A Practical Approach to the Estimattion of Pollutant Load From Optical Sensor Measurement

Abstract

Reduction and control of pollutant load from non-point sources is a key issue to improve the water condition in lakes. In the monitoring of water quality in rivers for this purpose, continuous measurement is required because the pollutant load from non-point sources is highly variable in a process of rain runoff. In practice, however, this requirement is often contradictory to the requirement of data accuracy. This paper proposes a new idea on the "concept of measurement" in order to develop a "dynamic measurement" of the pollutant load. The point of authors' idea is to use empirical relations found in the field as positively as possible, even if we don't have any scientific reason for the relation. Then, results of a field experiment on the pollutant load monitoring are presented to show the feasibility of the idea.

A Monitoring Method for Pollutant Load in Rivers by Using Artificial Neural Network

“…In order to acquire more information from the environment, we should make full use of existing knowledge including empirical and local ones rather than simply pursuing the perfect or universal knowledge of science. In the previous study 5) , the authors proposed an idea of utilizing empirical and local correlations to estimate time series of pollutant loads in rivers. Artificial Neural Network must be a power tool to model the correlations because we behave successfully in daily life by utilizing the empirical and local correlations stored in our human neural networks.…”

“…Considering the above mentioned conditions, the authors proposed a technique of water quality measurement in which signals from an optical measurement device are converted to time series of COD, T-P and T-N being based on empirical correlations obtained in each river 5) . The technique was extended by introducing ANN (artificial neural network) for modeling the empirical relations 6) .…”