Calibrating and verifying the model

The advection-dispersion of solutes depends largely on the simulated flows and fluxes calculated by the MIKE SHE flow model. After your first AD simula­tions, you will usually have to go back and improve the calibration of your flow model. Rarely, can the simulated concentrations and mass fluxes be cali­brated to the measured concentrations by tuning only the solute transport model.

It is important to recognise that a transport model must be calibrated. This is true for all applications larger than the laboratory scale since model output cannot necessarily be compared directly to measured values. Measurements are mostly point measurements at a certain time whereas results often are mean values over larger volumes and longer times.

The purpose of the calibration is to tune the model so that it is able to repro­duce measured conditions for a particular period in a satisfactory way. This period - known as the calibration period - should be chosen long enough to include events of similar kind as the ones you are going to investigate.

A satisfactory calibration is reached when the model is able to reproduce the measured values taking the following conditions into account:

·         uncertainty in the measurements (time, space, equipment)

·         representativeness of measurements (point/average grid values)

·         differences between your conceptual model and nature

·         uncertainty in other model parameters and data (source description etc.)

In general, it is impossible to specify an exact level of divergence between measured data and computed results before the model is satisfactorily cali­brated. In each application you have to consider all factors influencing your result.

After the calibration, you should verify your model by running one or more simulations for which measurements are available without changing your model parameters. If the model is able to reproduce the validation measure­ments you can consider your calibration to be successful. This ensures that simulations can be made for any period similar to the calibration and the veri­fication period with satisfactory results.