In Western Europe, the incidence of DST is likely the highest globally, posing a significant threat with prolonged bans on shellfish harvesting, mainly caused by species of the dinoflagellate genus Dinophysis. Using a time series from 2014 to 2020, our study aimed (i) to determine the concentration of D. acuminata in water at which shellfish toxin levels could surpass the regulatory limit (160 µg OA equiv kg-1) and (ii) to assess the predictability of toxic events for timely mitigation actions, especially concerning potential harvesting bans. The analysis considered factors such as (i) overdispersion in the data, (ii) distinct periods of presence and absence, (iii) the persistence of cells, and (iv) the temporal lag between cells in the water and toxins in shellfish. Four generalized additive models were tested, with the Tweedie (TW-GAM) model showing superior performance (>85%) and lower complexity. The results suggest existing thresholds currently employed (200 and 500 cells L-1) are well-suited for the Portuguese coast, supported by empirical evidence (54-79% accuracy). The developed algorithm allows for thresholds to be tailored on a case-by-case basis, offering flexibility for regional variations.
Keywords: Dinophysis; HABs; Tweedie distribution; diarrhetic shellfish toxins; early warning; food safety; generalized additive models; monitoring; okadaic acid; time series.