A demographic model to forecast Dinophysis acuminata harmful algal blooms

Harmful algal blooms (HABs) in marine environments have significant adverse effects on public health, aquaculture and recreational activities. Surges of certain phytoplanktonic toxin-producing microalgae (mostly dinoflagellates or diatoms species) can induce Amnesic, Diarrhetic or Paralytic Shellfish Poisoning (ASP, DSP and PSP). Among HAB species, the genus Dinophysis leads to DSP in human consumers; this being the most recurrent problem in the Iberian Peninsula with the biggest economic impact on clam production and harvesting. While complete elimination of HABs is not feasible, timely implementation of appropriate measures can prevent their negative consequences. This is critical for aquaculture. Research on D. acuminata (dominant Dinophysis species in the North Atlantic) has been focused on ecophysiology and population dynamics, although with few modelling attempts. Weekly monitoring along the Portuguese coast since 2006 has revealed that D. acuminata thrives under spring/summer photosynthetically active radiation (PAR) coupled with water temperatures below 20°C, which typically coincide with the local upwelling regime. In order to advance this knowledge numerically, we developed a demographic model linking D. acuminata growth rate to PAR and sea surface temperature (SST). The 13-year (1-Jan-2006 to 31-Dec-2018) time-series of observations was closely fit by model forecasts. However, the model demonstrated limitations in issuing timely warnings of harmful proliferation of D. acuminata, failing to do so in 50% of cases, and issuing incorrect warnings in 5% of the cases. Furthermore, improving the odds of emitting timely warnings always worsened the odds of emitting false warnings, and vice-versa. To simultaneously improve both aspects, the modelling results clearly indicated the need of implementing both census/projection intervals smaller than 7 days and a laboratory detection limit below 20 cell/L. The time resolution of the census and of the model proved to be the most limiting factor that must be addressed in order to improve numerical forecasting of HABs.