Abstract
Harmful algae bloom (HAB) forecasting has developed rapidly over recent decades, but predicting harmful levels of marine biotoxins in shellfish is still a challenge. New discoveries suggest that predator‐prey interactions may be an important driver in the formation of HABs. Key species of harmful algae respond to copepod infochemicals with increased toxin production. In addition, copepods feed selectively on less defended prey, which may further promote harmful taxa. Here we explore if eavesdropping on predator‐prey dynamics by monitoring zooplankton can improve HAB forecasting. We first examine an 8‐yr time series including copepod biomass, harmful algae cells ( Dinophysis spp.), and diarrhetic shellfish toxins in blue mussels ( Mytilus edulis ) using generalized additive models. Models including copepod biomass more accurately predicted okadaic acid in mussels than phytoplankton alone. We then apply this connection more narrowly by analyzing the specific copepod exudates known to induce toxin production, copepodamides, from the mussels sampled in biotoxin monitoring. Adding copepodamide data from shellfish extracts increased model performance compared to copepod biomass. Results suggest that including grazing effects through copepodamide measurements may provide a cost‐efficient way to improve accuracy and lead time for predicting the accumulation of microalgal toxins in shellfish.