Geographical Information Systems (GIS) are used for several types of spatial planning but so far they have not been used for optimal patch design. Optimal patch design is a generic spatial problem in which the objective is to design spatially explicit landuse maps when both the composition and configuration of patches are important criteria. There are many applications in conservation, forestry management, watershed management and the management of large military estates. This thesis describes a new autonomous computer program, the genetic algorithm for optimal patch design (GAPD). GAPD combines four components: a genetic search algorithm, a parameterised region growing (PRG) program, raster GIS measurement functions and multi-criteria decision-making methods. The key component is the PRG which translates between the aspatial domain of the search algorithm and the spatial domain of the GIS. GAPD generates landuse maps that optimise the configuration and composition of patches to meet multiple objectives for a given set of input maps and criteria. The theories of landscape ecology are used to establish a framework for formulating optimal patch design problems. The thesis describes the conceptual design of GAPD and its implementation and test, first as a prototype for solving single patch problems and then as a fully functional system for solving multi-objective multi-patch problems. The feasibility of GAPD was established by investigations of issues concerning the representation and measurement of configuration in raster data structures and by testing the efficiency and effectiveness of GAPD with simple problems. GAPD was further evaluated in five hypothetical problems designed to cover a range of different scenarios. The results are promising and show that GAPD has potential as a decision support tool. The final section recommends a number of topics for further research covering technical developments of GAPD, the application of GAPD to real problems and investigations of general issues of optimal patch design.