Probabilistic forecasting of harmful algal blooms in western Lake Erie

Abstract:

Over the past decade, there has been a dramatic rise in the magnitude of harmful algal blooms (HABs) in western Lake Erie. These cyanobacteria blooms have drawn attention to phosphorus loading, a common driver of freshwater productivity. However, it is unclear how much of the year-to-year variability in bloom size is explained by anthropogenic phosphorus loading, and how much variability is related to other factors, including weather/climate drivers and measurement error. Here, we aim to advance the state-of-the-art in HAB forecasting by explicitly quantifying uncertainties in late-summer bloom observations, and propagating them through a Bayesian modeling framework that relates bloom size to phosphorus load. Because of the need to accurately represent predictive uncertainty, different statistical formulations are critically evaluated through cross validation. A model based on a novel implementation of a gamma error distribution is found to provide the most realistic uncertainty characterization, as well as high predictive skill. Our results also underscore the benefits of a hierarchical approach that allows us to assimilate data sets from multiple sources. Finally, our modeling analysis suggests that Lake Erie has become increasingly susceptible to large cyanobacteria blooms. We explore the nature of this change and assess potential biophysical explanations.