Neglecting Spatiotemporal Rainfall Variability Underestimates Flood Hazard and Risk

Neglecting Spatiotemporal Rainfall Variability Underestimates Flood Hazard and Risk | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Neglecting Spatiotemporal Rainfall Variability Underestimates Flood Hazard and Risk John A. Baer, Antonia Sebastian, Lauren E. Grimley, James Doss-Gollin, and 3 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8593870/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Flood risk assessments underpin flood management and resilience efforts worldwide, including land-use planning, infrastructure design, and insurance requirements. Many of these assessments rely on design storms, which assume a linear relationship between the frequency of storms, flooding, and damage and neglect the spatial and temporal structure of rainfall. Here, we show that these assumptions can lead to systematic underestimates of flood hazard and risk. Using a coastal watershed in North Carolina, we compare design storm-based estimates with those from stochastic storm transposition, a probabilistic framework used to generate synthetic events with realistic rainfall fields. Though both methods produce similar basin-averaged rainfall statistics, design storms underestimate flood inundation frequency by 31 to 35% and expected 50-year damage by 42% relative to SST. These results reveal how nonlinear storm-flood-damage relationships amplify risk from smaller, more frequent storms and illustrate that accounting for spatiotemporal rainfall variability is crucial to risk assessment. Earth and environmental sciences/Climate sciences Earth and environmental sciences/Environmental sciences Earth and environmental sciences/Hydrology Earth and environmental sciences/Natural hazards Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Flooding is the costliest disaster in the United States, and its impacts are projected to increase under climate and land use change 1 , 2 . Public and private sector entities attempt to mitigate these impacts through insurance requirements, land use and development restrictions, and resilience plans, investing billions of dollars annually. To make these critical decisions, these entities rely on flood hazard maps and risk information, tying communities’ flood resilience directly to the accuracy of these data. Despite efforts to mitigate the impacts of flooding, U.S. flood losses are projected to reach $ 40.6 billion USD annually by 2050 3 , suggesting that current flood management practices may not be effectively addressing evolving flood risk. Together, these trends underscore the need to continue improving the fidelity of flood hazard and risk information, as their accuracy fundamentally shapes the long-term safety and resilience of communities. One of the most widely used sources of flood hazard estimates in the U.S. is the Federal Emergency Management Agency’s (FEMA) collection of Special Flood Hazard Area (SFHA) maps, ​​which delineate the area with a 1% – and occasionally 0.2% – or greater annual chance of inundation 4 . Other products covering a broader range of return periods and hazards (e.g., pluvial, compound) have been produced by private sector and academic ​​entities ​​​​ 5–7 . Although methods differ across providers, most flood hazard products rely on deterministic approaches, and all are produced using computer models, which simplify complex natural systems to make simulation feasible. For example, some models simulate river flooding through regional flood frequency analysis, which uses discharge statistics at select stream gauges to estimate flood hazard at the watershed scale 8 . Other models simulate rainfall-runoff processes using design storms, wherein a synthetic storm is constructed for a selected duration and probability of occurrence, with the storm’s total rainfall derived from intensity-duration-frequency (IDF) curves, such as those developed by the National Oceanic and Atmospheric Administration’s (NOAA) Atlas 14 9,10 . The design storm’s rainfall is simulated as uniform in space and idealized in time, neglecting the complex and variable behavior of real storms. Design storm approaches also simplify probabilistic relationships between rainfall, flood hazard, and flood risk. Recently, the reliability of these simplified, design storm-based approaches has been questioned. Empirical studies show that a large share of flood damage occurs outside mapped floodplains 11 – 15 , indicating that existing hazard products may underestimate risk. Researchers have also challenged the key assumptions underlying these methods, particularly their simplifications of rainfall and probability. In the design storm approach, the return period associated with the design storm’s rainfall is assumed to equal that of the resulting inundation and damage. However, contrary to the assumption that risk scales linearly with storm size, events with return periods less than 20 years can generate disproportionately large amounts of damage 16 . Similarly, the relationship between extreme precipitation and the resulting flooding has been shown to be more complex and nonlinear than is assumed by design storm approaches 17 , 18 . Naturally occurring storms are heterogeneous across space and time, and watershed responses vary with soil moisture, landcover, and other dynamic properties. Consequently, storms that appear similar at the basin scale can trigger substantially different floods locally. These findings highlight the limitations of deterministic hazard products and suggest that accurate flood risk assessment requires methods capable of capturing spatiotemporal rainfall variability and probabilistic watershed responses. Stochastic storm transposition (SST) has emerged as a promising alternative to conventional design storm approaches that may help to overcome these limitations. SST is a probabilistic approach that preserves storm structure and captures the spatiotemporal variability in rainfall 19 – 21 . Previous studies have shown that coupling stochastic storm transposition (SST) with hydrologic models produces robust estimates of streamflow frequency 22 – 31 , often yielding larger peak discharge estimates than design storms (e.g., up to 22% greater 30 ) with meaningful implications for infrastructure design and risk assessment 32 . Perez et al. 33 extended SST to flood hazard and ​​​​exposure mapping for the first time, demonstrating that the relationships between the frequencies of rainfall, discharge, and inundation are complex and highly variable, and suggesting that design storm-based maps ​​​may not adequately reflect true hazard. By explicitly representing probabilistic variability and rainfall heterogeneity, SST directly addresses the limitations of conventional deterministic design storm approaches. Despite these advances, critical gaps remain. Current evidence suggests that existing hazard and risk products may underestimate the frequency with which damaging flood events occur 25 , with potential consequences for community preparedness and long-term resilience. Yet to our knowledge, no prior research has quantified the magnitude of these underestimates. Here, we address this crucial question by analyzing how the core assumptions of design storm approaches – particularly the presumed one-to-one-to-one relationship between storm frequency, flood frequency, and damage frequency – affects flood hazard maps or risk estimates. We hypothesize that design storm assumptions lead to systematic underestimates of both flood hazard and risk by oversimplifying rainfall-watershed interactions and underestimating the impacts of smaller, more frequent storms. Although prior studies have examined storm-discharge-hazard relationships, none have used SST to estimate flood risk (i.e., damages), nor has any study compared design storm-based hazard and risk estimates against those produced using SST. This study addresses these gaps by providing the first direct comparison of flood hazard maps and flood risk estimates derived from idealized design storms against those generated using SST. We pass radar-rainfall data through SST software to generate an ensemble of 1,842 synthetic storms with basin-averaged return periods between 1 and 500 years, then run these storms through a hydrodynamic model to produce hazard maps and risk estimates. We then compare these results against outputs from 27 idealized, spatially uniform design storms constructed from the National Oceanic and Atmospheric Administration’s (NOAA) Atlas-14 IDF curves and temporally distributed according to the Soil Conservation Service (SCS) curves (see Methods). The analysis is conducted in a 2,800 km 2 watershed surrounding the ​​city of New Bern, North Carolina, USA (pop. 32,980) (Fig. S1 ). We find that relative to SST, design storms underestimate inundation frequency by 31–35%, expected annual damage by $ 7.9 million (40%), and ​50-year damage by $ 250 million (42%). We find that by leveraging probabilistic simulations and high-resolution rainfall fields, SST better captures the complex, multivariate relationships between rainfall and watershed processes and produces a broader range of realistic flooding scenarios, offering communities more informative assessments of flood risk and strengthening the foundation for preparedness and risk-informed decision making.​​ Results Flood Hazard Comparison We directly compare flood hazard maps derived from SST with those produced using a conventional design storm approach. We compare differences in the estimated flood frequency between the two approaches by calculating the difference in return period (in years) of flooding (depths > = 0.1m) in each 5m pixel across the model domain (see Methods). Across almost three quarters (71.5%) of the model area susceptible to flooding (defined as the area that floods at least once in any simulation), SST yields flood hazard estimates that equal or exceed those produced with design storms (i.e., blue shaded regions in Fig. 1 ), including in highly populated and densely developed areas like downtown New Bern. These differences are often substantial – for instance, the difference in estimated return period exceeds 100 years in 10% of the model area susceptible to flooding and more than 20 years in 30% of the model area susceptible to flooding, indicating markedly more frequent flooding in simulations using SST (Fig. 1 ). Areas where the design storm approach produces higher hazard estimates are generally small and spatially isolated (i.e., red areas in Fig. 1 ), and are largely disconnected from riparian floodplains, suggesting dominance of localized pluvial flooding processes in these locations. To assess whether differences in flood hazard estimates are driven by storm statistics or storm structure, we compare rainfall frequency curves derived from the SST ensemble with those used to construct the design storms (Fig. S2). Treating the Atlas 14 rainfall frequency curve as the baseline values and the SST rainfall frequency curve as the modeled values, we find a mean absolute error of 18.95 mm and a bias of -12.55 mm, meaning that SST rainfall volumes are similar to, but on average smaller than, those of the design storms for the same return period. The close agreement between SST and design storm rainfall frequency curves – and the smaller overall rainfall volumes of the SST storms - indicate that the higher flood hazard estimated by SST arises primarily from the spatial and temporal structure of rainfall and its interaction with watershed properties, rather than from differences in basin-averaged rainfall volumes. On average, flood return periods derived from the SST ensemble mean, minimum, and maximum hazard maps are 19, 20, and 36 years lower, respectively, than those estimated using the corresponding design storm maps (Figs. S3-S5). For areas mapped as having a 25-, 50-, and 100-year return period under the design storm approach, SST instead predicts substantially more frequent flooding, with corresponding return periods of seven, 20, and 69 years (Table S1 ). These differences are driven primarily by a marked increase in the fraction of the watershed experiencing flooding more frequently than once every five years on average, as well as a smaller increase in the area flooded once every 10 to 25 years (Fig. S6). To further isolate the role of storm structure, we compare flood extent across all 1,842 synthetic storms and 27 design storms, defining flood extent as the fraction of the model domain inundated to depths exceeding 0.1 meters. For storms with basin-averaged rainfall return periods between one and five years, SST and design storms produce comparable ranges of flood extent (Fig. S7). For storms with 10-year rainfall return periods and greater, however, design storms tend to generate larger total flood extents than SST storms of equivalent basin-averaged rainfall, with this divergence increasing at longer return periods. These results demonstrate that applying SST does not lead to an increase in overall flood extent but instead increases the frequency with which many areas across the watershed are inundated. Flood Risk Comparison SST enables direct quantification of nonlinear relationships between the storm frequency and flood damage, resulting in substantially greater uncertainty in damage estimates relative to the design storm approach. This increased uncertainty is evident across storms of all magnitudes, but is most pronounced for smaller, more frequent events (e.g., 1–25 years; Fig. 2 ). For storms with basin-averaged rainfall return periods of 25 years or less, the upper-bound damage estimates produced by SST are often nearly double those produced by design storms and, in some cases, comparable to damage estimates associated with the mean 100-year design storm (Fig. S8). The most damaging event in the SST ensemble – associated with a return period of 500 years – is produced by a storm whose basin-averaged rainfall has a return period of only 200 years. Despite its higher annual probability of occurrence (0.05% vs 0.02%), this storm generates damages comparable to that of the mean 500-year design storm, possibly because its most intense rainfall is concentrated over urban areas with more impervious surfaces, and because the majority of rainfall occurs in the second half of the storm time-series, after soils have begun to saturate. These results demonstrate that basin-averaged rainfall frequency alone is a poor predictor of flood damage magnitude. Differences between SST and design storm risk estimates compound over longer time horizons (Fig. 3 ). SST produces a mean annual damage of $ 19.8 million, exceeding the design storm estimate by $ 7.9 million. More notably, SST yields a maximum annual damage estimate of $ 115 million – more than eight times the upper bound estimate derived from the design storms. Over a 50-year time horizon, SST produces a mean cumulative damage estimate of $ 842 million, approximately $ 248 million (40%) greater than the design storm estimate and equivalent to nearly 10% of the total building value within the model domain. The