Modeling Concepts - Coastal Lagoons: Ecosystem Processes and Modeling for Sustainable Use and Development - page 291

case properties in the fresh water are the most convenient boundary condition.
When multiplied by the river discharge the condition, in fact, becomes a flux
condition.

6.4.4 I NTERNAL C OEFFICIENTS : C ALIBRATION AND V ALIDATION

Internal coefficients are used to parameterize empirical closures of the processes
simulated by the model. These parameters have to be fitted using field data. This
procedure is called calibration. The range of variation of each parameter is usually
known from the literature. When the calibration procedure suggests that a parameter
value is out of that range, a scientific explanation has to be found. In fact the most
common reason for calibration values being out of range is the need to include
additional processes in the model.
The calibration effort increases in a nonlinear way with the number of parameters
(e.g., biological models). In the case of models with simple spatial grids the com-
putational time is small and some automatic procedures can be established to select
the best values of the parameters. In real systems horizontal transport simulation is
required and a small number of trials can be done.
After calibration the model must be validated using a data set not used in the
calibration process. This validation process will guarantee that the parameters are
adequate for a range of conditions representative of those found in the system being
studied.

6.5 MODEL ANALYSIS

6.5.1 M ODEL R ESTRICTIONS

Can natural processes be adequately modeled? This question could be argued at
great length within the abstract concept of nature cognoscibility. 41 A practical answer,
however, is that some natural processes can indeed be modeled adequately, but only
within restricted limits and a predefined accuracy. This section discusses some of
the restrictions that arise when attempting to model lagoon processes ( Figure 6.18) .
Physical restrictions arise when the equations used in the model are not complete,
i.e., they do not include all factors influencing the process under investigation. For
example, the exclusion of rotation effects of the Earth when modeling small-scale
dynamics or the assumption that density is independent of depth when studying
shallow-water motions.
Numerical restrictions, on the other hand, are introduced by the numerical
schemes used to solve the equations. These schemes require spatial and temporal
discretizations, which in turn impose restrictions on the computational grid size and
the integration time step.
Mixed restrictions arise when the physics of a process is altered by computational
constraints. An example is the phenomenon of subgrid processes. All processes
occurring at spatial scales smaller than a computational grid size (or subgrid) are
not simulated explicitly but are represented in the primitive momentum and transport
equations by some empirical eddy diffusion/dispersion coefficients.