High urban flood risk and no shelter access disproportionately impacts vulnerable communities in the USA

The study addresses escalating urban flood risks in the U.S. driven by climate change and urbanization, which increasingly result in severe societal, infrastructural, and health impacts. Vulnerable and historically marginalized communities—often situated in low-lying, flood-prone neighborhoods due to legacies of redlining and environmental racism—face disproportionate exposure and reduced benefit from flood protection and risk management. The authors posit that assessing flood risk without accounting for access to emergency shelters can allow natural hazards to become human disasters. The research question is how to integrate shelter accessibility into flood risk assessment to identify priority areas and populations most at risk, thereby enabling more equitable planning, preparedness, and resource allocation. The purpose is to provide a practical, scalable framework that combines FEMA’s Riverine Flood Risk Index with access to emergency shelters at the neighborhood scale and to synthesize demographic characteristics of populations exposed. The importance lies in revealing inequities in both risk and shelter access, informing targeted interventions, and supporting more equitable, community-centric disaster planning and policy.

Prior work has examined flood risk across spatially and socially vulnerable populations and the health and economic impacts of floods, but fully characterizing impacts on vulnerable groups remains challenging. Studies have explored shelter suitability and effectiveness, their role in community resilience, optimal shelter siting for evacuation, shelter response capacity, uneven shelter access for vulnerable housing, and community-based disaster mitigation strategies. Research also highlights how urban growth patterns and climate change exacerbate flood impacts, with environmental and social inequalities influencing services, infrastructure, and resource access. Existing national risk assessments include components such as expected annual loss, social vulnerability, and community resilience; however, reliance on property value-based loss can perpetuate inequities by prioritizing higher-value areas. The literature underscores the need to incorporate access to shelters and community characteristics into risk assessments to support equitable, effective disaster preparedness and response.

Study areas and data: Eight U.S. cities with high fluvial/pluvial flood risk were selected based on First Street Foundation’s Flood Factor: Chicago (IL), Cincinnati (OH), Detroit (MI), Fresno (CA), Indianapolis (IN), Nashville (TN), Pittsburgh (PA), and San Antonio (TX). Three core datasets were used: (1) FEMA National Risk Index (NRI) for riverine flood risk at the census tract level (risk defined via expected annual loss, social vulnerability, and community resilience); (2) Homeland Infrastructure Foundation-Level Data (HIFLD) National Shelter System Facilities for locations of potential emergency shelters (educational, community, health, civic, religious centers); and (3) 2015–2019 ACS 5-year estimates at the block group level for demographics and socioeconomic attributes, including seven socially vulnerable cohorts: disabled individuals, elderly, carless, low income, Hispanics, Asians, and African Americans. Access to shelters: Auto-based cumulative opportunities access to shelters was computed for each block group using travel-time thresholds of 30 and 60 minutes. For each block group i, access Ai equals the count of shelters Oj reachable within threshold t from the population-weighted centroid, using a binary impedance function f(Cij) = 1 if travel time Cij ≤ t and 0 otherwise. Spatial clustering: Bivariate Local Indicators of Spatial Autocorrelation (BiLISA; bivariate Local Moran’s I) were applied at the block group level to assess spatial association between the standardized flood risk in a core block group and the spatial lag (average in neighboring block groups under a first-order queen contiguity weights matrix) of shelter access. Significance was assessed via 2,000 conditional permutations with pseudo p-values at a 90% level. Significant locations were classified into High-High (HH), Low-Low (LL), Low-High (LH), and High-Low (HL), where the first term refers to local flood risk relative to the city mean and the second refers to neighboring shelter access relative to the city mean; Not Significant (NS) indicates no statistically significant local association. Analyses focused primarily on 30-minute access, with 60-minute results provided for comparison. Population synthesis: For each city and cluster class, the shares of land area and population were calculated, including breakdowns by the seven vulnerable cohorts to identify disparities in exposure to combined flood risk and shelter access.

