ARTIFICIAL NEURAL NETWORKS AS EMPIRICAL-MODELS FOR ESTIMATING PHYTOPLANKTON PRODUCTION

Many empirical models have been developed in order to obtain phytoplan kton production estimates from other variables that are easier to meas ure. These empirical models are usually based on regression of phytopl ankton production against biomass and other variables. They are partic ularly useful to fully exploit data sets acquired by both in situ inst rumental measurements and remote sensing. Two conventional empirical m odels were compared with a new approach, based on artificial neural ne tworks. Although very simple neural networks were used, they provided a much better fit to observed data than conventional models do and the y seem a very promising tool for phytoplankton production modeling.