Analyzing the Supply and Detecting Spatial Patterns of Urban Green Spaces via Optimization

Green spaces in urban areas offer great possibilities of recreation, provided that they are easily accessible. Therefore, an ideal city should offer large green spaces close to where its residents live. Although there are several measures for the assessment of urban green spaces, the existing measures usually focus either on the total size of all green spaces or on their accessibility. Hence, in this paper, we present a new methodology for assessing green-space provision and accessibility in an integrated way. The core of our methodology is an algorithm based on linear programming that computes an optimal assignment between residential areas and green spaces. In a basic setting, it assigns green spaces of a prescribed size exclusively to each resident, such that an objective function that, in particular, considers the average distance between residents and assigned green spaces is optimized. We contribute a detailed presentation on how to engineer an assignment-based method, such that it yields plausible results (e.g., by considering distances in the road network) and becomes efficient enough for the analysis of large metropolitan areas (e.g., we were able to process an instance of Berlin with about 130,000 polygons representing green spaces, 18,000 polygons representing residential areas, and 6 million road segments). Furthermore, we show that the optimal assignments resulting from our method enable a subsequent analysis that reveals both interesting global properties of a city as well as spatial patterns. For example, our method allows us to identify neighbourhoods with a shortage of green spaces, which will help spatial planners in their decision-making.