Spatial diagnostics of area-level disparities in calculated x-minute accessibility: a multimodal, multitemporal, and multiscale study of Groningen

Conclusions about area-level accessibility disparities depend on how access, time, and geography are defined. This study examines calculated accessibility in the province of Groningen, the Netherlands, across 15 service categories grouped into overall, education, health, recreation, and retail/other accessibility outcomes. Forty measures include walking and cycling at 15 minutes and public transport at 15 and 30 minutes in rush-hour, off-peak, and weekend periods, at neighborhood, H3-hexagon, and district scales. Ordinary least squares (OLS) estimates adjusted province-wide associations with socio-demographic and built-environment characteristics; multiscale geographically weighted regression (MGWR) estimates their geographical variation; and a two-stage screen built on local bivariate Lee’s L locates persistently low accessibility and tests its group composition. Social loneliness showed the clearest negative adjusted association, recurring across all outcomes, travel definitions (modes, time budgets, departure periods), models, and scales, and largely surviving Benjamini-Hochberg false-discovery-rate control. Very-low income negative associations were also recurrent, particularly regarding healthcare accessibility, whereas findings for families with school-age children were limited to associations with two retail/other accessibility contexts. MGWR identified local amplification and six geographically restricted negative contexts undetected by neighborhood-scale OLS, five involving immigrants and one low income. Persistently low accessibility clustered in 23 neighborhoods in eastern Groningen, holding 5,769 homes, 19 of them regardless of the radius considered. Low income and fragile social health among older residents recurred in 18 of them and social loneliness in 19, though no group exceeded what a spatially structured null produces. The framework separates adjusted relationships, local variation, and unadjusted co-location; its findings are diagnostic area-level signals, not evidence of individual deprivation, unmet need, or causal effects.