Delineating Neighborhoods: An Approach Combining Urban Morphology with Point and Flow Datasets

Although neighborhoods are a widely used analytical concept in urban geography, they are often proxied using grids or statistical sectors in empirical research. The rationales underlying these proxies are often separated from the theoretical considerations of what makes a neighborhood a neighborhood, casting shadows over their relevance and applicability. In this article, we identify two specific challenges separating empirical operationalizations from theoretical considerations in neighborhood delineations: (1) not incorporating key built environment elements and (2) monodimensional approaches. We develop a method that addresses this double challenge by (1) creating morphological basic spatial units (BSUs) and (2) aggregating them into neighborhoods using multilayer community detection (MLCD) drawing on datasets used in both formal and functional regionalization approaches. We illustrate this method for the case of Leuven, Belgium, by (1) using street blocks as BSUs and (2) focusing on proximity, land use, and social interactions. Through a comparative analysis, we show that our results align with theoretical considerations and perform as well as, and perhaps better, than statistical sectors and grids as neighborhood representations. We therefore argue that this flexible method can bridge formal and functional regionalization approaches making the case for its adoption in neighborhood delineation exercises.