Composition of the bacterial population in sand-filter columns receiving artificial wastewater, evaluated by soft independent modelling of class analogy (SIMCA)

The objective of the present study was to develop a test system for clustering and characterizing the functions of the bacterial populations in wastewater infiltration systems. To create different bacterial ecosystems, sand-filter columns were loaded with a low loading (2.9 ± 0.1 cm d−1) and a high loading (8.7 ± 0.2 cm d−1) of an artificial wastewater for 114 days. Before loading onto the columns the wastewater was pretreated in a three-chambered container for 24 h. The low loaded filters attained a high and steady rate of mineralization and complete nitrification, whereas the high loaded filters developed a clogging surface, and incomplete nitrification despite a high rate of mineralization. On day 60 the sand surface was sampled, and the bacteria were extracted and enumdrated. From each column system 88 bacterial isolates were collected. Each isolate was subjected to 51 physiological and biochemical tests. The results were analyzed using the SIMCA statistical package. Using partial least-square regression discriminant analysis we developed a model that revealed five major clusters of bacteria among the 176 isolates. Two clusters consisted of bacteria from the high loaded filters, and another two clusters consisted of bacteria from the low loaded filters. The fifth cluster represented a mixed population containing members from both loading levels. The results were used to characterize the bacterial populations, and a hypothesis as to how these populations developed in their respective system was proposed. The model could be used for classifying the bacterial isolates into low or high loaded system. Only six isolates from each loading level were misclassified. The model was stable as indicated by the fact that both the numbers of tests and the numbers of isolates entered in the model could be reduced without any significant loss of information.