Dimensioning urbanization – An advanced procedure for characterizing human settlement properties and patterns using spatial network analysis

Abstract The ongoing global phenomenon of people migrating to cities is referred to as urbanization and primarily manifests itself in the continuous and often rapid spatial expansion of urban agglomerations. Nevertheless, the dimension and structuring behind this process can be considered as a spatial continuum ranging from rural to urban settlements. Accordingly, gathering a detailed global knowledge about the size, form (e.g., compact or spread) and spatial distribution (e.g., dispersed or nucleated) of different types of settlements represents a major issue to better understand urbanization and develop effective mitigation, adaptation and management strategies. In such context, this paper introduces a novel procedure specifically designed to characterize settlement properties and patterns, which can be applied at high spatial resolution (hence being capable of accounting even for single villages and dwellings) from the local up to continental or global scale using binary maps (i.e., figure-ground diagrams describing the spatial distribution of built-up and non-built-up areas) derived from Earth observation (EO) products. Starting from a binary mask depicting built-up and non-built-up areas over a given study region, the proposed method first delineates settlement objects. Next, a spatial network is created where the nodes correspond to the centroids of the extracted objects and the edges connect neighboring objects lying within a prefixed Euclidean distance from each other. Suitable attributes describing the geometrical properties of the associated objects are then computed for all the nodes and specific weights of interest are assigned to the edges of the network. Finally, indexes modeling the relationships between different nodes are calculated to properly characterize the relevance of different settlements within the spatial network. Several experimental results obtained on the basis of figure-ground diagrams derived from existing EO-based geo-information layers from local to continental level assess the capabilities of the presented approach and demonstrate its potential to provide key information to quantitatively and qualitatively characterize settlement properties and patterns in any spatial detail (depending on the spatial resolution of the input data) and at arbitrary spatial scales.