Prediction

Once the numerical model has been successfully calibrated it can be used to make predictions.  The greatest confidence can be placed in these predictions if they are made for the same or similar conditions as those used during the calibration period.  Less confidence can be placed in the predictions when input is very different.

If no data is available for calibration, sensitivity analysis should be carried out to establish the sensitivity of the model results to variation in the input parameters.  During the sensitivity analysis the input variables about which least is known should be varied within physically reasonable bounds.  Sensitivity tests should include the impact of small changes in roughness, alternative sediment transport equations and sediment transport calibration coefficient.

Sensitivity tests should also be undertaken when predictions using inputs very different from calibration conditions are required.