PART IV: ^^jMERICAL MODEL CALIBRA.TION 



Procedure 



27. The three models in this study were calibrated against the tides of 

 the Carolina coast. The tide along the Outer Banks is semidiurnal in nature, 

 with the M„ constituent predominating. To calibrate the computational models, 

 four parameters were adjusted in WIFM: (u; the depths or elevations assigned 

 to each cell, (b) the Chezy friction coefficients for each cell, (c) the 

 choice of cells used as marigram stations, and (d) the boundary condition 

 imposed on the grids. The water depth or land elevation of a cell was esti- 

 mated from maps, and only a few depths were changed slightly during calibra- 

 tion. Except in shallow water ana at flow constrictions, the models' hydro- 

 dynamics vere fairly insensitive to clianges in Chezy coefficients. Tne choice 

 of which cells to serve as marigram stations was sometimes arbitrary due to 

 shoreline approximations and the size of cells relative to hydrodynamic 

 details in the prototype. For a model that was adjusted to the point where 

 only one WIFM gage disagreed slightly with prototype data, this discrepancy 

 was removed by changing the gage's placement in the grid. 



28. Calibrations largely involved the development of proper boundary 

 conditions for the models. For c tide, seawaru boundary conditions were 

 estimated by wave speed calculations and shoaling factors. With initial 

 estimates, the development of correct boundary conditions proceeded through 

 three steps: (a) computations with WIFM using the latest estimate, (b) com- 

 parison of computed marigrams to prototype data, and (c) refinement of the 

 boundary conditions. Successive iterations vere performed to natch computed 

 tides against prototype data. The other parameters in WIFM were also adjusted 

 during the refinement of the boundary conditions. 



29. The model's ability to accurately reproduce prototype data was 

 limited by cell sizes used in the grid (which governed how well topography 

 was simulated) and by any approximations used for the boundary conditions. In 

 this study, tlie prototype tides were approximated with just the M_ constituent. 

 For the North Carolina coast, this approximation provided a good representa- 

 tion of the entire tide while keeping the expenses of calibration to a mini- 

 mum. The number of iterations needed in the boundary condition refinement 

 process soars when multicmstituent tides are used, because of the increased 



21 



