Development of an areal object based network pattern classification in road extension simulation

Road development has been widely recognized as an important driving factor of land use/land cover change. A dynamic integrated road growth simulation is expected to enhance the performance of land change modeling, especially for a long-term projection. Lack of a systematic quantitative classification of road network patterns hinders the implementation of pattern control in road extension models, and the validation of model performance. This three-article dissertation examines the influence of pattern parameters in a road extension simulation and develops an innovative areal object based classification method to detect and distinguish various network patterns.;The first article develops a road extension simulation model. The road extension model is designed to simulate road growth following certain spatial arrangement. The challenges encountered in this model development leads to awareness of the need for empirically parameterizing the pattern control. The second article responds to the requirement for a systematic classification method. This paper proposes an innovative algorithm in detecting and distinguishing network patterns. Pattern is defined as a higher order phenomenon emerging from the repetition of a set of geographical qualities over space. As an assembled object, pattern is characterized not only by attributes of individual primitive objects but also the partonomic relationship between parts. In sum, the classification method consists of three steps: partitioning the space into small parcels, grouping parcels into pattern components, and finally merging similar components to form a distinct pattern. The third article presents an application of the classification method in a real network in Santa Cruz Department, Bolivia. The classification method creates a series of pattern maps in Santa Cruz Department. Each map presents the prominent patterns detected in corresponding scale. The series unfolds the hierarchical formation of patterns. The case study demonstrates the capability of the object-based classification for distinguishing more complex road networks in the real world, and provides insight into the understanding of the relationship between network pattern and socio-economic, political and environmental factors. In sum, the three articles of this dissertation systematically discuss the theoretical and practical roots, the algorithm, and the application of an areal object based road network pattern classification method.