Modeling Concepts - Coastal Lagoons: Ecosystem Processes and Modeling for Sustainable Use and Development - page 299

currents measured at an arbitrary depth y 0 can be converted to a depth-averaged
value by using the following relation: 37

y
D

gD
kMD

uu = 1 + (ln

0

+ ⋅ 1 ) (6.59)

0 23

⋅ ⋅

where u is the mean velocity, u 0 is the measured velocity, y 0 is the measuring depth
above seabed, D is the instantaneous water depth, g is the gravity, k is Von Karman's
constant ( 0.4), and M is the Manning number.

The difficulty associated with a calibration procedure that utilizes a large number
of parameters is that it is not clear which parameters are responsible for the change
observed in independent variables. This can be avoided by limiting the number of
parameters to those identified by a sensitivity analysis.

≈

6.5.4 V ALIDATION

Validation of a model is the comparison of its results to a set of independent obser-
vations not used for calibration purposes. 44 This means that two data sets are needed
in the model domain: one for calibration purposes and the other for validation
purposes. Good model validation results will convince potential users (managers,
planners, scientists, or engineers) to use the validated model to solve a problem. 48
The criterion for a good model validation is that its numerical solution reproduces,
within the prescribed accuracy, the system's variability during the simulation period.
A first validation stage should reproduce normal conditions over a period covering
at least one cycle of the system's natural variation. This cycle may be a day, a season,
a year, a climatic period, or any other cycle of the problem under study.
A second validation stage should cover those extreme conditions that have not
necessarily been encountered during the calibration stage. Those extreme conditions
for a lagoon can be high river inputs, stormy winds, and large volume fluxes at the
ocean-lagoon boundary. Field observations of these extreme conditions must be
available for a good validation exercise.
The comparison of model results to observed data during the validation process
provides one of the following outcomes:

• The simulation yields accurate results
• The simulation overestimates or underestimates the reality
•

The simulation yields confusing results despite all efforts

When a model validation process is inconclusive, the calibration process may
be repeated with parts of the data used for validation. It should be mentioned that
it is very rare that a single data set of internal parameters will include normal and
extreme situations. A reasonable model validation can be attempted with different
sets of internal parameters for different situations or time scales.
Ultimately, a model simulation must be checked for volume, mass, and energy
conservation. For example, the sum of incoming and outcoming volumes, heat, salt,
or conservative pollutants must be equal through a cycle of variability.