flow rate - Q
Concentration - C
0

Constant Head Cell
C=C
0


Recharge Cell
Flow rate - Q
Concentration - C
0

PHYSICAL AQUIFER
CHARACTERISTICS
1. Transmissivity (ft
2
/s) VPRM
2. Thickness (ft) THCK
3. Dispersivity (ft)
Longitudinal BETA
Ratio DLTRAT = T
xx
/T
yy

4. Porosity POROS
5. Storativity S
NOTE
For transient problems
TIMX increment multiplier
TINIT size of initial time step

MOC REACTION PARAMETERS
NREACT
Flag to instruct MOC to expect reaction data
0 - no reactions
1 - reactions taking place
expect card # 4 free format

Two types of reaction:
RETARDATION
KD - Distribution coefficient
RHOB - Bulk density
RADIOACTIVE DECAY
THALF - Half life of solute
INPUT PARAMETERS AFFECTING
ACCURACY FOR HYDRAULIC
CALCULATIONS
ITMAX
Maximum allowable number of iterations: 100-200
Increase ITMAX if hydraulic mass balance error is > 1%
NITP
Number of iteration parameters
USE 7
TOL
Convergence criteria: <0.01
Decrease TOL to get less hydraulic mass balance error
PARAMETERS AFFECTING
ACCURACY OF TRANSPORT
NPTPND - Number of particles in a cell






NPMAX - Maximum number of particles
= NX NY NPTPND
STABILITY CRITERIA FOR MOC
MOC may require dividing NTIM or PINT into smaller move time
steps

At minimum of

Dispersion


Mixing


Advection
( ) ( )
2 2
5 . 0
dy
D
dx
D
yy
xx
+
( )
k j i
k j i
W
nb
, ,
, ,
( )
max
x
V
x A
( )
max
y
V
y A
INPUT PARAMETERS AFFECTING
STABILITY OF MOC
CELDIS - max distance per move
If CELDIS < space between particles MOC will oscillate for
N yrs BUT gives smallest Mass Balance errors for T>N
If CELDIS = Stability Criteria DO a sensitivity analysis on
CELDIS
NPTPND - initial # of particles
Accuracy of MOC directly proportional to NPTPND
Runtime inversely proportional to NPTPND

RULE OF THUMB
Initially set NPTPND=4 or 5 and CELDIS=0.75 or 1
For final runs use NPTPND=9 and CELDIS=0.5
Output control
NPNTMV
Number of particle moves after which output is requested. Use 0 to
print at end of time steps

NPNTVL
Printing velocities
0 - do not print
1 - print for first time step
2 - print for all time steps

NPNTD
Print dispersion equation coefficients

NPDELC
Print changes in concentration

NPNCHV
Do not use this option. Always set to 0. It is used to request cards to be
punched.

Output control (cont.)
Concentrations
Calibration,Validation, and Sensitivity
Analysis

Calibration is the process of making the model match real-
world data. Involves making several model runs, varying
parameters until the best fit is achieved.

Validation is the process of confirming the validity of your
calibration by using the model to fit an independent set of data.

Sensitivity Analysis is the process of changing parameters to
see the effects on the model results. The most sensitive parameters
need to be checked for accuracy to ensure the best model.