Spatial heterogeneity assessment of factors affecting sewer pipe blockages and predictions.

Abstract Efficient management of sewer blockages requires increased preventive maintenance planning. Conventional approaches to the management of blockages in sewer pipe networks constitute largely unplanned maintenance stemming from a lack of adequate information and diagnosis of blockage causative mechanisms. This study mainly investigated a spatial statistical approach to determine the influence of explanatory factors on increased blockage propensity in sewers based on spatial heterogeneity. The approach consisted of the network K-function analysis, which provided an understanding of the significance of the spatial variation of blockages. A geographically-weighted Poisson regression then showed the degree of influence that explanatory factors had on increased blockage propensity in differentiated segments of the sewer pipe network. Lastly, blockage recurrence predictions were carried out with Random Forest ensembles. This approach was applied to three municipalities. Explanatory factors such as material type, number of service connections, self-cleaning velocity, sagging pipes, root intrusion risk, closed-circuit television inspection grade and distance to restaurants showed significant spatial heterogeneity and varying impacts on blockage propensity. The Random Forest ensemble predicted blockage recurrence with 60–80% accuracy for data from two municipalities and below 50% for the last. This approach provides knowledge that supports proactive maintenance planning in the management of blockages in sewer pipe networks.