Demographics and risk of isolation due to sea level rise in the United States

The study investigates how sea level rise (SLR) can isolate communities by disrupting transportation access to essential services, a burden that may not be captured by traditional inundation-focused assessments. The authors pose two key questions: (1) How does the risk of isolation vary among racial and ethnic groups along U.S. coastal areas? and (2) How does the risk of isolation correlate with socio-demographic characteristics linked to social vulnerability (age, income, renter status, racial composition)? Motivated by evidence that disadvantaged populations (racial minorities, older adults, low-income groups, renters) have fewer resources to adapt and are often located in under-resourced, less resilient built environments, the paper argues that focusing solely on inundation neglects complex, cascading burdens—particularly isolation—that can reinforce existing inequities. The purpose is to quantify and map disparities in isolation risk under multiple SLR scenarios to inform equitable adaptation planning.

Prior research and practice have emphasized direct inundation to assess SLR risk, overlooking the impacts of infrastructure disruption and loss of access to services. Studies have documented uneven exposure and inequitable adaptation in coastal cities (e.g., New York City), driven by historic discriminatory development and underinvestment in infrastructure in marginalized communities. Social disadvantage—linked to race, age, renter status, and income—often compounds vulnerability to climate hazards. The literature indicates communities with higher social disadvantage may be less protected or abandoned in adaptation decisions. This body of work underscores the need to move beyond parcel-level flooding metrics to consider network-based accessibility and isolation as critical dimensions of risk and equity.

Risk of isolation is measured by intersecting U.S. OpenStreetMap (OSM) road networks with NOAA Mean Higher High Water (MHHW) global SLR scenarios from 1 to 10 ft (one-foot increments). To capture spatially varying sea level change, relative sea-level rise (RSLR) is computed using tide gauge data and SLR projections, assigning each census block centroid to the nearest gauge. A census block is considered at risk of isolation if, at MHHW for a given SLR scenario, there is no available (non-inundated) route from the block centroid to any essential service, defined as primary schools and fire stations (used as proxies for essential services). Road networks were obtained via geofabrik.de; essential service locations come from DHS HIFLD. Populations at risk are estimated using 2020 U.S. Census block-level data. Socio-demographic attributes (median household income, median age, renter households, and racial composition for Black, Hispanic, and White alone not Hispanic) are drawn from ACS 2019 5-year estimates at tract and county levels for analyses above block scale. For timing analyses, intermediate (1.0 m by 2100) and high (2.0 m by 2100) SLR scenarios are used, with isolation determined in 10-year intervals (2030–2150) when projected SLR at the nearest gauge meets or exceeds the isolation threshold for a block. Multivariate analysis employs Generalized Linear Models (GLMs) at the tract level to relate: (1) percent population at risk of isolation, (2) percent at risk of inundation (where a block’s centroid intersects MHHW), and (3) percent “missing” population (isolation minus inundation divided by isolation) to SLR level and socio-demographics: %Isolation or %Inundation or %Missing = β0 + β1 SLR + β2 MedianAge + β3 ln(MedianIncome) + β4 %Black + β5 %Hispanic + β6 %RenterHouseholds. GLMs use observations across SLR levels 1–10 ft; example model sizes are on the order of ~53,000 tract-SLR observations. Code and an interactive dashboard are publicly available.

- Disparities by race and SLR level: Aggregated nationally, Hispanic populations are overrepresented in the isolated population beginning at 4 ft SLR; Black populations are overrepresented after approximately 6 ft; White populations become underrepresented after 5 ft. - Scenario timing: Under an intermediate SLR scenario (1.0 m by 2100), disproportionate isolation emerges by ~2120; under a high scenario (2.0 m by 2100), disparate effects appear as early as ~2090. - State-level disproportion at low SLR: At 3 ft SLR, Black populations in Pennsylvania (12.7% of state population) face disproportionate isolation; Hispanic populations face disproportionate isolation at 3 ft in Florida (27.1%), Louisiana (5.8%), Mississippi (3.6%), and Maine (2.1%). - County-level spread with rising SLR: The number of coastal counties where minorities are disproportionately at risk increases with SLR. For Hispanics, about 39.5% of counties at 5 ft and >50% at 10 ft show disproportionate isolation; for Black populations, ~24% at 5 ft and ~30% at 10 ft. - Localization and severity: At the tract level, some tracts show Black or Hispanic isolation risk exceeding ten times their representation in the county population, indicating highly localized inequities. - Regional patterns: Black populations disproportionately affected primarily in parts of the Northeast and California; Hispanic populations disproportionately affected across the study area, notably Northern California, Louisiana, and much of the Northeast. Thirty-four counties show disproportionate isolation for both Black and Hispanic populations at 3 ft SLR. - Socio-demographic correlates (GLM): Higher median age, higher %Black, higher %Hispanic, higher ln(income), higher %renters, and higher SLR are all positively associated with a larger fraction of tract population at risk of isolation and at risk of inundation. For the “missing” population (those captured by isolation but not inundation), younger age, higher income, more renters, and lower minority percentages are associated with larger missed fractions. - Magnitude of burden: Isolation-based risk identifies 30–90% more population at risk than inundation-only metrics and, in some locations, decades earlier.

The findings directly address the research questions by demonstrating that isolation due to SLR disproportionately affects historically marginalized racial and ethnic groups, and that this burden is intertwined with age, renter status, and income. Disparities become evident at relatively low to intermediate SLR levels and intensify with rising seas, with timing sensitive to global emissions pathways. The results highlight critical equity considerations: older adults face heightened isolation risks, exacerbating challenges to aging in place and access to healthcare; renters are particularly vulnerable due to higher eviction risks and frequent exclusion from flood mitigation programs (e.g., buyouts). Moreover, relying solely on inundation metrics systematically underestimates affected populations—especially in tracts with younger, higher-income, higher-renter, and lower-minority shares—potentially skewing planning and resource allocation. Incorporating isolation into risk assessments provides a more comprehensive and just basis for adaptation, with implications for public health access, transportation resilience, and potential secondary effects such as climate-related migration and gentrification pressures.

This study introduces and applies a network-based, service-access metric of isolation to quantify inequitable burdens of SLR across U.S. coastal communities. It shows that Hispanic and Black populations, as well as tracts with older residents and higher renter shares, face disproportionate isolation, sometimes decades before inundation-only assessments would indicate risk. The work underscores the necessity of integrating isolation metrics within explicit equity frameworks in climate adaptation and long-range urban planning (e.g., DAPP), to better target investments in transportation and essential service accessibility. Future research should examine dynamic interactions among SLR-driven isolation, infrastructure adaptation, demographic change, migration, economic investment, and gentrification, as well as test sensitivity to future road network modifications and evolving demographics.

- Spatial representation: Assumes entire census block is isolated based on the centroid’s routable access; finer intra-block disparities are not captured. - Static infrastructure: Uses current road networks and does not incorporate future adaptations (elevations, new links) or maintenance changes. - Data uncertainty: Census and ACS data carry margins of error, potentially leading to over- or under-estimation of populations and characteristics. - Demographic dynamics: Does not account for future demographic shifts (migration, age structure changes, tenure changes), which could alter future risk distributions. - Service proxies: Isolation is defined via access to primary schools and fire stations as proxies; other critical services may yield different access patterns.