A spatial data-driven urban pattern language framework for design and planning

Urban design and planning practices traditionally use a survey-analysis-design paradigm. Data are collected through site surveys followed by comprehensive analysis to generate knowledge for a context-specific design scheme with different professionals working on varying levels of detail. Nowadays, the practice has entered the age of digitalisation with the transition from traditional, empirically driven approaches to more dynamic, data-informed strategies where new forms of data and technology enable the analysis of urban forms at varying scales. This shift is not merely technological but philosophical in the design and planning practices, heralding a new era where urban design reflects and anticipates human and sustainability requirements. Quantitative representation and understanding of urban form and its cross-scale analysis have become an essential premise for this new practice. Building upon this foundation, this thesis, drawing on Christopher Alexander’s idea of a pattern language, proposes a novel framework for urban design analysis, advocating for a multi-scale pattern approach. This framework aspires to bridge the gap between abstract urban planning theories, the tangible realities of urban development, and the emerging new technology and data. At the heart of this framework lies the adoption of a pattern language approach, which aims to distil complex urban environments into universally workable abstractions through quantitative urban patterns, facilitating easier analysis and understanding. Hence, it promises to enhance how urban landscapes are depicted, analysed, and designed.
Built on the fundamental rationale of urban morphology that physical forms of urban elements are important manifestations of social and economic processes, the thesis deploys a range of quantitative techniques, including network science and machine learning, to explore the complexity of urban morphology, caters to urban professionals and encompasses three key scales of urban patterns: Macro, Meso, and Micro. At the macro level, policymakers and urban planners use spatial and master plans to shape the metropolis, focusing on urban elements like density and compactness, which are essential for sustainable and resilient development. The meso scale dives into district or neighbourhood planning, where urban designers prioritise street layouts and the connectivity of urban spaces to foster spatial harmony. On the micro-scale, urban designers and architects dive into the block level to create functional and aesthetically pleasing buildings and public spaces, utilising patterns like the topology of urban blocks for detailed design and analysis. Collectively, these scales together offer a coherent and comprehensive view of urban form, from the broad distribution of city elements to the intricate designs of individual buildings and their surroundings.
The thesis unfolds across several chapters, each dedicated to a specific scale of analysis and a unique set of research objectives. The research spans various case studies of cosmopolitan cities around the world. They are selected based on the diversity of urban morphology they represent and data availability.
Beginning with the macro scale, the study examines urban spatial structures through the lens of polycentricity, utilising urban mobility data to simulate and reveal the polycentric nature of urban environments with the network science and spatial interaction model. This part of the thesis evaluates the master plan based on a polycentric structure derived for both current and anticipated future scenarios. We present a framework leveraging quantitative measurement and open data to evaluate a master plan’s influence on urban polycentricity. The results demonstrate that Singapore’s master plan has effectively reinforced its sub-centres, while its central business district remains the dominant centre. This exploration highlights the distribution of urban elements and reveals the role of polycentricity as the macro scale pattern in fostering balanced and sustainable urban development.
Moving on to the meso scale, the thesis investigates identifying and quantifying street patterns with two machine-learning approaches, supervised and unsupervised learning, to classify and analyse complex street networks with help from an adaptive unit of analysis: the street-based local area (SLA). Supervised learning follows a standard machine-learning classification based on pre-defined patterns; they enabled the identification of street patterns at large scales with efficiency. The unsupervised learning provides a data-driven analysis of street morphology, offering insights into street patterns beyond conventional understanding and is adaptive to varying urban contexts. In this thesis, supervised machine learning classifies street networks into Gridiron, Organic, Hybrid, and Cul-de-sacs patterns. Findings show distinctive spatial structures, such as ring formations and urban cores, indicating stages of urban development and socio-economic narratives. The unsupervised machine learning suggests a complex hierarchy of street patterns, challenging traditional classifications and offering a more comprehensive categorisation that captures the subtle variances in street morphology. This research phase brings to light the diversity of street configurations and their profound influence on the urban spatial structure, emphasising the potential of technology-driven methodologies to revolutionise urban morphology analysis.
At the micro-scale, the study applies the SpaceMatrix method in conjunction with clustering techniques to dissect individual plots and structures. It investigates paradigm shifts in plot-level planning through an in-depth case study of Singapore; we categorise Singapore’s towns into four distinct clusters: Suburban, Balanced Mix, Dense Urban, and Vertical Growth, each reflecting unique density patterns and building forms. This clustering reveals how Singapore’s planning ideologies have transitioned from maximising space utilisation to prioritising sustainability and quality of living. The insights from the clustering analysis enhance our understanding of Singapore’s exceptional urban path and offer valuable perspectives for urban morphology, informing micro-level design and policy decisions with a solid data-driven foundation.
The research of this thesis culminates in creating a detailed urban morphological framework that combines insights from various urban scales, utilising a pattern language approach to provide a cohesive understanding of complex urban forms and their interrelationships. Through a case study comparison of two distinct cities, the research identifies predefined and linked urban patterns at varying scales and addresses two hypotheses: first, that urban landscapes can be deciphered into discernible patterns that are not random but follow specific rules; second, that the relationships between these patterns are unique to each city, informed by their individual histories and needs. The study relies on detailed geoinformation of urban elements and data science-based techniques. It could elucidate the rules governing the unique arrangement and interrelation of urban patterns more comprehensively than before in different cities.
The final section of the thesis recognises that the pattern language approach for urban analysis is still evolving. It suggests that this approach could be enriched by including a more comprehensive array of urban elements, such as dynamic human activities, to paint a fuller picture of urban morphology. Additionally, the approach could be extended to assess urban form performance, building on the performance anecdotes provided. In essence, the thesis contributes to urban analytics, design, and planning by promoting a data-centric, digitally forward method that lays the groundwork for future urban planning initiatives to foster a sustainable and lively built environment.