This dissertation documents the districting requirements for collection units for taking the
Canadian Census and provides a spatial decision support system for their automatic creation.
In the context of the literature on autodistricting, this problem falls under the general category
of creating districts for monitoring, surveillance and inventory applications since the Census
is essentially an spatial inventory exercise. The basic requirement is to create an area-based
categorical coverage such that the workload is equitably distributed amongst Census
Representatives within the limits of a large number of constraints and conditions.
A new omnibus automated districting process that combines a 3-stage cascading selection
procedure for identifying sub-blockface, blockface and block level collection units with a 4-
stage heuristic solution procedure for grouping blocks (termed 'assigns', 'annexes', 're-assigns'
and 'adjusts') is contributed by this research to provide a systematic response to varying
The resulting spatial decision support system for autodistricting has been tested on test data
sets and on one of the larger urban population centres of Canada. The set of test pattern sites
mimicking typical settlement patterns was generated to ensure that the various alternative
assignment or block grouping methods (i.e., unidirectional and bidirectional tessellations based
on circular and rectangular grids and regular, random and 'extrema-based' seeds) performed
as designed and specified. The Census Subdivision of Laval (in the Census Metropolitan Area
of Montreal) was selected as the test site for comparing the performance of the autodistricting
capacity to the actual, manually created, results from the 1986 Census.
To permit the comparison of results from classical manual and automated processes, a set of
satisficing evaluation functions that vary in accordance with data availability was implemented
in the context of a competing set of districting objectives. The most sophisticated of these
evaluation functions incorporates a composite index that combines the distribution and a
measure of the 'density' of the dwellings with the length of the route that must be followed to
complete the collection activity (including travel time to the start of the route and between route
To assess the continued acceptability of the districting from the previous Census, and/or to
select between alternative results generated by computer-assisted approaches, a set of
objective functions is provided that vary depending upon the available amount of geographic,
cartographic or statistical data.