A MONITORING METHOD FOR POLLUTANT LOAD IN RIVERS BY USING ARTIFICIAL NEURAL NETWORK

This paper proposes a new method for continuous measurement of pollutant load in rivers without much cost. The basic idea is making the most of “empirical correlations which exist in the target” in order to relate what we can measure to what we want to know. In a field experiment presented here, signals from two types of optical sensors were used to estimate the loads of chemical oxygen demand (COD), total nitrogen (T-N) and total phosphorus (T-P), and artificial neural network (ANN) models were trained to fix “the empirical correlations” among them. The field data were collected in seven rivers located in the watershed of Lake Kasumigaura. The experimental results showed that the three items of water quality were stably estimated with good accuracy for rather long time without too much training data.