The methodology is best described by a walk through of the cartographic model linked at the bottom of this page.
A region was established to the west
of the Moraine in which boreholes were to be examined. Boreholes in the
area were manually queried for sand and/or gravel underlying clay at the
surface by examining them in the Borehole Mapper accessory for MapInfo
and were entered into a new database. Each of the boreholes in this database
were in turn queried to determine to thickness of the sand/gravel unit
which was then entered as a new attribute. Where there was more than one
borehole at a coordinate (see Data Sources and Concerns)
data was aggragated. The number of bores at a point was given an interger
value and the thickness of the sand/gravel units was averaged. This ensured
that there was only one data entry for each coordinate. The original database
of all boreholes in the area was also aggragated to give an integer value
of the number of boreholes at each coordinate. This produces two data tables,
one of the boreholes selected for sand/gravel and one of all the boreholes.
These two tables were converted to
IDRISI vector format from MapInfo TAB format using FME suite.
Two point data vector files were assigned
with the thickness and number of coincident points of the boreholes selected
for sand/gravel. The point data vector file of all the boreholes in the
area was assigned with the number of data entries at each point. These
three vector files were rasterized to a 30x36 grid as this size grid was
found to best represent gaps in the available data as well as incorperating
more than one point data entry per raster cell to generalize the data somewhat
and hopefully minimize the error. In the process of rasterizing it was
neccessary to overlay the sum of the values of all the points in each cell
with the number of points which fell into that cell to provide an average
value for the cell.
Three surfaces were created from the
rasterized data to produce weighting surfaces for use in a multi-criteria
evaluation. Surfaces were created to show: the proportion of boreholes
in each cell with sand/gravel in relation to the total number of boreholes
for that cell, the average thickness of the sand/gravel units in each cell,
and the density of available data for each cell, ie. the total number of
boreholes in each cell.
The proportion of boreholes in each
cell with sand/gravel was calculated by dividing the number of boreholes
with sand/gravel in each cell by the total number of boreholes in each
cell. The resulting values were multiplied by a boolean surface with values
of 1 for cells with boreholes with sand/gravel and zero where there are
none in order to give a value of zero where there are no boreholes with
sand/gravel.
The average value of the thickness
of the sand/gravel units in each cell was calculated by dividing the sum
of the thicknesses by the total number of boreholes with sand/gravel. The
resulting values were then multiplied by the boolean surface of presence/absence
of boreholes with sand/gravel to give a value of zero where there are no
boreholes with sand/gravel.
The total number of boreholes in each
cell was calculated by rasterizing the point data of the total number of
boreholes at each point and summing the values.
The proportion of boreholes in each
cell with sand/gravel was fuzzed to give a byte value to all the cells
so that they could be used as a weighting surface in a multi-criteria evaluation.
The function by which the values were transformed was chosen as user-defined
in order to make the best use of the data. The histogram of the data was
used to determine the values used to define the curve.
The average value of the thickness
of the sand/gravel units was also fuzzed using a user-defined function.
Values less than one metre were given very small weights because they do
not represent a likely area where transport of material occured. Values
greater than two metres were given higher weights increasing linearly to
the maximum as two metres represents the smallest likely thickness representing
an area where transport likely occured. The histogram was again used to
asses the distribution of the data and aid in the creation of the transfomation
curve.
The total number of boreholes in each
cell was fuzzed using a user-defined function where values for the transformation
curve were determined from the histogram. Since almost all values were
less than 20, but ranged as high as 980, values below 20 were weighted
relatively heavily to make the best use of the unevenly distributed data.
The relative weights of the three
surfaces to be used in the multi-criteria evaluation were done in three
ways. For the first trial, relative thickness was given the highest weight
in a scenario where the thickness of the units most heavily influenced
the likelyhood that transport took place within that cell. The second trial
used even weights for all three surfaces to serve as a sort of baseline.
The last trial used the proportion of boreholes with sand/gravel as the
most heavilly weighted surface where this proportion had the greatest influence
on the likelyhood of transport having occured within that cell.
These weights and surfaces were then
used in a multi-criteria evaluation using a constraint where cells with
no boreholes containing sand/gravel were eliminated from the evaluation.