THE PREDICTION OF FLOW-RATE AND NUTRIENT LOAD IN ERGENE RIVER BASIN THROUGH ARTIFICIAL NEURAL NETWORKS

Abstract

This study aims to predict the highest rate of monthly average flow and load change in Ergene River, one of the most contaminated rivers of Turkey and having a high flood frequency. For this purpose, the Flow Observation Station (FOS) of Luleburgaz district was chosen for modelling as it is located at a point in the middle of the basin, where domestic and industrial wastes of the region with the population density of basin reach and seasonal floods are observed. An artificial neural networks method, the Feed-Forward Back Propagation Neural Networks (FFBPNN), method was used to evaluate the relation among hydro-meteorological data of Luleburgaz FOS recorded for 168 months between 1997 and 2010, and the flow-rate of Ergene River Luleburgaz Station was predicted monthly for the year of 2011. The load change in the river was observed with direct calculation method on the basis of the acquired flow-rate values and long-term nutrient concentration averages.