Assessing future heat stress risk across Swedish residential areas using a combined GIS and indicator-based approach

Sweden's housing stock, designed predominantly for winter thermal performance, is increasingly exposed to summer heat extremes following record heatwaves in 2014, 2018, and 2022. Despite projections of intensifying heat risk and the near-total absence of mechanical cooling in Swedish residential buildings, no prior study has delivered a nationally comprehensive, indicator-based compound risk assessment integrating hazard, exposure, and vulnerability at the municipal scale. This study addresses that gap through an integrated framework covering all 290 Swedish municipalities, combining GIS-based climate data processing, standardised Poisson regression for empirical indicator selection, and CA-Markov land cover projections under RCP 2.6, 4.5, and 8.5 scenarios to 2040. Thermal hazard is characterised through maximum seasonal air temperature, consecutive summer days, and tropical day frequency derived from UERRA-HARMONIE reanalysis and SMHI regional projections; vulnerability is constructed from seven demographic, socioeconomic, and built environment indicators; and exposure is represented by SCB municipal population forecasts. The composite index is externally validated against observed 2018 excess summer mortality across all 290 municipalities, with the equal-weighted specification outperforming PCA- and regression-beta-weighted alternatives (Pseudo R² = 0.117, p = 0.007); the index is interpreted as a spatial prioritisation tool rather than a mortality prediction model. The 2040 projections are framed as a counterfactual stress test, applying projected thermal forcing to a fixed sociodemographic baseline. Results identify a pronounced south-north risk gradient, with Stockholm, Göteborg, Malmö, Burlöv, and Solna remaining persistently classified as high risk under every scenario. By 2040, over 230 of 290 municipalities are reclassified to medium risk or above, driven primarily by committed demographic ageing rather than emissions pathway, with RCP 2.6 and RCP 8.5 producing near-identical patterns. Elderly fraction, low-income share, building density, and vegetation proximity emerge as the significant mortality predictors, providing an evidential basis for spatially targeted adaptation centred on green-blue infrastructure, equity-oriented cooling access, and differentiated heat-health warning systems.