University of North Dakota, North Dakota, United States
Abstract Submission: Accurate river discharge prediction is essential for effective water resource management. Traditional hydrological models have been used to simulate river discharge, but these methods often require extensive data inputs and complex calibration, limiting their applicability in extreme weather regions. In recent years, machine learning models such as Long Short-term models (LSTM) have emerged as an alternative for time-series forecasting. LSTM networks capture relationships in hydrological data, making them promising tools for river discharge prediction. This research aims to develop a hydrological model for river discharge in the Upper turtle river using LSTM and compare its performance with a traditional hydrological model. The study will use historical hydrological data, such as precipitation, temperature, and discharge measurements, to train and validate both models. Key performance metrics such as Root Mean Square Error (RMSE), and Nash-Sutcliffe Efficiency (NSE) will be used to evaluate the robustness of the models. By comparing the strengths and limitations of LSTM and traditional models, this research will provide valuable insights into the potential of machine learning for hydrological forecasting. The outcomes are expected to advance river discharge prediction techniques and offer practical recommendations for applying LSTM-based models across different river basins and hydrological conditions.
Learning Objectives/Expected Outcome (Optional) : A fully trained LSTM-based model capable of predicting river discharge with high temporal resolution. A calibrated traditional hydrological model for the same river basin. A comprehensive comparison of the LSTM and traditional models in terms of accuracy, robustness, and efficiency. Insights into the potential benefits and challenges of using LSTM for hydrological modeling in comparison to traditional methods. Recommendations for applying LSTM models to other river basins or integrating them with traditional models for enhanced hydrological forecasting.
