Authors: Savalan Naser Neisary – The University of Alabama
Title: Improving NextGen Streamflow Simulations Using a Post-Processing Machine Learning-based Framework
Presentation Type: Poster
Abstract: Accurate streamflow prediction is critical for effective water resources management, particularly in the drought-prone western US. The Next Generation Water Resources Modeling Framework (NextGen), a modular, open-source successor to the National Water Model, can generate streamflow predictions across 2.7 million reaches in CONUS. However, it has limitations in drought-prone watersheds with extensive water resources infrastructure, as well as modeling dominant hydrological processes, such as snowmelt. To address the limitations of NextGen, we developed a post-processing machine learning (PP-ML) framework using the XGBoost algorithm to enhance NextGen’s ability to capture dominant hydrological processes and account for water regulation effects from 1990 to 2020 in the Great Salt Lake Basin. Preliminary results indicate significant improvements in Mean Percentage Error, Percent Bias, and Kling-Gupta Efficiency. These findings highlight the potential of data-driven approaches and high spatial resolution SWE data to improve NWM forecasts by capturing water regulations and dominant hydrological processes.