Urban Green and Blue Spaces as Atmospheric Environment Modifiers: Tract-Level Satellite Exposure and Multiscale Mental Health Associations

Urban green and blue spaces (UGBS) modify local atmospheric conditions, reducing urban heat island intensity, moderating particulate concentrations, and providing thermal comfort, with documented links to population mental health. Prior citywide analyses (Ipede et al., 2026) assign uniform exposure values to every census tract within a city, obscuring neighbourhood-level variation critical for targeted planning. This study computed population-weighted greenspace and bluespace exposure independently for each census tract using Google Earth Engine, applying spectral unmixing of Sentinel-2 imagery for greenspace and a dual-sensor Sentinel-2 MNDWI and Sentinel-1 SAR fusion for bluespace. Atlanta, Miami, and Los Angeles served as study counties. Frequent mental distress from CDC PLACES 2023 was modelled using ordinary least squares regression(OLS), single-bandwidth Geographically Weighted Regression(GWR), and Multiscale GWR(MGWR), extending the framework established by Ipede et al. (2026) from the city level to the census tract level. MGWR converged for Atlanta and Miami, confirming that greenspace and bluespace operate at distinct spatial scales. Model comparison by corrected AIC favored the added local flexibility in Miami (ΔAICc = 75.7) but favored the simpler single-bandwidth model in Atlanta (ΔAICc = 22.5). In Los Angeles, extreme spatial clustering of bluespace exposure prevented a stable local estimate under adaptive and fixed kernel specifications. Dual sensor bluespace classification accuracy varied substantially by hydrological context across the three counties, from substantial agreement in Miami (κ = 0.656) to fair agreement in Los Angeles (κ = 0.278) to slight agreement in Atlanta (κ = 0.026–0.150 across four tested configurations). Independent validation of the greenspace exposure metric against NAIP reference imagery showed a consistent positive bias across all three counties (Pearson r = 0.241–0.654). Tract-level exposure modelling and MGWR reveal spatially heterogeneous, city-specific associations between green and blue space and mental distress that uniform city-level analysis cannot detect, offering a reproducible and transparently bounded framework directly applicable to urban atmospheric quality planning and public health investment decisions.