Modelling flood hazards at large scales – both uniform frequency hazard maps and event simulations whose frequency varies in space – is a relatively new scientific endeavour. Data and computation constraints have historically necessitated either a more local focus to modelling efforts, or the building of proof-of-concept global-scale models whose fidelity inhibits most practical applications. Here, we present a global climate-conditioned flood catastrophe model; the culmination of decades of research into scaling inundation modelling, the incorporation of climate change, and synthetic event generation. 30 m resolution global maps representing fluvial, pluvial, and coastal flooding for given return periods were simulated using a hydrodynamic model with sub-grid channels whose inputs were defined using regional flood frequency analyses. Change factors from climate model cascades were flexibly used to perturb the local flood frequency a given flood map represents. Separately, a 10,000-year-long set of synthetic events were simulated using a conditional multivariate statistical model fitted to global fluvial-pluvial-coastal reanalysis data. The empirical return period of a given event is used to sample the corresponding flood map return period in order to build a long synthetic series of floods. With a global exposure model built using a top-down approach – downscaling capital stock models to high-resolution satellite-derived land-use and building height data – and a global vulnerability model derived from an extensive review of modelling and engineering literature, we demonstrate the calibration and validation of the global risk model. We also show the software challenges overcome to run this model, as well as to enable end-users to flexibly calculate the flood risk of their own exposures in the Oasis Loss Modelling Framework.

Intensity–duration–frequency (IDF) curves are representations of the probability that a given rainfall intensity over a given duration [GU1] will be exceeded [GU2] within a given period. To construct IDF curves, rainfall observations are required, ideally at the sub-daily temporal resolution. Unfortunately, such measurements are available only for a few locations world-wide. This poses a major challenge for simulations of global pluvial flood hazard and risk which require information of intensity, duration, and probability as boundary conditions.As an alternative to global IDF curves created from remotely sensed rainfall, we here propose a bottom-up approach which departs at the gauge level and employs machine-learning for regionalizing information on IDF curves from gauged to ungauged areas.To that end, we use available quality-controlled sub-daily precipitation data from the GSDR data set to derive Simplified Metastatistical Extreme Value (SMEV) parameters at around 10,000 locations world-wide. After combining these parameters with globally available data of precipitation drivers, a random forest regression model is applied. Results indicate that some SMEV parameters can be better regionalized than others. With globally available SMEV parameters, it is possible to obtain rainfall intensity for any combination of duration and frequency.We then evaluated these IDF maps against analytical intensities derived at the GSDR stations directly. Results show overall good agreement, yet the tails of the distributions are not entirely represented in our simulated intensities. Additionally, we benchmarked our intensity maps against similar datasets such as PPDIST and GPEX. Last, we assessed practical implications by comparing flood maps created with the various datasets used as pluvial boundary condition. While there are fundamental differences in how each of the datasets is derived, our analysis indicates overall similar spatial patterns and distributions of rainfall intensities.While such data-driven approaches clearly depend on the quality and quantity of available sub-daily rainfall observations, our proposed bottom-up approach seems to be able to scale local data to global data applicable in both flood risk research and practice.

Climate change is exacerbating flood risks globally. In the U.S., Federal Emergency Management Agency (FEMA) Flood Insurance Rate Maps (FIRMs) delineate areas at high flood risk (i.e., 100-year floodplains), however, FIRMs are incomplete and outdated. We utilize risk estimates from advanced flood modeling and sociodemographic data at the census tract level to examine inequities in risk to federally-overlooked 100-year flooding across the conterminous United States (CONUS). Using multivariable statistics that account for clustering and non-normality, we analyze inequities by flood type (coastal, pluvial, and fluvial) and along the urban-rural continuum. Results indicate that 26 million CONUS residents live in federally-overlooked 100-year floodplains. Lower income is associated with elevated risk of federally-overlooked 100-year flooding for all flood types across the urban-rural continuum. We find inconsistent relationships for neighborhood racial/ethnic composition, yet increased Black composition is associated with greater risk of federally-overlooked 100-year pluvial flooding in metropolitan areas, where 47% of the CONUS population at risk to federally-overlooked flooding resides. In suburban and rural contexts, higher Hispanic/Latinx composition is associated with increased risk of federally-overlooked fluvial flooding. The large scale and inequities of federally-overlooked flood risk we find have major public policy, planning and environmental justice implications that demand flood resilience interventions.

