Water Resources Research Act Program

Details for Project ID ND-2024-Chu-1

A Novel Hybrid Machine Learning Method for Streamflow Forecasting in Cold Regions

Institute: North Dakota
Year Established: 2024 Start Date: 2024-09-01 End Date: 2025-08-31
Total Federal Funds: $9,400 Total Non-Federal Funds: $9,795

Principal Investigators: Xuefeng Chu

Project Summary: Snowmelt plays a crucial role in the hydrologic systems of the northern United States, significantly impacting water supplies. Scholars are now engaged in the development of accurate models using machine learning (ML) models that comprehensively capture the intricate relationships of snowmelt, streamflow, and other related hydrologic processes. Although promising, the use of ML for streamflow prediction has been focused on rainfall-dominated watersheds. This study endeavors to develop a novel intelligent hybrid ML model for streamflow prediction in cold regions under the influence of snowmelt. This endeavor follows the identification of the most effective feature selection method between the cross-correlation and Baruta-SHAP methodologies, with a parallel focus on enhancing the input data architecture. The new modeling approach will be compared with other robust models to demonstrate its improvements in computational efficiency and accuracy. It will be tested in a watershed in North Dakota to highlight its broad implications in hydrology.