Skip to content
No. 024On holdDisaster early warningUpdated Nov 14, 2026

Predicting the heat that people actually face inside tin-roofed informal homes from ordinary weather forecasts, so that heat warnings fire when those homes become dangerous rather than only when the official weather station is hot

Good it would do
3 / 5
Can AI help
3 / 5
How neglected
4 / 5
Can we check it
4 / 5

Why it's on hold

Parked: the gap is real, but quick checks on the open data suggest station temperature plus a roof-type offset captures most of the predictable indoor heat, and what decides whether anyone benefits (an indoor danger threshold, an agency that would change when it warns, open indoor data from African homes) is outside compute's reach. Funded and published work (HABVIA, the Ghana hospital forecasting group, Sunny Lives) sits close by.

Why it's stuck

Heat warnings are triggered by forecasts built on weather-station readings, and stations do not measure what happens inside informal housing. Ramsay and others (iScience 2021), who placed 778 temperature and humidity loggers indoors and outdoors in 12 informal settlements in Makassar, Indonesia, found chronic heat stress conditions 'which are underestimated by weather stations'.

What we'd try

Combine many building-physics simulations of typical informal dwellings (corrugated iron roof, with and without a ceiling, shading, ventilation) with machine learning, to learn a mapping from forecast weather plus roof and neighbourhood features (some visible from satellite images) to indoor heat stress. Then turn that into warning triggers that could sit on top of an existing national forecast.

We'd look again if

Any one of: (1) a meteorological service, city heat action plan team or Red Cross/Red Crescent (e.g. the Dhaka heatwave protocol owners) says an indoor-adjusted trigger would change when they act, not only where; (2) HABVIA or IIED/Slum Dwellers International release indoor, station and housing data from African informal homes, or the Ghana group shares its 47 living rooms; (3) a health-validated indoor heat threshold for tropical homes is published. Then run the forecast-time benchmark in the dossier on the CC0 South Asia dataset with GEFSv12 reforecasts.

Source: https://bridges.monash.edu/articles/dataset/Chronic_heat_stress_in_tropical_urban_informal_settlements/12924035

Think we got it wrong?

Ruled-out problems can come back. Tell us what we missed.

Members can read the full record for this problem, check its sources and comment. Open the member view →