Probabilistic Urban Futures: Combining expert knowledge and data in Bayesian Network models for Urban Growth

Primary supervisor

Steven Mascaro

Co-supervisors

  • Bradley Rasmussen

As cities face unprecedented growth, the need for tools that can integrate diverse knowledge sources ranging from geospatial data to the nuanced intuition of urban planners is critical. This research will explore how Bayesian Networks can be adapted to serve as configurable, transparent models that empower decision-makers to weigh alternatives involving complex factors such as development yield, urban zoning, locality to services and infrastructure capacity.

The project aims to bridge the gap between complex data science and practical urban planning by developing a robust, probabilistic Bayesian Network (BN) framework for predicting urban growth. This research addresses the need for tools that can adapt to the limitations of representing complex systems. By validating how expert judgement can be encoded into BN models, and influences their conclusions, this project will contribute to a new generation of decision-support tools that are interpretable, configurable, and capable of informing evidence-based decision-making in real-world planning scenarios.

    Aim/outline

    Bayesian Networks have traditionally been used in land use modelling to handle uncertainty and causal relationships. There are interesting theoretical questions regarding how to best balance "hard" data (which has many limitations) with "soft" expert judgement (which captures tacit knowledge) in these models. Importantly, there is still little understanding of how to interpret the conclusions from models with diverse knowledge sources for decision-making.

    An initial base deterministic urban development BN model has already been established. The challenge of this research is to 1) evolve this foundation into a fully probabilistic model that accurately reflects the stochastic nature of urban systems and 2) develop techniques for understanding how the data and expert input affect the uncertainty in model conclusions and the implications for decision-making.

    Key Research Tasks:

    • Methodological Investigation: Investigate and compare two distinct approaches to populating the BN
      • Expert Elicitation: Designing protocols to capture structure and priors through structured interviews with urban planners (incorporating domain expertise).
      • Data Learning: Utilizing machine learning techniques to learn structure and parameters directly from existing geospatial and open datasets (incorporating data-driven evidence).
    • Sensitivity Analysis & Robustness: Determine which variables (e.g., heritage detractors vs. transport accessibility) most significantly impact the "propensity to develop" and assess the overall robustness of the model. Develop techniques for describing how model outcomes are affected by expert uncertainty and data variability

    Required knowledge

    • Background in Data Science, Computer Science, Geospatial Information Systems (GIS), or Urban Planning
    • Background in applied statistical methods and analysis
    • Interest in probabilistic graphical models (Bayesian Networks) and machine learning
    • Strong programming skills (Python/R)
    • Ability to communicate complex technical concepts to non-technical stakeholders (facilitating expert elicitation)