Modeling Concepts - Coastal Lagoons: Ecosystem Processes and Modeling for Sustainable Use and Development - page 296

Ideally, the best open boundary conditions for a model domain are field mea-
surements of dependent parameters at these boundaries. An optimal choice can be
a combination of some “typical” and some “extreme” boundary conditions measure-
ments. In the absence of field measurements, one must rely on dynamical boundary
conditions. Some examples are the (wave) radiation, the absorbent, the flow relax-
ation, the clamped, or the gradient open boundary conditions. 45
Since the majority of lagoon problems are significantly boundary dependent, the
accuracy of the model solution will depend on the accuracy of the boundary conditions.
For example, the simulation of the water quality in a lagoon will be dependent on the
boundary nutrients loading. In some cases, one must also include nonpoint sources of
nutrient (diffusive) loading ( see Chapter 9.4, the example of the Vistula Lagoon case
study). In summary, the accuracy of the model solution will depend, among other factors,
on the accuracy of the boundary condition measurements. However, when measurement
facilities are limited, the relationship between uncertainties in both boundary conditions
and model solutions must be studied using sensitivity or error analysis. This will dictate
the minimal measurement requirements for the desired solution accuracy.
A last type of input data restrictions relates to internal parameter specification.
The internal parameters are the independent variables that are normally required to
calculate the dependent variables. These parameters are the bottom topography, the
basin geometry and boundaries, and the flow resistance properties within the water
body, at the water surface and at the bottom. 4a
The bathymetry or bottom topography and the basin geometry are conservative
independent variables that do not vary during a hydrodynamic simulation. Initially,
a few depth points may need to be changed in order to smooth out the bathymetry
during model calibration, when erroneous data are suspected, or in order to avoid
instability problems near open boundaries (e.g., lagoon or river entrances, dikes).
Bottom topography and basin geometry are the main factors that influence the spatial
structure of a numerical solution. As far as model computations are concerned, the
bottom structure and basin geometry should not be prescribed in more detail than
the computational grid resolution. However, practice shows that this information
should be as precise as possible.
Resistance data are also needed to estimate the frictional forces that dissipate kinetic
energy at the surface, at the bottom, and in the water body. Resistance data include drag
coefficients, bottom roughness coefficients, wind drag coefficients, and other parameters
related to turbulence production and dissipation. In addition, transport equations also
contain dispersion parameters as well as other kinetic transformation constants .
These internal parameters may be used for model calibration. However, consider-
ing their large number, and the probable uncertainty associated with some of them, it
is suggested that this number be reduced following a sensitivity analysis. Only those
parameters that significantly affect the solution may be kept as calibration parameters.

6.5.2 S ENSITIVITY A NALYSIS

Sensitivity analysis is the process by which the response of different elements in
the model to external influences is investigated. 44 It is customary to compare the
model performance to some test cases that have analytical solutions.