for example, the combination of parameters producing the best match of 

 calculated and measured values. This provides an objective measure of 

 goodness of fit, whereas visual inspection is somewhat subjective. However, a 

 mathematically based criterion should always be checked by visual inspection 

 of shoreline position plots as cancellation of errors is prone to happen for 

 sinuous shorelines and may produce a misleading measure of goodness of fit. 



Sensitivity Testing 



92. Sensitivity testing refers to the process of examining changes in 

 the output of a model resulting from intentional changes in the input. If 

 large variations in model predictions are produced by small changes in the 

 input, calculated results will depend greatly on the quality of the verifica- 

 tion, which is usually in some degree of doubt in practical applications. A 

 second reason for conducting sensitivity tests concerns the natural varia- 

 bility existing in the nearshore system, as discussed in a previous section. 

 No single model prediction can be expected to provide the correct answer, and 

 a range of predictions should be made and judgment exercised to select the 

 most probable or reasonable result. If the model is oversensitive to small 

 changes in input values, the range of predictions will be too broad and, in 

 essence, provide no information. Experience has shown that GENESIS is usually 

 insensitive to small changes in parameter values. Nevertheless, sensitivity 

 testing should always be done. 



Interpretation of Results 



93. Results should always be checked for general reasonability . In 

 this regard, an overview of regional and local coastal processes and the 

 sediment budget calculation or first-order modeling discussed previously 

 should be employed to judge model results. For example, is the overall trend 

 of the calculated shoreline position correct and not just the dominant 

 feature? Do the magnitude and direction of the calculated longshore sand 

 transport rate agree with independent estimates? Experience gained in the 

 verification, sensitivity analysis, and modeling of alternative plans will 



45 



