A state-of-the-art review for the prediction of overflow in urban sewer systems

Sewer overflow (SO) is becoming a concerning issue since discharged wastewater contains toxic substances and debris resulting in hazardous pollution to the surrounding environment and water quality degradation; and spilled stormwater may cause localized flooding and even back-up into buildings. Therefore, it is necessary to predict the occurrence of SO in advance, which enables the utilities to post warnings, prioritize the resource allocation and take proactive measures to minimize negative effects on environment and society. This paper aims to provide a state-of-the-art review for the prediction of sewer overflow which is lacking in literature, including bibliometric survey, scientometric analysis, in-depth systematic review, and elucidation of the existing research gaps and the potential future research directions. The findings reveal that the majority focuses on combined sewer overflow (CSO), and artificial intelligence-based models are the most popular ones. The input factors vary widely among three model categories. Volume, likelihood of occurrence and water level are the three mostly adopted output factors. Further research directions are recommended to fill these gaps (e.g., consider socioeconomic factors and pipe properties, deploy IoT facilities to reduce false alarms, distinguish between regular and extreme weather conditions). This state-of-the-art review fills the gap of few endeavors focusing on SO prediction, and could provide the scholars and engineers with inclusive hindsight in dealing with harmful incidents.