Chapter 6
Development and Application of Predictive River
Ecosystem Models Based on Classification Trees
and Artificial Neural Networks
P. Goethals' A. Dedecker' W. Gabriels' N. De Pauw
6.1
I ntrod uction
Prediction of freshwater organisms based on machine learning techniques is
becoming more and more reliable due to the availability of appropriate datasets
and modelling techniques. Artificial neural networks (Lek and Guegan 1999),
fuzzy logic (Barros et al. 2000), evolutionary algorithms (Caldarelli et al. 1998),
cellular automata (Gronewold and Sonnenschein 1998), etc. proved to be powerful
tools to perform ecological modelling, especially when large datasets are involved.
Models have several interesting applications in river management. They allow for
a better interpretation of the results, easing the cause-allocation of the actual river
status and increasing the insight needed to improve assessment systems (Fig. 6.1.).
Models also allow for simulating the effect of potential management options and
thus supporting decision-making. The development of effective and efficient
monitoring networks based on models is probably another important advantage.
The 'River Invertebrate Prediction and Classification System' (RIVPACS)
approach, based on statistical modelling, is currently one of the best available
techniques for assessing the biological quality of running waters because it offers
the ability to use environmental variables to predict species that are expected to
occur at a site if it is unstressed. The expected fauna is then compared with the
observed community of macroinvertebrates in order to assess the river quality
(Wright et al. 2000). However, biological communities are dynamic and the nature
of RIVP ACS would need to be altered in order to predict a change in faunal
composition in response to new environmental conditions at a given site (De Pauw
2000).
Development and Application of Predictive River
Ecosystem Models Based on Classification Trees
and Artificial Neural Networks
P. Goethals' A. Dedecker' W. Gabriels' N. De Pauw
6.1
I ntrod uction
Prediction of freshwater organisms based on machine learning techniques is
becoming more and more reliable due to the availability of appropriate datasets
and modelling techniques. Artificial neural networks (Lek and Guegan 1999),
fuzzy logic (Barros et al. 2000), evolutionary algorithms (Caldarelli et al. 1998),
cellular automata (Gronewold and Sonnenschein 1998), etc. proved to be powerful
tools to perform ecological modelling, especially when large datasets are involved.
Models have several interesting applications in river management. They allow for
a better interpretation of the results, easing the cause-allocation of the actual river
status and increasing the insight needed to improve assessment systems (Fig. 6.1.).
Models also allow for simulating the effect of potential management options and
thus supporting decision-making. The development of effective and efficient
monitoring networks based on models is probably another important advantage.
The 'River Invertebrate Prediction and Classification System' (RIVPACS)
approach, based on statistical modelling, is currently one of the best available
techniques for assessing the biological quality of running waters because it offers
the ability to use environmental variables to predict species that are expected to
occur at a site if it is unstressed. The expected fauna is then compared with the
observed community of macroinvertebrates in order to assess the river quality
(Wright et al. 2000). However, biological communities are dynamic and the nature
of RIVP ACS would need to be altered in order to predict a change in faunal
composition in response to new environmental conditions at a given site (De Pauw
2000).