Neighbourhood vulnerability is an interaction, not a temperature.
We construct and validate an open, reproducible heat-vulnerability index that combines surface temperature, night-time persistence, housing, mobility, canopy and service access.
The index is evaluated across 12.4 million urban blocks in 42 metropolitan areas. It improves identification of high-risk neighbourhoods compared with temperature-only approaches and remains directionally stable across six climate zones.
All processing steps, source transformations and sensitivity tests are published as a versioned data package.
Three results change the operating picture.
Higher risk when low canopy overlaps with poor housing performance.
Of high-risk blocks are missed by citywide temperature thresholds alone.
Directional stability across alternative index weightings.
A transparent, versioned measurement chain.
Variables were selected through literature review, expert consultation and preregistered sensitivity criteria. Each source is transformed to a common block-level grid and normalised within climate zones.
Cross-sectional comparative geospatial analysis with out-of-sample validation.
Urban block, aggregated to neighbourhood for public reporting.
12.4 million blocks across 42 metropolitan areas.
Emergency health utilisation, reported indoor exposure and field sensor subsets.
1,000 weighting combinations plus source-removal tests.
View preregistered analysis deviations
Two source layers were replaced after coverage checks. Both substitutions were documented before outcome validation and did not alter primary conclusions.
Reproduce the record.
Equors heat vulnerability package v2.1
Processing scripts, source manifest, derived block index and validation notebook.