Heat
Why warm nights matter more
A body recovers from heat overnight. When the night does not cool, it does not recover, and consecutive hot nights compound into the physiological state that kills people. Across the fourteen stations this site reads over thirty-five years, the count of nights that never fall below 28°C is the measure that moves upward most consistently — more consistently than the hottest afternoon, which is close to flat at most of them.
Recovery is the thing being lost
Heat illness is cumulative. The daytime maximum sets the load; the night decides whether it clears.
Through a hot day the body sheds heat by sweating and by moving blood to the skin. Both have costs — fluid, salt, cardiac work — and both are paid back at night, when air temperature falls below skin temperature and the body can lose heat passively. A night that stays warm removes that repayment.
This is why heat mortality tracks multi-day episodes rather than single peaks, and why it concentrates in housing that holds heat: a tin roof, a single room, no cross-ventilation. The building keeps releasing the day's heat long after the air outside has cooled, so indoor night temperature can exceed outdoor night temperature for hours.
It is also why the daytime criterion misses so much. IMD's heatwave rule is built on the maximum, and IMD handles night temperature under a separate warm-night criterion that is not what a declared heatwave refers to.
- Warm night
- IMD defines one as a departure of the minimum temperature above its station normal. The 28°C figure used on this site is a different thing — see below.
- Diurnal range
- The gap between the day's maximum and the night's minimum. A narrowing range is a warming night, not a cooling day.
- Apparent temperature
- What the heat does to a body once humidity is accounted for. Rising at more of these stations than the dry-bulb maximum is.
What moves, across fourteen stations and thirty-five years
How many stations move each way between the first and second halves of the record. Published as a count of stations, because averaging fourteen grid points does not produce a national trend.
Stations reading
14 of 14
ERA5 reanalysis, 1991 onward
Nights above 28°C — stations trending up
9 stations
ERA5, first half vs second half
— trending down
3 stations
ERA5, first half vs second half
Peak daytime maximum — stations trending up
6 stations
ERA5, first half vs second half
— trending down
6 stations
ERA5, first half vs second half
Peak apparent temperature — stations trending up
8 stations
ERA5, first half vs second half
There is no legal limit on a night temperature
None
No statute sets one
Read the first pair against the second. The hottest afternoon is split almost evenly between stations getting hotter and stations getting cooler; the warm-night count is not. That divergence is the finding, and it is the reason a heat policy written around the daytime maximum can be met while the danger increases.
The 28°C threshold is this site's own choice and not IMD's. IMD's warm-night criterion is a departure from each station's normal minimum, which is the right definition for a declaration and a poor one for a thirty-five-year comparison, because the normal itself is a moving reference. A fixed absolute threshold is comparable across the whole record by construction. It is stated here rather than left implied.
How to read it
Three things this series is, and one thing it is not.
It is modelled. ERA5 is a reanalysis, not a station thermometer, and every figure here is marked accordingly. A reanalysis also smooths local extremes hardest at exactly the tail this metric lives in.
It contains no city. A grid point carries no urban heat island, and the cities on this list have in many cases doubled their built area over the record. Where the real night temperature has risen fastest is precisely where this method sees least.
Fourteen points are not a country. The counts above are counts of stations moving, deliberately, rather than a national average that would imply a coverage this does not have.
What it is not is a reason to relax about daytime heat. Three of the four things that make heat dangerous — humidity, night-time recovery and duration — are simply not dry-bulb maximum, and a flat maximum is evidence about one of the four.
What this number cannot tell you
- It cannot tell you who is exposed. Housing, occupation and access to cooling decide the outcome, and none is in a temperature series.
- It cannot be compared with IMD's own warm-night declarations, which use a departure from normal rather than a fixed threshold.
- It cannot resolve a single bad night. It is a count of nights per season.
- It carries no mortality. India's heat death figures come from a different series and are widely accepted to be undercounts.
Delhi, first half against second
Delhi's own record shows the pattern in miniature. Between 1991–2008 and 2009–2026 its mean count of heatwave days fell slightly and its mean peak maximum fell slightly — while its mean count of nights above 28°C rose. Neither movement is large on its own; the direction of the difference between them is the point.
Where this comes from
Every figure above is read out of one of these, at the address printed beside it.
- ERA5 reanalysis archive ECMWF, served by Open-Meteo · Thirty-five years of daily maxima, minima and apparent temperature at fourteen stations
- Heat wave and warm night criteria India Meteorological Department · IMD's own definitions, which differ from the fixed threshold used here
- Accidental Deaths & Suicides in India National Crime Records Bureau · India's official series for deaths attributed to heat
Reuse freely — CC BY 4.0. The grant covers this page. Each source above keeps its own terms, which is why every one of them is named.
Next
The reading this explains, the explanations next to it, and the one place to stand in it.