Abstract Description: Unpredictability and variability in rainfall, snowpack, and temperatures are increasing and complicating the work of dam safety engineers and water management operators. Though most hydrologic prediction models still rely on historical relationships between inputs and outputs, machine learning hydrologic models are consistently outperforming traditional models. However, there is a gap in integrating these technological advancements into dam safety operations for managing extreme weather events, reducing risk, and monitoring construction. In this presentation we will share the results in the form of accuracy benchmarks, which show operational machine learning models (HydroForecast in this case) outperforming traditional approaches in different hydrologic regimes at 1-10 days ahead. We will use a case study from Hurricane Helene to discuss how HydroForecast increases the accuracy and density of critical real-time weather and streamflow data during extreme events.
Learning Objectives:
Learn about machine learning hydrologic forecasting models for dam safety applications.
Understand basic differences between machine learning and traditional hydrologic models.
Learn how machine learning hydrology is suited for managing through extremes and climate change.
