The eventual goal of this syudy was to develop integrated water quality indices used for the effective wastewater management. As a first step to accomplish this goal, various polluted streams were pattern-recognized by analyzing microbial, algal and invertebrate communities through artificial neural networks (ANNs) techniques. The sampling sites were selected based on different sources of pollution: farming, livestock, domestic and industrial sites. Physico-chemical environmental factors were measured at all the sampling sites. Grouping of collected microbial and benthic macro-invertebrate communities revealed the impact of sources of pollution. Microbial taxa were diverse at the clean sites while microbial communities were tolerant at the polluted sites. SOM results showed that the sampling sites were essentially categorized into a few areas of SOM map: reference site-dominant area, polluted site-dominant area and mixed area of the two different sites. The eubacterial populations and their distribution in the SOM map appeared to be significantly affected by organic substrates and nutrients in the sites. If the identification of eubacterial communities were completed, more specific relations of the microbial communities with environmental and other biological data would be revealed. In macro-invertebrate communities, species richness was high and the taxa were diversely distributed at the clean sites. At the polluted sites, in contrast, a few species tolerant to organic pollution were dominantly present. The taxa occurring at the intermediate range of pollution were distributed differently depending upon the sources of pollution. For algal community analysis, the population was identified up to a genus level and the data showed a specific distribution according to the sampling sites and dates, to a certain extent.
However, it may be possible to obtain more specific data useful for the community patterning if the population could be identified up to species level. Once the biological and environmental data are integratively processed using a neural network, multilayer receptron (MLP), a target integrative water quality index(e.g., species richness) will be developed. The index will be quite useful for an ecosystem-wise monitoring of the water quality status in various kinds of water bodies.