upper bound of the 90% confidence interval for the SST-based 50-year damage is nearly $ 1.02 billion, roughly 1.5 times the corresponding design storm estimate. Storm-Flood-Damage Frequency Analysis Analysis of extreme events within the SST ensemble reveals a systematic decoupling between the frequencies of rainfall, flooding, and damage (Fig. S9). Of the ten largest flood extents simulated with SST, all with return periods exceeding 285 years, seven are produced by storms whose basin-averaged rainfall return periods are less than 200 years, including one event with a rainfall return period of just 70 years (Fig. S10). The single largest flood extent – associated with a 500-year return period – is generated by a storm with a basin-averaged rainfall return period of 285 years. Rainfall for this storm is relatively evenly distributed in space, but with the heaviest bands of rainfall slightly to the east, near more densely developed regions within the model domain, and with more rainfall occurring later in the storm. Flood damage frequency is similarly decoupled from the frequency of rainfall. Of the ten most damaging events, all but the largest are produced by storms with basin-averaged total rainfalls that have return periods less than 200 years, and six by storms with return periods less than 150 years. Several high-damage events are generated by relatively small and frequent storms that produce modest total flood extents. For example, one storm with a 40-year return period generates flooding with a 20-year extent but results in damage with an estimated return period of 154 years (Table S2). We find that storms that generate disproportionately large impacts tend to concentrate intense rainfall over urban areas or late in the 24-hour storm duration, suggesting that the observed decoupling between the frequency of rainfall, flooding, and damage in the simulations with SST is likely driven by the spatiotemporal variability of rainfall. SST allows both for local rainfall extremes – which are averaged out in design storm rainfall – and for scenarios in which these local extremes fall over less pervious and/or more saturated landcover, leading to more intense flooding locally. Different storm structures may then trigger these local pockets of flooding in different locations and at different times, altering the way sub-catchment outflows combine at confluences and leading to changes in flood extent and depth further downstream. These dynamics lead to greater variability in flooding depth and extent at the local scale, particularly for smaller storms, whose basin-averaged rainfall may be far lower than the local maximum rainfall intensity. Since the spatial distribution of property values is non-uniform, the increased uncertainty in local flood frequency leads to even greater variability in the total resulting damage. This phenomenon is demonstrated on Fig. 4 , which provides five examples of storms that all generate similar damage, but have vastly different spatiotemporal structures, flood extent, and damage profiles. These results demonstrate that storms with similar basin-averaged rainfall or flood extent can nevertheless produce damage of vastly different magnitudes, depending on their spatiotemporal rainfall structure and interaction with the built environment. Discussion Our results suggest that the relationships between the frequency of storms and the resulting watershed responses are not one-to-one, as assumed by design storm approaches, but rather complex and nonlinear, consistent with findings from past studies that have assessed the frequency relationships between rainfall and discharge 23 , 27 , 28 , 30 , 34 . A key mechanism underlying this disconnect is the localized intensity of storms. High-intensity rainfall over impervious surfaces or saturated soils generates disproportionately high runoff, while identical rainfall distributed over permeable or forested areas may produce minimal flooding. Variations in storm heading, forward velocity, and spatial rainfall pattern may further alter the timing of flood peaks downstream, leading to a wide range of possible peak water levels across the watershed. These dynamics are not captured by design storm approaches, which simplify rainfall spatially and temporally, leading to systematic underestimation of localized pluvial flooding and nonlinear flood responses. Our results also indicate that the relationship between rainfall, hazard and risk is fundamentally nonlinear. Storms that produce disproportionately high damage tend to concentrate intense rainfall over urban areas or in the latter half of the 24-hour period, consistent with previous studies of flood hazard 27 , 30 , 33 , 35 . Even when basin-averaged rainfall is identical, differences in the spatial and temporal distribution of rainfall alter runoff generation, infiltration, and flood routing, producing substantial local variation in inundation depth and extent. Consequently, smaller and more frequent (e.g., < 25-year return period) storms can generate outsized flooding and damage, particularly in densely developed areas. Thus, the design storm approach does not yield conservative hazard or risk estimates, contrary to recent assertions 36 , 37 , suggesting that in order to accurately capture flood hazard and risk, modelers must represent local-scale extrema of flood drivers and other variables. These local variabilities explain why the frequency associated with basin-averaged rainfall does not translate linearly to frequency of flooding or damage, violating a core assumption of design storm-based flood hazard assessment. These findings have important implications for widely used flood hazard and risk products – including FEMA’s National Flood Hazard Layer (NFHL) and First Street’s Flood Factor® (V4.0) – which rely heavily on design storms and regional flood frequency analysis (RFFA). Our results suggest that by neglecting the effects of spatiotemporal variability and nonlinear storm-flood-damage relationships, these approaches omit critical risks. While RFFA partially addresses limitations of design storms by using historical discharge data from stream gauges, it remains constrained by short record lengths, sparse, point-scale observations, assumptions of regional homogeneity, and limited representation of non-stationarity in climate or land use change. In contrast, SST leverages high-resolution radar-rainfall data, accounts for current land use patterns, and simulates nonlinear rainfall-flood-damage interactions across diverse landscapes, capturing variability overlooked by both design storm approaches and RFFA and potentially leading to more accurate assessments of hazard and risk. Our findings underscore the value of probabilistic, high-resolution hazard modeling for informing planning, emergency preparedness, and resilience investments. In particular, our results support FEMA’s Future of Flood Risk Data (FFRD) initiative 38 , which proposes to use SST-based frameworks to update flood hazard maps and risk estimates across the United States. Probabilistic methods such as SST produce spatially continuous hazard estimates and allow uncertainty to be characterized at the building scale 39 – 41 , enabling more risk-informed decision making. These high-resolution analyses may support more effective allocation of resources for emergency response and recovery and enable targeted programs for mandating insurance purchases, development guidance, building elevation standards, infrastructure design, and mitigation funding, ultimately contributing to improved community preparedness. Although our results demonstrate that SST improves hazard and risk estimation, our analysis is limited to storms sampled from a relatively short historical record (i.e., 2002–2024) and does not fully incorporate probabilistic antecedent conditions (e.g., soil moisture), upstream inflows, or coastal boundary conditions that could further amplify differences relative to design storm approaches, both in terms of expected values and uncertainty. Moreover, between-storm variability in watershed conditions may further influence flood response, but capturing interarrival times and their effects on antecedent conditions remains challenging and continuous simulations over long time scales are computationally expensive. Future studies could expand probabilistic inputs to include variability in tides, storm surge, and river boundary conditions (and their compounding effects), or could consider storms of different durations (e.g., 12- or 72-hour events) that might drive different responses at the watershed scale (i.e., pluvial vs fluvial). Incorporating these factors would likely amplify the divergence between SST and design storm estimates and further accentuate nonlinear storm-flood-damage relationships. Our research demonstrates that flood hazard and risk estimates that neglect spatiotemporal rainfall variability underestimate both hazard and damage by failing to capture nonlinear relationships between rainfall, inundation, and flood impacts driven by local-scale watershed response. High-resolution, probabilistic modeling with realistic rainfall representation provides a more comprehensive basis for flood hazard and risk assessment than deterministic design storm approaches. These improved estimates enable better-informed decisions related to emergency management, infrastructure planning, and community resilience. Given the substantial and growing economic and social consequences of flooding, transitioning from design storm-based approaches to probabilistic frameworks represents a critical step toward more accurate and actionable flood risk assessment. Methods Synthetic Storm Database SST was performed in the open-source RainyDay software package 21 using the National Oceanic and Atmospheric Administration’s Stage IV data 42 from January 1, 2002 through December 31, 2023. These data were prepared for RainyDay by resampling them onto a rectilinear grid via a mass-conserving regridding scheme in the xesmf Python package 43 . SST requires that users delineate a meteorologically homogeneous region known as the transposition domain (Appendix A), which we developed quantitatively via the Spatial L-Moments of Annual Maxima (SLAM) approach 44 . SLAM domains are specific to the selected watershed (Fig. A1) and storm duration of interest and can be adjusted based on a user-specified global significance level (GSL). For our watershed, SST was not particularly sensitive to changes in the GSL (i.e., extreme rainfall characteristics are homogeneous within the region), and so we selected the largest transposition domain (where p = 0.25; Fig. A2) in order to maximize the number of potential parent storms. We then used RainyDay to identify 220 parent storms within our transposition domain from the regridded Stage IV data. The parent storm duration was chosen as 24 hours, a duration commonly selected for modeling with design storms and relevant to the times of concentration experienced in our watershed. Rainfall estimates from thirteen of these parent storms contained radar artifacts such as bright-band contamination and blockage and were subsequently removed from consideration. We ultimately generated 10 ensemble members each comprised of 500 synthetic years of rainfall, with up to 18 storm arrivals per year. These realizations were then postprocessed to apply an alternative transposition scheme to parent storms identified as tropical cyclones, in order to enforce physically realistic transpositions. A description of the postprocessing method and validation results are included in Appendix B. Design storms were developed for each of the 1-, 2-, 5-, 10-, 25-, 50-, 100-, 200-, and 500-year return period, 24-hour storms. Total rainfall was derived from NOAA Atlas 14 9 . Rainfall is distributed in time via the Soil Conservation Service (SCS) Type III curve 45 and is uniform in space. For each return period, we generated a storm for the mean rainfall intensity, as well as for the upper and lower bounds of the 90% confidence interval of the IDF curve for Coastal Carolina Regional Airport (CCRA). This resulted in a total of 27 design storms used for comparison. No areal reduction factors were applied in constructing the design storms, and thus the storm intensities and resulting flooding are likely to be conservative estimates. Hydrodynamic Modeling We built and validated a high-resolution, 2D hydrodynamic model in order to simulate the watershed response to the SST and design storms. The model was developed in the Super-Fast Inundation of Coasts (SFINCS) model code 46 , 47 , a 2D, reduced-physics hydrodynamic modeling platform that can simulate fluvial, pluvial, storm surge, and compound flooding. SFINCS solves continuity and momentum equations based on the local inertial equations in LISFLOOD-FP 48 , plus optional wind drag and advection terms in the momentum equation. SFINCS can simulate processes including spatially varying friction, infiltration (here, via the SCS curve-number method, with recovery), rainfall-runoff, and wind shear. Inputs to the SFINCS model include gridded and spatially uniform time-series precipitation datasets, gridded wind datasets, and time-series or constant water level boundary conditions. The model developed in this study was adapted from an existing SFINCS model of the Carolinas 49 and was constructed using a high-resolution digital elevation model (including channel bathymetry), gridded Manning’s roughness coefficients ( n ) that vary spatially based on land use/land cover, and rasters of spatially varying maximum soil moisture and saturated hydraulic conductivity. Further discussion of SFINCS model development is provided in Appendix C. The model uses a 100-meter grid resolution and a 5-meter subgrid which allows the model to run at a lower native resolution while still accounting for finer-scale variations in topography and surface roughness, thereby maximizing computational efficiency 47 . The model was validated against five storm events of varying intensity and duration, including Hurricanes Matthew and Florence. Simulated hydrographs generally matched observations for all five validation events, resulting in an average Kling-Gupta efficiency (KGE) of 0.72 and an average peak elevation bias and root-mean-square error (RMSE) of 0.38 and 1.14 meters, respectively. For the two tropical cyclone validation events (Hurricanes Florence and Matthew), the model closely recreated observed high watermarks (HWMs) in the densely developed areas around New Bern, with an overall RMSE of 0.43 meters and a bias of 0.02 meters. Overall, our model performs similarly to other hydrodynamic models used for hazard and risk modeling and demonstrates high skill for storms of varying magnitude and frequency. Further description of the model validation process and model performance is provided in Appendix C. Probabilistic Flood Hazard Comparison We simulated the watershed response to the 27 design storms using the validated hydrodynamic model, along with a subset of 1,842 of the SST storm realizations that had mean basin-wide precipitation greater