The severity of flood impacts is influenced by social vulnerability, which stems from marginalization processes that depress a community’s ability to mitigate and recover from flood events. Understanding how social vulnerability operates in different flood contexts informs who is most susceptible to which types of impacts. This study examines the empirical relationship between social vulnerability and flood risk and how that relationship varies by element at risk and flood magnitude. Using inputs of social vulnerability indicators and flood risk to crops and buildings, we employed spatial clustering and spatial regression to determine which social vulnerability characteristics are most associated with economic risk. Regions with high crop risk are associated with more natural resource-based employment and housing tenure, while low-risk regions are less linguistically isolated. For buildings, high-risk regions have higher proportions of renters and lower proportions of racial minorities, while low-risk areas are associated with mobile homes and vacant housing. Overall, housing tenure and natural resource dependence were consistently correlated with building and crop risk. This study advances scientific knowledge by highlighting how specific social vulnerability dimensions relate to flood risks across sectors and geographies.

The role climate change plays in increasing the burden placed on governments and insurers to pay for recovery has not been extensively explored and is the focus of this study. This study examines the impacts of climate change attributed flooding on federal disaster aid disbursement in Harris County, Texas following Hurricane Harvey in 2017. Our approach uses flood models to estimate the amount of flood damages attributable and not attributable to climate change under two climate change attribution scenarios from peer reviewed studies: 20% and 38% increases in rainfall associated with the hurricane due to climate change. These estimates are combined with census tract-level disbursement data for FEMA’s National Flood Insurance Program (NFIP) and the Individual Assistance (IA) part of the Individuals and Households Program. We employ spatial lag regression models with direct and spatial spillover effects to analyze the relationship between a tract’s flood damages—both attributed and not attributed to climate change—and federal disaster aid. We find that both types of flood damage shape federal aid disbursements, but that climate change attributed damages tend to have larger effect sizes (elasticities) especially for IA. Specifically, for a 1% increase in additional climate change attributed damages per household in a census tract (under the 20% scenario), expected NFIP levels in that census tract are 0.26% higher and IA levels are 0.3% higher. Implications center on federal funding in an era of climate change.

Catastrophe (cat) models are tools, typically used in the (re)insurance industry, that evaluate the risks to a given portfolio by modelling the impact of thousands of years of synthetic hazard events. Of particular interest to users is an evaluation of the low probability (tail) risks. This includes asking questions such as, “what is the worst loss event that will be exceeded, on average, every 200 years?” An assessment of tail risks is inherently uncertain. This is compounded by a large number of uncertain or free parameters throughout the modelling chain which may be set via expert (subjective) judgement or via a process of calibration. The calibration process would take a given portfolio with known historical losses and adjust some of the free parameters to match the historical losses. This process may be reframed as creating a structured ensemble of catastrophe models with a range of each of the free or uncertain parameters. The process would then compare the modelled losses from each of the ensemble members to the known historical record and select the model that best represents the historical losses. A major limitation of the ensemble approach to catastrophe model calibration is the short historical record from which to select the most representative model. This work uses a flood catastrophe model ensemble to explore the calibration process by creating a short synthetic loss record from a single ensemble member and examining the downstream effects of using this loss record for model selection.

In recent years the precision and skill of global flood hazard models has increased dramatically. This, alongside developments allowing for hazard model conversion to stochastic event sets and the open-sourcing of catastrophe modeling software, have opened up the possibilities of developing detailed and skillful global flood catastrophe models; assessing not just average risk but also the possible impacts of major flood events and the probability distribution of annual losses. In order to realize these possibilities, it is necessary to develop a global vulnerability framework that appropriately represents the state of the art in vulnerability modeling whilst being flexible to user inputs and faithfully representing uncertainties. Here, we present a framework for implementing a flexible vulnerability module within a global flood catastrophe model. Vulnerability curves are derived for a variety of occupancies (residential, commercial, industrial), for both building and contents losses. The mean loss ratio curves are derived from literature and commercial datasets before being normalized and fit to a family of logarithmic functions of depth, which can be adjusted for varying property characteristics. Uncertainty distributions are parameterised using a 4 parameter beta model and derived from a large insurance claims dataset (~2 million claims). Finally, using the same large claims dataset, we explore the event-level correlation of the quantiles sampled within our uncertainty distribution. Specifically, we evaluate the extent to which the quantiles sampled of the uncertainty distribution, in a Monte Carlo approach, should be clustered for each event. This is vital for correctly estimating the losses from rare, high-impact events and allows for a realistic representation of vulnerability uncertainty in aggregate loss estimates.

