Scientific Methodology and Model Documentation

Temperature baselines are derived from the ERA5 reanalysis product published by the Copernicus Climate Change Service (C3S) at ECMWF (Hersbach et al. 2020). The reference climatology is computed over the 1991–2020 standard normal period, accessed via the Open-Meteo Historical Weather API.

Future projections use a two-model CMIP6 ensemble: MRI-AGCM3-2-S and MPI-ESM1-2-XR, accessed via the Open-Meteo Climate API. Projections are strictly capped at 2050 — the validated horizon for these CMIP6 model outputs. No post-2050 extrapolation is performed.

The primary climate variable is the annual maximum five-day consecutive temperature mean (TX5d) at 31 km spatial resolution. Relative humidity uses the daily mean (relative_humidity_2m_mean) from the ERA5 ensemble.

The mortality model follows the dose-response methodology of Gasparrini et al. (2017, Lancet Planetary Health), who pooled 395 studies across 197 countries to derive a chronic temperature–mortality coefficient.

The relative risk of mortality at temperature excess $\Delta T$ above the local historical threshold is:

The attributable fraction (AF) — the fraction of deaths attributable to heat on heatwave days — is:

Total attributable deaths per year:

Uncertainty: mortality estimates carry a ±15% confidence interval, reflecting parameter uncertainty in β, population data quality, and CMIP6 ensemble spread. The Gasparrini coefficient is most accurate for sustained exposure events (12–44 days); it systematically undershoots acute events of fewer than 9 days.

The economic model uses a hybrid bipartite approach combining the Burke quadratic GDP-temperature relationship with ILO labor productivity estimates for extreme-heat days.

The Burke et al. (2018, Nature) quadratic specifies a global optimum temperature of approximately 13°C, above which GDP growth is reduced:

For days exceeding the 34°C labor-stress threshold, an additional ILO (2019) shock is applied: 40% of the workforce at 20% reduced productivity per heatwave day. The two components are summed to produce total annual GDP-at-risk.

Economic estimates carry a ±8% confidence interval, reflecting GDP data quality and model uncertainty.

Wet-bulb temperature (WBT) is the physiological survivability limit metric. The empirical formula of Stull (2011, J. Applied Meteorology and Climatology):

WBT is capped at 35°C per the survivability limit established by Sherwood & Huber (2010, PNAS), who showed that sustained WBT above 35°C is incompatible with human thermoregulation regardless of acclimatisation.

The canopy expansion and cool-roof albedo scenarios are directional illustrations only. They are not calibrated intervention models. Canopy offset reduces the effective TX5d by approximately 0.8°C per 10% canopy increase in urban areas, based on the urban heat island meta-analysis of Bowler et al. (2010) and Santamouris (2015).

Cool-roof albedo offset applies a surface energy balance reduction of approximately 0.4°C per 10% albedo increase. These offsets feed back through the full mortality and economic model chain.

In highly humid coastal biomes (wet-bulb temperature baseline above 28°C), canopy expansion may theoretically trap surface humidity and increase wet-bulb exposure. This edge case is flagged automatically.

Population, GDP, death rate, and healthcare access data are sourced from the OpenPlanet Verified City Vault — hardcoded 2023–2024 values for 59 major cities drawn from UN Population Division estimates, World Bank national accounts, and most recent national census publications.

A census validator rejects any population value outside the range [10,000 – 35,000,000]. For cities not in the Vault, live Open-Meteo Geocoding results are combined with World Bank country-level statistics; a critical alert is logged for anomalous values.

All projections are research-grade estimates under specific emissions scenarios (SSP2-4.5 or SSP5-8.5). They are directional indicators for planning and analysis — not deterministic forecasts and not investment advice.

Quantified uncertainties: mortality ±15% CI, economics ±8% CI. Unquantified uncertainty sources include: CMIP6 model spread (two-model ensemble), downscaling bias at 31 km resolution, urban-rural temperature gradients, and future adaptation behaviour.

The Gasparrini β coefficient is derived from historical exposure data and may not fully represent future adaptation responses. Economic estimates apply a global coefficient to individual cities; local economic structure is not captured.