Fan, Xiangning and Atkinson, Peter and Whyatt, Duncan and Blackburn, George (2025) Exploring the Dynamics of Rapid Urbanization in Megacities using Remotely Sensed Time-Series and Object-Based Graph Structures. PhD thesis, Lancaster University.
![[thumbnail of 2025XiangningPhd]](https://eprints.lancs.ac.uk/style/images/fileicons/text.png)
Abstract
Urbanization has accelerated globally in recent decades, producing dramatic urban land expansion. With cities attracting attention from scientists from different fields, urban growth is a key topic, commonly studied using remotely sensed raster imagery. Urban entities such as cities are more readily thought of as spatial objects. Despite this, object-based methods are rarely applied in research on urban growth. This research utilized newly available remotely sensed annual land cover time-series data coupled with a novel object-based approach involving (i) raster-to-vector conversion, (ii) careful temporal linking of objects, (iii) comprehensive specification of the possible urban growth states (introduction, establishment, coalescence and no change) and (iv) the creation of a spatial graph structure linking neighbouring objects, to study urban growth. The stated object-based graph structure facilitated analysis of the state of urban objects based on previous states of the object and its neighbours, and the spatial-temporal links between them. First, the unprecedented scale of urban expansion between 1992 and 2014 was quantified across 13 regional capitals and their surrounding cities in China. By characterizing urban growth based on urban objects in different buffers at the regional level, the results suggested that core cities doubled or tripled in size, with synchronized growth patterns at specific times potentially driven by national and regional policies. Regional disparities were also observed which highlight the impacts of regional governance and local policy interventions. Second, a conceptual framework characterizing urban growth events was proposed including introduction (including through dispersal), establishment, coalescence and no change. Applying a rule-based approach to identify these events and quantifying their spatial-temporal changes, synchronous temporal trends in growth events in the core and buffer regions at the landscape level were observed. However, a specific logical sequence of these events at the population (or landscape) level was not obvious. The results show concurrent events with shifting dominance of specific events over time, thus, providing insights into urban growth processes and reflecting the complexity of urban growth processes. Third, a Bayesian linear mixed-effects model was integrated with a spatial-temporal graph of urban objects to model the states of urban objects. It was found that the coalescence state of urban objects is influenced by their prior object states, proximity to neighbouring objects, and the states of neighbouring objects in a defined small buffer. The growth state (i.e., growth or unchanged) is related to its previous state and the dynamics of neighbouring objects. The area of objects that have grown was found to be influenced by the largest interactions with neighbouring objects, with the magnitude of these effects varying by object size. By modelling explicitly the relationships between urban objects on a graph, the developed object-based approach provides valuable insights into the dynamics of urban objects and their relationships within megacities, using cities in China as examples. This research advances the understanding of urban growth by quantifying spatial-temporal patterns, building spatial-temporal links between urban objects, and explicitly modelling the relationships between objects. It provides a new perspective for studying urban dynamics and may contribute to better urban development, governance strategies and sustainable environment management in future.