Elucidation and Prediction of Aquatic Ecosystems by Artificial Neuronal Networks

Abstract

Models in aquatic ecology are needed for hypothesis testing (elucidation) and management (predictions) of changing properties in estuaries, lakes, wetlands, and rivers. Two modelling approaches are distinguished to achieve these aims: inductive and deductive modelling. Inductive modelling is considered to be the result of structuring, aggregation, or pattern extraction of ecological data (see Fig. 10.1). The most comon techniques available for inductive modelling are regression analysis and neuronal network training. Deductive modelling goes much further towards integration of structured and aggregated ecological data into relevant ecological theory (see Fig. 10.1). Deductive modelling is normally based on physical mass balances for food webs and nutrient cycles, or heuristic rule sets.