than or equal to the 1-year, 24-hour design storm rainfall for New Bern. Each SST storm realization was assigned water level boundary conditions based on the historical median water elevations for the month in which that realization’s parent storm occurred. Since design storms are statistically derived from the entire period of record and thus cannot be associated with observed storms, they could not be associated with water levels from a particular month, and so they were instead assigned water level boundary conditions based on the overall historical median. All runs neglected wind and tidal influences and assumed an initial soil moisture equal to half of the maximum storage capacity. Three hazard maps were generated for each of the design storm return periods, one for each of the mean, confidence interval upper bound, and confidence interval lower bound rainfall estimates. To create the hazard maps, the maximum flood elevation from each run (which are output on the model’s 100-meter grid) was first downscaled onto the model’s five-meter subgrid via bilinear interpolation, then converted to a flood depth raster by subtracting out the subgrid elevation profile. Areas with flood depths greater than or equal to 0.1 meters were assigned a flooding return period equal to that of the storm that generated that flood, in keeping with the one-to-one return period equivalency assumed by design storm approaches. The resulting hazard rasters were then combined together to create a single hazard map. Ten hazard maps were generated for SST, with one map per SST ensemble member. For each ensemble member, an SST hazard map was created by first stacking the flood elevations from all of that member’s model runs into a three-dimensional array, then sorting them from smallest to largest for each model grid cell. The sorted flood elevations were then assigned return periods between 1 and 500 years. We iterated through the sorted flood elevations from smallest to largest return period, downscaling each onto the model subgrid and converting them to flood depths. Once flood depths were greater than or equal to 0.1 m in a given pixel, the return period for flooding in that pixel was set equal to the return period of the current iteration. Ensemble statistics could then be calculated from these maps to understand the minimum, mean, and maximum exceedance probability for each pixel, as well as the standard deviation. Flood Risk Comparison Flood elevations were calculated at each building in the model domain by extracting the maximum flood elevation for each storm event at polygon centroids from the North Carolina Statewide Building Footprint dataset (2020–2022) available from North Carolina Emergency Management 50 . The flood depth at each building was calculated as the difference between the maximum simulated water surface elevation and the building’s first floor elevation (inside the FEMA SFHA) or ground surface elevation from the model subgrid at the building centroid (outside of the SFHA). Each building was then reclassified into one of the specific occupancy codes used by FEMA’s Hazus 51 program’s Flood Assessment Structure Tool lookup tables 52 based on number of stories, presence or absence of basement, and occupancy type. These specific occupancy codes were then used to assign a corresponding Hazus depth-damage function – which relates the flood depth at a structure to the expected damage – to each building. Finally, the resulting damage was calculated by entering the simulated flood depths at each building into that building’s depth-damage functions. Damage at each building could then be summed to calculate total damage for that storm. Storm-specific damages were then aggregated into 1- and 50-year flood damage estimates, the latter of which was chosen to estimate damage over the typical lifespan of building stock 53 . For the design storm approach, long-term damage was simulated deterministically. Similar to the approach described in Lendering et al. 54 , annual damage was calculated by taking the product of each storm’s annual exceedance probability (AEP; the inverse of its return period) and its damage and then summing across all design storms. The 50-year damage was then assumed to be 50 times the annual damage. This process was repeated for the mean, lower-bound, and upper-bound design storms, producing a range of expected annual damages. Damage for SST was calculated probabilistically. First, the distribution of annual damages was estimated by randomly sampling 10,000 years of damage. Each year was assumed to have k storm arrivals, where k was drawn from a Poisson distribution with λ = 1.87 storms/year, ensuring that the average number of annual storm arrivals matched that of the design storms, whose AEPs sum to 1.87. For each year, k storms were sampled from our database of SST storms, and their corresponding flood damages summed. To calculate 50-year damage, 50 annual damage estimates were randomly sampled (with replacement) from the 10,000 annual damage estimates, and their values summed. Storm-Flood-Damage Frequency Analysis To assess whether the relationships between the return periods of storms, floods, and damages are one-to-one (as assumed by design storm approaches), we calculated the frequency of the mean total rainfall, flood extent (represented as the fraction of our watershed with an inundation depth of at least 0.1 meters), and total damage for each of our 1,842 SST storms. For each of these outputs, frequency was calculated using the same methods as RainyDay. For flood extent and damage, it was also assumed that the minimum and maximum return periods of these variables were identical to those of the storms. Similar to Wright et al. (2014) 27 and Yu et al. 28 , we then used Spearman rank correlation ( ρ s ) as a quantitative measure of the correlation between the return periods of rainfall, flood extents, and flood damage. Finally, we qualitatively analyzed the spatiotemporal structures of storms that caused extreme damage events in order to understand which storm characteristics influence nonlinearities in storm-flood-damage recurrence intervals. Declarations Competing Interests The authors declare no competing interests. Author Contribution J.A.B – conceptualization, methodology, software, validation, formal analysis, writing-original draft, visualization; A.S. – conceptualization, writing-original draft, supervision, funding acquisition; L.G. – conceptualization, methodology, writing-review & editing; J.D.G. – methodology, writing-review & editing; D.W. – methodology, software, writing-review & editing; M.A.H. – software, writing-review & editing; M.W. – writing-review & editing. Acknowledgement The authors would like to acknowledge B. FitzGerald for providing the transposition domain, R. Luettich for providing wind field data for the validation events, and G. Karlovits and D. Rosa for their insights and discussion during the project. Data Availability Stage IV precipitation data can be downloaded from https://data.eol.ucar.edu/dataset/21.093 . OWI wind and pressure data used to create the inputs for Hurricanes Florence and Matthew are proprietary data from OceanWeather Inc. All other data used to build and validate the models in this study are publicly available through USGS and NOAA. Replication data for the figures included in this manuscript are available via NHERI’s DesignSafe repository https://doi.org/10.17603/ds2-27r7-jv83 55 . Code Availability The SFINCS model documentation and source code can be downloaded from https://sfincs.readthedocs.io/en/latest/index.html . The RainyDay documentation and source code is available from https://github.com/HydroclimateExtremesGroup/RainyDay . Model files and code used to analyze these data are available upon request to the authors. References National Academies of Sciences, Engineering, and Medicine. Framing the Challenge of Urban Flooding in the United States . https://www.nap.edu/catalog/25381 (2019) doi: 10.17226/25381 . United States Government Accountability Office. FEMA Flood Maps: Better Planning and Analysis Needed to Address Current and Future Flood Hazards . https://www.gao.gov/assets/gao-22-104079.pdf (2021). Wing, O. E. J. et al. Inequitable patterns of US flood risk in the Anthropocene. Nat. Clim. Chang. 12, 156–162 (2022). Federal Emergency Management Agency. Glossary: Flood Zones. https://www.fema.gov/about/glossary/flood-zones (2020). First Street. First Street Flood Model Version 4.0 Technical Methodology . (2025). Sanders, B. F. et al. Large and inequitable flood risks in Los Angeles, California. Nat Sustain 6, 47–57 (2022). Wing, O. E. J. et al. A 30 m Global Flood Inundation Model for Any Climate Scenario. Water Resources Research 60, e2023WR036460 (2024). Smith, A., Sampson, C. & Bates, P. Regional flood frequency analysis at the global scale. Water Resour. Res. 51, 539–553 (2015). Bonnin, G. M., Martin, D., Lin, B., Yekta, M. & Riley, D. NOAA Atlas 14 Precipitation-Frequency Atlas of the United States Volume 2 Version 3.0: Delaware, District of Columbia, Illinois, Indiana, Kentucky, Maryland, New Jersey, North Carolina, Ohio, Pennsylvania, South Carolina, Tennessee, Virginia, West Virginia . (2006). Lanciotti, S., Ridolfi, E., Russo, F. & Napolitano, F. Intensity–Duration–Frequency Curves in a Data-Rich Era: A Review. Water 14, 3705 (2022). Blessing, R., Sebastian, A. & Brody, S. D. Flood Risk Delineation in the United States: How Much Loss Are We Capturing? Natural Hazards Review 18, 04017002 (2017). Brody, S., Blessing, R., Sebastian, A. & Bedient, P. Delineating the Reality of Flood Risk and Loss in Southeast Texas. Natural Hazards Review 14, 89–97 (2013). Brody, S. d., Sebastian, A., Blessing, R. & Bedient, P. b. Case study results from southeast Houston, Texas: identifying the impacts of residential location on flood risk and loss. Journal of Flood Risk Management 11, S110–S120 (2018). Garcia, H. M. et al. Reconstructing Repetitive Flood Exposure Across 78 Events From 1996 to 2020 in North Carolina, USA. Earth’s Future 13, e2025EF006026 (2025). Highfield, W. E., Norman, S. A. & Brody, S. D. Examining the 100-Year Floodplain as a Metric of Risk, Loss, and Household Adjustment. Risk Analysis 33, 186–191 (2013). Nayak, A., Gentine, P. & Lall, U. Financial losses associated with US floods occur with frequent low-return-period precipitation. Nat Water 3, 1256–1267 (2025). Do, H. X., Mei, Y. & Gronewold, A. D. To What Extent Are Changes in Flood Magnitude Related to Changes in Precipitation Extremes? Geophysical Research Letters 47, e2020GL088684 (2020). Tuyls, D. M., Thorndahl, S. & Rasmussen, M. R. Return period assessment of urban pluvial floods through modelling of rainfall–flood response. Journal of Hydroinformatics 20, 829–845 (2018). Foufoula-Georgiou, E. A probabilistic storm transposition approach for estimating exceedance probabilities of extreme precipitation depths. Water Resources Research 25, 799–815 (1989). Wilson, L. L. & Foufoula-Georgiou, E. Regional Rainfall Frequency Analysis via Stochastic Storm Transposition. J. Hydraul. Eng. 116, 859–880 (1990). Wright, D. B., Mantilla, R. & Peters-Lidard, C. D. A Remote Sensing-Based Tool for Assessing Rainfall-Driven Hazards. Environ Model Softw 90, 34–54 (2017). England, J. F., Julien, P. Y. & Velleux, M. L. Physically-based extreme flood frequency with stochastic storm transposition and paleoflood data on large watersheds. Journal of Hydrology 510, 228–245 (2014). Franchini, M., Helmlinger, K. R., Foufoula-Georgiou, E. & Todini, E. Stochastic storm transposition coupled with rainfall—runoff modeling for estimation of exceedance probabilities of design floods. Journal of Hydrology 175, 511–532 (1996). Lawler, S. et al. Application of stochastic storm transposition for hydrologic modeling in the mountainous western US. Stoch Environ Res Risk Assess 39, 109–127 (2025a). Lompi, M., Caporali, E., Mediero, L. & Mazzanti, B. Improving flash flood risk assessment using a simple approach for extreme rainfall scaling and storms transposition. J Flood Risk Management 15, e12796 (2022). Perez, G., Gomez-Velez, J. D., Mantilla, R., Wright, D. B. & Li, Z. The Effect of Storm Direction on Flood Frequency Analysis. Geophysical Research Letters 48, e2020GL091918 (2021). Wright, D. B., Smith, J. A. & Baeck, M. L. Flood frequency analysis using radar rainfall fields and stochastic storm transposition. Water Resources Research 50, 1592–1615 (2014). Yu, G., Wright, D. B., Zhu, Z., Smith, C. & Holman, K. D. Process-based flood frequency analysis in an agricultural watershed exhibiting nonstationary flood seasonality. Hydrol. Earth Syst. Sci. 23, 2225–2243 (2019). Yu, G., Wright, D. B. & Holman, K. D. Connecting Hydrometeorological Processes to Low-Probability Floods in the Mountainous Colorado Front Range. Water Resources Research 57, e2021WR029768 (2021). Zhou, Z. et al. The impact of the spatiotemporal structure of rainfall on flood frequency over a small urban watershed: an approach coupling stochastic storm transposition and hydrologic modeling. Hydrol. Earth Syst. Sci. 25, 4701–4717 (2021). Zhu, Z., Wright, D. B. & Yu, G. The Impact of Rainfall Space-Time Structure in Flood Frequency Analysis. Water Resources Research 54, 8983–8998 (2018). Wright, D. B., Yu, G. & England, J. F. Six decades of rainfall and flood frequency analysis using stochastic storm transposition: Review, progress, and prospects. Journal of Hydrology 585, 124816 (2020). Perez, G., Coon, E. & Rathore, S. Advancing process-based flood frequency analysis for assessing flood hazard and population flood exposure. Journal of Hydrology 28, (2024). Vangelis, H., Zotou, I., Kourtis, I. M., Bellos, V. & Tsihrintzis, V. A. Relationship of Rainfall and Flood Return Periods through Hydrologic and Hydraulic Modeling. Water 14, 3618 (2022). Asher, M. et al. Extreme rainfall and temporal loading in Great Britain: Analysis of present and future trends using a convection-permitting climate model. Journal of Hydrology: Regional Studies 62, 102750 (2025). Cache, T., Bevacqua, E., Zscheischler, J., Müller-Thomy, H. & Peleg, N. Simulating Realistic Design Storms: A Joint Return Period Approach. Water Resources Research 61, e2024WR039739 (2025). Cea, L. et al. Recent advances and future challenges in urban pluvial flood modelling. Urban Water Journal 22, 149–173 (2025). Lehman, W. FEMA’s Future of Flood Risk Data Initiative. (2023). Alfonso, L., Mukolwe, M. M. & Di Baldassarre, G. Probabilistic Flood Maps to support decision-making: Mapping the Value of Information. Water Resources Research 52, 1026–1043 (2016). Di Baldassarre, G., Schumann, G., Bates, P. D., Freer, J. E. & Beven, K. J. Flood-plain mapping: a critical discussion of deterministic and probabilistic approaches. Hydrological Sciences Journal 55, 364–376 (2010). Diehl, R. M., Gourevitch, J. D., Drago, S. & Wemple, B. C. Improving flood hazard datasets using a low-complexity, probabilistic floodplain mapping approach. PLoS ONE 16, e0248683 (2021). Du, J. NCEP/EMC 4KM Gridded Data (GRIB) Stage IV Data. 1.0. (2011). Zhuang, J., Dussin, R., Huard, D., & Bougault, P. xESMF. (2024). FitzGerald, B., Wright, D. B., Yan, L., Dietrich, A. & Sebastian, A. An