Predicting Biochemical Oxygen Demand As Indicator Of River Pollution Using Artificial Neural Networks

Abstract

Artificial Neural Networks (ANNs) are frequently used to predict various ecological processes and phenomenon related to water resources. Various ANN applications involve the prediction of water quality using various environmental parameters. This paper discusses the use of two environmental data sets from two sampling sites, merged together and applied to forecasting a dependent variable, namely biological oxygen in demand (BOD) using ANN modelling. ANN was applied to map the relationships between physical, chemical and biological time-series data of Sungai Air Itam, Pulau Pinang, Malaysia. This river is part of the Sungai Pinang river basin and is considered as highly eutrophic. Sungai Pinang and its tributaries are the main rivers flowing through the state of Penang, Malaysia. The water quality has been increasingly deteriorated by both natural and anthropogenic effects. The purpose of this study is to investigate the application of ANN to predict the biochemical oxygen demand as a measure of eutrophication status of rivers. The results of the study show that it is possible to forecast 1 month ahead BOD for Sungai Air Itam using a simple ANN with 16-4-1 architecture. The most important input for this predictive model is phosphate, and the sensitivity of the ANN models to the inputs is also dependent on the training datasets.