Most probable pollution source Identification in rivers by neural networks

Abstract

In this study, a certain characteristic of neural networks called Self Organizing Feature Maps (SOFM's) was applied to pollution source identification in the Kor and Sivand Rivers located inFars Province,Iran. Wastewater quality data from significant industrial pollution sources to these rivers (mainly factories located upstream) were given. Observed sets of water quality data in sampling stations, downstream from the pollution sources, were utilized to identify the most probable pollution source that may have contributed to pollution in these rivers. With the aid of partial semantic maps generated by SOFM's, different patterns with different likelihoods were recognized in the pollution data. Certain patterns matched that of the pollution sources very closely. In other words, the fingerprints of all pollution sources (which were studied) were identified in the pollution data. Therefore, it is possible to use the maps as an aid to the management and decision support system of these rivers.