Data structures and methods for reproducible street network analysis: overview and implementations in R

• DOI: 10.31219/osf.io/78yub
• Publication Date: 2020-09-25T15:13:51Z
• Authors: Andrea Gilardi; Robin Lovelace; Mark Padgham
• Keywords: Programming Languages and Compilers; Databases and Information Systems; Computer Sciences; 11. Sustainability; Physical Sciences and Mathematics; 0101 mathematics; 01 natural sciences

Emerging datasets, new methods and shifting transport planning priorities have sparked a renewed interest in the analysis of street networks, linear geometries and attributes representing roads and other transport ways at local, city and regional scales. This paper outlines data structures, vital pre-processing steps and analysis methods using open source software. While the concepts and datasets presented are language agnostic, we implement them in R, an increasingly popular language for geographical and network data analysis. Two approaches are compared: one based on the stplanr package, which contains igraph and sf objects, and the other based on the dodgr package, which defines the dodgr_streetnet class, extending R’s mature data.frame class. A range of datasets, from a roundabout to an entire city network, are used to highlight the main features of each approach. We find that both enable street network analysis tasks, including shortest path and centrality measures. We conclude that the merits and potential pitfalls of the graph-based and data.frame-based approaches should be considered before embarking on street network analysis or software development. Finally, the discussion provides a basis for research using these data structures.