Bayesian mechanistic modeling elucidates controls on bloom timing and magnitude in Western Lake Erie

Session: Harmful Algal Blooms and Their Toxicity: Remote Sensing and Modeling Approaches (1)

Dario Del Giudice, NC State University, [email protected]
Shiqi Fang, North Carolina State University, [email protected]
Donald Scavia, University of Michigan, [email protected]
Daniel Obenour, NC State University, [email protected]

Abstract

Maumee Bay is one of the areas of Lake Erie most affected by harmful algal blooms. While nutrient loads are known to be an important driver of overall bloom size, it is unclear how various environmental influences synergistically control the timing of cyanobacteria growth and decay. To address this gap, we develop a parsimonious mechanistic model describing June-October dynamics of chlorophyll, nitrogen (N), and phosphorus (P) in the bay-area. We calibrate the model to a new, geostatistically-derived dataset of daily water quality attributes, spanning 2002-2017. Calibration is conducted in a rigorous Bayesian framework, which probabilistically characterizes system kinetics, such as summertime denitrification rates. The model reproduces intraseasonal and interannual patterns well, with r² exceeding 0.74 for N. Results show that N typically becomes limiting in September after blooms peak. Interestingly, we find an approximately one-month gap between the beginning of nutrient depletion (late June) and the acceleration of algal growth, and explore different causes such as shifts in species dominance. Our results further indicate denitrification is an important removal pathway, while sediment release is a substantial P source. In conclusion, conceptual modeling with Bayesian calibration represents an effective approach to better understand short and long-term eutrophication patterns in lakes.