qPCR-based prediction of low-level microcystin-LR using mcyE and passive sampling across multiple lakes and years

• Passive sampling and qPCR targeting mcyE were used to monitor low-level MC-LR over three years in ten lakes. • Nutrients and NOM did not explain variation in MC-LR detection across lakes. • mcyE outperformed all other predictors in identifying MC-LR presence. • mcyE concentrations require location- and time-specific interpretation. Microcystin-LR (MC-LR) is a cyanobacterial hepatotoxin that poses health risks even at low concentrations. Because quantitative analysis of MC-LR is costly and time-consuming, water managers rely on early warning tools to determine when confirmatory testing is warranted. Quantitative PCR (qPCR) targeting the mcy genes has emerged as one such tool, but its reliability across lakes and seasons — particularly at low toxin concentrations — remains unclear. In this study, we used passive sampling to detect low concentrations (< 1 µg L −1 ) of MC-LR and paired this with qPCR monitoring of mcyE to assess whether mcyE alone can serve as a reliable indicator of low-level MC-LR presence over three years across ten lakes (total of n = 893 distinct samples). We developed location- and year-specific hierarchical Bayesian models to estimate the probability of MC-LR detection from mcyE concentrations. We also included environmental covariates to determine if their inclusion improved model performance. Although mcyE was the strongest overall predictor, its relationship with MC-LR varied substantially by location and year, and these hierarchical models were essential in capturing this variability. These findings highlight the promise of mcyE -based early warning systems for low-concentrations of MC-LR but emphasize that interpretation must be tailored to local ecological and seasonal conditions. Created in https://BioRender.com