How the index is calculated

Surface is Landsat land surface temperature. Thermal resolution is about 100 metres. The supporting bands are 30 metres.

The month is a median composite of the usable passes. It is not thirty daily readings, and it is not MODIS, Sentinel-3 or ECOSTRESS unless a later cut says so.

Pixels are clipped to urban land in OpenStreetMap: residential, commercial, industrial, retail. Rice, lagoon or market garden that remains inside a polygon after that clip still enters the score.

Air, humidity, rain and wind come from that country’s official station network: the national meteorological service.

Each zone takes the nearest station that has that variable. There is no interpolated humidity or rain surface.

Station names and identifiers sit on the city cut. They do not define the product.

For a published hot-month cut the score is: z(LST) + 0.4 × z(air) + 0.3 × z(humidity) − 0.15 × z(rain). Each z is the value standardized across the zones of that same cut. Lower is more comfortable.

Land surface temperature dominates because its weight is 1. Humidity and rain only move a zone when that zone uses a different station from its neighbours.

A January cut with 5°C of surface is not scored with this comfort rule. No cold-month cut is published yet.

Neighbourhood polygons and metro municipalities are two cuts. They never share a table or a URL.

A city enters the product when it has neighbourhood polygons, at least one official station with data in that month, and usable Landsat.