Chapter 8· Analysis of Stream Macroinvertebrate Communities
161
The sensitivity tests were also able to show variations for the different taxa of
community. Densities of Chironomidae, for example, varied greatly in response to
different input ranges. This could be observed for all the variables in the data of
July 1997 (Fig. 8.17a).
Densities of Hirudinea were also sensitive to
environmental variables in the data of July 1997. The higher sensitivity of
Hirudinea was also observed in the data of November 1997 without strong
environmental effects (Fig. 8.17b). Densities of Oligochaeta, in contrast, were
characteristieally insensitive to input variables, especially in July 1997. This
indieated that, in this field study, the density of Oligochaeta was not greatly
affected by environmental variables during the flooding period in comparison with
other dominant taxa such as Chironomidae and Chironomus sp. (Chon et al.
2000b)
This study examined the feasibility of the recurrent artificial neural network in
extracting information out of temporal development. The results showed that the
dynamies of sets of multivariate data about communities could be rapidly
pattemed and forecasted by the network.
8.4
Patterning Organizational Aspects of Community
8.4.1
Relationships among Hierarchical Levels in Communities
Useful ecologieal informaties resides in community organization, especially in
"associations" among different levels in communities such as taxonomieal or
functional groups. In these associative relationships, the complex community
usually develops a hierarchy, which is a good subject for understanding system
behavior of the target ecosystem (Allen and Starr 1982; O'neill et al. 1986).
Benthie macroinvertebrate communities in streams usually have clear taxonomie
hierarchies and functional groups (e.g., collectors, shredders, etc), and these are
essential to verify organizational characteristics in community compositions
(Cummins et al. 1973; Cummins 1974). By understanding associative information
on community organization, a comprehensive view on ecosystem could be
established, and this would help to prepare reliable strategies for achieving the
sustainable management of ecosystems. By applying the counterpropagation
network, we tried to elaborate the feasibility of artificial neural network to extract
information of interrelationships among hierarchical levels in communities of
benthie macroinvertebrates in streams (Park et al. 2001a).
The counterpropagation network (Hecht-Nielson 1987) is a type of hybrid
model consisting of the two artificial neural networks: the Kohonen selforganizing map (Kohonen 1989) and the Grossberg outstar (Grossberg 1969,
1982) (Figs. 8.18). The network is eventually designed to approximate continuous
functional associations between variables, and serves as a statistically optimal selfprogramming lookup-table (Hecht-Nielson 1990). The input data are arranged in
161
The sensitivity tests were also able to show variations for the different taxa of
community. Densities of Chironomidae, for example, varied greatly in response to
different input ranges. This could be observed for all the variables in the data of
July 1997 (Fig. 8.17a).
Densities of Hirudinea were also sensitive to
environmental variables in the data of July 1997. The higher sensitivity of
Hirudinea was also observed in the data of November 1997 without strong
environmental effects (Fig. 8.17b). Densities of Oligochaeta, in contrast, were
characteristieally insensitive to input variables, especially in July 1997. This
indieated that, in this field study, the density of Oligochaeta was not greatly
affected by environmental variables during the flooding period in comparison with
other dominant taxa such as Chironomidae and Chironomus sp. (Chon et al.
2000b)
This study examined the feasibility of the recurrent artificial neural network in
extracting information out of temporal development. The results showed that the
dynamies of sets of multivariate data about communities could be rapidly
pattemed and forecasted by the network.
8.4
Patterning Organizational Aspects of Community
8.4.1
Relationships among Hierarchical Levels in Communities
Useful ecologieal informaties resides in community organization, especially in
"associations" among different levels in communities such as taxonomieal or
functional groups. In these associative relationships, the complex community
usually develops a hierarchy, which is a good subject for understanding system
behavior of the target ecosystem (Allen and Starr 1982; O'neill et al. 1986).
Benthie macroinvertebrate communities in streams usually have clear taxonomie
hierarchies and functional groups (e.g., collectors, shredders, etc), and these are
essential to verify organizational characteristics in community compositions
(Cummins et al. 1973; Cummins 1974). By understanding associative information
on community organization, a comprehensive view on ecosystem could be
established, and this would help to prepare reliable strategies for achieving the
sustainable management of ecosystems. By applying the counterpropagation
network, we tried to elaborate the feasibility of artificial neural network to extract
information of interrelationships among hierarchical levels in communities of
benthie macroinvertebrates in streams (Park et al. 2001a).
The counterpropagation network (Hecht-Nielson 1987) is a type of hybrid
model consisting of the two artificial neural networks: the Kohonen selforganizing map (Kohonen 1989) and the Grossberg outstar (Grossberg 1969,
1982) (Figs. 8.18). The network is eventually designed to approximate continuous
functional associations between variables, and serves as a statistically optimal selfprogramming lookup-table (Hecht-Nielson 1990). The input data are arranged in