A Naïve Bayesian network approach to determine the potential drivers of the toxic dinoflagellate Coolia monotis (Meunier, 1919) in the Gulf of Gabès, Tunisia

The blooms of the toxic epibenthic dinoflagellate Coolia monotis can be predicted with accuracy derived from knowledge of the main forcing variables. A naive Bayesian network modeling framework was developed to predict the most impacting variables on both Coolia monotis occurrences and blooms. The proposed model took into account the physical environment effects (salinity, temperature, and tide amplitude), the meteorological constraints (evaporation, air temperature, insolation, rainfall, atmospheric pressure, and humidity), the phytoplankton community structure (diatoms, dinoflagellates, Cyanobacteria, and Euglenophyceae) and the sampling months and stations on both C. monotis occurrences and blooms. The study was based on an 11-year survey of the presumed toxic species at 15 sampling stations monitored in the framework of the national phytoplankton monitoring program along the Gulf of Gabès coast. C. monotis occurred mainly in the northern and the southern Gulf during winter, spring, and autumn. The blooms (concentrations up to 10 4 cells dm −3 ) were recorded almost exclusively in three sampling stations, which constitute the hotspot of Coolia monotis blooms in the Gulf of Gabès with reduced spread over the surrounding areas. The blooms occurred in spring, winter, and summer. The shift to the highest salinity, associated with reduced rainfall, low atmospheric pressure, low tide amplitude, and low water and air temperature are the most favorable conditions for the species blooms and occurrences. This study is useful for the management of this ecosystem so as to plan for the setup of an early warning system for the prediction of potential toxic events.