Verification of Water Quality Models

The basic concepts of water quality models are reviewed and the need to recognize calibration and verification of models with observed data is stressed. Post auditing of models after environmental control procedures are implemented is necessary to determine true model prediction ability. The direct questions in the decision-making arena related to model credibility are examined by several quantitative measures of model performance: regression analysis between observed and computed values, relative and root mean square errors and comparison of means. An analysis of 15 DO water quality models indicated median relative error of 10%. For a eutrophication model of Lake Ontario, application of quantitative verification measures indicated the credibility of the calculations decreased as the model was applied to smaller scale phenomena. Inclusion of more complex kinetics reduced median relative error across five variables and over a 10 year modeling framework from 32% to 22%. It is suggested that quantitative measures of water quality model credibility be integrated into the modeling analysis whenever possible.