E 
NSIN 
tern Wisconsin 
uifer thickness, 
analyzed using 
ion in a spatial 
ets in a suitable 
cilitated by the 
e GIS to model 
/here they were 
er levels and 
t the effect of 
ifer on ground- 
del used in this 
inite-difference 
monly referred 
and Harbaugh, 
the study area 
e cells of finite 
| the better the 
ning measured 
1s are available 
aulic properties 
y solving the 
er flow in each 
vater levels and 
1 is considered 
ed water levels 
with measured 
input values of 
aquifer system 
bration process 
, of the model 
With the advent 
put-data sets for 
been greatly 
es how a GIS, 
stems Research 
Institute Inc., 1991), was used to assemble 
input-data sets for and analyze results from a 
finite-difference model of ground-water flow in 
northeastern Wisconsin. 
The study area is a 26,000 km? area 
which includes Lake Winnebago, part of Lake 
Michigan, and the major ground-water 
pumping centers of Green Bay and the Fox 
Cities. The area was discretized using 
commands within the GIS and a Fortran 
program (D.O. Winkless and J.M. Kernodle, 
U.S. Geological Survey, written communica- 
tion, 1994). The gridded area is georeferenced 
and is represented as point and polygon data 
layers. The grid spacing is smaller in areas of 
large withdrawals of ground water to improve 
simulation of the effects of pumping on 
ground-water levels and flows. Row and 
column numbers, and a sequence number are 
associated with each cell and are used to move 
data between data layers and the model. The 
model grid used in this study contains 141 
rows and 100 columns (14,100 cells). The 
smallest cell area is approximately 1 km?; the 
largest cell area is 165 km?. The grid is 
oriented so that rows are parallel to ground- 
water flow. 
The model requires information about 
each cell including cell type (active, inactive, 
boundary cell properties), an initial water level, 
thickness of the aquifer units, leakances 
between aquifer layers, recharge rate, and, if 
necessary, river properties and well pumpage. 
This information is read by the model as two- 
dimensional arrays for properties, such as 
thickness or water levels, and as single values 
identified by model layer, row and column for 
data sets containing well and river information. 
2. METHODS 
Hydrogeologic data were entered into the 
GIS by digitizing maps, modifying existing 
GIS data layers for the study area, and entering 
point data from USGS data bases. By 
intersecting data layers containing the 
hydrogeologic data with the grid layer, each 
cell was assigned hydrogeologic properties. 
69 
Calculations for each cell were made within 
the GIS to compute thickness from altitudes, or 
leakances from area, thickness and hydraulic 
conductivity values. The graphic capabilities of 
the GIS provided the opportunity to check the 
spatial distribution and magnitude of aquifer 
properties and edit the data layers, if necessary, 
before building the model input-data sets. Once 
the correct values were assigned to each cell, 
files in a format suitable for MODFLOW input 
were created from the GIS. Two dimensional 
arrays were written to a file using a 
combination of GIS commands and a Fortran 
program. Files for river and well data were 
made by outputting the layer, row, column, and 
values using GIS commands. 
3. RESULTS 
Using a GIS greatly aids in compiling 
data for use in regional ground-water flow 
models. The ability to graphically display and 
interactively edit the information provides an 
efficient method to create large data sets that 
are spatially correct. Transferring contour data 
from paper maps to a gridded digital format is 
easily accomplished and verified. Calculations 
of model input parameters are efficiently 
performed in the grid layer. 
Assigning an accurate altitude to river 
reaches within cells, which is required in 
MODFLOW, was the most difficult step to 
automate because in some locations the 
altitudes assigned from a digital elevation 
model (DEM) resulted in a gradient of the river 
that is opposite of the actual gradient. The 
altitude of the river was assigned to each cell 
containing a river reach by converting the lines 
representing rivers to points and assigning an 
altitude to the points from a DEM of the land 
surface. The altitudes of all reaches within a 
cell were compared and the minimum altitude 
for each cell was assigned to the river reaches 
in the cell. The incorrect altitude was the result 
of sample density of the DEM, cell size or 
complexity of topography. A low sample 
density of the DEM may not provide the 
resolution necessary to accurately represent the 
gradient of linear features such as rivers.