A Neural Network Based Method for Modelling Spatio-Temporal Behaviour of Environmental Pollutants

Uncertainty due to spatial and temporal variability of different pollutants found in soil, water and air is an important issue in the modelling of environmental impacts. Recognition of spatial environmental patterns is relatively straightforward to achieve by applying most of the well-known artificial neural networks or adopting a stochastic approach. However, it is the temporal component of the environmental variables that makes modelling of the environmental pollutants particularly complicated and challenging. This paper presents the results of a case study investigating the artificial neural network performance when modelling the spatio-temporal behaviour of environmental pollutants. The test case involved the modelling and prediction of the distribution of water quality indicators along a river from the point of discharge of treated mine effluents. The behaviour of river water quality indicators is rather complex and shows a lot of fluctuations due to a number of hydrochemical, hydrobiological and hydrodynamic factors. Results obtained in this study indicate that artificial neural networks provide a very powerful and robust method for simultaneous modelling of spatio-temporal behaviour of environmental pollutants.