Model Calibration, Validation, and Verification

• Chart 1 of 3



• Chart 2 of 3



• Chart 3 of 3

Most environmental models include parameters which must be tuned or adjusted to obtain reasonable match between model predictions and observed conditions. All models require checking and testing to evaluate how well they perform. The first activity is referred to as model calibration, and the latter as model validation. Without calibration and validation, a modeling application is only an educated guess. This may be sufficient for some scoping applications, but generally not for management decisions.

Model calibration and validation can be defined as follows:

Model calibration involves minimization of the deviation between measured field conditions and model output by adjusting parameters of the model. Data required for this step are a set of known input values along with corresponding field observation results. Calibration typically includes a sensitivity analysis, which provides information as to which parameters have the greatest effect on output. Careful consideration should be given to adjustment of model parameters to ensure that values are within the range of reasonable possibility. EPA provides guidance on reasonable values of parameters for water quality models.

Model validation involves the use of a second, independent set of information to check the model calibration. The data used for validation should consist of field measurements of the same type as the data output from the model. Specific features such as mean values, variability, extreme values, or all predicted values may be of interest to the modeler and require testing.

Model validation is sometimes referred to as verification. Under current usage, this terminology is discouraged. Instead, model verification is used to refer to another process, the examination of the numerical technique and computer code to ascertain that it truly represents the conceptual model and that there are no inherent numerical problems with obtaining a solution.

A variety of statistical tests are available for assessing model goodness of fit during calibration and validation. A useful introductory summary is provided in:

Reckhow, K.H., J.T. Clements, and R.C. Dodd. 1990. Statistical evaluation of mechanistic water quality models. J. Environ. Eng., 116(2): 250-268.