Propagation of Uncertainties in CMIP6-Based Multi-Model Streamflow Projections Across the Conterminous US

Changing climate and socioeconomic dynamics can significantly impact future hydrology. To understand these impacts and project future water availability, downscaled and bias-corrected climate model projections are often used to drive hydrologic models. However, this process involves significant uncertainties that require comprehensive characterization and quantification to effectively inform stakeholder decision-making. The need is particularly critical in large-scale applications with diverse hydro-climatological and geographical conditions, such as the Conterminous United States (CONUS). This study examines a high-resolution streamflow projections dataset (https://doi.org/10.13139/OLCF/2318650) developed under a multi-model framework. The dataset employs multiple downscaled Coupled Models Intercomparison Project phase 6 (CMIP6) global climate models for the baseline (1980–2019), near–future (2020–2059), and far–future (2060–2099) periods under four emission scenarios (SSP585, SSP370, SSP245, and SSP126). Both statistical and dynamical downscaling approaches (DBCCA and RegCM) are utilized, based on two reference meteorological forcings (Daymet and Livneh). These downscaled climate forcings then drive two calibrated hydrologic models (VIC and PRMS) to simulate total runoff. The Routing Application for Parallel computatIon of Discharge (RAPID) model routes the simulated runoff to produce a CONUS-wide multi-model ensemble of streamflow projections across the 2.7 million NHDPlusV2 stream reaches. To quantify and attribute the sources of uncertainty in these streamflow projections, we apply several simple yet effective forms of Analysis of Variance (ANOVA). Specifically, we quantify the uncertainties within our modeling framework related to changes in different flow regimes, including high flow, mean flow, and low flow. Furthermore, reach-scale flood frequency analysis will be demonstrated for one of the USGS hydrologic sub-regions (HUC02) to assess the shift in projected flood frequencies and underlying uncertainties. Regional variability of these projected changes and associated uncertainties will then be explored. Such comprehensive assessments will be crucial for regional water resource planning, management, and the development of effective climate resilience strategies.