Descriptive and prediction models of phytoplankton in the northern Adriatic

The northern Adriatic is one of the most productive parts of the Mediterranean Sea due to the nutrient discharges of the Po River. The northwestern part of the northern Adriatic exhibits eutrophic to mesotrophic characteristics with recurrent algal blooms. To contribute to the understanding of eutrophication trends in the northern Adriatic a machine learning tool for induction of models in form of regression trees and in form of set of rules was applied on a data set comprising physical, chemical and biological parameters measured at six stations on the profile from the Po River delta (Italy) to the city of Rovinj on the western coast of Istria (Croatia). Two types of models were successfully elaborated, i.e. (1) the descriptive model to explain the dynamics of phytoplankton concentration in the northern Adriatic as a result of independent environmental variables, and (2) the predictive model to predict the phytoplankton concentration. The descriptive model for phytoplankton dynamics integrates and presents in a user-friendly way the knowledge collected through measurements over a period of 35 years. Such presentation contributes to a better understanding of the ecosystem functioning. The predictive model for phytoplankton concentration successfully forecasts the phytoplankton concentration 14 days in advance.

► We used machine learning methods to elaborate descriptive and predictive model of phytoplankton dynamics in northeren Adriatic.
► The phytoplankton descriptive model presents the knowledge collected over a period of 35 years in a tree like easy-to-read structure.
► The predictive model successfully makes a 15 days forecast of the phytoplankton concentration in north Adriatic.