Across 42 cities, the difference between nearby neighbourhoods frequently exceeded the average warming projected for the entire metropolitan region.
The pattern is consistent: dense surfaces and limited canopy raise exposure, but the largest harms appear where those conditions overlap with poor housing quality, long outdoor commutes and low access to cooling.
Higher risk of dangerous indoor heat where low canopy and low-quality housing occur together.
This changes the policy unit. A citywide alert is necessary, but it does not identify who can act on it, who can reach a cooling centre, or which buildings remain unsafe after sunset.
EXPOSURE
Exposure accumulates across systems.
Heat is commonly reported as a meteorological event. For residents, it is a chain of infrastructure conditions: whether the home releases heat overnight, whether the journey to work is shaded, and whether healthcare and power remain reliable under pressure.
Night-time indoor temperatures remained elevated for up to six hours after outdoor conditions improved.
Building performance and household resources mediate the health impact of the same external temperature.
Comparable indoor sensor data remains sparse in low-income neighbourhoods and informal settlements.
Design policy around the person, not the average.
The strongest programmes combine immediate operations with capital planning. They use block-level thresholds for outreach, treat cooling access as a transport problem, and connect housing retrofit data to public-health response.
- Trigger locally. Use neighbourhood thresholds and night-time persistence—not only a citywide maximum.
- Coordinate services. Align health, transport, power and housing teams around the same operating picture.
- Publish performance. Show which interventions reduced exposure, for whom, and under which conditions.
Measuring neighbourhood heat vulnerability with open urban data.
Methods, code, datasets, peer review and version history are available in the linked record.
Open research recordEvidence trail.
Equors separates reported facts, editorial interpretation and unresolved limitations. This article was checked against linked datasets, public methods and expert review notes.