Flood event set generation, as employed in catastrophe risk models, relies on gauge information that is not available in data-scarce regions. To overcome this limitation, we develop a stochastic fluvial and pluvial flood model of Southeast Asia, using freely and globally available discharge data from the global hydrological model GloFAS and rainfall from the ERA5 reanalysis. We use a conditional multivariate statistical model to produce a synthetic catalog of 10,000 years of flood events. We calculate the flood population exposure associated with each flood event using freely available population data from WorldPop and generate exposure probability exceedance curves. We validate the population exposure curves against observed flood disaster data from EM-DAT, showing that our methodology provides exposure estimates that are in line with historical observations. We find that there is a 1% probability that more than 30 million people will be exposed to flooding in a given year according to our event set. This number is roughly half the population living in the 100-year return period flood zone of Fathom's hazard maps, suggesting most studies based on static flood maps overestimate exposure. This analysis provides significant progress over previous non-stochastic studies which are only able to compute total or average exposure within a given floodplain area and demonstrates that a reanalysis-based stochastic flood model can be designed to generate reliable estimates of population exposure probability exceedance. This study is a step toward a fully global catastrophe model for floods capable of providing exposure and loss estimates worldwide. Global hydrological models can be used to drive a large-scale stochastic flood inundation model in Southeast Asia A reanalysis-based stochastic flood model generates realistic flood events The computed flood exposure exceedance curve for Southeast Asia compares well to the EM-DAT database

Global flood mapping has developed rapidly over the past decade, but previous approaches have limited scope, function, and accuracy. These limitations restrict the applicability and fundamental science questions that can be answered with existing model frameworks. Harnessing recently available data and modelling methods, this paper presents a new global ~30 m resolution Global Flood Map (GFM) with complete coverage of fluvial, pluvial, and coastal perils, for any return period or climate scenario, including accounting for uncertainty. With an extensive compilation of global benchmark case studies – ranging from locally collected event water levels, to national inventories of engineering flood maps – we execute a comprehensive validation of the new GFM. For flood extent comparisons, we demonstrate that the GFM achieves a critical success index of ~0.75. In the more discriminatory tests of flood water levels, the GFM deviates from observations by ~0.6 m on average. Results indicating this level of global model fidelity are unprecedented in the literature. With an optimistic scenario of future warming (SSP1-2.6), we show end-of-century global flood hazard increases are limited to 9% (likely range -6–29%); this is within the likely climatological uncertainty of -8–12% in the current hazard estimate. In contrast, pessimistic scenario (SSP5-8.5) hazard changes emerge from the background noise in the 2040s, rising to a 49% (likely range of 7–109%) increase by 2100. This work verifies the fitness-for-purpose of this new-generation GFM for impact analyses with a variety of beneficial applications across policymaking, planning, and commercial risk assessment.

Abstract Damage and disruption from flooding have rapidly escalated over recent decades. Knowing who and what is at risk, how these risks are changing, and what is driving these changes is of immense importance to flood management and policy. Accurate predictions of flood risk are also critical to public safety. However, many high‐profile research studies reporting risks at national and global scales rely upon a significant oversimplification of how floods behave—as a level pool—an approach known as bathtub modeling that is avoided in flood management practice due to known biases (e.g., >200% error in flood area) compared to physics‐based modeling. With publicity by news media, findings that would likely not be trusted by flood management professionals are thus widely communicated to policy makers and the public, scientific credibility is put at risk, and maladaptation becomes more likely. Here, we call upon researchers to abandon the practice of bathtub modeling in flood risk studies, and for those involved in the peer‐review process to ensure the conclusions of impact analyses are consistent with the limitations of the assumed flood physics. We document biases and uncertainties from bathtub modeling in both coastal and inland geographies, and we present examples of physics‐based modeling approaches suited to large‐scale applications. Reducing biases and uncertainties in flood hazard estimates will sharpen scientific understanding of changing risks, better serve the needs of policy makers, enable news media to more objectively report present and future risks to the public, and better inform adaptation planning.

