Are India's heat deaths being counted?

India is running hotter than the climate it was built for

India’s average temperature is now well above the baseline its cities, crops and hospitals were designed around. The annual temperature anomalyTemperature anomalyThink of a fever thermometer showing how much hotter or colder you are than normal. A temperature anomaly does the same for a place – it tells you how much today’s temperature differs from a long-term average, usually a 30-year period like 1991–2020. A positive anomaly means warmer than that reference; negative means cooler. It is not the absolute temperature, so 0.5°C above normal is a big shift for a whole country’s average, even if it sounds small. It does not tell you the highest temperature of the day or what it feels like.The whole warming story begins with anomalies. Without this measure, we cannot say whether India is heating up or just having a few hot years. The charts use anomalies to show the long-term trend clearly., measured against the 1991–2020 normal, has climbed from around –0.92°C in 1940 to +0.07°C in 2025. That half-degree swing does not sound dramatic, but a national average smooths out local extremes that kill.On a bad week a single city can sit several degrees above its own historical norm while the national line barely flickers. The chart makes the direction clear: the ten hottest years on record fall after the mid-2000s, and the line is rising. One visible reason in this data is that cumulative greenhouse emissions are pushing the whole climate system warmer, and India sits squarely in a heat-prone belt. This is the backdrop every death number has to be read against.

A warming century, one stripe per year

Sometimes a number needs no axis. This chart, the famous climate stripes view, turns each year into a coloured bar, blue for cooler than the 1991–2020 norm, red for warmer. India’s stripes start deep blue in the 1940s, then gradually bleach, then flood red in the last two decades.

There is no dispute about the trend: the planet is warming, and India is warming with it. The stripes are a visual census of a century of heat, and they make the same point as the line chart: the climate India faces today is not the climate its grandparents knew. The heat is real, and it is accelerating.

Each decade has been hotter than the one before

Smooth out the year-to-year noise by averaging into decades, and the warming becomes a staircase, not a wobble. The 1950s averaged roughly 0.5°C below the 1991–2020 baseline; by the 2010s, the average climbed to about 0.2°C above it. The latest incomplete decade is on track to be warmer still. This decade-by-decade rise matters because it shows the warming is systematic, not an accident of a few scorching years. The body of evidence, spanning surface stations, satellites and reanalysis, points in one direction: India is heating at a pace that challenges every assumption built into its housing, farm calendars and hospital load planning.

The warming is not shared evenly across the map

A single national number hides geography. This map colours each state by how much it warmed between the 1951–1980 baseline and the 2015–2024 period. The Himalaya belt, including Ladakh, Himachal and Uttarakhand, emerges as the fastest-warming region, with Ladakh approaching 1.8°C of warming in this state-level ERA5 cut.

That is still far above the all-India average. But the deadliest heat, as later charts will make clear, lands not in the sparsely populated mountains but on the crowded, humid plains. The warming map is a reminder: exposure is decided not just by degrees, but by how many people live where the thermometer climbs.

And it is not only the heat, it is the humidity

A dry 40°C is hard; a humid 40°C is dangerous. The body cools itself by sweating, and when the air is already heavy with moisture, sweat does not evaporate, the internal thermostat fails. India’s average relative humidity, pulled from the ERA5 reanalysis, has risen from about 51.5% in 1940 to 64.2% in 2024. Add rising humidity to rising temperature, and the combined load on the body, the heat index or feels-like temperature, climbs faster than the thermometer alone. This is the hidden amplifier: the heat that reaches the skin is getting wetter, and wet heat is what drives the health emergency.

Days that feel dangerously hot, now and to 2100

The heat index counts days when the temperature-humidity combination hits a dangerous threshold, here 39°C. In the 1950s, India averaged about 5 such days a year. By 2014, that had nearly tripled to 13.87 days. And the future path depends on emissions. Under the middle-road scenario (SSP2-4.5), dangerous heat-index days climb to around 57 by 2100; under the high-emissions scenario (SSP5-8.5), they exceed 132, more than a third of the year.

This is the exposure projection that underlies every excess-death model. Even the low-emissions pathway, the best-case outcome, pushes the count to 32 days. The line for the last decade of the century is terrifyingly high, but it is not a forecast; it is a warning of what happens if the world does not cut emissions.

What actually happened after 2014, where the models stop

The older model series ended in 2014, but the real heat did not pause. ERA5 reanalysis fills the gap with observed counts of hot days (max ≥ 40°C) and warm nights (min ≥ 26°C) for India. Every year from 2015 to 2025, the country saw 10 to 30 extremely hot days and a staggering 70 to 85 warm nights on average. The latest point, 2025, logged 13.1 hot days and 70.3 warm nights. Warm nights are the hidden killer, because without a cool night the body never recovers from daytime stress.

