| Abstrakt: |
This study examines the quality of effluents discharged from the Century paper and pulp mill into the Gola River to assess its suitability for irrigation. Water quality depends on the assessment of the many parameters determined by chemical methods. A number of water quality parameters such as pH, electrical conductivity, turbidity (Tb), alkalinity, acidity, total hardness, total dissolved solids, calcium, magnesium, potassium, chloride, sodium, nitrate-nitrogen (NO3-N), sulfate, phosphate, biochemical oxygen demand (BOD), and chemical oxygen demand (COD) were determined for the effluents in Gola River. For this study, samples from 20 locations along the length of the Gola River in Uttarakhand were collected. The result of water quality analysis showed that the concentration of Tb, NO3-N, COD, and BOD in the effluent water from the Century paper and pulp mill was beyond the permissible limit for its discharge in the water body or river. The concentration of all the parameters was decreasing along the length of the channel. A high concentration of COD, BOD, and NO3-N was found in downstream areas as compared to upstream discharge areas of effluent channels. The determination of biological water parameters such as BOD and COD is often tedious, time consuming, and error prone. Moreover, the relationships that exist between different water quality parameters to predict BOD and COD are complex and nonlinear. Artificial neural networks (ANNs) are accurate models for predicting nonlinear and complex processes. The main objective of this study is to estimate BOD and COD accurately with limited significant water quality parameters as input data using the ANN model and compare the performance of ANN with the multiple linear regression (MLR) models. In this study, a three-layer feed-forward back-propagation-type ANN was used to model the relationship between the water quality parameters and BOD and COD. Training, validation, and testing of ANN models were done by splitting data into various proportions. An optimum ANN model was decided based on the number of trial and error processes using hidden nodes, epochs, and other training parameters. Performance of the models was evaluated by root mean squared error, mean absolute error, coefficient of determination, and ratio of output by target. Comparison of results showed that the ANN model gave reasonable estimates for the BOD and COD prediction. Hence, it is concluded that the ANNs were able to show remarkable performance to predicting the BOD and COD in Gola River. [ABSTRACT FROM AUTHOR] |
Full text from SpringerLink