Increasing recognition of the extent and speed 
of habitat fragmentation and loss, particularly 
in the urban fringe, is driving the need to 
qualitatively and quantitatively analyze regional 
landscape structure for decision support in land 
use planning and environmental policy 
implementation. In this thesis, a framework for 
addressing such needs is proposed by introducing 
an extension to the spatial data models currently 
in use for geographic spatial analysis. Existing 
approaches from the Geographic Information Systems 
(GIS) and Computer Aided Drawing (CAD) literature 
are reviewed and found wanting in the kinds of 
spatial analysis tasks required in this problem 
domain. A methodology is proposed to address a 
specific need for the rational design of a landscape 
scale conservation network. In this approach the 
problem is framed as a multi-objective combinatorial 
optimization problem. A prototype of this methodology 
is developed which serves as both a demonstration 
application of the optimization techniques and an 
implementation of parts of the proposed spatial data 
model on a real world problem with real world data. 
Specifically, a Non-dominated Sorting Genetic Algorithm 
(NSGA) is applied to the structured data to generate 
a set of Pareto-optimal solutions where each solution 
represents a different configuration of features on 
the landscape and the objective functions consist of 
metrics derived from principles of design suggested in 
the landscape ecology literature. A key theme in the 
work is translating between representations. Data is 
transformed from the attributed planar graphs of the 
source data, to these graphs' attributed dual (adjacency) 
graph, to an indexed string (character vector/hyperedge) 
and back to an attributed planar graph. Each representation 
is the most natural or useful for a particular stage of 
analysis. For example, the current configuration of 
ecological and social features is coded as colourings on 
the vertices of the dual graph of the source data which 
are available as planar graphs that cartographically 
represent area features. The prototype implementation of 
this framework was created and tested on a set of small, 
medium and large sized input data. The performance of the 
framework and its implementation are evaluated.