A method for the reduction of lake model prediction error

Recent work on methods of error analysis to accompany the application of lake models has not enjoyed great acceptance in part because of the magnitude of the error term. For the models that have undergone a rigorous error analysis (generally the single equation cross-sectional regression models), prediction errors for various water quality variables are often ± 30% or more. A procedure is proposed herein for the reduction of the error associated with the prediction of lake phosphorus concentration from land use and hydrologic data. Existing lake quality data are used in the prediction, and the model is employed only to project changes from the present state. This obviates the need to project all land use impacts with the model: only those proposed to change are projected. The result is a substantial reduction in prediction error for many planning scenarios. © 1983.

Duke Scholars

Author Kenneth H. Reckhow Environmental Social Systems