Water Resources Research Act Program

Details for Project ID ND-2025-Chu-1

Enhancing Daily Streamflow Forecasting: A Coupled GPR-EBBO Hybrid Machine Learning Method

Institute: North Dakota
Year Established: 2025 Start Date: 2026-01-01 End Date: 2026-12-31
Total Federal Funds: $12,978 Total Non-Federal Funds: $13,040

Principal Investigators: Xuefeng Chu

Project Summary: Accurate streamflow prediction is essential in hydrology for flood warnings, water resource management, and disaster readiness. More researchers are now engaged in the development of accurate machine learning (ML) models. However, there are some limitations in such ML models, including issues associated with analyzing high-dimensional time series, generalizing over diverse time horizons, and optimizing suitable predictors. The goal of this study is to develop an innovative hybrid ML framework for streamflow forecasting. To forecast streamflow using daily discharge and simulated input data based on precipitation, a novel hybrid ML model, GPR-EBBO, will be developed. The research will employ two decomposition techniques for breaking down signals to deal with complexity and non-stationarity. It will also utilize feature selection for selecting the best inputs to reduce dimensions. The proposed framework will be evaluated by comparing with various robust ML models to highlight its improvements in computational efficiency and accuracy for predicting streamflow.