L-Moments-Based Hypothesis Test to Identify Homogeneous Storm Transposition Regions. Journal of Hydrology (in review). United States Department of Agriculture. Urban Hydrology for Small Watersheds TR-55 . (1986). Leijnse, T., Van Ormondt, M., Nederhoff, K. & Van Dongeren, A. Modeling compound flooding in coastal systems using a computationally efficient reduced-physics solver: Including fluvial, pluvial, tidal, wind- and wave-driven processes. Coastal Engineering 163, 103796 (2021). van Ormondt, M., Leijnse, T., de Goede, R., Nederhoff, K. & van Dongeren, A. A subgrid method for the linear inertial equations of a compound flood model. EGUsphere 1–36 (2024) doi: 10.5194/egusphere-2024-1839 . Bates, P. D., Horritt, M. S. & Fewtrell, T. J. A simple inertial formulation of the shallow water equations for efficient two-dimensional flood inundation modelling. Journal of Hydrology 387, 33–45 (2010). Grimley, L. E. et al. Determining the Relative Contributions of Runoff, Coastal, and Compound Processes to Flood Exposure Across the Carolinas During Hurricane Florence. Water Resources Research 61, e2023WR036727 (2025). NC Emergency Management GIS Section. Countywide 2020–2022 Building Footprints. Federal Emergency Management Agency. Hazus 6.1 Flood Model Technical Manual . https://www.fema.gov/sites/default/files/documents/fema_hazus-flood-model-technical-manual-6-1.pdf (2024). Burns, J. N., Sharma, U., Lindeman, C. & Raines, J. Hazus Flood Assessment Structure Tool. Federal Emergency Management Agency (2021). Andersen, R. & Negendahl, K. Lifespan prediction of existing building typologies. Journal of Building Engineering 65, 105696 (2023). Lendering, K., Sebastian, A., Jonkman, S. N. & Kok, M. Framework for assessing the performance of flood adaptation innovations using a Risk-Based approach. Journal of Flood Risk Management 12, (2018). Jack Baer et al. Neglecting Spatiotemporal Rainfall Variability Underestimates Flood Hazard and Risk. DesignSafe-CI https://doi.org/10.17603/ds2-27r7-jv83 (2026). Additional Declarations No competing interests reported. Supplementary Files BaerNeglectingSpatiotemporalRainfallSI.pdf Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 06 Feb, 2026 Reviews received at journal 02 Feb, 2026 Reviews received at journal 02 Feb, 2026 Reviewers agreed at journal 23 Jan, 2026 Reviewers agreed at journal 22 Jan, 2026 Reviewers invited by journal 21 Jan, 2026 Editor assigned by journal 18 Jan, 2026 Submission checks completed at journal 18 Jan, 2026 First submitted to journal 13 Jan, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8593870","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":578943147,"identity":"a94dd9be-1648-4cfc-ab18-71792c397796","order_by":0,"name":"John A. Baer","email":"","orcid":"","institution":"University of North Carolina at Chapel Hill","correspondingAuthor":false,"prefix":"","firstName":"John","middleName":"A.","lastName":"Baer","suffix":""},{"id":578943148,"identity":"f53ae31a-9e56-40d5-9982-a0c2815aa088","order_by":1,"name":"Antonia Sebastian","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0ElEQVRIiWNgGAWjYHACNhDBw8DAfADCP0CcFgOgFrYE0rSALDIgTgt//+FjD378+SPDP7vnm3RBDYMc340E/FokbqSlG/a2GfBI3Dm7TXrGMQZjSUJaGG7wmEnwNgD9ciN3mzRvA0PiBkJa5M+f/yb5548Bj/yNnGcgLfUEtRgcyGGT5mEz4DG4AWQAtSQYENJieCPN3Fi2zZgHyDC25jkmYTjzzAP8WuTOH3728M0fOXu5G8kPb/PU2MjzHSdgCzqQIE35KBgFo2AUjALsAADcHkF4lw0E2wAAAABJRU5ErkJggg==","orcid":"","institution":"University of North Carolina at Chapel Hill","correspondingAuthor":true,"prefix":"","firstName":"Antonia","middleName":"","lastName":"Sebastian","suffix":""},{"id":578943149,"identity":"06db29a2-f93d-4956-ad22-e6f950ea8835","order_by":2,"name":"Lauren E. Grimley","email":"","orcid":"","institution":"University of North Carolina at Chapel Hill","correspondingAuthor":false,"prefix":"","firstName":"Lauren","middleName":"E.","lastName":"Grimley","suffix":""},{"id":578943151,"identity":"cb90e930-caf3-4475-a07b-d7135de19fb7","order_by":3,"name":"James Doss-Gollin","email":"","orcid":"","institution":"Rice University","correspondingAuthor":false,"prefix":"","firstName":"James","middleName":"","lastName":"Doss-Gollin","suffix":""},{"id":578943152,"identity":"7b3a6596-d5f5-4559-b71a-b51c057942fb","order_by":4,"name":"Daniel B. Wright","email":"","orcid":"","institution":"University of Wisconsin-Madison","correspondingAuthor":false,"prefix":"","firstName":"Daniel","middleName":"B.","lastName":"Wright","suffix":""},{"id":578943153,"identity":"4b2ff209-3eb5-4b71-99a9-ef0f80cd7090","order_by":5,"name":"Mohammad Ashar Hussain","email":"","orcid":"","institution":"University of Wisconsin-Madison","correspondingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"Ashar","lastName":"Hussain","suffix":""},{"id":578943154,"identity":"8669c041-2887-4da4-a921-c9794435742d","order_by":6,"name":"Marissa Webber","email":"","orcid":"","institution":"University of North Carolina at Chapel Hill","correspondingAuthor":false,"prefix":"","firstName":"Marissa","middleName":"","lastName":"Webber","suffix":""}],"badges":[],"createdAt":"2026-01-13 15:39:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8593870/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8593870/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101203067,"identity":"033fafd5-28d2-4813-be3a-8f19b92a5122","added_by":"auto","created_at":"2026-01-27 09:38:41","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":2544124,"visible":true,"origin":"","legend":"","description":"","filename":"BaerNeglectingSpatiotemporalRainfallMain.docx","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/e568c0a9bffb88f082fb63de.docx"},{"id":100973134,"identity":"8160971d-3c6c-4c73-8a70-77a8e1b7334f","added_by":"auto","created_at":"2026-01-23 10:33:34","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":8607,"visible":true,"origin":"","legend":"","description":"","filename":"8c124a83eb09478b9075909e23241899.json","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/aee5065a5fde81e98a927b31.json"},{"id":101202714,"identity":"8d82c218-9fcf-44d9-84cb-429ffc4fc7f6","added_by":"auto","created_at":"2026-01-27 09:37:19","extension":"pdf","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":4086830,"visible":true,"origin":"","legend":"","description":"","filename":"BaerNeglectingSpatiotemporalRainfallSI.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/3101a0823fd64d6ad836b2eb.pdf"},{"id":101203978,"identity":"9271aed8-b89d-4c89-a527-8aa4fcd2c851","added_by":"auto","created_at":"2026-01-27 09:41:12","extension":"xml","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":116393,"visible":true,"origin":"","legend":"","description":"","filename":"8c124a83eb09478b9075909e232418991enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/dd714ca4c151f500fac8f836.xml"},{"id":100973141,"identity":"e9f90327-5e5d-45cb-8b38-d06a5e8e856f","added_by":"auto","created_at":"2026-01-23 10:33:34","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":335750,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/5541d388065d8465b11a05b2.png"},{"id":100973136,"identity":"77148437-9619-4469-8fe0-9afe57d4b249","added_by":"auto","created_at":"2026-01-23 10:33:34","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":49561,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/8cb1d89f12347d0a16cf3be7.png"},{"id":101203187,"identity":"73da0960-b9fc-451c-98f8-4452663512cc","added_by":"auto","created_at":"2026-01-27 09:39:01","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":34344,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/447a3c9ede51f13423b43ca5.png"},{"id":101203003,"identity":"47a548a8-c46a-44c9-87df-aeb84aaecd28","added_by":"auto","created_at":"2026-01-27 09:38:29","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":140980,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/d1ac772f594dc6011f2a67e1.png"},{"id":100973142,"identity":"2a6c37cd-0537-4cfb-8d62-0b1f272c085a","added_by":"auto","created_at":"2026-01-23 10:33:34","extension":"xml","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":114750,"visible":true,"origin":"","legend":"","description":"","filename":"8c124a83eb09478b9075909e232418991structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/2c91492999749f7a44f6088b.xml"},{"id":100973139,"identity":"470ea7a9-512a-40fc-b4f3-18ad8357e167","added_by":"auto","created_at":"2026-01-23 10:33:34","extension":"html","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":126856,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/f1c150987a9fae55a5b1065f.html"},{"id":101296703,"identity":"bb248cb7-f863-4700-a481-321c10ac4505","added_by":"auto","created_at":"2026-01-28 09:19:09","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":1673059,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial distribution of differences in flood hazard between the SST ensemble mean and the mean design storm estimates at a 5-m resolution. The upper panel shows the full model domain; inset panels highlight downtown New Bern (lower right) and the Trent River floodplain near Trenton (lower left). Blue pixels indicate locations where SST predicts more frequent flooding (shorter return periods); red pixels indicate locations where design storms predict more frequent flooding, and gray pixels indicate similar hazard estimates. Transparent areas were not flooded under either approach. SST produces substantially higher flood hazard across most of the watershed, including in densely developed areas, while higher design storm hazard estimates are confined to small, isolated areas dominated by localized pluvial flooding.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/fc17cc759244918374602243.png"},{"id":100973131,"identity":"d802a3c0-507f-4a4f-830f-fc4c626c7551","added_by":"auto","created_at":"2026-01-23 10:33:34","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":184091,"visible":true,"origin":"","legend":"\u003cp\u003eScatterplot of total 24-hour basin-averaged rainfall versus flood damage for SST storms and design storms. Black error bars denote the 90% confidence interval of design storm damage estimates. SST produces substantially greater variability in flood damage for a given rainfall return period, particularly for smaller and more frequent storms, resulting in upper-bound risk estimates that substantially exceed those from design storms.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/4fb22f90c2bbb656de68f6f2.png"},{"id":100973133,"identity":"65f84197-cef7-49b6-bd17-a4c20e342cda","added_by":"auto","created_at":"2026-01-23 10:33:34","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":107347,"visible":true,"origin":"","legend":"\u003cp\u003eDistributions of annual flood damage (left) and cumulative 50-year damage (right) produced by each modeling approach, with zero-damage years excluded from annual estimates. SST yields higher mean damage estimates and substantially broader distributions at both time scales, reflecting greater uncertainty relative to deterministic design storm methods. At a 50-year time horizon, the mean of the SST ensemble exceeds the maximum of the design storm interval.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/f7f79a2f35b709c61aa929f6.png"},{"id":100973135,"identity":"6dff53dc-562e-4cca-acb3-c7814ec2c6ec","added_by":"auto","created_at":"2026-01-23 10:33:34","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":486240,"visible":true,"origin":"","legend":"\u003cp\u003eExamples of four SST storms that produce approximately 100-year damage events, compared with the 100-year design storm. Columns show (from left to right): 24-hour hyetographs associated with each storm, the spatial distribution of total rainfall and estimated rainfall return period, downscaled flood depths at maximum inundation and estimated flood extent return periods, and spatial distributions of building damage and estimated damage return period. Despite large differences in storm structure, rainfall frequency, and flood extent, these events produce comparable total damage, illustrating the nonlinear and non-unique relationship between the frequencies of rainfall, flooding, and damage.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/51dfc5b8b1838697eb042699.png"},{"id":101298950,"identity":"6d8129e9-f3c7-4085-b3d6-8dd55b9f04d3","added_by":"auto","created_at":"2026-01-28 09:38:15","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2863548,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/9a04c1ed-c602-4acd-be72-0e1be9d2e072.pdf"},{"id":100973143,"identity":"bb561982-e40b-45f7-b7b2-d08394c14c1f","added_by":"auto","created_at":"2026-01-23 10:33:34","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":4086830,"visible":true,"origin":"","legend":"","description":"","filename":"BaerNeglectingSpatiotemporalRainfallSI.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8593870/v1/f5e48603b49887087569d5d9.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Neglecting Spatiotemporal Rainfall Variability Underestimates Flood Hazard and Risk","fulltext":[{"header":"Introduction","content":"\u003cp\u003eFlooding is the costliest disaster in the United States, and its impacts are projected to increase under climate and land use change \u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e. Public and private sector entities attempt to mitigate these impacts through insurance requirements, land use and development restrictions, and resilience plans, investing billions of dollars annually. To make these critical decisions, these entities rely on flood hazard maps and risk information, tying communities\u0026rsquo; flood resilience directly to the accuracy of these data. Despite efforts to mitigate the impacts of flooding, U.S. flood losses are projected to reach \u003cspan\u003e$\u003c/span\u003e40.6\u0026nbsp;billion USD annually by 2050 \u003csup\u003e3\u003c/sup\u003e, suggesting that current flood management practices may not be effectively addressing evolving flood risk. Together, these trends underscore the need to continue improving the fidelity of flood hazard and risk information, as their accuracy fundamentally shapes the long-term safety and resilience of communities.\u003c/p\u003e \u003cp\u003eOne of the most widely used sources of flood hazard estimates in the U.S. is the Federal Emergency Management Agency\u0026rsquo;s (FEMA) collection of Special Flood Hazard Area (SFHA) maps, ​​which delineate the area with a 1% \u0026ndash; and occasionally 0.2% \u0026ndash; or greater annual chance of inundation \u003csup\u003e\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e. Other products covering a broader range of return periods and hazards (e.g., pluvial, compound) have been produced by private sector and academic ​​entities ​​​​\u003csup\u003e5\u0026ndash;7\u003c/sup\u003e. Although methods differ across providers, most flood hazard products rely on deterministic approaches, and all are produced using computer models, which simplify complex natural systems to make simulation feasible. For example, some models simulate river flooding through regional flood frequency analysis, which uses discharge statistics at select stream gauges to estimate flood hazard at the watershed scale \u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e. Other models simulate rainfall-runoff processes using design storms, wherein a synthetic storm is constructed for a selected duration and probability of occurrence, with the storm\u0026rsquo;s total rainfall derived from intensity-duration-frequency (IDF) curves, such as those developed by the National Oceanic and Atmospheric Administration\u0026rsquo;s (NOAA) Atlas 14 \u003csup\u003e9,10\u003c/sup\u003e. The design storm\u0026rsquo;s rainfall is simulated as uniform in space and idealized in time, neglecting the complex and variable behavior of real storms. Design storm approaches also simplify probabilistic relationships between rainfall, flood hazard, and flood risk.