A Multi-Scale Flexible Framework for Urban Modelling

Abstract

The configuration of urban areas, and of infrastructures which serve them is central to managing the urbanisation process. Integrated assessment frameworks aim to inform decisions regarding planning, policy, and design to coordinate projects across sectors. Development of such models poses a number of challenges; (i) scenario generation, (ii) intelligibility to stakeholders, (iii) validity, (iv) control and feedback, (v) execution time, (vi) data requirements, (vii) uncertainties and, (viii) flexibility/reusability. This research has developed a multi-scale flexible framework which disaggregates projected regional employment to ward-level population, and further to rasterised development. This comprises; (i) transport network generalised cost, (ii) cost composition, (iii) spatial interaction incorporating transport accessibility, (iv) development zoning, (v) multi-criteria evaluation of development suitability, and (vi) cellular development. The framework is generically implemented, each model being specified in terms of inputs, outputs, and parameters. Modellinkage is via input/output chaining, providing the opportunity to experiment with alternative solutions. Execution is flexible/configurable to perform multiple model runs whilst varying parameters and propagating metadata through stages. Python controls execution flow, C++ provides performance, PostgreSQL manages data, and QGIS assists input/output. The framework is deployed in baseline scenarios for London and Innsbruck, and in more detailed scenario/uncertainty exploration for London. The framework’s utility is judged by criteria corresponding to the above challenges and is found to be favourable, with performance, flexibility and uncertainty support as key attributes. The framework executes models for London in ~52 seconds on modest hardware (1.6GHz, 8GB). This involves costweighted Dijkstra - 4 transport networks (~42s), cost composition and accessibility conversion (~4s), spatial interaction - 633 wards (~2s), rasterised 4-hectare development zones (~1s), 7 criteria development suitability evaluation (~1s), and cellular development - 100m scale (~2s). Combinatorial uncertainties are accommodated by a flexible, modular structure which promotes reuse, and records run configuration as well as model parameters in chained metadata