AXA proposes a novel continental-scale generator of synthetic gridded rainfall daily timeseries (10km resolution) with applications to cross-country risk assessment under current and future climate scenarios. Europe serves as a case-study to demonstrate and assess its performance in terms of hazard modelling and extrapolation to unobserved extreme local and regional events. This generator belongs to the class of time and space reshuffling Stochastic Weather Generators (SWGs) and generates unobserved events by re-sequencing historical multisite timeseries (E-OBS). Consistency at continental scale is ensured by relying on weather regimes and atmospheric situations characterized from the ERA5 reanalysis over Europe. The use of atmospheric drivers and dry-wet alternating cycles allows for the determination of both precipitation-prone situations or on the contrary drier spells, while preserving the physics of the atmospheric water cycle. Spatial reshuffling is introduced by regional differentiation. Transitions between regimes can be either calibrated from the historical data or extrapolated to represent future states of the climate along with an appropriate uplifting of the humidity-related variables. This generator is operationally used at AXA as part of a European flood risk model and serves as the main input to an hydrological and hydraulic model.

The flooding brought about by compound coastal flooding events can be devastating. Before, during, and immediately following these events, flood inundation maps (FIMs) can provide essential information to emergency management. However, there are a number of frameworks capable of estimating FIMs during flood events. In this article, we evaluate FIMs derived from three such frameworks in the context of Hurricane Harvey. Our analysis reveals that each of the three FIM frameworks provides different FIMs that differ in their level of accuracy. Each of the three FIMs also produces different exposure and consequence estimates because of their spatial differences. This investigation highlights the need for a centralized means of vetting and adjudicating multiple FIMs during compound flood events empowered by the ability to distribute FIMs as geographic information system (GIS) services and coalesce FIMs into a common operating picture. Furthermore, we provide evidence that the ability to produce multi-model estimates of FIMs to create probabilistic FIMs may provide a better product than the use of a lone FIM.

The past decade has seen considerable advances in the field of global flood modelling. In the 2010s, it began as a niche academic endeavour building models of the order 103 m horizontal resolution. In the 2020s, it is maturing into an established scientific discipline and yields profitable commercial ventures, with global models emerging of the order 101 m resolution.Building on the original 102 resolution global inland flood model of Sampson et al. (2015) – with a hydraulic engine based on the sub-grid version of the LISFLOOD-FP local inertial formulation of the shallow water equations (Bates et al., 2010; Neal et al., 2012) – we present the critical advances required to create a ~30 m resolution model of considerably greater fidelity and functionality:Using FABDEM as the underlying elevation grid, a machine-learning correction of the Copernicus global digital surface model to a digital terrain model (Hawker et al., 2022). Representing river hydrography with MERIT-Hydro (Yamazaki et al., 2019), ensuring the correct alignment of river channels with valley bottoms. Estimating river bathymetry prior to inundation modelling with a gradually varied flow solver (Neal et al., 2021). Updating boundary condition generation models with new hydrometric datasets and machine-learning hydrologic regionalization techniques (e.g. Zhao et al., 2021). Driving a global coastal flood model with a tide–surge–wave regional frequency analysis using tide gauges and reanalyses (Sweet et al., 2020). Implementing known and estimated flood protection measures as a rapid and adaptable post-process. Generating global climate change factors for fluvial, pluvial, and coastal floods for any plausible 21st century climate state. Applying climate change factors as a tractable post-process to a set of multi-frequency flood maps. These updates form the third version of Fathom's global flood maps. We show that these herald a new era of global flood modelling precision and accuracy, with additional utility wrought from linking climate projections to high-resolution true hydrodynamic models at the global scale for the first time. We also chart the road ahead for global flood modelling: outlining the significant data and modelling challenges our community must address to continue on this unprecedented development trajectory. References:Bates, P., et al. (2010) https://doi.org/10.1016/j.jhydrol.2010.03.027Hawker, L. & Uhe, P., et al. (2022) https://doi.org/10.1088/1748-9326/ac4d4fNeal, J., et al. (2012) https://doi.org/10.1029/2012WR012514Neal, J., et al. (2021) https://doi.org/10.1029/2020WR028301Sampson, C., et al. (2015) https://doi.org/10.1002/2015WR016954Sweet, W., et al. (2020) https://doi.org/10.3389/fmars.2020.581769Yamazaki, D., et al. (2019) https://doi.org/10.1029/2019WR024873Zhao, G., et al. (2021) https://doi.org/10.5194/hess-25-5981-202

