Environmental Engineering Reference
In-Depth Information
Table 23.1 (continued)
Topic and modelling
approach
Model evaluation approach
Reference
Sustainable production of
mussel aquaculture; Box
ecosystem model
coupled with a
hydrodynamic model
Set-up of 2 year experiments to compare
key variables (mussel growth, nitrate,
phytoplankton concentrations) from
model results with experimental data
Grant et al. (2007)
Visualization of plotted data from
experiments and models
Hatching of eggs and
survival time of
herbivore soil-dwelling
insects; Partial
differential equations
Application of the model to other insect
taxa
Johnson et al. (2007)
Visualization of plotted data from
experiments and models
Dynamics of a woodpecker
population; IBM
Qualitative assessment of model
performance as location used by the
birds are not completely independent
(e.g. part of time series)
Schiegg et al. (2005)
Comparison of secondary model
predictions (e.g. natal dispersal
distance, population structure) with
descriptive statistics
Prediction of distribution
and density of badger
sets; (SDM)
Statistical validation with ROC curve
using density data from two
independent sites
Jepsen et al. (2005)
Model accuracy as a proportion of
correctly classified cells (locations)
Fig. 23.3 Illustration of the development of a fictional plant population. Crosses represent
measurement data, the solid line model output and the arrows precipitation events: (a)( left ) The
measurement points were used for model development, the fit it relatively good. (b)( right ) shows
an independent dataset with a different precipitation regime ( arrows ) which was not used during
model development. The fit is less good and indicates, that for model validation the precipitation
response might be reconsidered to improve the application range of the model. For instance, Rupp
and Rupp (2010) (Fig. 7) illustrate the application of the approach with empirical data
validation also other approaches which make privileged use of independent datasets
for validation can strongly reduce the risk of asking the wrong questions or following
hypotheses which seem to be reasonably suggested by the original data (i.e. getting
the right answer to the wrong question, type III errors, see Mosteller 1948).
Search WWH ::




Custom Search