Spatial Clustering Analysis of Water Main Break Events

In this paper, we consider drinking water pipe breakage as a process in space and aim to identify patterns that can subsequently be used to improve the understanding and modeling of the breakage process. For instance, patterns obtained by spatial clustering algorithms might indicate spatial correlation or lead to the identification of good predictors of breakage within clusters. More specifically, our goal is to assess the efficacy of applying spatial clustering methods to the exploratory analysis of water distribution pipe breakage. This study indicates that the DBSCAN algorithm, which is a density based clustering method, when adapted to consider network shortest path distances, is suitable for identifying spatial clusters of breaks, based on the density of the spatial distribution of break events. An analysis of the limitations created by the subjective nature of parameter choice, and the potential usefulness of the clustering results, is provided along with future work directions.