Developing an empirical model of phytoplankton primary production: a neural network case study

摘要: We describe the development of a neural network model for estimating primary production phytoplankton. Data from an enriched estuary in eastern United States, Chesapeake Bay, were used to train, validate and test model. Two error backpropagation multilayer perceptrons trained: simpler one (3-5-1) more complex (12-5-1). Both networks outperformed conventional empirical models, even though only latter, which exploits larger suite predictive variables, provided truly accurate outputs. The application this is thoroughly discussed results sensitivity analysis are also presented.