March, 1982                                               786085

The equations formulated thus far apply only to a con-
fined aquifer. The program simulates unconfined and semi-
unconfined aquifers by addition of a step to account for
the change in transmissivity as the head at a given node
rises or falls. The equivalent transmissivity in the x and
y directions can be estimated by taking the geometric mean
of the transmissivities between the node in question and
the next adjacent node, as follows:
Ti,j,x = Ki yv(hi,j - BOTij) * (hi+1,j - BOTi+1,j)     (3.2)
Ti,j,y =  i,jy'vfhi,j - BOTi,j) * (hij+1 - BOTij+1)   (3.3)
By correcting the transmissivities at each timestep a rea-
sonable approximation of *the behavior of an unconfined
aquifer can be obtained. This approximation holds in areas
where the phreatic surface is not excessively steep and the
assumption of horizontal flow in the aquifer is bonored.
The predicted drawdowns from a pumping well are too large
in the immediate vicinity of the well but are very close to
the theoretical drawdowns a short distance from the well,
as shown in Section 5.0, Program Verification.
If the timesteps chosen for a simulation are small
with respect to the time frame of pumping or injection
sequences, the response of the aquifer is fairly consistent
through time.    That is, if the head at a given node is
dropping, then it is a good assumption that it will con-
tinue to drop during the next timestep.      This assumption
can be employed by use of a head prediction step prior to
solution of the finite difference equations at each time-
step, thereby significantly decreasing the iterations re-
quired for convergence.    It should be noted that the same

Golder Associates

786085

March, 1982