An artificial neural networks approach and hybrid method with wavelet transform to investigate the quality of Tallo River, Indonesia

Water contamination has always been one of the greatest intense environmental issues. Rivers are more polluted than the other surface and underground water resources, since passing through different areas. The current study aimed to examine the exactitude of artificial neural networks (ANN) and wavelet-ANN (WANN) models in estimating the concentrations of pollutants including Cl, EC, Mg, and TDS by comparing the results of the observed data. Tallo River in Indonesia was selected as the case study. The concentrations of pollutant parameters Cl, EC, Mg, and TDS were available and used between 2010 and 2022. Then 70% (100 months) of the data were considered as training data, while 30% (44 months) were supposed to be the testing ones. ANN and WANN models were examined to evaluate and predict the concentrations of pollutants in river water. The results of each model were compared to the observed data, and the models' accuracy was assessed. The results demonstrated that applying wavelet transform improved the precision of simulation. All efficiency criteria associated with the WANN model yielded superior results compared to the ANN model. The findings indicated that using the hybrid method with wavelet transformation ameliorated the ANN model's exactitude by 10% during training and 16% during testing. Finally, the findings exhibited that the WANN method is better than ANN; consequently, the former has performed more exactitude modeling in the estimation of water quality.