Authors: Ian Mathews, Gustavious Williams, Jim Nelson, Riley Hales – Brigham Young University
Title: An Alternative Method to Calibration for Improving Large-Scale Models Locally
Presentation Type: Lightning Talk
Abstract: Global hydrologic models like GEOGLOWS and the National Water Model (NWM) offer global river discharge hindcasts and daily forecasts, especially valuable in data-scarce regions. However, these models frequently have biased flow magnitudes, limiting their local utility. Since fully calibrating a global model is impractical if not impossible, post-processing bias correction is applied. The Stream Analysis for Bias Estimation and Reduction (SABER) is a postprocessor designed for correcting discharge from large hydrologic models. SABER uses geospatial analysis and statistics to generalize bias patterns observed at gauged basins to ungauged ones. This bias correction enhances historic data and can improve forecasts for local decision-making. Our ongoing work includes collaborating with the Colorado Basin River Forecast Center (CBRFC) to integrate their local data with GEOGLOWS and the NWM. We plan to use Machine Learning (ML) and SABER to directly enhance the bias correction process, using data from the CBRFC as “true” values for training, aiming to apply these ML-enhanced methods as post-processors to GEOGLOWS and the NWM, increasing their effectiveness for local forecasting. This process can then be applied to other forecasting agencies, enhancing their ability to use global models.