Development of a logistic regression model for the prediction of toxigenic Pseudo-nitzschia blooms in Monterey Bay, California

Blooms of the diatom genus Pseudo-nitzschia have been recognized as a public health issue in California since 1991 when domoic acid, the neurotoxin produced by toxigenic species of Pseudo-nitzschia, was first detected in local shellfish. Although these blooms are recurring and recognized hazards, the factors driving bloom proliferation remain poorly understood. The lack of long-term field studies and/or deficiencies in the scope of environmental data included within them hinders the development of robust forecasting tools. For this study, we successfully developed predictive logistic models of toxigenic Pseudo-nitzschia blooms in Monterey Bay, California, from a multi-project dataset representing 8.3 yr of sampling effort. Models were developed for year-round (annual model) or seasonal use (spring and fall-winter models). The consideration of seasonality was significant: chlorophyll a (chl a) and silicic acid were predictors in all models, but period-specific inclusions of temperature, upwelling index, river discharge, and/or nitrate provided significant model refinement. Predictive power for ’unknown’ (future) bloom cases was demonstrated at >= 75% for all models, out-performing a chl a anomaly model, and performing comparably to, or better than, previously described statistical models for Pseudo-nitzschia blooms or toxicity. The models presented here are the first to have been developed from long-term (>1.5 yr) monitoring efforts, and the first to have been developed for bloom prediction of toxigenic Pseudo-nitzschia species. The descriptive capacity of our models places historical and recent observations into greater ecological context, which could help to resolve historical alternation between the implication of freshwater discharge and upwelling processes in bloom dynamics.