CS25E - Concurrent Session 25E: Assessing Change and Uncertainty in Future Flood Estimates with Large-Ensemble Streamflow Projections

Abstract Description: The magnitude and frequency of extreme weather events are projected to intensify under changing hydroclimate conditions, posing significant implications for dam safety. This is particularly concerning as many dams in the conterminous US (CONUS) are rated poorly in terms of flood regulation capacity, structural condition, public safety, and resilience. Additionally, the changes of extreme streamflow are expected to be inconsistent and non-homogeneous across the CONUS, further complicating risk prioritization and emergency preparedness. Since many energy-water stakeholders lack the analytical and computational resources to assess climate-driven flood risks, a large-scale, nonstationary flood risk assessment is needed to identify the most critical streams and regions at risk.

Critical to this assessment is quantifying the impact of future extreme events at the basin scale, including both flood hazards and potential dam failure consequences. To achieve this, we implement a multi-model, uncertainty-aware framework to evaluate projected changes in flood frequencies across ~2.7 million NHDPlusV2 river reaches across the CONUS. The framework leverages a large ensemble of future streamflow projections derived from multiple downscaled and bias-corrected global climate model outputs from the Climate Model Intercomparison Project Phase 6 (CMIP6). We develop CONUS-wide flood frequency estimates using spatially consistent, L-moment based regional flood frequency approach, assess projected changes, and characterize uncertainties using a simple yet efficient Analysis of Variance (ANOVA) method.
The initial results suggest that a large majority of stream reaches in the CONUS exhibit significant increase in projected flood frequencies. While climate model selection remains the dominant source of uncertainty in projected flood frequency changes, other factors, such as the choice of hydrologic model, downscaling methods, and probabilistic distribution, also influence projections. Furthermore, we employ nonstationary methods—often absent from the conventional dam safety risk assessments—to account for evolving climate dynamics, land use land cover changes, and socio-economic factors. We demonstrate a proof of concept for selected hydropower reservoirs across the CONUS, laying the groundwork for a large-scale consequence modelling as the next step of the improved dam safety risk assessment framework. Overall, this effort aims to better support decision-making across multiple sectors, including dam owners and managers.