Data-driven approaches to derive parameters for lot-scale urban development models

For assessing the performance of urban infrastructures over long time horizons of 30-90 years, urban development models are desirable that explicitly represent the physical layout of the city, while confining model complexity to an appropriate level. Such models have recently appeared in the literature, but parameters were often defined ad-hoc or without documentation. This paper presents approaches to derive important parameters for such models based on commonly available data. We apply logit regression models considering four high-level characteristics of the urban landscape (distance from main station, accessibility to motorway, accessibility to marine and green spaces) to predict location of urban development for different building types, estimate characteristic building footprint and floor space areas for different building types depending on their location in the city, derive building coverage ratios using Voronoi polygons and estimate the number of buildings in new developments using hierarchical clustering. The applicability of all methods is demonstrated in a case study in Odense, Denmark. The derived parameters are case-specific, while the methods can easily be transferred to different case studies.