Long-term prediction of multiple river water quality indexes based on hybrid deep learning models

Rivers are an important part of the natural water cycle, but they are facing serious pollution problems due to a variety of human activities. Long-term prediction of river water quality indexes (WQI) is important for the protection of river water environment. Currently, data-driven deep learning models are effective in the task of long-term prediction of WQI, especially the transformer structure-based models have achieved advanced prediction results on a variety of water quality datasets. However, the high computational complexity of transformer models and their insensitivity to anomalous data have limited the application of the models. In this study, we propose a channel independent linear transformer model that has higher prediction accuracy and computational efficiency than the transformer model. We conducted long-term predictions of two WQI, dissolved oxygen and chlorophyll concentration, in the Liaohe River Basin and compared them with a variety of different advanced models. The experimental results show that our model has the best prediction results among all comparative models, and the proposed method for long-term prediction of river WQI provides effective technical support for the establishment of a river water environment monitoring system.