COS 45-5 - CANCELLED - Legacies of residential segregation shape urban biodiversity data collection and understanding

Abstract: Citizen science projects can engage millions of people in biodiversity research and advance conservation worldwide, but sampling bias can unwittingly shape data collection and research outcomes. Sampling bias associated with the global biodiversity program eBird is most often studied at large spatial scales, yet evidence is accumulating that at local scales, sampling bias intersects with social factors, as observations overrepresent areas with greater wealth and lower racial diversity. Simultaneously, residential segregation practices, like ‘redlining’ in the US, have caused inequitable distributions of environmental attributes like pollution and tree cover. Further, the “luxury effect,” a pattern of higher biodiversity in affluent neighborhoods, has been observed across many cities. Here, we aimed to quantify whether sampling bias associated with legacies of discriminatory residential policies and persisting inequity results in biased inference and predictions of avian biodiversity from eBird data. Additionally, we investigated whether the luxury effect hypothesis was supported after accounting for sampling bias. We used preferential-sampling methods and landscape and social variables to jointly model sampling intensity and avian species richness using eBird data in three US cities. We included US census-derived socioeconomic variables and indices of residential segregation as predictors to disentangle how uneven sampling related to these variables influences model-based predictions of species richness. The proportion of canopy cover and the intensity of building development were positively and negatively associated with species richness, respectively. We found that population density and the local Shannon-index of racial diversity, rather than the proportions of distinct racial groups or household income, were most strongly associated with sampling intensity, with greater sampling in more homogeneous, sparsely populated neighborhoods. Additionally, a positive relationship between socioeconomics and species richness remained after accounting for sampling bias, suggesting that social inequity influences both urban ecological processes and where people participate in citizen science. The result that racial heterogeneity better explains eBird sampling processes and species richness than does specific racial or ethnic group proportions suggests that larger systemic structures, more than individual neighborhood demographics, underlie urban ecological patterns like the luxury effect. Our approach can improve predictions of avian biodiversity in cities, and we show how inequitable sampling shapes our interpretation of biological patterns. Our results amplify the need to support citizen science engagement in marginalized communities and encourage researchers to consider sampling inequities when using citizen science data to ensure that research and conservation do not perpetuate inequities.