Environmental Engineering Reference
In-Depth Information
Fig. 23.1 Basic steps to ensure model accuracy. The sequence of steps may be changed depending
on model purpose and assessment routines
a preceding assessment of the specific range of contexts for which the model can
provide reliable results. Discussing the numerical correctness of the result is an
important point of model evaluation. In most (practically all) of the ecologically
relevant cases, the mathematical correctness of the executed simulation cannot be
mathematically proven (Hawking 1988; Oreskes et al. 1994). Therefore, it is a
matter of plausibility analysis to estimate whether the results appear sufficiently
reliable. This is of specific importance for differential equation models, but also
plays a role for other modelling techniques. The following approaches are fre-
quently used to investigate the numerical correctness - they give hints to increase
our trust in the model results but without representing a proof of correctness:
l
A relatively laborious way would be to implement the same model on a different
hardware- and software platform (on a different computer and/or with a different
simulation software, compiler).
It is less laborious to change step width of model calculation and/or the integra-
tion method. This will usually lead to slightly different results. If the differences
are not meaningful, the numerical approximation can be considered as appropri-
ate for the aim of the investigation.
l
