Abstract

Total coliforms, as a microbiological indicator of water quality, have been tested on the basis of condition, dynamics, as well as on the dependence on other physicochemical and biological parameters, by methods and models of data mining. Using a combination of intelligent approaches, cluster analysis and classification, total coliforms have been analyzed and modeled on the examples of the Gruža and the Grošnica reservoirs. These reservoirs have different morphometric characteristics, different trophic status as well as dominant bacterial communities. The study is based on the existing information system and automated data analyses for the period of 10 years. The system determines the accuracy of analyses by validity percentage. The analyses show that the number of total coliforms is connected to anthropogenic activity, the amount of organic mater, as well as to the presence of bacterial community which is not dominant or characteristic for the specific reservoir.

Key words: Total coliforms, reservoir, water quality, data mining, cluster analysis, classification.

Abbreviation

TC, total coliforms; H, heterotrophs; Hm, heterotrophs (mesophile); FO, facultative oligotrophs; BOD₅, 5-day biochemical oxygen demand; Mn, manganese; COD, chemical oxygen demand; TP, total phosphate; EC, conductivity; Fe, iron; NH₄⁺,ammonia; Chl-a, chlorophyll a; Cl^-, chloride; TSS, total suspended solids; MPN, Most Probable Number; cfu, colony forming units; CA, cluster analysis.