\u003c/p\u003e \u003cp\u003eRecently, the reliability of these simplified, design storm-based approaches has been questioned. Empirical studies show that a large share of flood damage occurs outside mapped floodplains \u003csup\u003e\u003cspan additionalcitationids=\"CR12 CR13 CR14\" citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e\u003c/sup\u003e, indicating that existing hazard products may underestimate risk. Researchers have also challenged the key assumptions underlying these methods, particularly their simplifications of rainfall and probability. In the design storm approach, the return period associated with the design storm\u0026rsquo;s rainfall is assumed to equal that of the resulting inundation and damage. However, contrary to the assumption that risk scales linearly with storm size, events with return periods less than 20 years can generate disproportionately large amounts of damage \u003csup\u003e\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e. Similarly, the relationship between extreme precipitation and the resulting flooding has been shown to be more complex and nonlinear than is assumed by design storm approaches \u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e. Naturally occurring storms are heterogeneous across space and time, and watershed responses vary with soil moisture, landcover, and other dynamic properties. Consequently, storms that appear similar at the basin scale can trigger substantially different floods locally. These findings highlight the limitations of deterministic hazard products and suggest that accurate flood risk assessment requires methods capable of capturing spatiotemporal rainfall variability and probabilistic watershed responses.\u003c/p\u003e \u003cp\u003eStochastic storm transposition (SST) has emerged as a promising alternative to conventional design storm approaches that may help to overcome these limitations. SST is a probabilistic approach that preserves storm structure and captures the spatiotemporal variability in rainfall \u003csup\u003e\u003cspan additionalcitationids=\"CR20\" citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e. Previous studies have shown that coupling stochastic storm transposition (SST) with hydrologic models produces robust estimates of streamflow frequency \u003csup\u003e\u003cspan additionalcitationids=\"CR23 CR24 CR25 CR26 CR27 CR28 CR29 CR30\" citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e, often yielding larger peak discharge estimates than design storms (e.g., up to 22% greater \u003csup\u003e\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e\u003c/sup\u003e) with meaningful implications for infrastructure design and risk assessment \u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e. Perez et al. \u003csup\u003e33\u003c/sup\u003e extended SST to flood hazard and ​​​​exposure mapping for the first time, demonstrating that the relationships between the frequencies of rainfall, discharge, and inundation are complex and highly variable, and suggesting that design storm-based maps ​​​may not adequately reflect true hazard. By explicitly representing probabilistic variability and rainfall heterogeneity, SST directly addresses the limitations of conventional deterministic design storm approaches.\u003c/p\u003e \u003cp\u003eDespite these advances, critical gaps remain. Current evidence suggests that existing hazard and risk products may underestimate the frequency with which damaging flood events occur \u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e\u003c/sup\u003e, with potential consequences for community preparedness and long-term resilience. Yet to our knowledge, no prior research has quantified the magnitude of these underestimates. Here, we address this crucial question by analyzing how the core assumptions of design storm approaches \u0026ndash; particularly the presumed one-to-one-to-one relationship between storm frequency, flood frequency, and damage frequency \u0026ndash; affects flood hazard maps or risk estimates. We hypothesize that design storm assumptions lead to systematic underestimates of both flood hazard and risk by oversimplifying rainfall-watershed interactions and underestimating the impacts of smaller, more frequent storms. Although prior studies have examined storm-discharge-hazard relationships, none have used SST to estimate flood risk (i.e., damages), nor has any study compared design storm-based hazard and risk estimates against those produced using SST.\u003c/p\u003e \u003cp\u003eThis study addresses these gaps by providing the first direct comparison of flood hazard maps and flood risk estimates derived from idealized design storms against those generated using SST. We pass radar-rainfall data through SST software to generate an ensemble of 1,842 synthetic storms with basin-averaged return periods between 1 and 500 years, then run these storms through a hydrodynamic model to produce hazard maps and risk estimates. We then compare these results against outputs from 27 idealized, spatially uniform design storms constructed from the National Oceanic and Atmospheric Administration\u0026rsquo;s (NOAA) Atlas-14 IDF curves and temporally distributed according to the Soil Conservation Service (SCS) curves (see Methods). The analysis is conducted in a 2,800 km\u003csup\u003e2\u003c/sup\u003e watershed surrounding the ​​city of New Bern, North Carolina, USA (pop. 32,980) (Fig. \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). We find that relative to SST, design storms underestimate inundation frequency by 31\u0026ndash;35%, expected annual damage by \u003cspan\u003e$\u003c/span\u003e7.9\u0026nbsp;million (40%), and ​50-year damage by \u003cspan\u003e$\u003c/span\u003e250\u0026nbsp;million (42%). We find that by leveraging probabilistic simulations and high-resolution rainfall fields, SST better captures the complex, multivariate relationships between rainfall and watershed processes and produces a broader range of realistic flooding scenarios, offering communities more informative assessments of flood risk and strengthening the foundation for preparedness and risk-informed decision making.​​\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eFlood Hazard Comparison\u003c/h2\u003e \u003cp\u003eWe directly compare flood hazard maps derived from SST with those produced using a conventional design storm approach. We compare differences in the estimated flood frequency between the two approaches by calculating the difference in return period (in years) of flooding (depths\u0026thinsp;\u0026gt;\u0026thinsp;=\u0026thinsp;0.1m) in each 5m pixel across the model domain (see Methods). Across almost three quarters (71.5%) of the model area susceptible to flooding (defined as the area that floods at least once in any simulation), SST yields flood hazard estimates that equal or exceed those produced with design storms (i.e., blue shaded regions in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), including in highly populated and densely developed areas like downtown New Bern. These differences are often substantial \u0026ndash; for instance, the difference in estimated return period exceeds 100 years in 10% of the model area susceptible to flooding and more than 20 years in 30% of the model area susceptible to flooding, indicating markedly more frequent flooding in simulations using SST (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Areas where the design storm approach produces higher hazard estimates are generally small and spatially isolated (i.e., red areas in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), and are largely disconnected from riparian floodplains, suggesting dominance of localized pluvial flooding processes in these locations.\u003c/p\u003e \u003cp\u003eTo assess whether differences in flood hazard estimates are driven by storm statistics or storm structure, we compare rainfall frequency curves derived from the SST ensemble with those used to construct the design storms (Fig. S2). Treating the Atlas 14 rainfall frequency curve as the baseline values and the SST rainfall frequency curve as the modeled values, we find a mean absolute error of 18.95 mm and a bias of -12.55 mm, meaning that SST rainfall volumes are similar to, but on average smaller than, those of the design storms for the same return period. The close agreement between SST and design storm rainfall frequency curves \u0026ndash; and the smaller overall rainfall volumes of the SST storms - indicate that the higher flood hazard estimated by SST arises primarily from the spatial and temporal structure of rainfall and its interaction with watershed properties, rather than from differences in basin-averaged rainfall volumes.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eOn average, flood return periods derived from the SST ensemble mean, minimum, and maximum hazard maps are 19, 20, and 36 years lower, respectively, than those estimated using the corresponding design storm maps (Figs. S3-S5). For areas mapped as having a 25-, 50-, and 100-year return period under the design storm approach, SST instead predicts substantially more frequent flooding, with corresponding return periods of seven, 20, and 69 years (Table \u003cspan refid=\"MOESM1\" class=\"InternalRef\"\u003eS1\u003c/span\u003e). These differences are driven primarily by a marked increase in the fraction of the watershed experiencing flooding more frequently than once every five years on average, as well as a smaller increase in the area flooded once every 10 to 25 years (Fig. S6).\u003c/p\u003e \u003cp\u003eTo further isolate the role of storm structure, we compare flood extent across all 1,842 synthetic storms and 27 design storms, defining flood extent as the fraction of the model domain inundated to depths exceeding 0.1 meters. For storms with basin-averaged rainfall return periods between one and five years, SST and design storms produce comparable ranges of flood extent (Fig. S7). For storms with 10-year rainfall return periods and greater, however, design storms tend to generate larger total flood extents than SST storms of equivalent basin-averaged rainfall, with this divergence increasing at longer return periods. These results demonstrate that applying SST does not lead to an increase in overall flood extent but instead increases the frequency with which many areas across the watershed are inundated.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eFlood Risk Comparison\u003c/h3\u003e\n\u003cp\u003eSST enables direct quantification of nonlinear relationships between the storm frequency and flood damage, resulting in substantially greater uncertainty in damage estimates relative to the design storm approach. This increased uncertainty is evident across storms of all magnitudes, but is most pronounced for smaller, more frequent events (e.g., 1\u0026ndash;25 years; Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). For storms with basin-averaged rainfall return periods of 25 years or less, the upper-bound damage estimates produced by SST are often nearly double those produced by design storms and, in some cases, comparable to damage estimates associated with the mean 100-year design storm (Fig. S8).\u003c/p\u003e \u003cp\u003eThe most damaging event in the SST ensemble \u0026ndash; associated with a return period of 500 years \u0026ndash; is produced by a storm whose basin-averaged rainfall has a return period of only 200 years. Despite its higher annual probability of occurrence (0.05% vs 0.02%), this storm generates damages comparable to that of the mean 500-year design storm, possibly because its most intense rainfall is concentrated over urban areas with more impervious surfaces, and because the majority of rainfall occurs in the second half of the storm time-series, after soils have begun to saturate. These results demonstrate that basin-averaged rainfall frequency alone is a poor predictor of flood damage magnitude.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eDifferences between SST and design storm risk estimates compound over longer time horizons (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). SST produces a mean annual damage of \u003cspan\u003e$\u003c/span\u003e19.8\u0026nbsp;million, exceeding the design storm estimate by \u003cspan\u003e$\u003c/span\u003e7.9\u0026nbsp;million. More notably, SST yields a maximum annual damage estimate of \u003cspan\u003e$\u003c/span\u003e115\u0026nbsp;million \u0026ndash; more than eight times the upper bound estimate derived from the design storms. Over a 50-year time horizon, SST produces a mean cumulative damage estimate of \u003cspan\u003e$\u003c/span\u003e842\u0026nbsp;million, approximately \u003cspan\u003e$\u003c/span\u003e248\u0026nbsp;million (40%) greater than the design storm estimate and equivalent to nearly 10% of the total building value within the model domain. The upper bound of the 90% confidence interval for the SST-based 50-year damage is nearly \u003cspan\u003e$\u003c/span\u003e1.02\u0026nbsp;billion, roughly 1.5 times the corresponding design storm estimate.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e\n\u003ch3\u003eStorm-Flood-Damage Frequency Analysis\u003c/h3\u003e\n\u003cp\u003eAnalysis of extreme events within the SST ensemble reveals a systematic decoupling between the frequencies of rainfall, flooding, and damage (Fig. S9). Of the ten largest flood extents simulated with SST, all with return periods exceeding 285 years, seven are produced by storms whose basin-averaged rainfall return periods are less than 200 years, including one event with a rainfall return period of just 70 years (Fig. S10). The single largest flood extent \u0026ndash; associated with a 500-year return period \u0026ndash; is generated by a storm with a basin-averaged rainfall return period of 285 years. Rainfall for this storm is relatively evenly distributed in space, but with the heaviest bands of rainfall slightly to the east, near more densely developed regions within the model domain, and with more rainfall occurring later in the storm.