Abstract
Principal component analysis, cluster analysis and neural network were applied to assess the pollution levels along the Yoshinogawa River in Tokushima, Japan, by using water quality data (Dissolved Oxygen (DO), Biochemical Oxygen Demand (BOD), Chemical Oxygen Demand (COD),and Total Phosphorus (TP)) measured in 2002. Because there are a few defects in the data, we estimated them by linear equation and neural network to avoid information loss.
In the Yoshinogawa River, there are four branching bays (Figure 1). Though it is usually difficult to find the relationships between the water quality of the Yoshinogawa River and the distances from their estuaries, the first principal components consisting of five parameters explained the nature of the river well.
The water quality at Heiwabashi of the Shinmachigawa River was the most contaminated in the Yoshinogawa River (Figure 6, Table 12). This suggests that there is a source of water pollution in the upper region of Heiwabashi and that there is a source to improve water quality in the lower region of Heiwabashi.
