Doctor of Philosophy
School of Earth and Environmental Sciences
Wickramasuriya Denagamage, Rohan Chandralal, Developing and testing new spatio-temporal modelling techniques to explore the dynamics of changing rural landscapes in Australia, Doctor of Philosophy thesis, School of Earth and Environmental Sciences, University of Wollongong, 2011. http://ro.uow.edu.au/theses/3515
Amenity-led counterurbanization is driving a multifunctional transition in rural lands of Australia. Fundamental to this transition is the subdivision of large farming properties as traditional farmers exit agriculture, acquisition of subdivided lots by affluent and mobile urban populations that seek a lifestyle change in rural areas, and diverse land management activities adopted by an increasingly heterogeneous mix of land owners. This transition poses new socio-economic and environmental challenges. A better understanding of these changes and their implications is essential to manage the transition with the least negative impact on socio-economic and environmental conditions. This is quite challenging given the complexity of processes and heterogeneity of actors involved in changing rural landscapes.
Spatial simulation modelling techniques, particularly agent-based land use/land cover (LULC) modelling have been applied elsewhere to understand such complex dynamics. However, existing models cannot be directly applied to study Australian amenity landscapes mainly because of the inadequacy of structure, system elements and encompassed processes that limit the ability of such models to sufficiently describe amenity landscapes and their dynamics. Based on these premises, the broad aims of this thesis were to 1) develop, verify and validate new tools and methods to address critical gaps in existing spatial simulation modelling practices that limit how amenity migration researchers and land use modellers are able to frame their research questions, and 2) demonstrate the use of newly developed tools and methods for amenity migration research in particular and LULC modelling research in general.
This thesis first developed a novel, fully-automated land subdivision tool that uses vector data and is capable of generating complete subdivision layouts with both lot and street arrangements for land parcels of any shape. When the new streets are generally parallel to each other and lots are of approximately the same size, the simulated subdivision layouts very closely resemble observed subdivision patterns in the south-eastern Australia study area. From this validation exercise, the opportunities to improve the subdivision tool for a next version were also identified.
The thesis then presents an innovative method to simulate land subdivision in LULC change models using subdivision layouts generated either by the tool developed in this thesis or by alternative tools. This method is demonstrated with a prototype agent-based LULC change model developed for an amenity landscape (Windellama, New South Wales, Australia). Central to the success of the proposed method is implementing a hierarchical landscape where adjacent cells of the same LULC type form patches, patches form properties and properties form the landscape. The second key element of the method is incorporating real subdivision layouts. The necessary addition of new streets during a subdivision changes both the LULC type and the LULC patches to which some cells belong. An innovative queue-based modified flood-fill algorithm is used to reset LULC patches following a subdivision. Results show that this algorithm is computationally efficient even as spatial resolution is increased. Achieving the right balance between computational efficiency and detail of representation is crucial in LULC change models, and the results show that this balance is reached at 50m resolution for the modelled area.
Finally in the thesis, a comprehensive agent-based LULC change model is developed around the previously described prototype model. This model includes six agent types (graziers, hobby farmers, green lifestylers, non-farming retirees, absentees and real estate agents), simulates realistic land subdivision, and incorporates a detailed endogenous land market. The agent types show clear differences in the socioeconomic attributes and land management strategies they adopt. Verification shows that this model is robust, while validation gives confidence that the representation of land use dynamics for the main LULC classes is sufficiently realistic. Long term experimental simulations with the model reveal that internal buying and nonmonetary conditions imposed on land transactions have a significant effect on LULC and demographic change trajectories in amenity landscapes. Experimental results also suggest that the size of lots decided at the subdivision stage can alter the ownership composition in amenity landscapes in the long run, and thereby affect the course of LULC change.
In summary, the tools, methods and models developed in this study lay the groundwork for amenity migration researchers to gain new insights into changing rural landscapes by enabling them to frame research questions that would have been impractical to answer if not for these key innovations. Importantly, the technical innovations introduced in the thesis are pertinent to entire land use modelling discipline, thus demonstrating the broad applicability of the findings of this thesis.