\u003c/p\u003e \u003cp\u003eFlood damage frequency is similarly decoupled from the frequency of rainfall. Of the ten most damaging events, all but the largest are produced by storms with basin-averaged total rainfalls that have return periods less than 200 years, and six by storms with return periods less than 150 years. Several high-damage events are generated by relatively small and frequent storms that produce modest total flood extents. For example, one storm with a 40-year return period generates flooding with a 20-year extent but results in damage with an estimated return period of 154 years (Table S2).\u003c/p\u003e \u003cp\u003eWe find that storms that generate disproportionately large impacts tend to concentrate intense rainfall over urban areas or late in the 24-hour storm duration, suggesting that the observed decoupling between the frequency of rainfall, flooding, and damage in the simulations with SST is likely driven by the spatiotemporal variability of rainfall. SST allows both for local rainfall extremes \u0026ndash; which are averaged out in design storm rainfall \u0026ndash; and for scenarios in which these local extremes fall over less pervious and/or more saturated landcover, leading to more intense flooding locally. Different storm structures may then trigger these local pockets of flooding in different locations and at different times, altering the way sub-catchment outflows combine at confluences and leading to changes in flood extent and depth further downstream. These dynamics lead to greater variability in flooding depth and extent at the local scale, particularly for smaller storms, whose basin-averaged rainfall may be far lower than the local maximum rainfall intensity. Since the spatial distribution of property values is non-uniform, the increased uncertainty in local flood frequency leads to even greater variability in the total resulting damage. This phenomenon is demonstrated on Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, which provides five examples of storms that all generate similar damage, but have vastly different spatiotemporal structures, flood extent, and damage profiles. These results demonstrate that storms with similar basin-averaged rainfall or flood extent can nevertheless produce damage of vastly different magnitudes, depending on their spatiotemporal rainfall structure and interaction with the built environment.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eOur results suggest that the relationships between the frequency of storms and the resulting watershed responses are not one-to-one, as assumed by design storm approaches, but rather complex and nonlinear, consistent with findings from past studies that have assessed the frequency relationships between rainfall and discharge \u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e,\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e34\u003c/span\u003e\u003c/sup\u003e. A key mechanism underlying this disconnect is the localized intensity of storms. High-intensity rainfall over impervious surfaces or saturated soils generates disproportionately high runoff, while identical rainfall distributed over permeable or forested areas may produce minimal flooding. Variations in storm heading, forward velocity, and spatial rainfall pattern may further alter the timing of flood peaks downstream, leading to a wide range of possible peak water levels across the watershed. These dynamics are not captured by design storm approaches, which simplify rainfall spatially and temporally, leading to systematic underestimation of localized pluvial flooding and nonlinear flood responses.\u003c/p\u003e \u003cp\u003eOur results also indicate that the relationship between rainfall, hazard and risk is fundamentally nonlinear. Storms that produce disproportionately high damage tend to concentrate intense rainfall over urban areas or in the latter half of the 24-hour period, consistent with previous studies of flood hazard \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e,\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e,\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e\u003c/sup\u003e. Even when basin-averaged rainfall is identical, differences in the spatial and temporal distribution of rainfall alter runoff generation, infiltration, and flood routing, producing substantial local variation in inundation depth and extent. Consequently, smaller and more frequent (e.g., \u0026lt;\u0026thinsp;25-year return period) storms can generate outsized flooding and damage, particularly in densely developed areas. Thus, the design storm approach does not yield conservative hazard or risk estimates, contrary to recent assertions \u003csup\u003e\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e,\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e\u003c/sup\u003e, suggesting that in order to accurately capture flood hazard and risk, modelers must represent local-scale extrema of flood drivers and other variables. These local variabilities explain why the frequency associated with basin-averaged rainfall does not translate linearly to frequency of flooding or damage, violating a core assumption of design storm-based flood hazard assessment.\u003c/p\u003e \u003cp\u003eThese findings have important implications for widely used flood hazard and risk products \u0026ndash; including FEMA\u0026rsquo;s National Flood Hazard Layer (NFHL) and First Street\u0026rsquo;s Flood Factor\u0026reg; (V4.0) \u0026ndash; which rely heavily on design storms and regional flood frequency analysis (RFFA). Our results suggest that by neglecting the effects of spatiotemporal variability and nonlinear storm-flood-damage relationships, these approaches omit critical risks. While RFFA partially addresses limitations of design storms by using historical discharge data from stream gauges, it remains constrained by short record lengths, sparse, point-scale observations, assumptions of regional homogeneity, and limited representation of non-stationarity in climate or land use change. In contrast, SST leverages high-resolution radar-rainfall data, accounts for current land use patterns, and simulates nonlinear rainfall-flood-damage interactions across diverse landscapes, capturing variability overlooked by both design storm approaches and RFFA and potentially leading to more accurate assessments of hazard and risk.\u003c/p\u003e \u003cp\u003eOur findings underscore the value of probabilistic, high-resolution hazard modeling for informing planning, emergency preparedness, and resilience investments. In particular, our results support FEMA\u0026rsquo;s Future of Flood Risk Data (FFRD) initiative \u003csup\u003e\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e, which proposes to use SST-based frameworks to update flood hazard maps and risk estimates across the United States. Probabilistic methods such as SST produce spatially continuous hazard estimates and allow uncertainty to be characterized at the building scale \u003csup\u003e\u003cspan additionalcitationids=\"CR40\" citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e41\u003c/span\u003e\u003c/sup\u003e, enabling more risk-informed decision making. These high-resolution analyses may support more effective allocation of resources for emergency response and recovery and enable targeted programs for mandating insurance purchases, development guidance, building elevation standards, infrastructure design, and mitigation funding, ultimately contributing to improved community preparedness.\u003c/p\u003e \u003cp\u003eAlthough our results demonstrate that SST improves hazard and risk estimation, our analysis is limited to storms sampled from a relatively short historical record (i.e., 2002\u0026ndash;2024) and does not fully incorporate probabilistic antecedent conditions (e.g., soil moisture), upstream inflows, or coastal boundary conditions that could further amplify differences relative to design storm approaches, both in terms of expected values and uncertainty. Moreover, between-storm variability in watershed conditions may further influence flood response, but capturing interarrival times and their effects on antecedent conditions remains challenging and continuous simulations over long time scales are computationally expensive. Future studies could expand probabilistic inputs to include variability in tides, storm surge, and river boundary conditions (and their compounding effects), or could consider storms of different durations (e.g., 12- or 72-hour events) that might drive different responses at the watershed scale (i.e., pluvial vs fluvial). Incorporating these factors would likely amplify the divergence between SST and design storm estimates and further accentuate nonlinear storm-flood-damage relationships.\u003c/p\u003e \u003cp\u003eOur research demonstrates that flood hazard and risk estimates that neglect spatiotemporal rainfall variability underestimate both hazard and damage by failing to capture nonlinear relationships between rainfall, inundation, and flood impacts driven by local-scale watershed response. High-resolution, probabilistic modeling with realistic rainfall representation provides a more comprehensive basis for flood hazard and risk assessment than deterministic design storm approaches. These improved estimates enable better-informed decisions related to emergency management, infrastructure planning, and community resilience. Given the substantial and growing economic and social consequences of flooding, transitioning from design storm-based approaches to probabilistic frameworks represents a critical step toward more accurate and actionable flood risk assessment.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003eSynthetic Storm Database\u003c/h2\u003e \u003cp\u003eSST was performed in the open-source RainyDay software package \u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e\u003c/sup\u003e using the National Oceanic and Atmospheric Administration\u0026rsquo;s Stage IV data \u003csup\u003e\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e42\u003c/span\u003e\u003c/sup\u003e from January 1, 2002 through December 31, 2023. These data were prepared for RainyDay by resampling them onto a rectilinear grid via a mass-conserving regridding scheme in the xesmf Python package \u003csup\u003e\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e43\u003c/span\u003e\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eSST requires that users delineate a meteorologically homogeneous region known as the transposition domain (Appendix A), which we developed quantitatively via the Spatial L-Moments of Annual Maxima (SLAM) approach \u003csup\u003e\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e44\u003c/span\u003e\u003c/sup\u003e. SLAM domains are specific to the selected watershed (Fig. A1) and storm duration of interest and can be adjusted based on a user-specified global significance level (GSL). For our watershed, SST was not particularly sensitive to changes in the GSL (i.e., extreme rainfall characteristics are homogeneous within the region), and so we selected the largest transposition domain (where \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;0.25; Fig. A2) in order to maximize the number of potential parent storms.\u003c/p\u003e \u003cp\u003eWe then used RainyDay to identify 220 parent storms within our transposition domain from the regridded Stage IV data. The parent storm duration was chosen as 24 hours, a duration commonly selected for modeling with design storms and relevant to the times of concentration experienced in our watershed. Rainfall estimates from thirteen of these parent storms contained radar artifacts such as bright-band contamination and blockage and were subsequently removed from consideration. We ultimately generated 10 ensemble members each comprised of 500 synthetic years of rainfall, with up to 18 storm arrivals per year. These realizations were then postprocessed to apply an alternative transposition scheme to parent storms identified as tropical cyclones, in order to enforce physically realistic transpositions. A description of the postprocessing method and validation results are included in Appendix B.\u003c/p\u003e \u003cp\u003eDesign storms were developed for each of the 1-, 2-, 5-, 10-, 25-, 50-, 100-, 200-, and 500-year return period, 24-hour storms. Total rainfall was derived from NOAA Atlas 14 \u003csup\u003e9\u003c/sup\u003e. Rainfall is distributed in time via the Soil Conservation Service (SCS) Type III curve \u003csup\u003e\u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e45\u003c/span\u003e\u003c/sup\u003e and is uniform in space. For each return period, we generated a storm for the mean rainfall intensity, as well as for the upper and lower bounds of the 90% confidence interval of the IDF curve for Coastal Carolina Regional Airport (CCRA). This resulted in a total of 27 design storms used for comparison. No areal reduction factors were applied in constructing the design storms, and thus the storm intensities and resulting flooding are likely to be conservative estimates.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eHydrodynamic Modeling\u003c/h3\u003e\n\u003cp\u003eWe built and validated a high-resolution, 2D hydrodynamic model in order to simulate the watershed response to the SST and design storms. The model was developed in the Super-Fast Inundation of Coasts (SFINCS) model code \u003csup\u003e\u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e46\u003c/span\u003e,\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e, a 2D, reduced-physics hydrodynamic modeling platform that can simulate fluvial, pluvial, storm surge, and compound flooding. SFINCS solves continuity and momentum equations based on the local inertial equations in LISFLOOD-FP \u003csup\u003e48\u003c/sup\u003e, plus optional wind drag and advection terms in the momentum equation.\u003c/p\u003e \u003cp\u003eSFINCS can simulate processes including spatially varying friction, infiltration (here, via the SCS curve-number method, with recovery), rainfall-runoff, and wind shear. Inputs to the SFINCS model include gridded and spatially uniform time-series precipitation datasets, gridded wind datasets, and time-series or constant water level boundary conditions. The model developed in this study was adapted from an existing SFINCS model of the Carolinas \u003csup\u003e\u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e49\u003c/span\u003e\u003c/sup\u003e and was constructed using a high-resolution digital elevation model (including channel bathymetry), gridded Manning\u0026rsquo;s roughness coefficients (\u003cem\u003en\u003c/em\u003e) that vary spatially based on land use/land cover, and rasters of spatially varying maximum soil moisture and saturated hydraulic conductivity. Further discussion of SFINCS model development is provided in Appendix C.