Dissecting the Spatial Structure of Cities from Human Mobility Patterns to Define Functional Urban Boundaries

TL;DR

This paper introduces a spectral analysis-based method to define functional urban boundaries using human mobility data, revealing distinct scaling behaviors in urban and rural areas and improving city delineation beyond administrative borders.

Contribution

It presents a novel network centrality approach to distinguish urban from rural land-uses based on human mobility patterns, enhancing city boundary definitions.

Findings

Urban areas exhibit near-linear scaling of trips with population.

The method identifies urban boundaries by clustering places with significant mobility flows.

Chilean cities show clear differences in trip scaling between urban and rural regions.

Abstract

Since the industrial revolution, accelerated urban growth has overflown administrative divisions, merged cities into large built extensions, and blurred the boundaries between urban and rural land-uses. These traits, present in most of contemporary metropolis, complicate the definition of cities, a crucial issue considering that objective and comparable metrics are the basic inputs needed for the planning and design of sustainable urban environments. In this context, city definitions that respond to administrative or political criteria usually overlook human dynamics, a key factor that could help to make cities comparable across the urban fabric of diverse social, cultural and economic realities. Using a technique based on the spectral analysis of complex networks, we rank places in 11 of the major Chilean urban regions from a high-resolution human mobility dataset: Official origin-destination (OD) surveys. We propose a method for further distinguishing urban and rural land-uses within these regions, by means of a network centrality measure from which we construct a spectre of geographic places. This spectre, constructed from the ranking of locations as measured by their approximate number of embedded human flows, allows us to probe several urban boundaries. From the analysis of the urban scaling exponent of trips in relation to the population across these city delineations, we identify two clearly distinct scaling regimes occurring in urban and rural areas. The comparison of our results with land cover derived from remote sensing suggests that, for the case of trips, the scaling exponent in urban areas is close to linear. We conclude with estimations for well-formed cities in the Chilean urban system, which according to our analysis could emerge from clusters composed by places that capture at least ~138 trips (over the expectation) of the underlying mobility network.