<abstract xmlns="http://www.w3.org/1999/xhtml">The Riva River is a water basin located within the borders of Istanbul in the Marmara Region (Turkey) in the south-north direction. Water samples were taken for the 35 km drainage area of the Riva River Basin before the river flows into the Black Sea at 4 stations on the Riva River every month and analyses were carried out. Changes were observed in the quality of water from upstream to downstream. For this purpose, the spatial and temporal variations of water quality were investigated using 13 water quality variables with the ANOVA test. It was observed that COD, <italic>DO</italic>, S and BOD were important in determining the spatial variation. On the other hand, it was found out that all the variables were effective in determining the temporal variation. Moreover, the correlation analysis which was carried out in order to assess the relations between water quality variables showed that the variables of BOD-COD, BOD<italic>-EC</italic>, COD<italic>-EC,</italic> BOD<italic>-T</italic> and COD<italic>-T</italic> were correlated and the regression analysis showed that COD, <italic>TKN</italic> and NH4-N explained BOD and BOD, NH4-N, <italic>T</italic> and <italic>TSS</italic> explained COD by approximately 80 %. Consequently, the Artificial Neural Network (<italic>ANN</italic>), Decision Tree and Logistic Regression models were developed using the data of training set in order to predict the water quality classes of the variables of COD, BOD and NH4-N. Quality classes were predicted for the variables by inputting the data of testing set into the developed models. According to these results, it was seen that the <italic>ANN</italic> was the best prediction model for COD, the Decision Tree for BOD and the <italic>ANN</italic> and Decision Tree for NH4-N.</abstract>

The Riva River is a water basin located within the borders of Istanbul in the Marmara Region (Turkey) in the south-north direction. Water samples were taken for the 35 km drainage area of the Riva River Basin before the river flows into the Black Sea at 4 stations on the Riva River every month and analyses were carried out. Changes were observed in the quality of water from upstream to downstream. For this purpose, the spatial and temporal variations of water quality were investigated using 13 water quality variables with the ANOVA test. It was observed that COD, DO, S and BOD were important in determining the spatial variation. On the other hand, it was found out that all the variables were effective in determining the temporal variation. Moreover, the correlation analysis which was carried out in order to assess the relations between water quality variables showed that the variables of BOD-COD, BOD-EC, COD-EC, BOD-T and COD-T were correlated and the regression analysis showed that COD, TKN and NH4-N explained BOD and BOD, NH4-N, T and TSS explained COD by approximately 80 %. Consequently, the Artificial Neural Network (ANN), Decision Tree and Logistic Regression models were developed using the data of training set in order to predict the water quality classes of the variables of COD, BOD and NH4-N. Quality classes were predicted for the variables by inputting the data of testing set into the developed models. According to these results, it was seen that the ANN was the best prediction model for COD, the Decision Tree for BOD and the ANN and Decision Tree for NH4-N.

Prediction of Water Quality in Riva River Watershed

Ecological Chemistry and Engineering S

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

{"article-title":"Prediction of Water Quality in Riva River Watershed"}

The Riva River is a water basin located within the borders of Istanbul in the Marmara Region (Turkey) in the south-north direction. Water samples were taken...

Prediction of Water Quality in Riva River Watershed

Pubblicato online: 21 gen 2020

Pagine: 727 - 742

DOI: https://doi.org/10.1515/eces-2019-0051

Parole chiave
Riva river, water quality, artificial neural network (), decision tree, regression models

© 2019 Nurtac Oz et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Prediction of Water Quality in Riva River Watershed

Pubblicato online: 21 gen 2020

Pagine: 727 - 742

DOI: https://doi.org/10.1515/eces-2019-0051

Parole chiaveRiva river, water quality, artificial neural network (), decision tree, regression models

© 2019 Nurtac Oz et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.

Parole chiave
Riva river, water quality, artificial neural network (), decision tree, regression models