Microcystin in western Lake Erie: co-occurrence with biomass indicators, and the effects of abiotic factors

Identifying factors useful for microcystin (MC) concentration prediction in freshwater remains challenging as algal biomass indicators alone do not fully capture the conditions leading to elevated toxin concentrations. We evaluated whether integrating biomass indicators (chlorophyll a nd phycocyanin) with abiotic factors improves prediction of MC concentration in western Lake Erie using 12 years of water quality data (2012−2023). We first analyzed concordance among established thresholds for biomass indicators and MC, revealing that all three indicators were concordant in only 42% of sampling events, with simultaneous threshold exceedance occurring in 15.5% of cases. We then compared four Bayesian additive regression trees (BART) model configurations: abiotic factors alone (AF), abiotic factors with chlorophyll (AF + Chl), abiotic factors with phycocyanin (AF + Phyco), and all variables combined (AF + Chl + Phyco). The AF + Phyco model demonstrated superior performance (R 2 = 0.70), resolving an additional 10% of MC variance compared to the AF-only model (R 2 = 0.60). The most complex model showed similar performance, suggesting phycocyanin is the most informative biomass indicator when combined with abiotic factors. Temperature emerged as the primary predictor, with phycocyanin ranking second at comparable importance levels. Partial dependence analysis revealed critical thresholds at 20 °C for temperature and 100 μg L −1 for total phosphorus, above which MC concentrations increased markedly. Our findings support proactive monitoring strategies that prioritize phycocyanin-based measurements combined with key abiotic factors for MC risk assessment in eutrophic lakes like Lake Erie.