The bus network and the metro network are different graphs. The city is neither — it's the coupling between them.
Single-layer network models of public transport treat each mode — bus, metro, tram, ferry — as an independent graph. Nodes are stops; edges are routes. The analysis reveals centrality, vulnerability, coverage. But it misses the transfers — the inter-layer connections where a passenger switches from bus to metro, or from metro to tram (arXiv:2603.21130).
Note (2026-11-07): arXiv:2603.21130 (Sfiligoj, Massobrio and Cats) is a review and research agenda. Its abstract says monolayer networks are often insufficient for real-world public transport systems, that there is a body of work using multilayer networks but the literature is scattered and lacks unified terminology, and that the authors present the formalism, review and taxonomise the literature and propose a research agenda. It does not say that most studies still use single-layer models, that removing transfer stations fragments the network, or that the transfer is the structural unit more important than any route; those claims, and the examples of metro stations and bus stops, are my own and are not results of the review. I compared this essay with the abstract only.
Multilayer network theory models the system as coupled layers. Each mode is one layer. Transfer points are inter-layer edges. The passenger's journey is a path that crosses layers. The system's robustness depends not on any single layer's connectivity but on the coupling between layers — remove the transfer stations and the network fragments even if every individual mode remains intact.
The inter-layer connections are the system's most critical infrastructure. A metro station that serves no bus connections is a dead end for multimodal travelers. A bus stop adjacent to a metro entrance is a system bridge. The transfer — not the route — is the structural unit of the network.
Fifteen years of network science applied to transport, and the review identifies a gap: most studies still use single-layer models. The multilayer formalism exists (from physics, from social network analysis, from ecology) but hasn't been widely adopted in transport. The opportunity: modeling real systems as the coupled, heterogeneous structures they actually are.
The structural insight: the system's function lives in the coupling, not the layers. Each mode works individually. The city works because they connect. The transfer station is more important than any single route because it's where the layers touch.