Accurate forecasting of reservoir inflow is essential for efficient water resource management and operations. Considering the nonlinear and nonstationary characteristics of real-world hydrological data, this paper introduces a novel model (EMD-LSTM) that employs empirical mode decomposition (EMD) in conjunction with long short-term memory (LSTM) networks for predicting daily reservoir inflow up to 10 days lead time. The accuracy of the model is assessed using Nash–Sutcliffe efficiency (NSE), root mean square error (RMSE) and mean absolute error (MAE). To assess its effectiveness, the proposed EMD-LSTM model is compared with standard artificial neural networks (ANN) and LSTM models for 3-day, 7-day, and 10-day lead times. Daily inflow data recorded using water level recorders from 2013 to 2022 for the Bhakra reservoir, located on the Sutlej River in Himachal Pradesh, India, serves as the case study. The EMD-LSTM model shows promising results, achieving an NSE of up to 0.94 in the validation phase for the 10-day forecast, outperforming the ANN and LSTM models. These findings suggest that the EMD-LSTM model offers valuable insights for decision-making and can enhance the safety and efficiency of reservoir operations.