6.0 MODEL CALIBRATION
Due to the variable nature of the aquifer over the
entire study site, some parameters in some areas of the
model must be inferred. This is done by modifying the pa-
rameters in a logical fashion until the predicted results
adequately match those observed in the field. This process
is called calibration and is critical to the accuracy of
the predictions based on the calibrated model. If too many
parameters are inferred or set through calibration an un-
realistic parameter set could be selected. This could re-
sult in a model calibration which matches the predicted and
the observed results adequately but behaves erroneously
when changes are imposed.    Therefore, measured field data
should be used as extensively as possible and unsupportable
calibration refinements should be avoided.
The head simulation and the seepage transport algo-
rithms may require calibration before using the model. The
head solution is calibrated by determining as many of the
aquifer parameters as possible through pumping tests, ex-
ploration borings, regional water balances, and other ap-
plicable  sources.    The remaining parameters   are set to
their most likely value. The model can then be calibrated
by modifying parameters until the predicted head levels
match observed heads in a steady state mode.       The final
head configuration is usually stored on a RESTART file to
use as the initial heads in subsequent simulations.       The
transient behavior of the model should be calibrated if
adequate observed data is available. Transient phenomenon
such as spring snow melt or pumping wells are commonly
used.   If transient observed data is not available the
model should still be checked with a reasonable estimate of
seasonal variation of infiltration to insure that no ex-
treme fluctuations take place in the model.

Golder Associates

786085

March, 1982