Dissolved Oxygen Prediction in Rivers Based on RF-LSTMModel: A Case Study of Dujiangyan

Accurate prediction of dissolved oxygen levels in river systems is essential for effective water quality management. A novel hybrid model combining Random Forest (RF) and Long Short-Term Memory (LSTM) networks is presented for dissolved oxygen forecasting in river ecosystems. The methodology utilizes RF to analyze feature importance and select relevant variables, reducing input dimensions and eliminating parameters with minimal influence on dissolved oxygen. LSTM networks then model the temporal relationships between the selected water quality parameters and dissolved oxygen levels. The model was validated using real-world river water quality data. Evaluation results show the RF-LSTM model achieves performance metrics of MSE (0.0028), RMSE (0.0529), MAE (0.0405) and R² (0.9890). The research demonstrates LSTM networks’ effectiveness for DO prediction and highlights the importance of data preprocessing and feature selection in environmental modeling.