Large lake models - uses, abuses, and future

Mathematical modeling has played and should continue to play an important role in Great Lakes management and scientific development. Great Lakes modeling is entering a phase of relative maturity in which expectations are more realistic than in the past. For example, it is now realized that the modeling process itself is valuable even if the resulting models are not immediately useful for management. The major thrust in the past has been water quality (eutrophication) modeling, but there has been a recent shift toward developing toxic substances models. Modelers and model users have been limited by a lack of knowledge of Great Lakes processes, limited data availability, and incomplete or improper validation. In the future, greater emphasis is needed on specifying prediction uncertainty and conducting proper model validation - including calibration, verification, and post-audits. Among the Great Lakes modeling activities likely to have the greatest payoff in the near future are (1) the development and refinement of toxic substances models, (2) post-auditing and improvement of eutrophication models, and (3) the adaptation of models for use on personal computers to allow greater model utilization.