Abstract
Generally, three-dimensional hydrodynamics and water quality models tend to have low reproducibility in a tidal river due to the complex fluctuation of flow field and external forces. In this study, we forecasted variations of water quality, such as dissolved oxygen and electric conductivity, in a tidal river by applying deep learning models, which were recently proposed application models of artificial neural network. Consequently, the results showed that a deep learning model can predict the variation in the water quality with high precision compared to normal feedforward neural network models and multiple regression analysis. In addition, the result showed the deep learning model is suitable for the regression of the phenomena regardless of seasonal variation.
