Associate Professor University of Toledo, United States
Lake Erie's Western Basin is impacted by yearly harmful algal blooms (HABs) caused by the cyanobacteria Microcystis aeruginosa. Effectively monitoring the Microcystis toxin, microcystin (MC), is crucial to ensure the safety of the public. The most common method for estimating MC concentrations in the Western Basin is the calibration-based Enzyme-Linked Immunosorbent Assay (ELISA). The ELISA is a calibration curve method, relating MC concentration to an easy-to-measure response through an empirical model. The statistical uncertainty associated with such method is high because the empirical model is based on a small sample size. We propose an alternative approach for developing the empirical model using a Bayesian hierarchical model (BHM). The BHM method leverages information from existing data from multiple ELISA tests to stabilize the model coefficient, thereby improving estimation accuracy. For our study, we demonstrate the effectiveness of BHM in HABs monitoring by recalibrating MC concentrations from the Lake Erie HABs monitoring program operated by The National Oceanic and Atmospheric Administration Great Lakes Environmental Research Laboratory (NOAA GLERL). The estimation uncertainty is measured by 95% credible intervals and standard deviation of the measured concentrations. In this presentation, we demonstrate the effectiveness of the BHM approach in improving measurement accuracy.
Using the quality control samples with a known concentration of 0.78 µg/L, we compare estimated concentrations from the BHM and the standard ELISA method. Our results show BHM method reduces the estimation standard deviation by a factor of ~10. Such a magnitude of reduction in measurement uncertainty in MC would have a drastic impact on water quality monitoring, and all calibration-curve-based measurement methods, accounting for over 90% of all chemical analytic work.