Stream Flow Prediction: Narmada Basin

In river basins that are prone to flooding and drought, streamflow forecasting is essential for managing water resources. In this study, we used two machine learning methods, LSTM and Prophet, to estimate streamflow at a river location in the Narmada Basin, India. A forecasting model known as the Prophet was built expressly to forecast time series data that include seasonal and trend components. A recurrent neural network that can identify long-term patterns in data is the LSTM model. We tested and trained the LSTM and Prophet models using historical streamflow data from the Narmada Basin. The results showed that both models could accurately project streamflow at the river site, especially during the monsoon season. But overall, the LSTM model barely outperformed the Prophet model. We also found that the LSTM model could account for the non-linear relationship between streamflow and meteorological factors like rainfall and temperature. Because of the Narmada Basin's unique hydrology and climate, the LSTM model is a valuable tool for predicting streamflow there.