“Connecting Microbial Community Assembly Processes to Water Resource Recovery Facility Performance,” Francis L. de Los Reyes III, Ph.D., BCEEM, F. WEF Department Civil Construction and Environmental Engineering, North Carolina State University, Raleigh, NC

The MWRD hosts a seminar series at the Stickney Water Reclamation Plant that is open to the public. These seminars are eligible for Professional Development Credits/CEUs.

High-throughput sequencing has enabled rapid and thorough characterization of water resource recovery facility (WRRF) microbial communities, yet the translation of these data into process improvements remains unclear. Microbial community data are highly complex, multidimensional, and noisy, making them difficult to interpret without a robust framework.

While machine learning (ML) approaches are promising for linking microbial community dynamics to WRRF performance that might be missed by our a priori biases, their usefulness is limited when applied without mechanistic and/or ecological context. Microbial community assembly (MCA) processes describe the processes through which microbial communities form, going beyond identifying which microbial community members are present. In this study, we integrated MCA processes as ecological constraints around ML models. Using a global dataset of 1,200 activated sludge samples from 269 WRRFs, we analyzed microbial diversity, niche breadth, and relative stochasticity. Random forest models incorporating WRRF metadata and MCA processes show that while the inclusion of MCA processes do not universally improve predictive accuracy for general WRRF performance (e.g., biological oxygen demand (BOD) removal), they enhance model precision (i.e., RMSE and MAE) and are important for predicting total nitrogen removal.

Feature importance analyses also indicate climate and WRRF design parameters as important model features. Further, MCA processes, when combined with alpha and beta diversity indices, provide deeper explanatory power into WRRF performance metrics, including BOD and total nitrogen removal. These findings suggest that ML approaches constrained by ecological frameworks offer a promising approach to understanding and optimizing WRRF performance through microbial ecology.

In person: Stickney Water Reclamation Plant, Lue-Hing R&D Complex, 6001 West Pershing Rd., Cicero, IL. Reservations must be made at least 24 hours in advance by emailing MnRSeminars@mwrd.org or calling 708-588-4264 or 708-588-4059.

Virtual: Visit Seminars on April 24 for the seminar link. 

The seminar is eligible for Professional Development Credits/CEUs.