\u003c/p\u003e \u003cp\u003eThe model uses a 100-meter grid resolution and a 5-meter subgrid which allows the model to run at a lower native resolution while still accounting for finer-scale variations in topography and surface roughness, thereby maximizing computational efficiency \u003csup\u003e\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e47\u003c/span\u003e\u003c/sup\u003e. The model was validated against five storm events of varying intensity and duration, including Hurricanes Matthew and Florence. Simulated hydrographs generally matched observations for all five validation events, resulting in an average Kling-Gupta efficiency (KGE) of 0.72 and an average peak elevation bias and root-mean-square error (RMSE) of 0.38 and 1.14 meters, respectively. For the two tropical cyclone validation events (Hurricanes Florence and Matthew), the model closely recreated observed high watermarks (HWMs) in the densely developed areas around New Bern, with an overall RMSE of 0.43 meters and a bias of 0.02 meters. Overall, our model performs similarly to other hydrodynamic models used for hazard and risk modeling and demonstrates high skill for storms of varying magnitude and frequency. Further description of the model validation process and model performance is provided in Appendix C.\u003c/p\u003e\n\u003ch3\u003eProbabilistic Flood Hazard Comparison\u003c/h3\u003e\n\u003cp\u003eWe simulated the watershed response to the 27 design storms using the validated hydrodynamic model, along with a subset of 1,842 of the SST storm realizations that had mean basin-wide precipitation greater than or equal to the 1-year, 24-hour design storm rainfall for New Bern. Each SST storm realization was assigned water level boundary conditions based on the historical median water elevations for the month in which that realization\u0026rsquo;s parent storm occurred. Since design storms are statistically derived from the entire period of record and thus cannot be associated with observed storms, they could not be associated with water levels from a particular month, and so they were instead assigned water level boundary conditions based on the overall historical median. All runs neglected wind and tidal influences and assumed an initial soil moisture equal to half of the maximum storage capacity.\u003c/p\u003e \u003cp\u003eThree hazard maps were generated for each of the design storm return periods, one for each of the mean, confidence interval upper bound, and confidence interval lower bound rainfall estimates. To create the hazard maps, the maximum flood elevation from each run (which are output on the model\u0026rsquo;s 100-meter grid) was first downscaled onto the model\u0026rsquo;s five-meter subgrid via bilinear interpolation, then converted to a flood depth raster by subtracting out the subgrid elevation profile. Areas with flood depths greater than or equal to 0.1 meters were assigned a flooding return period equal to that of the storm that generated that flood, in keeping with the one-to-one return period equivalency assumed by design storm approaches. The resulting hazard rasters were then combined together to create a single hazard map.\u003c/p\u003e \u003cp\u003eTen hazard maps were generated for SST, with one map per SST ensemble member. For each ensemble member, an SST hazard map was created by first stacking the flood elevations from all of that member\u0026rsquo;s model runs into a three-dimensional array, then sorting them from smallest to largest for each model grid cell. The sorted flood elevations were then assigned return periods between 1 and 500 years. We iterated through the sorted flood elevations from smallest to largest return period, downscaling each onto the model subgrid and converting them to flood depths. Once flood depths were greater than or equal to 0.1 m in a given pixel, the return period for flooding in that pixel was set equal to the return period of the current iteration. Ensemble statistics could then be calculated from these maps to understand the minimum, mean, and maximum exceedance probability for each pixel, as well as the standard deviation.\u003c/p\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eFlood Risk Comparison\u003c/h2\u003e \u003cp\u003eFlood elevations were calculated at each building in the model domain by extracting the maximum flood elevation for each storm event at polygon centroids from the North Carolina Statewide Building Footprint dataset (2020\u0026ndash;2022) available from North Carolina Emergency Management \u003csup\u003e\u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e50\u003c/span\u003e\u003c/sup\u003e. The flood depth at each building was calculated as the difference between the maximum simulated water surface elevation and the building\u0026rsquo;s first floor elevation (inside the FEMA SFHA) or ground surface elevation from the model subgrid at the building centroid (outside of the SFHA). Each building was then reclassified into one of the specific occupancy codes used by FEMA\u0026rsquo;s Hazus \u003csup\u003e\u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e51\u003c/span\u003e\u003c/sup\u003e program\u0026rsquo;s Flood Assessment Structure Tool lookup tables \u003csup\u003e\u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e52\u003c/span\u003e\u003c/sup\u003e based on number of stories, presence or absence of basement, and occupancy type. These specific occupancy codes were then used to assign a corresponding Hazus depth-damage function \u0026ndash; which relates the flood depth at a structure to the expected damage \u0026ndash; to each building. Finally, the resulting damage was calculated by entering the simulated flood depths at each building into that building\u0026rsquo;s depth-damage functions. Damage at each building could then be summed to calculate total damage for that storm.\u003c/p\u003e \u003cp\u003eStorm-specific damages were then aggregated into 1- and 50-year flood damage estimates, the latter of which was chosen to estimate damage over the typical lifespan of building stock \u003csup\u003e\u003cspan citationid=\"CR53\" class=\"CitationRef\"\u003e53\u003c/span\u003e\u003c/sup\u003e. For the design storm approach, long-term damage was simulated deterministically. Similar to the approach described in Lendering et al. \u003csup\u003e54\u003c/sup\u003e, annual damage was calculated by taking the product of each storm\u0026rsquo;s annual exceedance probability (AEP; the inverse of its return period) and its damage and then summing across all design storms. The 50-year damage was then assumed to be 50 times the annual damage. This process was repeated for the mean, lower-bound, and upper-bound design storms, producing a range of expected annual damages.\u003c/p\u003e \u003cp\u003eDamage for SST was calculated probabilistically. First, the distribution of annual damages was estimated by randomly sampling 10,000 years of damage. Each year was assumed to have \u003cem\u003ek\u003c/em\u003e storm arrivals, where \u003cem\u003ek\u003c/em\u003e was drawn from a Poisson distribution with λ\u0026thinsp;=\u0026thinsp;1.87 storms/year, ensuring that the average number of annual storm arrivals matched that of the design storms, whose AEPs sum to 1.87. For each year, \u003cem\u003ek\u003c/em\u003e storms were sampled from our database of SST storms, and their corresponding flood damages summed. To calculate 50-year damage, 50 annual damage estimates were randomly sampled (with replacement) from the 10,000 annual damage estimates, and their values summed.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eStorm-Flood-Damage Frequency Analysis\u003c/h2\u003e \u003cp\u003eTo assess whether the relationships between the return periods of storms, floods, and damages are one-to-one (as assumed by design storm approaches), we calculated the frequency of the mean total rainfall, flood extent (represented as the fraction of our watershed with an inundation depth of at least 0.1 meters), and total damage for each of our 1,842 SST storms. For each of these outputs, frequency was calculated using the same methods as RainyDay. For flood extent and damage, it was also assumed that the minimum and maximum return periods of these variables were identical to those of the storms.\u003c/p\u003e \u003cp\u003eSimilar to Wright et al. (2014) \u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e\u003c/sup\u003e and Yu et al. \u003csup\u003e28\u003c/sup\u003e, we then used Spearman rank correlation (\u003cem\u003eρ\u003c/em\u003e\u003csub\u003e\u003cem\u003es\u003c/em\u003e\u003c/sub\u003e) as a quantitative measure of the correlation between the return periods of rainfall, flood extents, and flood damage. Finally, we qualitatively analyzed the spatiotemporal structures of storms that caused extreme damage events in order to understand which storm characteristics influence nonlinearities in storm-flood-damage recurrence intervals.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003ch2\u003eCompeting Interests\u003c/h2\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eJ.A.B \u0026ndash; conceptualization, methodology, software, validation, formal analysis, writing-original draft, visualization; A.S. \u0026ndash; conceptualization, writing-original draft, supervision, funding acquisition; L.G. \u0026ndash; conceptualization, methodology, writing-review \u0026amp; editing; J.D.G. \u0026ndash; methodology, writing-review \u0026amp; editing; D.W. \u0026ndash; methodology, software, writing-review \u0026amp; editing; M.A.H. \u0026ndash; software, writing-review \u0026amp; editing; M.W. \u0026ndash; writing-review \u0026amp; editing.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003eThe authors would like to acknowledge B. FitzGerald for providing the transposition domain, R. Luettich for providing wind field data for the validation events, and G. Karlovits and D. Rosa for their insights and discussion during the project.\u003c/p\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003eData Availability\u003c/h2\u003e \u003cp\u003eStage IV precipitation data can be downloaded from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://data.eol.ucar.edu/dataset/21.093\u003c/span\u003e\u003cspan address=\"https://data.eol.ucar.edu/dataset/21.093\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. OWI wind and pressure data used to create the inputs for Hurricanes Florence and Matthew are proprietary data from OceanWeather Inc. All other data used to build and validate the models in this study are publicly available through USGS and NOAA. Replication data for the figures included in this manuscript are available via NHERI\u0026rsquo;s DesignSafe repository \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.17603/ds2-27r7-jv83\u003c/span\u003e\u003cspan address=\"10.17603/ds2-27r7-jv83\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e \u003csup\u003e55\u003c/sup\u003e.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eCode Availability\u003c/h2\u003e \u003cp\u003eThe SFINCS model documentation and source code can be downloaded from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://sfincs.readthedocs.io/en/latest/index.html\u003c/span\u003e\u003cspan address=\"https://sfincs.readthedocs.io/en/latest/index.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. The RainyDay documentation and source code is available from \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://github.com/HydroclimateExtremesGroup/RainyDay\u003c/span\u003e\u003cspan address=\"https://github.com/HydroclimateExtremesGroup/RainyDay\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Model files and code used to analyze these data are available upon request to the authors.\u003c/p\u003e \u003c/div\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eNational Academies of Sciences, Engineering, and Medicine. \u003cem\u003eFraming the Challenge of Urban Flooding in the United States\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.nap.edu/catalog/25381\u003c/span\u003e\u003cspan address=\"https://www.nap.edu/catalog/25381\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2019) doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.17226/25381\u003c/span\u003e\u003cspan address=\"10.17226/25381\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUnited States Government Accountability Office. \u003cem\u003eFEMA Flood Maps: Better Planning and Analysis Needed to Address Current and Future Flood Hazards\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.gao.gov/assets/gao-22-104079.pdf\u003c/span\u003e\u003cspan address=\"https://www.gao.gov/assets/gao-22-104079.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWing, O. E. J. \u003cem\u003eet al.\u003c/em\u003e Inequitable patterns of US flood risk in the Anthropocene. \u003cem\u003eNat. Clim. Chang.\u003c/em\u003e 12, 156\u0026ndash;162 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFederal Emergency Management Agency. Glossary: Flood Zones. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.fema.gov/about/glossary/flood-zones\u003c/span\u003e\u003cspan address=\"https://www.fema.gov/about/glossary/flood-zones\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFirst Street. \u003cem\u003eFirst Street Flood Model Version 4.0 Technical Methodology\u003c/em\u003e. (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSanders, B. F. \u003cem\u003eet al.\u003c/em\u003e Large and inequitable flood risks in Los Angeles, California. \u003cem\u003eNat Sustain\u003c/em\u003e 6, 47\u0026ndash;57 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWing, O. E. J. \u003cem\u003eet al.\u003c/em\u003e A 30 m Global Flood Inundation Model for Any Climate Scenario. \u003cem\u003eWater Resources Research\u003c/em\u003e 60, e2023WR036460 (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSmith, A., Sampson, C. \u0026amp; Bates, P. Regional flood frequency analysis at the global scale. \u003cem\u003eWater Resour. Res.\u003c/em\u003e 51, 539\u0026ndash;553 (2015).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBonnin, G. M., Martin, D., Lin, B., Yekta, M. \u0026amp; Riley, D. \u003cem\u003eNOAA Atlas 14 Precipitation-Frequency Atlas of the United States Volume 2 Version 3.0: Delaware, District of Columbia, Illinois, Indiana, Kentucky, Maryland, New Jersey, North Carolina, Ohio, Pennsylvania, South Carolina, Tennessee, Virginia, West Virginia\u003c/em\u003e. (2006).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLanciotti, S., Ridolfi, E., Russo, F. \u0026amp; Napolitano, F. Intensity\u0026ndash;Duration\u0026ndash;Frequency Curves in a Data-Rich Era: A Review. \u003cem\u003eWater\u003c/em\u003e 14, 3705 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlessing, R., Sebastian, A. \u0026amp; Brody, S. D. Flood Risk Delineation in the United States: How Much Loss Are We Capturing? \u003cem\u003eNatural Hazards Review\u003c/em\u003e 18, 04017002 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrody, S., Blessing, R., Sebastian, A. \u0026amp; Bedient, P. Delineating the Reality of Flood Risk and Loss in Southeast Texas. \u003cem\u003eNatural Hazards Review\u003c/em\u003e 14, 89\u0026ndash;97 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBrody, S. d., Sebastian, A., Blessing, R. \u0026amp; Bedient, P. b. Case study results from southeast Houston, Texas: identifying the impacts of residential location on flood risk and loss. \u003cem\u003eJournal of Flood Risk Management\u003c/em\u003e 11, S110\u0026ndash;S120 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGarcia, H. M. \u003cem\u003eet al.