36
heat values per polygon, per day
Acknowledgements
Built and maintained by CrisisReady, with support from the Harvard Data Science Initiative and the Harvard Salata Institute for Climate and Sustainability.
HeatReady turns the global weather record (hourly reanalysis built from satellite, station, and climate-model data, the same record the IPCC and WMO work from) and satellite imagery into daily heat metrics for any boundary a project draws: a district in Dhaka, a county in Arizona, a farming village in northern Ghana, a census block in Paris. A new project is ready in about half an hour and updates itself every morning.
Loading the reveal…
Most heat data comes on a 9 to 28 km grid, like this ERA5 view of the Northeast. Useful at the scale of a region.
ERA5-Land reanalysis (Copernicus), uncorrected 0.1° grid, daily peak heat index · Basemap © Esri
Boston and Cambridge, New York, Paris, and Barcelona, mapped neighborhood by neighborhood for the same days of summer 2026. By afternoon temperature, by heat index, and by nights that never cooled, the four cities fall in a different order each time.
Every map value is a served HeatReady value over the 91 days all four cities share. A night without relief is one whose low never fell below 20°C (68°F), the WMO tropical-night convention. Death tolls are named with their source where they appear and are not directly comparable. Methodology on the Evidence page →
Every polygon, every day, gets its value from the best validated source available. We try four methods, in order. A weather station inside the boundary comes first. Then the model correction, refined with nearby stations. Then the model correction on its own. If none of these has passed validation for that climate zone, we serve the raw ERA5 grid value.
A real station observation inside the polygon, overriding every modeled value.
The model correction refined against nearby real weather stations.
The model's correction, applied only where it has passed validation for that zone.
The raw reanalysis grid value, served when no validated correction applies.
Full per-zone validation results are on the Evidence page; what each field means is in the API docs.
The global model will serve the best validated data by climate zone, however, they necessarily wont overfit to any specific city. In Seoul, we were able to leverage a municipal network of 756 sensors to further enrich the global model. We then tested each data source against a held out selection of neighborhood in the city. Against the raw ERA5 grid, HeatReady’s global model cut the error by almost half. A model trained on Seoul’s own sensors cut it by two thirds. Cities don’t need to have a dense sensor system like Seoul. Even a handful of local sensor systems over time can help improve local performance. However, even without this local enrichment, if a project serves the global model, it means that it outperforms ERA5.
Email us to request API credentials. The account is set up and the first project walked through the same day. Journalists and researchers reach the team at the same address.
Built and maintained by CrisisReady, with support from the Harvard Data Science Initiative and the Harvard Salata Institute for Climate and Sustainability.