<p>Intensity–duration–frequency<strong> </strong>(IDF) curves are representations of the probability that a given rainfall intensity will occur within a given period. At the global scale, however, only for a few locations sub-daily rain gauge data is available from which global IDF curves could be derived. This poses a major challenge for simulations of global pluvial flood hazard and risk which require information of intensity, duration, and probability as boundary conditions. Therefore, efficient yet accurate means for scaling the locally available data to the global extent need to be found.</p> <p>Consequently, we use available quality-controlled sub-daily precipitation data from the GSDR data set to derive growth curve parameters at around 10,000 locations world-wide. After combining these scale and shape parameters with globally available data of main precipitation drivers, a regionalized machine learning model is first trained and tested and then applied to produce global IDF maps.</p> <p>Finally, we evaluated these maps against an ensemble of openly available local IDF curves found in literature. By selecting locations spread across the globe, we try to ensure to include as much variability as possible in the evaluation. Additionally, the global IDF curves were benchmarked against available more bespoke IDF data in the USA and UK.</p> <p>While such data-driven approaches clearly depend on the quality and quantity of available sub-daily rainfall observations, the method still shows to capabilities of current data-driven modelling approaches to scale local data to global data applicable in both flood risk research and practice.</p>

We present a transparent and validated climate-conditioned catastrophe flood model for the UK, that simulates pluvial, fluvial and coastal flood risks at 1 arcsec spatial resolution (∼ 20–25 m). Hazard layers for 10 different return periods are produced over the whole UK for historic, 2020, 2030, 2050 and 2070 conditions using the UK Climate Projections 2018 (UKCP18) climate simulations. From these, monetary losses are computed for five specific global warming levels above pre-industrial values (0.6, 1.1, 1.8, 2.5 and 3.3 ∘C). The analysis contains a greater level of detail and nuance compared to previous work, and represents our current best understanding of the UK's changing flood risk landscape. Validation against historical national return period flood maps yielded critical success index values of 0.65 and 0.76 for England and Wales, respectively, and maximum water levels for the Carlisle 2005 flood were replicated to a root mean square error (RMSE) of 0.41 m without calibration. This level of skill is similar to local modelling with site-specific data. Expected annual damage in 2020 was GBP 730 million, which compares favourably to the observed value of GBP 714 million reported by the Association of British Insurers. Previous UK flood loss estimates based on government data are ∼ 3× higher, and lie well outside our modelled loss distribution, which is plausibly centred on the observations. We estimate that UK 1 % annual probability flood losses were ∼ 6 % greater for the average climate conditions of 2020 (∼ 1.1 ∘C of warming) compared to those of 1990 (∼ 0.6 ∘C of warming), and this increase can be kept to around ∼ 8 % if all countries' COP26 2030 carbon emission reduction pledges and “net zero” commitments are implemented in full. Implementing only the COP26 pledges increases UK 1 % annual probability flood losses by 23 % above average 1990 values, and potentially 37 % in a “worst case” scenario where carbon reduction targets are missed and climate sensitivity is high.

In the United States, the Federal Emergency Management Agency (FEMA) delineates 100-year flood zones to define risks, regulate flood insurance premiums, and inform flood management. Evidence indicates that FEMA flood maps are incomplete, calling much of our current knowledge of U.S. flood hazard inequities into question. We use a state-of-the-art flood hazard model and census tract-level dasymetrically mapped sociodemographic data to examine flood risk inequities in the Greater Houston area, where increasingly frequent and damaging flood events are occurring. We innovate by analyzing federally overlooked 100-year flood risks (100-year flood zones delineated by the flood hazard model that are outside of FEMA 100-year flood zones). Results indicate that nearly 1 million Greater Houston residents live in federally overlooked 100-year flood zones. Black and Asian neighborhoods experience disproportionate risk in federally overlooked pluvial and fluvial flood zones, and Hispanic neighborhoods experience disproportionate risk in all federally overlooked zones (coastal, pluvial, and fluvial). High flood risk and the relative lack of protective resources in federally overlooked 100-year flood zones doubly jeopardizes racial and ethnic minority communities. Our findings and recent flood disasters suggest that future flood impacts in Greater Houston will be catastrophic and unjust unless FEMA revises their risk mapping and management approach to promote long-term public safety and social equity.