\u003c/em\u003e Reconstructing Repetitive Flood Exposure Across 78 Events From 1996 to 2020 in North Carolina, USA. \u003cem\u003eEarth\u0026rsquo;s Future\u003c/em\u003e 13, e2025EF006026 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHighfield, W. E., Norman, S. A. \u0026amp; Brody, S. D. Examining the 100-Year Floodplain as a Metric of Risk, Loss, and Household Adjustment. \u003cem\u003eRisk Analysis\u003c/em\u003e 33, 186\u0026ndash;191 (2013).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNayak, A., Gentine, P. \u0026amp; Lall, U. Financial losses associated with US floods occur with frequent low-return-period precipitation. \u003cem\u003eNat Water\u003c/em\u003e 3, 1256\u0026ndash;1267 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDo, H. X., Mei, Y. \u0026amp; Gronewold, A. D. To What Extent Are Changes in Flood Magnitude Related to Changes in Precipitation Extremes? \u003cem\u003eGeophysical Research Letters\u003c/em\u003e 47, e2020GL088684 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTuyls, D. M., Thorndahl, S. \u0026amp; Rasmussen, M. R. Return period assessment of urban pluvial floods through modelling of rainfall\u0026ndash;flood response. \u003cem\u003eJournal of Hydroinformatics\u003c/em\u003e 20, 829\u0026ndash;845 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFoufoula-Georgiou, E. A probabilistic storm transposition approach for estimating exceedance probabilities of extreme precipitation depths. \u003cem\u003eWater Resources Research\u003c/em\u003e 25, 799\u0026ndash;815 (1989).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWilson, L. L. \u0026amp; Foufoula-Georgiou, E. Regional Rainfall Frequency Analysis via Stochastic Storm Transposition. \u003cem\u003eJ. Hydraul. Eng.\u003c/em\u003e 116, 859\u0026ndash;880 (1990).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWright, D. B., Mantilla, R. \u0026amp; Peters-Lidard, C. D. A Remote Sensing-Based Tool for Assessing Rainfall-Driven Hazards. \u003cem\u003eEnviron Model Softw\u003c/em\u003e 90, 34\u0026ndash;54 (2017).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eEngland, J. F., Julien, P. Y. \u0026amp; Velleux, M. L. Physically-based extreme flood frequency with stochastic storm transposition and paleoflood data on large watersheds. \u003cem\u003eJournal of Hydrology\u003c/em\u003e 510, 228\u0026ndash;245 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFranchini, M., Helmlinger, K. R., Foufoula-Georgiou, E. \u0026amp; Todini, E. Stochastic storm transposition coupled with rainfall\u0026mdash;runoff modeling for estimation of exceedance probabilities of design floods. \u003cem\u003eJournal of Hydrology\u003c/em\u003e 175, 511\u0026ndash;532 (1996).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLawler, S. \u003cem\u003eet al.\u003c/em\u003e Application of stochastic storm transposition for hydrologic modeling in the mountainous western US. \u003cem\u003eStoch Environ Res Risk Assess\u003c/em\u003e 39, 109\u0026ndash;127 (2025a).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLompi, M., Caporali, E., Mediero, L. \u0026amp; Mazzanti, B. Improving flash flood risk assessment using a simple approach for extreme rainfall scaling and storms transposition. \u003cem\u003eJ Flood Risk Management\u003c/em\u003e 15, e12796 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerez, G., Gomez-Velez, J. D., Mantilla, R., Wright, D. B. \u0026amp; Li, Z. The Effect of Storm Direction on Flood Frequency Analysis. \u003cem\u003eGeophysical Research Letters\u003c/em\u003e 48, e2020GL091918 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWright, D. B., Smith, J. A. \u0026amp; Baeck, M. L. Flood frequency analysis using radar rainfall fields and stochastic storm transposition. \u003cem\u003eWater Resources Research\u003c/em\u003e 50, 1592\u0026ndash;1615 (2014).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu, G., Wright, D. B., Zhu, Z., Smith, C. \u0026amp; Holman, K. D. Process-based flood frequency analysis in an agricultural watershed exhibiting nonstationary flood seasonality. \u003cem\u003eHydrol. Earth Syst. Sci.\u003c/em\u003e 23, 2225\u0026ndash;2243 (2019).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYu, G., Wright, D. B. \u0026amp; Holman, K. D. Connecting Hydrometeorological Processes to Low-Probability Floods in the Mountainous Colorado Front Range. \u003cem\u003eWater Resources Research\u003c/em\u003e 57, e2021WR029768 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou, Z. \u003cem\u003eet al.\u003c/em\u003e The impact of the spatiotemporal structure of rainfall on flood frequency over a small urban watershed: an approach coupling stochastic storm transposition and hydrologic modeling. \u003cem\u003eHydrol. Earth Syst. Sci.\u003c/em\u003e 25, 4701\u0026ndash;4717 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhu, Z., Wright, D. B. \u0026amp; Yu, G. The Impact of Rainfall Space-Time Structure in Flood Frequency Analysis. \u003cem\u003eWater Resources Research\u003c/em\u003e 54, 8983\u0026ndash;8998 (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWright, D. B., Yu, G. \u0026amp; England, J. F. Six decades of rainfall and flood frequency analysis using stochastic storm transposition: Review, progress, and prospects. \u003cem\u003eJournal of Hydrology\u003c/em\u003e 585, 124816 (2020).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePerez, G., Coon, E. \u0026amp; Rathore, S. Advancing process-based flood frequency analysis for assessing flood hazard and population flood exposure. \u003cem\u003eJournal of Hydrology\u003c/em\u003e 28, (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVangelis, H., Zotou, I., Kourtis, I. M., Bellos, V. \u0026amp; Tsihrintzis, V. A. Relationship of Rainfall and Flood Return Periods through Hydrologic and Hydraulic Modeling. \u003cem\u003eWater\u003c/em\u003e 14, 3618 (2022).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAsher, M. \u003cem\u003eet al.\u003c/em\u003e Extreme rainfall and temporal loading in Great Britain: Analysis of present and future trends using a convection-permitting climate model. \u003cem\u003eJournal of Hydrology: Regional Studies\u003c/em\u003e 62, 102750 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCache, T., Bevacqua, E., Zscheischler, J., M\u0026uuml;ller-Thomy, H. \u0026amp; Peleg, N. Simulating Realistic Design Storms: A Joint Return Period Approach. \u003cem\u003eWater Resources Research\u003c/em\u003e 61, e2024WR039739 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCea, L. \u003cem\u003eet al.\u003c/em\u003e Recent advances and future challenges in urban pluvial flood modelling. \u003cem\u003eUrban Water Journal\u003c/em\u003e 22, 149\u0026ndash;173 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLehman, W. FEMA\u0026rsquo;s Future of Flood Risk Data Initiative. (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlfonso, L., Mukolwe, M. M. \u0026amp; Di Baldassarre, G. Probabilistic Flood Maps to support decision-making: Mapping the Value of Information. \u003cem\u003eWater Resources Research\u003c/em\u003e 52, 1026\u0026ndash;1043 (2016).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDi Baldassarre, G., Schumann, G., Bates, P. D., Freer, J. E. \u0026amp; Beven, K. J. Flood-plain mapping: a critical discussion of deterministic and probabilistic approaches. \u003cem\u003eHydrological Sciences Journal\u003c/em\u003e 55, 364\u0026ndash;376 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDiehl, R. M., Gourevitch, J. D., Drago, S. \u0026amp; Wemple, B. C. Improving flood hazard datasets using a low-complexity, probabilistic floodplain mapping approach. \u003cem\u003ePLoS ONE\u003c/em\u003e 16, e0248683 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDu, J. NCEP/EMC 4KM Gridded Data (GRIB) Stage IV Data. 1.0. (2011).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhuang, J., Dussin, R., Huard, D., \u0026amp; Bougault, P. xESMF. (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFitzGerald, B., Wright, D. B., Yan, L., Dietrich, A. \u0026amp; Sebastian, A. An L-Moments-Based Hypothesis Test to Identify Homogeneous Storm Transposition Regions. \u003cem\u003eJournal of Hydrology\u003c/em\u003e (in review).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUnited States Department of Agriculture. \u003cem\u003eUrban Hydrology for Small Watersheds TR-55\u003c/em\u003e. (1986).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLeijnse, T., Van Ormondt, M., Nederhoff, K. \u0026amp; Van Dongeren, A. Modeling compound flooding in coastal systems using a computationally efficient reduced-physics solver: Including fluvial, pluvial, tidal, wind- and wave-driven processes. \u003cem\u003eCoastal Engineering\u003c/em\u003e 163, 103796 (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan Ormondt, M., Leijnse, T., de Goede, R., Nederhoff, K. \u0026amp; van Dongeren, A. A subgrid method for the linear inertial equations of a compound flood model. \u003cem\u003eEGUsphere\u003c/em\u003e 1\u0026ndash;36 (2024) doi:\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.5194/egusphere-2024-1839\u003c/span\u003e\u003cspan address=\"10.5194/egusphere-2024-1839\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBates, P. D., Horritt, M. S. \u0026amp; Fewtrell, T. J. A simple inertial formulation of the shallow water equations for efficient two-dimensional flood inundation modelling. \u003cem\u003eJournal of Hydrology\u003c/em\u003e 387, 33\u0026ndash;45 (2010).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrimley, L. E. \u003cem\u003eet al.\u003c/em\u003e Determining the Relative Contributions of Runoff, Coastal, and Compound Processes to Flood Exposure Across the Carolinas During Hurricane Florence. \u003cem\u003eWater Resources Research\u003c/em\u003e 61, e2023WR036727 (2025).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNC Emergency Management GIS Section. Countywide 2020\u0026ndash;2022 Building Footprints.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFederal Emergency Management Agency. \u003cem\u003eHazus 6.1 Flood Model Technical Manual\u003c/em\u003e. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.fema.gov/sites/default/files/documents/fema_hazus-flood-model-technical-manual-6-1.pdf\u003c/span\u003e\u003cspan address=\"https://www.fema.gov/sites/default/files/documents/fema_hazus-flood-model-technical-manual-6-1.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2024).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBurns, J. N., Sharma, U., Lindeman, C. \u0026amp; Raines, J. Hazus Flood Assessment Structure Tool. Federal Emergency Management Agency (2021).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAndersen, R. \u0026amp; Negendahl, K. Lifespan prediction of existing building typologies. \u003cem\u003eJournal of Building Engineering\u003c/em\u003e 65, 105696 (2023).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLendering, K., Sebastian, A., Jonkman, S. N. \u0026amp; Kok, M. Framework for assessing the performance of flood adaptation innovations using a Risk-Based approach. \u003cem\u003eJournal of Flood Risk Management\u003c/em\u003e 12, (2018).\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJack Baer \u003cem\u003eet al.\u003c/em\u003e Neglecting Spatiotemporal Rainfall Variability Underestimates Flood Hazard and Risk. DesignSafe-CI \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.17603/ds2-27r7-jv83\u003c/span\u003e\u003cspan address=\"10.17603/ds2-27r7-jv83\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (2026).\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"npj-natural-hazards","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Natural Hazards](https://www.nature.com/npjnathazards/)","snPcode":"44304","submissionUrl":"https://submission.springernature.com/new-submission/44304/3","title":"npj Natural Hazards","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8593870/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8593870/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eFlood risk assessments underpin flood management and resilience efforts worldwide, including land-use planning, infrastructure design, and insurance requirements. Many of these assessments rely on design storms, which assume a linear relationship between the frequency of storms, flooding, and damage and neglect the spatial and temporal structure of rainfall. Here, we show that these assumptions can lead to systematic underestimates of flood hazard and risk. Using a coastal watershed in North Carolina, we compare design storm-based estimates with those from stochastic storm transposition, a probabilistic framework used to generate synthetic events with realistic rainfall fields. Though both methods produce similar basin-averaged rainfall statistics, design storms underestimate flood inundation frequency by 31 to 35% and expected 50-year damage by 42% relative to SST. These results reveal how nonlinear storm-flood-damage relationships amplify risk from smaller, more frequent storms and illustrate that accounting for spatiotemporal rainfall variability is crucial to risk assessment.\u003c/p\u003e","manuscriptTitle":"Neglecting Spatiotemporal Rainfall Variability Underestimates Flood Hazard and Risk","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-23 10:33:19","doi":"10.21203/rs.3.rs-8593870/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-02-06T15:37:10+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-02T17:11:56+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2026-02-02T10:33:32+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"81118892096872774076260592285675681243","date":"2026-01-23T12:48:18+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"142887247469761839793474151408568624245","date":"2026-01-22T15:22:35+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-01-21T12:40:13+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-01-18T07:44:49+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-01-18T07:33:03+00:00","index":"","fulltext":""},{"type":"submitted","content":"npj Natural Hazards","date":"2026-01-13T15:32:34+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"npj-natural-hazards","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [npj Natural Hazards](https://www.nature.com/npjnathazards/)","snPcode":"44304","submissionUrl":"https://submission.springernature.com/new-submission/44304/3","title":"npj Natural Hazards","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"NPJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5faa21f7-f7f2-49b6-999e-da8061f943e9","owner":[],"postedDate":"January 23rd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":61600669,"name":"Earth and environmental sciences/Climate sciences"},{"id":61600670,"name":"Earth and environmental sciences/Environmental sciences"},{"id":61600671,"name":"Earth and environmental sciences/Hydrology"},{"id":61600672,"name":"Earth and environmental sciences/Natural hazards"}],"tags":[],"updatedAt":"2026-04-06T01:38:37+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-23 10:33:19","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8593870","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8593870","identity":"rs-8593870","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}