Environmental Engineering Reference
In-Depth Information
system operates to generate this behaviour (Troitzsch 2004). This approach is
mostly applicable for model systems which allow to represent self-organization
processes and where higher-level properties emerge from the interaction of lower-
level entities (IBM, Chap. 12). When it can be ensured that no type III error is made
(see above), such a consistency check for key processes and dynamics on different
hierarchical levels increases the probability that the system behaviour has been
represented correctly and results are reliable within the applied conceptual system.
An example for a structural validation was presented by Laperri
ยด
re et al. (2009)
who modelled the spread of plague with an IBM. Although the processes leading to
disease outbreaks and the role of the different components of the infection system
are well known, the data for the investigated case on the island of Madagascar are
scarce with low temporal resolution, which is not uncommon for ecological data
(Jopp and Lange 2007). Therefore, the model validation process was performed
more in a conceptional way rather than testing a one-to-one accuracy: various input
scenarios (e.g. prevalence levels in recipients) were linked to model output and a
comparison of sufficiently general characteristics of the target systems (e.g. thresh-
old levels) was performed.
Structural validation allows to test theory-driven models. It is often performed with
expert knowledge or when the direct participation of stakeholders is desirable. In the
latter case, model results are evaluated by experts for their consistency. Another way
of doing this is to perform a Turing test, where the experts have to decide whether the
results derive from simulations or from real-world field investigations.
23.3.5 Limits of Validation and Validity Range
The described approaches help to learn more about the conditions under which a
model can be used and how eventually critical points can be avoided. Usually there
are no sharp boundaries of a model's validity range. It is more reasonable to consider a
gradual range of certainty depending on the given conditions. Therefore, frequently
models informmore about potential outcomes rather than providing a strict prognosis.
Experience shows that ecological prognoses are difficult. Models usually are stronger
in providing
aspecific
result can be expected. However, whether the given structure and parameter values
are actually the optimal description for a specific case remains difficult to foresee.
information:
the model assumptions hold,
if-then
if
then
23.4 Assessment of Higher-Level Model Implications
Validation attempts to directly relate model output with ecological observations. It
is also possible, and for some model approaches useful, to include more indirect and
conceptual analyses which investigate model implications on higher organizational
levels, and thereby, extend the possible range of validation applications.
