Predictive monitoring, identification, and control of Microcystis blooms in a drinking water source basin: An integrative artificial intelligence and bioinformatics approach

Harmful algal blooms (HABs) pose serious challenges for drinking water utilities; however, conventional monitoring in drinking water source basins remains largely reactive. To address this, we present an integrative framework that combines artificial intelligence (AI) forecasting with strain-resolved bioinformatic diagnostics to enable proactive monitoring of bloom-causing HAB in drinking water source. The AI model employed environmental input features and was optimized, which achieved high predictive accuracy (up to 0.9 R²) with early warning capabilities (1 week lead time). Explainable AI pinpointed organic and nitrogen compounds as top predictors for HAB events. Phylogenomic analysis using short- and long-read sequencing elucidated the population structure, toxicological traits, and ecological adaptation of the bloom-causing Microcystis population (NRERC-214). Field surveys with bioinformatic analysis at a full-scale drinking water treatment plant found complete removal of Microcystis. The relative strengths and weaknesses of the individual units in removing Microcystis in the plant was validated with the phenotypic/genotypic features of Microcystis bioinformatically determined in this study. Overall, this study suggested the integration of a predictive surveillance and diagnosis framework into a source water management scheme, supporting a paradigm shift from reactive post-event management to proactive, risk-based control for drinking water.