Vulnerability Component¶
The vulnerability component constructs a continuous, spatially explicit composite Social Vulnerability Index (SVI) quantifying the differential sensitivity of the exposed population to heat stress, on the same 100 m reference grid.
The baseline SVI layer is built in NB02 (cityheat.vulnerability_layer). The time-varying projected SVI — one layer per target year and scenario — is built in NB03 (cityheat.dynamic_vulnerability), which also writes year/scenario-specific exposure_with_vulnerability HDF5 files that NB04–NB08 read directly. Vulnerability therefore evolves consistently with hazard and exposure, rather than being held static after the baseline year.
Three SVI dimensions¶
The SVI combines three equal-weighted dimensions:
1. Thermal sensitivity of the built environment¶
- Proxy: building construction vintage from the GHS-AGE raster (2025 release, 100 m).
- Epoch-of-construction codes 1–10 (1 = pre-1975, 10 = post-2015).
- Older buildings have lower insulation standards → higher indoor heat accumulation.
- Mapping:
thermal_vuln = (10 − code) / 9so oldest stock = 1.0, newest ≈ 0.0.
2. Social marginality¶
- Proxy: share of resident population born abroad.
- Source: EUROSTAT Census 2021 rasters (
ESTAT_OBS-VALUE-OTH / ESTAT_OBS-VALUE-T). - Foreign-born residents may face language barriers to warning reception and limited cooling access.
3. Economic vulnerability¶
- Proxy: non-employment rate among working-age population:
1 − EMP / Y_1564. - Source: EUROSTAT Census 2021 employment and 15–64 population rasters.
- Non-employed individuals are more likely to be at home during peak-heat hours without access to air-conditioned workplaces.
Composite SVI¶
SVI = (thermal_vuln + social_marginality + economic_vulnerability) / 3
Components are normalised to [0, 1] before averaging, but the normalisation method is heterogeneous by design:
- Thermal: direct linear inversion
(10 − code) / 9maps codes 1–10 to [0, 1]. - Foreign-born share and non-employment rate: two versions are computed — (i) a percentile-rank normalisation used for legacy diagnostics, and (ii) a robust min–max scaling to [0, 1] (default quantile bounds q05–q95) used as the baseline-anchored absolute score for the dynamic projection engine.
The composite SVI is in [0, 1], where 1 = most vulnerable.
Dynamic projection (2021–2050)¶
The SVI is projected dynamically rather than held fixed at the 2021 static baseline (the baseline_year parameter in the config). The projection engine decomposes SVI change into component-level spatial processes controlled by two key city-specific parameters.
Short-run projection (2021 → 2030)¶
Anchored to NUTS3-level trend indicators from the DRMKC Risk Data Hub.
Long-run projection (2030 → 2050)¶
Follows national SSP trajectories from the Global Data Lab / Wittgenstein Centre database, downscaled back to 100 m. SSP4 trajectories are remapped to SSP3, and SSP5 to SSP2, where required by data availability.
City-specific parameters¶
Values below are from the four demonstrated-city configs (configs/<city>.yml, vulnerability.dynamic block):
| Parameter | Description | Rome | Athens | Lisbon | Copenhagen |
|---|---|---|---|---|---|
| k | Persistence — share of short-run shock carried to long run | 0.85 | 0.90 | 0.72 | 0.70 |
| φ₂₀₃₀ | Spatial anomaly retention in 2030 | 0.92 | 0.95 | 0.88 | 0.86 |
| φ₂₀₅₀ | Spatial anomaly retention in 2050 | 0.75 | 0.85 | 0.65 | 0.60 |
| Retrofit rate (yr⁻¹) | Annual building thermal upgrade rate | 1.38% | 0.93% | 1.39% | 0.65% |
| New-build vulnerability | Vulnerability anchor for new/retrofitted stock | 0.15 | 0.15 | 0.15 | 0.15 |
Higher φ values mean local spatial contrasts are preserved longer. Lower φ values allow faster convergence toward the national mean.
Thermal retrofit pathway¶
The thermal component follows a distinct, slower pathway: - City-specific annual retrofit rates (~0.65–1.4% yr⁻¹), derived from EC (2019) weighted renovation rates. - Newly built or retrofitted stock assigned a low vulnerability anchor of 0.15 (an EU-wide constant reflecting EPBD nZEB standards for new builds; nZEB overheating is treated as a UQ sensitivity). - Growth-sensitivity controls link retrofit tempo to local demographic trends.
Use in the framework¶
In URBADAPT-HEAT v1.0, the projected SVI is used for: - Distributional stratification of mortality benefits across vulnerability quintiles. - Equity diagnostics — tracking whether adaptation benefits accrue to the most vulnerable. - Policy-targeting variants — equity-weighted GVI allocation for the street trees pathway.
The SVI is not applied as a direct multiplicative modifier of the heat-mortality impact functions in the core mortality calculation in v1.0.
Input summary¶
| Input / Parameter | Source | Notes |
|---|---|---|
| Building vintage | GHS-AGE 2025, 100 m | Epoch codes 1–10 |
| Foreign-born share | ESTAT Census 2021 OTH/T | Min-max normalised |
| Non-employment rate | ESTAT Census 2021 EMP/Y_1564 | Inverted and normalised |
| Short-run projection | DRMKC NUTS3 indicators | 2020–2030 trend anchors |
| Long-run projection | GDL / Wittgenstein SSP database | 2030–2050 national envelopes |
| Persistence k | City-specific (YAML config) | — |
| Spatial smoothing φ | City-specific (YAML config) | φ₂₀₃₀ and φ₂₀₅₀ |
| Thermal retrofit pathway | City-specific (YAML config) | Rate + new-build anchor |
This page is maintained in the URBADAPT-HEAT wiki and synced automatically. Edit it there, not in the website repository.