Abstract
Artificial neural network (ANN) is applied to a predictive model for the intermittent occurrence of taste-and-odor problem in the source of drinking water in Kamafusa Reservoir, Japan. To predict the temporal variation of 2-methylisoborneol (MIB) concentration, which triggers taste-and-odor problem, ten-year data of water quality from continuous water quality monitor as well as using the microscopic analysis data of planktonic cyanobacteria. The model examined in this paper is capable of reproducing the trend of evolution of MIB concentration and hence the intermittent occurrence of taste-and-odor events observed in Kamafusa Reservoir. Thus the model can be used as a decision-making tool for reservoir management office in measuring and treating the quality of water.
