Analysis of short-term estuarine phytoplankton dynamics using neural networks

The artificial neural network (ANN) approach was investigated to model the short-term phytoplankton dynamics in the Skidaway River Estuary. The ability of ANN to model phytoplankton biomass and density of dominant species was evaluated using surface water sampling data collected during bloom and non-bloom periods. During the spring bloom period, the ANN models provided good accuracy for phytoplankton biomass and densities of rapidly growing species using salinity, nitrate, temperature and dissolved oxygen. During the non-bloom period, variation of phytoplankton was small and could not be modeled successfully using the same four environmental factors used to create spring bloom models. Lagged phytoplankton measurements can be added as inputs to increase accuracy of all phytoplankton models.