Projected losses of ecosystem services in the US disproportionately affect non-white and lower-income populations

The study addresses a critical gap in ecosystem services (ES) research: how ES benefits are distributed among different societal groups. While ES have often been measured in biophysical or aggregate economic terms, these metrics obscure distributional outcomes relevant to equity and environmental justice. With anticipated land cover change (loss of forests, grasslands, wetlands; expansion of cropland and urban areas) and population shifts (urban growth, rural decline, increasing racial diversity and socioeconomic segregation) in the US through 2100, the authors hypothesize growing mismatches between ES supply and demand that will disproportionately burden marginalized groups. They evaluate how projected changes in ES—clean air provision, protection from West Nile virus (WNV), and crop pollination—translate into benefits or losses across rural/urban status, income quintiles, racial/ethnic groups, and US regions under alternative future scenarios.

Prior research shows ES declines are projected globally with particularly severe impacts in developing regions, and local case studies indicate affluent groups often capture more ES benefits. In the US, environmental justice literature documents disproportionate exposure of minority and low-income populations to hazards like air pollution and natural disasters. However, distributional analyses of non-market ES benefits are rare. Conceptual frameworks distinguishing ES supply, demand, and benefits are useful to assess mismatches. Earlier ES studies commonly aggregate by geography (pixels, regions), potentially masking intra-regional inequities tied to race and class segregation. Emerging tools (social media data, agent-based models, social vulnerability indices) have been proposed to better disaggregate ES benefits among beneficiaries.

Study design: Projected changes in ES supply, demand, and benefits were modeled for all US conterminous counties from 2020 to 2100 for three services: air quality (clean air provision), vector-borne disease control (protection against WNV), and crop pollination. Benefits were defined as functions of both supply and demand, with no benefit if either is absent. Results were disaggregated by county density class (rural, suburban, urban), county income quintile (per capita income), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, Other), and broad US region (Midwest, Northeast, South, West). Scenarios and projections: Land cover projections used USGS FORE-SCE (1992–2100) and population projections used US EPA ICLUS, both aligned to IPCC SRES scenarios (A1B, A2, B1, B2). County-level income and demographic projections were from SSP-aligned datasets (Wear and Prestemon; Hauer 2019) mapped onto SRES following van Vuuren and Carter. Scenario assumptions vary in globalization, environmental regulation, population/GDP growth, technology, energy sources, and resource protection; A2 represents high population growth, regionalized economies, and heavy fossil fuel use; B2 emphasizes biodiversity protection and mixed energy with medium growth; A1B and B1 feature rapid tech innovation and higher GDP growth with differing environmental emphasis. Beneficiary disaggregation: Counties were binned as rural (<10,000 people), suburban (10,000–50,000), or urban (>50,000). Income groups used county per-capita income quintiles. Regions followed state groupings. Demographic groups were partitioned within counties by population shares (Black [non-Hispanic], Hispanic, Other [non-Hispanic American Indian/Alaska Native and Asian/Pacific Islander], White [non-Hispanic]); ES benefits were apportioned to groups by county population proportions. For crop pollination, farmer demographics were adjusted using USDA NASS (2012) to reflect farm operator race, finding operators on average 15% more White than the general county population; county-specific scalars adjusted demographic shares for pollination beneficiaries. Air quality model: The Intervention Model for Air Pollution (InMAP) was used to simulate annual average transport, transformation, and deposition of emissions. County-level emissions of NOx, NH3, SO2, and VOCs were estimated by multiplying areas of land cover classes by emissions factors (MEGAN v2.1 for biogenic VOCs and US National Emissions Inventory for others). InMAP assumes linear relationships between precursor emissions and ground-level PM2.5 and O3. Health damages were estimated by linking InMAP outputs of elevated PM2.5/O3 concentrations to county populations. Relative risk of premature all-cause mortality from PM2.5 exposure was computed using a Cox proportional hazards function; expected premature mortalities were calculated per Tessum et al. Supply metric: avoided PM2.5 emissions; demand: downwind exposed population; benefit: avoided mortalities. Crop pollination model: The Lonsdorf et al. model (LEM) generated a 0–1 index of relative wild-bee visitation (supply) using land-cover-specific floral and nesting values and an exponential distance-decay parameter (α) set to an average foraging distance of 600 m for temperate wild bees. Demand was the county area of pollinator-dependent crops, calculated from USDA NASS (2017) crop areas and median pollination dependency factors (Klein et al.). Future demand assumed constant dependency-weighted crop mix proportions by county scaled by projected cropland area. Benefit was the product of wild-bee visitation index within cropland and pollinator-dependent crop area. Vector-borne disease control model (WNV): County-level generalized linear models with Gamma distributions predicted WNV incidence (cases per 100,000 people) using proportions of land cover types as predictors, stratified by EPA Level 1 ecoregions to account for regional vector habitat differences. Models were trained on 2006–2016 human WNV incidence and NLCD land cover, with highly correlated predictors excluded and variable selection via AIC (glmulti in R). Cross-validated predictive accuracy averaged 62% (range 33%–97%). Supply was defined as avoided risk of exposure to WNV (inverse of realized incidence given land cover); demand was county human population; benefits were avoided WNV cases computed as supply × demand / 100,000. Computation and aggregation: ES supply, demand, and benefit were computed at county level for each year and scenario (2020–2100) and summarized as percent change 2020 to 2100. National aggregates were sums of county-level benefits. Distributions were then analyzed across the beneficiary groupings.

