Transient Ground Water Flow

Transient Equation
The Transient equation is an extension of the Poisson and Laplace equations we
considered before, and allows ‘accumulation’ (A in IPOLA terms). Not surprisingly, the
equation is a great deal like the heat/diffusion equation we worked with previously and
the solution methods are exactly the same. We’ll focus on an explicit solution.

Here we’ll consider only 2-D flow, but the approach for 3-D flows would be the same.

Derivation of Transient Ground Water Flow Equation

A non-steady state (or transient) ground water system has changes in water storage with
time (water accumulation or removal); hence A ≠ 0. Let’s ignore Production/Loss. So
( 0 ) simplifies to

 I  O   A (0)

Recall that accumulation of chemical mass in the control volume we used to study
diffusion was very easy: chemical mass was simply the concentration times the volume.
The volume was assumed constant. So, any change in concentration was immediately
reflected as a change in stored chemical mass.

In a similar way, the storage of water in a confined aquifer depends on the head:

Vw = -xy S h,

where S – the Storage Coefficient – is the proportionality factor. When the head changes
from 10 to 5, h = ht – ht-1 = 5-10 = - 5, and the volume Vw is released from storage. As
an instantaneous rate of accumulation we have

Vw h
  xyS (0)
t t
Considering a 3-D aquifer ‘cell’ of thickness b in map view, we have

qy|y+y

qx|x qx|x+x
y Vw

x b

qy|y
where qx|x indicates the flux (L3L-2T-1 or LT-1) in the x direction at the point x. In order to
satisfy ( 0 ), we must have

(qx|x+x - qx|x)yb + (qy|y+y - qy|y)xb = -Sxyh/t;

that is, the flux into each wall times the area over which it occurs is the flow into or out
of that wall and all of these flows must sum to Sxyh/t.

Ground water flows in response to head gradients (e.g., h/x = 0.001) in accordance
with Darcy’s law:

h
q  K (0)
x

where q is the flux and K is the hydraulic conductivity (LT-1).

Incorporating Darcy’s law into our flow balance equation, we have

 h h   h h  h
 K K  yb   K K  xb   Sxy (0)
 x x  x x x  
 y y  y
y 
y 
t
The product of the hydraulic conductivity K and the aquifer thickness b is known as the
transmissivity T and has units of L2T-1. Both K and b are considered constant here.
Incorporating T, dividing both sides by Txy and by -1

 h   h h 
h   
  
 x x  x x x  y
 y  y
y y 
  S h (0)

x y T t

We are taking the gradients of the gradients, and if we shrink x and y to differential
size we have

 h   h h 
h    
  
 x x x   y y y 
  S h (0)
 
x  x y  y

x y T t

or equivalently,

 2 h  2 h S h
  (0)
x 2 y 2 T t

This is the 2-D version of the transient flow equation.

Finite Difference Expression of Poisson Equation

 2 h h x  x  2h x  h x x
 (0)
x 2 x 2

and

2h h y  y
 2h y
h y  y
 (0)
y 2
y 2

We need a finite difference estimate for ∂h/∂t too. This is a little more complicated
because now we have to include time.
 t-t
x -x C|x, t-1 x +x

C/t|t-t/2 Estimate here

t C|x, t
x

The simplest approach is to use the ‘known’ head from the previous time step. At the
start of any numerical solution, the Initial Condition will specify the time 0 heads. So, we
can approximate the derivative ∂h/∂t by h/t or

h h hx , y ,t  hx , y ,t  t
  (0)
t t t

Therefore, the 2-D transient flow equation ( 0 ) will have the following finite difference
representation:

h x  x , y ,t  t
 2h x , y ,t  t
h x  x , y , t  t
h x , y  y ,t  t
 2h x , y ,t  t
h x , y  y , t  t S h x , y ,t  h x , y , t   t
 
x y ( 0
2 2
T )t

Note that we have chosen to compute the spatial gradients at the t – t time level.

We assume that x = y (i.e., we work on a simple isotropic grid). Then h|x,y,t-t appears
twice of the left side of the equation and can be combined so that we can simplify to

h x  x , y ,t  t
 4h x , y ,t  t
h x  x , y ,t  t
h x , y  y ,t  t
h x , y  y ,t  t S hx , y ,t  hx , y ,t  t
 (0)
x 2
T t

Finally, we can multiply both sides by –(Tt)/S and rearrange to solve for the head at x,y.

T t h x  x , y ,t  t
 4h x , y ,t   t
h x  x , y ,t  t
h x , y  y ,t  t
h x , y  y ,t  t

h x , y ,t  h x , y , t   t   (0)
S 
 x 2



The finite difference expression for the 1-D equation can be obtained by eliminating all
terms at y +/- y, and 2 of the h|x,y,t-t terms. We end up with
 Tt h x  x , y ,t  t
 2h x , y ,t  t
h x  x , y ,t  t

h x , y , t  h x , y , t  t    (0)
S 
 x 2



Notice that this is exactly the solution we had for the heat/diffusion equation in 1-D with
diffusion coefficient D = T/S. Hence the same stability criterion applies: Tt/(Sx2)  ½.

Analytical solution of 1-D Transient Ground Water Flow

h 
 h 0 ,t x 
2  h 100,t 0  h 0,t cos( n ) 
 nx  Tn2 2t /( SL2 )
h x ,t  h 0 , t  100,t  0

L
 
 n 1 n
sin 
 L 
e (0)

Sum the series until the change in the sum becomes insignificant.

Assignment
Modify your heat/diffusion code to solve the 1-D transient ground water flow equation.
We wish to simulate temporal changes in the confined aquifer heads between two
reservoirs (shown below). Assume the aquifer is 100 m long, has a T of 0.02 m2 min-1,
and a storage coefficient of 0.002. The head is initially uniform at 16 m (i.e., the initial
condition is h|x,0 = 16 m) and drops to 11 m at x = 100 at time 0 (i.e., the boundary
conditions are h|0,t = 16 m and h|100,t = 11 m).

Confining Layer

Aquifer

Figure 1. Wang and Anderson, 1982. Introduction to Groundwater Modeling. W. H. Freeman and
Company, San Francisco. 237 pp.

Plot the distance versus head for at least 4 times. Plot the numerical solution as open
symbols and the analytical solutions as lines.

References
Carslaw and Jaeger, 1959. Conduction of Heat in Solids. Oxford University
Press, New York. 510 pp.

Wang and Anderson, 1982. Introduction to Groundwater Modeling. W. H. Freeman and