Floods in the United States represent a significant natural hazard, causing substantial economic losses and fatalities. Flash floods, a particularly devastating type of flood, account for a considerable portion of these losses. Understanding how future floods will evolve under climate change is crucial for effective mitigation and adaptation strategies. Flood risks are strongly linked to increased precipitation extremes, a consequence of global warming. While the Clausius-Clapeyron equation suggests a theoretical increase in atmospheric water holding capacity, the actual increase in extreme precipitation may be even greater due to changes in storm structure, dynamics, and large-scale weather patterns. Global Climate Models (GCMs), due to their coarse resolutions, struggle to accurately represent mesoscale weather systems critical for flash flood formation. Convection-permitting models, operating at finer spatial scales (kilometer-scale), offer a more accurate simulation of sub-daily extreme precipitation rates and provide valuable insights into climate change impacts on flash flooding. This study leverages a 13-year retrospective convection-permitting climate simulation and its future climate counterpart to investigate changes in flash floods across the conterminous United States (CONUS). The retrospective simulation downscales ERA-Interim reanalysis data, while the future simulation employs the Pseudo Global Warming (PGW) approach, adding climate change perturbations derived from an ensemble of CMIP5 GCMs under the RCP8.5 high-emissions scenario. The PGW approach focuses on thermodynamic climate change impacts, excluding uncertainties related to large-scale weather pattern shifts. The study employs the Ensemble Framework For Flash Flood Forecasting (EF5), an operational model used by the U.S. National Weather Service, to translate precipitation changes into changes in flood flashiness. By analyzing rainfall-flood event associations, the study aims to provide quantitative assessments of future flood-producing storms and flashiness changes, including geographical shifts in flash flood hotspots.

Several studies have highlighted the devastating impacts of floods in the US, emphasizing the economic losses and fatalities associated with these events, particularly flash floods. The strong relationship between rainfall rates and flood flashiness has been well-established, underscoring the importance of accurately representing convective processes in climate models. Previous research has noted the limitations of GCMs in simulating mesoscale weather systems relevant to flash floods, leading to the increased use of higher-resolution convection-permitting models. These models have demonstrated improved simulations of extreme precipitation, offering a valuable tool for assessing climate change impacts on flash flooding. Existing literature also emphasizes the challenges in projecting future flood risks due to uncertainties in climate models, hydrologic model structures, and parameterizations. Studies have reported varying trends in flood frequencies and magnitudes, with some suggesting decreasing trends in low-end flood frequencies due to the role of antecedent catchment states, while others show increasing trends in extreme precipitation. The inconsistencies highlight the complexities in projecting future flood risks and the importance of using high-resolution models.

This study utilizes hourly data from a 13-year retrospective convection-permitting climate simulation (CTL) and a corresponding future climate simulation (PGW) at 4-km grid spacing. The CTL simulation downscales ERA-Interim reanalysis data from 2000-2013, while the PGW simulation applies the Pseudo Global Warming approach, adding climate change perturbations from CMIP5 GCMs under RCP8.5 to the ERA-Interim boundary conditions. Both simulations are used to drive the EF5 hydrologic model, simulating streamflow at a 1-km spatial resolution and hourly temporal resolution from 2001-2011. The study employs the Characteristic Points Method (CPM) to establish rainfall-flood event associations, enabling the extraction of relevant hydrological characteristics, including rainfall duration, peak rainfall rate, rainfall volume, flood duration, peak flow rate, flood volume, and rainfall-flood lag time. A flashiness index is calculated for each event using Equation 1, which considers peak flow, baseflow, drainage area, and flood rising time. Basin-level flashiness indices are determined as the median of event-based flashiness values. The retrospective simulation (CTL) is validated against Stage IV radar-gauge merged precipitation data and the Global Streamflow Characteristics Dataset, addressing potential biases in precipitation and runoff estimations. The study aggregates the results into 17 Bukovsky climate regions to allow for regional comparisons and interpretations of climate change impacts.

The study’s key findings are as follows: 1. **CONUS-wide increase in flashiness:** A 7.9% increase in flashiness across the CONUS is projected by the end of the century under the RCP8.5 scenario. This indicates that future floods will onset more rapidly with higher peak runoff and less time for early warning. 2. **Regional variations in flashiness increase:** While regions outside the North American Monsoon (NAM) experienced moderate flashiness increases (+4.1%), the Southwest showed a significantly larger increase (+10.5%). This disproportionate increase is attributed to the NAM’s influence and the “it never rains, but it pours” pattern typical of arid climates. 3. **Emergence of new flash flood hotspots:** The central US, historically less prone to flash floods, experienced an average 8.5% increase in flashiness. The Prairie and Deep South regions are identified as transitioning into new flash flood hotspots. 4. **Northward expansion of flash flood risk:** The study revealed a northward advancement of flashiness increases, impacting regions like the Northern Rockies, Northern Plains, and Prairie. This geographic shift in risk requires recalibration of local flood resilience measures. 5. **Changes in rainfall and flood characteristics:** The analysis revealed consistent decreases in rainfall and flood durations across most of the CONUS, while peak rainfall rates and flow rates showed positive changes, particularly in the western US. The decrease in lag time between rainfall and flood peaks indicates faster flood responses, contributing to increased flashiness. 6. **Relationship between rainfall changes and flashiness:** Peak rainfall rates and rainfall volumes are strongly positively correlated with flashiness, with increases exceeding 120% causing the flashiness index to plateau, indicating a physical limit in channel conveyance due to geomorphic and land cover constraints.

The findings of this study directly address the research question of how future flood flashiness will respond to climate change. The significant CONUS-wide increase in flashiness highlights the need for urgent adaptation measures. The identification of emerging flash flood hotspots, especially in the central US, necessitates reevaluation of current floodwater management strategies and infrastructure. The northward expansion of flash flood risk further underscores the challenges in adapting to a changing climate. The observed reduction in flood and rainfall durations, coupled with increases in peak values, points towards more intense but shorter-duration events, making early warning systems critical. The close relationship between rainfall characteristics (intensity and volume) and flashiness further emphasizes the importance of accurately predicting extreme precipitation in climate models. The limitations of the PGW approach in capturing changes in large-scale weather patterns should be considered when interpreting the results. The study provides a valuable baseline for future research, highlighting the need to incorporate other anthropogenic factors like land-use changes and urbanization to gain a comprehensive understanding of future flood risks.

This study provides compelling evidence of a significant increase in future flood flashiness across the CONUS under a high-emissions scenario. The findings identify existing and emerging flash flood hotspots, highlight regional variations in flashiness increases, and demonstrate the northward expansion of future flash flood risk. The study emphasizes the need for immediate adaptation and mitigation measures, urging the reassessment of current flood risk management strategies and infrastructure. Future research should investigate the seasonal cycles and spatial extent of storms and floods, explore the spatiotemporal correlation between the two, and incorporate the impacts of anthropogenic factors such as land-use change and river controls to refine flood risk projections.

The study's conclusions are based on a high-end emissions scenario (RCP8.5), which may not be realized in the future. The PGW approach primarily focuses on thermodynamic climate change impacts, neglecting the potential effects of changes in large-scale weather patterns. Anthropogenic influences like land-use changes and river regulations are not explicitly considered, although it is acknowledged that such factors can exacerbate flood risks. The model's limitations in simulating snowmelt and rain-on-snow events might affect the accuracy of predictions in regions where these events are significant contributors to flooding.