Predictivity of deterministic and neural network models for lake water quality management

In the framework of this paper it is distinguished between short-term and long-term predictions as common in lake water quality management Short-term predictions are necessary for tactical control of sudden escalating impacts on freshwater quality such as harmful algal blooms or toxic contamination. Long-term predictions are needed for strategic management of gradual escalating impacts on freshwater quality such as eutrophication or acidification. Two water quality models, SALMOSED and ANNA, are used to demonstrate fhat: (I) deterministic models are invaluable for strategic management to assess alternative control options before implementation; and (2) neural network models prove to become invaluable for tactical control to predict timing; magnitudes and succession of harmful impacts.