Determination and modeling of benzene micropollutant in surface waters: Orhaneli river study

Priority micro-pollutants are an important research topic today due to their increasing concentration and their resistant structure in water environments. Particularly, mining activities, the widespread of thermal power plants, the increase in the use of motor vehicles and the development of the industry, the resources that cause the priority micro-pollutant existence are increasing day by day. In this study, benzene pollutant, which is one of the primary micro-pollutants, was monitored for 12 months. Orhaneli river, which is located in an area where industrial and mining activities are carried out intensely and where thermal power plants are located, has been chosen as the study area. In the modeling study, an estimation model of the benzene pollutant was made using artificial neural networks; the effects of meteorological, some physical and chemical parameters on benzene concentrations were investigated. Artificial neural networks structure consists of an input layer, a hidden layer and an output layer. After basic component analysis application, artificial neural networks are used as input parameters, precipitation, temperature, humidity, pH, flow rate, dissolved oxygen, conductivity, total organic carbon and total nitrogen. While creating the network, 11 different back propagation algorithms were compared with mean square error values and Levenberg - Marquardt algorithm was determined as the best training algorithm. In the estimation of benzene concentration produced for the selected training algorithm, the number of optimized neurons to be used in the hidden layer was found to be 6. The correlation coefficient (R2) for the optimized three-layer artificial neural network model (9:6:1) was obtained as 0.9889. As a result of the study, it can be clearly seen that the results of artificial neural network based models are quite suitable for the concentration of benzene in surface waters.