Lessons Learned from Experiments in Auto-Calibration of Large Scale Environmental and Hydrological Models

Recent work by the authors has identified the need to efficiently auto-calibrate models of considerable complexity, in order to investigate future water quality and quality concerns in the Great Lakes’ Basin. Initial results in the use of standard models such as AGNPS and SWAT have yielded somewhat disappointing results. The calibration process has been cumbersome, and predictions based on outputs from, say, a watershed model input to a lake circulation model have generally proven problematic: If the flows are correct, the loadings are not, and so forth.