Confirmation of water quality models

Water quality simulation models, whether descriptive or predictive, must undergo confirmatory analyses if inferences drawn from the models are to be meaningful. Current practices in the confirmation of simulation models are examined and criticized from this perspective. In particular, labeling this process "verification" or "validation" (truth) probably contributes to the often inadequate efforts, since these states are unattainable. The evaluation of scientific hypotheses, or water quality simulation models, may proceed according to inductive logic, the hypothetico-deductive approach, or perhaps according to a falsification criterion. The result of successful testing is at best confirmation or corroboration, which is not truth but rather measured consistency with empirical evidence. On this basis a number of statistical tests are suggested for model confirmation. The major difficulty to overcome, before confirmation becomes meaningful, is the generally inadequate data for establishing rigorous statistical tests. © 1983.

Duke Scholars

Author Kenneth H. Reckhow Environmental Sciences and Policy