This chart shows that exposure did not pause when the projections took over, and the night-time heat vastly outnumbers the extreme hot days. The counting gap cannot be blamed on a lack of heat.

The local heat the national average hides

No one lives in a national average. This chart lines up each city’s 1940s baseline against its 2016–2025 average count of very hot days (max ≥ 35°C), then shows the change. The ranking is the point: Jodhpur, Ahmedabad, Delhi, Lucknow, Varanasi, Jaipur, Nagpur, Patna and Raipur sit in a different climate from Bengaluru or Srinagar. In the interior north and central plains, very hot days are not rare events; they are a long season. The baseline bars keep the trend visible, while the latest-decade bars show the burden people face now. This is the lived geography behind the death counts: heat is local, uneven and far more intense than the national average suggests.

And the nights are not cooling down either

After a brutal afternoon, the night is supposed to be a reprieve. This chart keeps the big metros and vulnerable-region city points separate, on the same scale, so the comparison stays honest. Chennai is the clearest metro warning: hot nights, defined here as minimum temperature at or above 28°C, rose from 22 in 1940 to 99 in 2025. Delhi remains high, Mumbai has climbed from a low base, and Bengaluru still barely registers on this threshold. In the second panel, Patna, Lucknow and Bhubaneswar show why night-time heat matters in the vulnerable east and north; Hyderabad, Raipur and Ranchi are included because they are part of the predefined regional context, not because they make the strongest lines. When the night refuses to cool, the body’s core temperature never drops, and that is when the heart and kidneys give out.

Heat is also a power bill and a grid problem

More heat means more cooling, more electricity, more strain on the grid. Cooling degree daysCooling degree daysImagine you own a building that needs air conditioning. Cooling degree days (CDD) measure how much and how often the temperature rises above a base comfort level, usually 18°C. Each degree above that base on each day counts as one cooling degree day. The more CDD, the more electricity you need to keep the building cool. It is a way to quantify cooling demand from weather alone, without looking at actual energy bills. It does not count how many air conditioners are running or how much power is used – only the climate pressure to cool.The CDD projections show that as the planet warms, the need for cooling will grow substantially, straining power grids exactly when people most need fans and ACs to survive. This is part of the vulnerability story: heat kills not only directly but also through power cuts during peak demand. (CDD), a measure of how much energy you need to cool a building, have climbed from around 4,767 in 1950 to 4,983 in 2014. By 2100, under the middle-road scenario, CDD rise to about 6,125; under high emissions, 7,695. The danger and the power cut can arrive together. A fan that cannot run is no defence, and this chart shows the grid stress is baked into the warming trajectory. When the load-shedding starts on the hottest evening, the mortality risk spikes for the poorest, who have no backup.

The future, mapped: where dangerous heat lands by 2100

These three maps show the geography of dangerous heat-index days per year by the 2080–2099 average, under low, middle and high emissions. They share one colour scale. In the high-emissions map, much of the hot, humid plains and coasts spend up to about 200 days a year, more than half the year, in conditions that feel dangerous. The low-emissions map is far lighter, but even there the Gangetic plain and the west coast stay burdened.

This is the future the counting systems must prepare for. The maps are CMIP6 model projections, not forecasts, and they carry the standard caveats, but the message is clear: how dangerous it gets is still a choice, and the range spans from manageable to catastrophic.

Most of India’s workers are out in the heat

Heat exposure is decided at work, and India’s job structure puts most workers in the sun with no way to stop. By the Periodic Labour Force Survey (2023–24), about 43.5% of workers are in agriculture, open fields under the sky, and another 24.9% are in industry, which includes open-air construction and brick kilns. Only 31.6% are in mostly-indoor services. For the majority, a heatwave is not something you can wait out at home.

Most workers cannot simply stay home

The next constraint is not the thermometer but the pay packet. PLFS says 58.4% of Indian workers are self-employed and 19.8% are casual labourers; only 21.7% have regular wage or salaried jobs. ILOSTAT puts informal employment at 87.2% overall and 98.6% in agriculture.

These bars should not be added, because worker status and informality overlap, but they point in the same direction: most workers have little formal protection between heat and hunger. A heat warning asks them to slow down. The labour market tells them they may not be paid if they do.