Shelter access vs flood risk: Shelter allocation is not aligned with flood risk. Clusters with low shelter access occur across both high and low flood risk (bundles of HL and LL), indicating residents in low-access regions experience flood risk disproportionately. Shelters are more accessible in central business districts and inner-city areas than at edges, suggesting edge residents may be more adversely affected during urban floods. A large share of block groups are Not Significant (NS), even under a lenient 90% significance threshold: Detroit 91%, Chicago 82%, Nashville 43%. City-level cluster proportions (examples): HL clusters: San Antonio 13% (highest) vs Detroit 2% (lowest). LH clusters: Nashville 26% (highest) vs Detroit 2% (lowest). HH clusters: Fresno 20% (highest) vs Detroit 2% (lowest). LL clusters: Fresno 20% (highest) vs Detroit 2% (lowest). On average across cities, HL has the greatest spatial extent (22.9%), followed by HH (10.3%), LL (9.5%), and LH (8.2%). Population exposure and equity: Population exposure is more critical than the count of block groups. There is an unequal distribution of residents in flood risk areas across cities. Indianapolis, Nashville, Fresno, and San Antonio show the highest inequalities across vulnerable cohorts; Pittsburgh, Cincinnati, Chicago, and Detroit the lowest. Contrary to prior findings emphasizing African Americans, combining flood risk with shelter access identifies Asians and the elderly as the most at-risk groups. Carless and disabled populations are also highly vulnerable due to mobility constraints that limit shelter reach even in high-access areas. Exposure statistics: Share of population in regions with flood risk: Indianapolis 63.2%, Nashville 56.7%, Fresno 55.2%, San Antonio 54.7%; Detroit as low as 9.7% (which also had the lowest relative flood-risk area). HL cluster shares of risk-prone lands: San Antonio 53.0%, Nashville 43.4%, Indianapolis 26.1%, Fresno 22.5%. Cities with higher LH shares: Cincinnati 11.1%, Pittsburgh 10.5%, Indianapolis 9.9%, Fresno 9.4%. Share of population living in HL areas: San Antonio 39.2%, Indianapolis 26.7%, Nashville 24.7%, Cincinnati 19.6%.

Integrating shelter access into flood risk assessments yields practical planning and policy benefits: (i) it shifts focus from purely geographic risk to community needs by incorporating socioeconomic vulnerabilities; (ii) it prioritizes resources to areas of greatest need, reducing equity gaps; and (iii) it catalyzes collaboration among planners, emergency managers, and policymakers. The authors illustrate funding implications via land-centric vs community-centric allocation: Pittsburgh has a larger flood-prone area, whereas Chicago has larger at-risk communities; incorporating shelter access and community exposure supports more equitable, community-centric allocations. The current FEMA NRI includes expected annual loss, which can bias resource allocation toward higher-value properties; augmenting risk assessments with community characteristics (e.g., household income within clusters) can balance mitigation objectives with equity goals. The integrated framework supports both long-term strategies (e.g., establishing permanent shelters in high-risk, low-access areas) and short-term strategies (e.g., temporary shelters during flood seasons, enhancing mobility options and evacuation transport for those without vehicles). It also motivates the development of interactive strategic planning tools to test scenarios, optimize resource allocation, inform policy, and improve community awareness, thereby advancing resilience and preparedness.

This study integrates access to emergency shelters into riverine flood risk assessment at the neighborhood scale for eight high-risk U.S. cities, classifying areas into HH, HL, LH, and LL clusters and synthesizing demographic characteristics of exposed populations. Findings reveal misalignment between shelter access and flood risk, higher access in inner-city areas, and disproportionate exposure of Asians and the elderly, as well as carless and disabled populations, in high-risk, low-access areas. By highlighting where vulnerable communities face compounded risk due to limited shelter access, the framework supports equitable planning, emergency response, and resource allocation. The approach is practical, scalable, and transferable to other urban regions experiencing flood risk, and it encourages a shift toward community-centric mitigation and preparedness. Future work can refine inputs (e.g., shelter operational status and suitability) and expand scenario-based tools to further inform adaptive, equitable resilience planning.

The analysis uses population-weighted centroids of block groups as origins for travel-time and access calculations, which may introduce spatial error, particularly in larger block groups. Shelter access was measured based on presence/location only; some shelters may not operate during floods or may be unsuitable for specific populations (e.g., individuals with disabilities or the elderly). More precise origin locations, real-time operational status, and suitability information would improve access estimates. The study provides a foundational framework rather than a definitive solution, encouraging multidisciplinary enhancements in future work.