Use of machine learning for determining phytoplankton dynamic on station RV001 in front of Rovinj (Northern Adriatic)


  • Goran Volf Faculty of Civil Engineering
  • Boris Kompare
  • Nevenka Ožanić


phytoplankton, Northern Adriatic, machine learning


The paper describes the use of machine learning (ML) and discusses various approaches in modeling phytoplankton based on data from station RV001 in front of Rovinj which well represents the main processes in the open northern Adriatic (NA). Station RV001 is an example of oligotrophic seawater in NA. In order to contribute to the understanding of phytoplankton dynamics at the observation station, based on data covering physical, biological and chemical parameters, ML techniques were used. The final result is a construction of models in the form of regression and model trees, respectively; there were models constructed to be used to explain the dynamics of phytoplankton concentrations at the mentioned station as a result of independent environmental variables. Models in an affordable way combine and show knowledge collected by measurements during 35 year period, which have greatly contributed to a better understanding of ecosystem functioning.