Assessing Environmental and Socioeconomic Complexities of Decentralized Wastewater Infrastructure using Machine Learning

Onsite wastewater treatment systems (OWTS), like septic systems, are used widely in the United States, with ~16.4% of households reliant on these systems. OWTS process ~4 billion gallons of wastewater per day, yet only approximately half of wastewater generated by OWTS is safety treated. There is an urgent need to identify major factors that contribute to impaired system functionality to create adaptive management strategies and protect environmental and human health. We used a machine learning approach and a unique dataset on OWTS in Athens-Clarke County, Georgia, to evaluate our ability to predict OWTS failure. We found that OWTS age was the highest predictor variable of impaired function, in addition to several environmental and system specific variables. Our work demonstrates the power of using machine learning to assess OWTS function with limited data and highlights the data types to prioritize in collection by local jurisdictions.