- National trends: Across nearly all scenarios and services, US ES benefits decline between 2020 and 2100, with the largest declines under SRES A2; declines are mitigated under B1 and B2. Under A1B, vector-borne disease control benefits (WNV) increase nationally by 4.5%. - County-level dynamics: Marked spatial variability in changes to ES supply and demand leads to frequent supply–demand mismatches (supply decreases while demand increases). For air quality, many counties see both supply and demand decrease, but urban counties experience increased demand (population growth) and larger declines in benefits. For crop pollination, demand generally increases (cropland expansion and more pollinator-dependent area) while supply declines (loss of forest/wetland habitats), creating widespread mismatches. For WNV control, demand decreases in most counties but increases in urban centers; supply often remains constant or increases, creating unrealized benefits; the direction of WNV risk response to land cover change varies by ecoregion. - Distributional outcomes by density: Rural counties gain benefits in air quality and WNV control but lose in crop pollination; urban counties exhibit the opposite (losses in air quality and WNV control benefits; gains in crop pollination), with suburban showing minimal change. - By income: Lowest-income quintile counties experience the greatest losses in air quality and WNV control benefits; highest-income quintile counties gain in air quality and WNV control but see declines in crop pollination benefits. Middle quintiles show smaller magnitude changes. - By race/ethnicity: Non-white groups (Black, Hispanic, Other) experience substantial losses in benefits across services, while White populations experience moderate gains. Averaged across scenarios: for non-white people, air quality, crop pollination, and WNV control benefits decrease by 224%, 118%, and 111%, respectively; for White people, these benefits increase by 10%, 35%, and 36%. - By region: Regional differences are comparatively small. Air quality benefits decline across all four regions. WNV control benefits increase in the Midwest and Northeast (under A1B and B1) and decrease in the South and West. Crop pollination changes are relatively large in the Midwest and Northeast. - Driver of inequities: Disproportionate losses for non-white and lower-income communities are largely driven by conversion of forests and wetlands to cropland and urban land in counties where these populations are projected to grow.

The findings support the hypothesis that ES benefit losses will disproportionately affect marginalized communities. Despite weak spatial patterns at regional or even county map scales, disaggregating by income and race reveals pronounced inequities: non-white, lower-income, and urban populations bear larger declines. This aligns with environmental justice evidence on disproportionate exposure to hazards and demonstrates that aggregated ES assessments can mask inequities. Land cover conversions (forest/wetland loss to cropland/urban) reduce ES supply (air quality, pollination), and projected demographic shifts concentrate demand among vulnerable groups. WNV dynamics are ecoregion-specific, complicating national generalizations. The study underscores the need to integrate equity into conservation and land-use policy. Responses to ES losses include direct damages (e.g., increased premature mortality from air pollution, reduced pollinator-dependent yields) and potential substitution (e.g., pesticide spraying for vector control), though cost-effectiveness and access to substitutes remain uncertain. Migration in response to ES declines can shift demand geographically and pose social justice challenges, especially when support is unequal. Current inequities in ES benefits arise from historical socio-political processes, and trends suggest disparities may worsen without intervention.

This work contributes a national, multi-ES, beneficiary-disaggregated projection showing that future ES benefit declines in the US will be uneven and regressive, disproportionately burdening non-white, lower-income, and urban populations. It advances ES assessment by linking supply, demand, and benefits across scenarios and explicitly partitioning outcomes by race, income, and urbanicity. Policy implications include spatially targeting conservation and land-use interventions to where ES supply is threatened and demand is high among vulnerable groups, integrating equity into conservation planning, promoting agroecological practices to support pollinators for farmers with limited access to substitutes, and adapting federal payment programs (e.g., Conservation Reserve Program) to consider vulnerability to ES loss. Future research should enhance ES model validation and uncertainty assessment, incorporate climate change impacts on land cover and ES, evaluate accessibility and equity of ES substitutes, and improve beneficiary disaggregation at finer spatial and social scales.

- Projections reflect scenario-based possibilities, not best estimates; land cover and population datasets (FORE-SCE, ICLUS) embody assumptions that may diverge from actual futures, and different land use models show divergent projections. - National-scale modeling required simplified, generalized approaches and parameters, favoring scalability over complexity; this may omit important local processes. - Climate change effects are not explicitly modeled in land cover, population, or ES components; exclusion likely underestimates future disparities and impacts. - Crop demand composition is assumed constant within counties over time (proportional scaling), potentially misrepresenting future pollination demand shifts. - Demographic partitioning of ES benefits assumes impacts scale with county-level population shares, not capturing within-county heterogeneity or differential exposure/sensitivity. - WNV models exclude climatic, socioeconomic, and fine-scale habitat variables; predictive accuracy varies by ecoregion (33%–97%). - Emissions estimates assume uniform emission factors for land cover classes across space; InMAP uses linear relationships that may not capture all atmospheric nonlinearities.