Demographic and socioeconomic gradients of projected heat exposure across contrasting European regions: a paired census-tract analysis of outermost and metropolitan territories
Regional Environmental Change · 2026-12 · 26 (4)
DOI: 10.1007/s10113-026-02719-3 ↗Kurz gefasst
Fachartikel in der Zeitschrift Regional Environmental Change, online erschienen am 21.11.2026. Verfasst von Francisco Armas und 4 weiteren Autor:innen. Die Arbeit nennt 59 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Universidad de la Laguna.
Abstract (Originalsprache)
Climate-adaptation planning across the European Union assumes that projected heat exposure follows comparable demographic and socioeconomic gradients in outermost-region archipelagos and in continental metropolitan regions, a transferability assumption rarely tested at fine spatial resolution. We apply an identical statistical pipeline to harmonised census-tract data for two structurally contrasted Spanish regions: the Canary Islands (1388 tracts; 2.10 million inhabitants; WRF 3-km dynamical downscaling) and the Madrid metropolitan area (3117 tracts; 4.29 million inhabitants; ADAPTECCA ESD-RegBA 5-km statistical downscaling), both with 2022 census-tract income and demographic indicators, under CMIP6 SSP1-2.6 and SSP5-8.5 for mid- and end-century horizons. Inference rests on percentile-based heat indices (days above the local historical 95th percentile of maximum temperature, heatwave days and warm-spell duration in the continental case), which are computed against each product’s own historical distribution and are therefore robust to the level bias that we document for absolute-threshold indices in both legs. Because the two climate products differ in downscaling method, resolution and baseline, cross-case inference is restricted to the sign, existence and geography of the social gradients. The two regions load on distinct axes of vulnerability, with asymmetric evidence for each. In the archipelago, the gradient is demographic: the elderly population share carries a standardised coefficient of + 0.283 [0.235, 0.332] on projected P95-day change and + 0.295 [0.247, 0.342] on heatwave-day change (both p < 0.001), is positive and significant in four of the five islands estimable separately and retains its sign and significance in all ten robustness variants tested (+ 0.131 to + 0.295, all p < 0.001). A deprivation index built from the low-income share and mean income is an order of magnitude weaker (+ 0.086 and + 0.069), changes sign across variants and vanishes under the reference pathway (+ 0.010, p = 0.60), so the insular economic null is scenario-dependent rather than general. In the metropolitan area, the gradient is economic and the demographic one is absent: deprivation carries + 0.256 [0.227, 0.285] on warm-spell duration and the elderly share + 0.033, with the economic coefficient rising to + 0.353 [0.310, 0.395] within the central municipality and remaining between + 0.205 and + 0.423 in nine of the ten variants tested, the exception being a night-time percentile index on which it is absent. A stacked model with case-by-axis interactions tests the contrast formally: the demographic coefficient differs between cases by − 0.262 [− 0.428, − 0.095] and the difference survives a wild cluster bootstrap over the fourteen analytical units ( p = 0.012), whereas the + 0.187 difference in the economic coefficient does not ( p = 0.15). The asymmetry is therefore established for one axis and suggestive for the other. Across the twelve supra-tract units, the unit-level deprivation coefficient increases with an ex-ante index of joint within-unit social and climatic heterogeneity, formalised during revision (slope + 0.105 [0.054, 0.156], p = 0.001 pooled, not significant within either case separately). Bivariate spatial-cluster analysis identifies populations under climatic-social double burden of comparable share but structurally contrasted geographies: 123,479 to 240,116 insular residents on the demographic axis, dispersed across five islands and 389,436 to 529,549 continental residents on the economic axis, in the southern arc of the metropolitan area; within each range the lower bound is obtained under Queen contiguity weights and the upper under eight-nearest-neighbour weights, both at p < 0.05. Separately, and derived from the gridded observational analysis rather than from the simulation, 596,984 Canarian residents (32.9% of the archipelago) already live in tracts recording 60 or more tropical nights per year. These figures describe today’s (2022) population composition exposed to end-century climate, not a demographic projection. The demographic and socioeconomic axes of vulnerability are not equivalent across territorial contexts, and neither is a single climate product: EU adaptation planning should test, before deployment, which axis is recoverable at the analytical unit and whether the exposure indicator is robust to the biases of the product available for that territory.
Förderung
- Universidad de la Laguna
Zitieren
APA
Armas, F., Herrera, A., Sosa, P., Díaz, J. P., & Corral, S. (2026). Demographic and socioeconomic gradients of projected heat exposure across contrasting European regions: a paired census-tract analysis of outermost and metropolitan territories. Regional Environmental Change, 26(4). https://doi.org/10.1007/s10113-026-02719-3
Vancouver
Armas F, Herrera A, Sosa P, Díaz JP, Corral S. Demographic and socioeconomic gradients of projected heat exposure across contrasting European regions: a paired census-tract analysis of outermost and metropolitan territories. Regional Environmental Change. 2026;26(4). doi:10.1007/s10113-026-02719-3
BibTeX
@article{Armas2026,
author = {Armas, Francisco and Herrera, Alejandro and Sosa, Paula and Díaz, Juan Pedro and Corral, Serafín},
title = {Demographic and socioeconomic gradients of projected heat exposure across contrasting European regions: a paired census-tract analysis of outermost and metropolitan territories},
journal = {Regional Environmental Change},
year = {2026},
volume = {26},
number = {4},
doi = {10.1007/s10113-026-02719-3}
}
EUVIMED