The flood event “Berndt” in North-Western Europe in July 2021 demonstrated to stakeholders the critical importance of understanding present-day events, their spatial-temporal coherence and magnitude, and the probability of their occurrence. The shear size of multi-country events such as “Berndt” necessitate a large scale and highly automated approach to the modelling of their characteristics.To this end, we (AXA, Fathom and Deltares) developed a modelling framework that can efficiently compute very large event sets and worked on improving its underlying skill by making the framework highly modular. The framework consists of the following modules: a synthetic weather generator, which may sample many thousands of years of continuous weather data at daily and sub-daily time scales; a gridded hydrological model forced by the synthetic weather that produces long-term hydrological timeseries and derived statistics, and a subcatchment-scale fluvial and pluvial flood model archive, produced from large amounts of simulations with the Fathom hydraulic flood model engine. Its modular character allows for exchanging components, improving existing components or modifying parameters to assess sensitivities or uncertainties.In last year’s contribution, we presented our first synthetic simulations of weather and hydrology. In this year’s contribution we have improved stochastic weather generation and have established a full 10,000-year event set and flood catalogue, both under the present climate and a future climate projection that encompasses several assumptions. We will show the statistics and spatial configuration of past events according to our catalogue, including “Berndt 2021”, and how the statistics of extreme events will change according to the defined climate scenario simulation.

Air can hold more moisture as temperature increases, leading to more extreme rainfall events. Yet, this does not necessarily result in larger river floods. Here we use model projections to explore differences in the response of the atmosphere and catchments to an increase in global mean temperature. For both extreme rainfall and flow, we compute relative changes per °C, often called change factors (CFs) or scaling factors. Unlike some other studies, the multidecadal temperature mean is used instead of the temperature during the extreme event. This allows us to use CFs to produce maps of future changes for any emission scenario and future period.We relied on rainfall projected by 4 high-resolution GCMs from CMIP6 HighResMIP post-processed using 3 levels spatial smoothing (low to high smoothing). We also used hydrological simulations from 3 global hydrological models (GHMs) forced by 4 GCMs and produced as part of the ISIMIP2b project. We computed changes in the median of annual maxima based on periods of 31 years on 0.25° (HighResMIP) and 0.5° (ISIMIP2b) global grids. Working with two 12-member ensembles enables us to assess uncertainties in future changes.We found that whilst extreme rainfall is projected to increase over 87% of the land area (ensemble median), only 69% of the land area is projected to show an increase in extreme flow magnitude. Importantly, while there is high model agreement (at least ¾ of the models agree) that extreme rainfall will increase over 76% of the land area, high agreement that future flows will increase is only found over 40% of the land area. We show that these discrepancies are caused by changes in soil moisture and snow pack projected by the GHMs, highlighting the importance of river flood drivers other than extreme rainfall.

Climate change is already increasing the severity of extreme weather events such as with rainfall during hurricanes. But little research to date investigates if, and to what extent, there are social inequalities in climate change-attributed extreme weather event impacts. Here, we use climate change attribution science paired with hydrological flood models to estimate climate change-attributed flood depths and damages during Hurricane Harvey in Harris County, Texas. Using detailed land-parcel and census tract socio-economic data, we then describe the socio-spatial characteristics associated with these climate change-induced impacts. We show that 30 to 50% of the flooded properties would not have flooded without climate change. Climate change-attributed impacts were particularly felt in Latina/x/o neighborhoods, and especially so in Latina/x/o neighborhoods that were low-income and among those located outside of FEMA’s 100-year floodplain. Our focus is thus on climate justice challenges that not only concern future climate change-induced risks, but are already affecting vulnerable populations disproportionately now.