Heat is already costing work hours

That inability to stop already has a measurable cost. Lancet Countdown's India 2025 data sheet estimates that heat exposure cost India 247 billion potential labour hours in 2024, equal to 419 hours per person and 124% more than the 1990-1999 average. The sector split is blunt: agriculture accounts for 66% of heat-related labour-hour losses, construction for another 20%, and all other sectors together for the remaining 14%.

These are labour-capacity estimates, not a count of deaths, observed absences or actual wages lost. But they explain why the official mortality record can look small while the lived damage is large. Heat first shows up as slower work, shorter safe workdays, lost income and exhausted bodies, especially in the same outdoor sectors where stopping work is hardest.

Electricity reached the home. Cooling did not.

Electricity is the first step, not the finish line. NFHS-6 says 98.3% of Indians live in households with electricity, and NFHS-5 says 88.3% of households own an electric fan. But a fan moves air; it does not make a dangerous room safe during humid heat or a power cut. The cooling ladder drops fast after that: NFHS-5’s combined air-conditioner/cooler category reaches 23.7% of households, while NSS 78 separates the appliances and finds 14.1% with an air cooler and just 4.9% with an air conditioner. The urban-rural split is sharper: AC ownership is 12.6% urban and 1.2% rural; cooler ownership is 21.9% urban and 10.4% rural.

India has mostly wired homes. It has not yet cooled them.

The cooling map does not match the heat map

Where India keeps air coolers is almost the opposite of where humid heat is most dangerous. This map shows state-wise air-cooler ownership: Punjab, Haryana, Rajasthan, Delhi and Chandigarh all cluster near or above 40–50%. These are dry-heat states, where evaporative coolers work. The humid south and east, Kerala, Tamil Nadu, West Bengal, Odisha, own almost none, because coolers barely work in damp air.

Yet these are precisely the regions where wet-bulb heat is deadliest, as the heat risk index later confirms. The tools people have for cooling are mismatched to the climate danger they face.

And who can least afford to cope: India’s poverty map

Heat is only as deadly as your ability to escape it. About 15% of Indians are multidimensionally poor, meaning they suffer deprivation in housing, cooking fuel, assets or health. This map, based on NITI Aayog’s National MPI 2023 (NFHS-5, 2019–21), shows poverty concentrated in the east and centre: Bihar 34%, Jharkhand 29%, Uttar Pradesh 23%, Madhya Pradesh 21%, while Kerala (0.55%) and Tamil Nadu (2.2%) are far lower. This poverty belt overlaps the humid, crowded plains where heat exposure is rising. A tin-roof home with no reliable power and an undernourished body turns a hot week into a fatal one.

Where monthly spending room is thinnest

The poverty map says who is deprived. HCES adds the rupee scale of the cushion they have. In 2023-24, rural monthly per capita consumption expenditure was lowest in Chhattisgarh at Rs 2,739, followed by Jharkhand at Rs 2,946, Odisha at Rs 3,357, Madhya Pradesh at Rs 3,441 and Uttar Pradesh at Rs 3,481. Bihar was Rs 3,670, while the all-India rural average was Rs 4,122. MPCE is consumption expenditure, not income, savings or money available for cooling.

It also excludes imputed value of free welfare items. Still, the point is hard to miss: in many heat-vulnerable states, the monthly spending room is thin before a heatwave arrives. Buying an AC, running a cooler, paying for transport to care, or losing a workday all demand cash that many rural households simply do not have.

Heat hits a body that is often already weakened

Heat does not kill at random. It finds bodies already under strain, and India’s are: by NFHS-5 (2019–21), 67.1% of young children, 57% of women and 25% of men are anaemic. Anaemia reduces the blood’s oxygen-carrying capacity and lowers heat tolerance. Widespread high blood pressure, diabetes and kidney disease add to the vulnerability. An ageing population, with the share over 65 at 7.15% in 2024, faces declining thermoregulation.

The same heatwave that a healthy adult shrugs off can be fatal for those with these underlying conditions. The death toll is not a random sample; it is a harvest of the already sick.

The deadliest heat is humid, not just hot

When CEEW folded heat, humidity, exposure and vulnerability into one risk index for all 734 districts, the result was a contrarian map. The highest risk states are not the dry record-breaking north, but the humid south and coasts: Kerala, Goa, coastal Andhra Pradesh, all of Maharashtra have 100% of districts in the high or very high risk category. That 100 is not a risk score; it means every district in that state falls into CEEW's high or very high category. The dry northern states, for all their scorching days, have lower shares because the risk from dry heat is mitigatable; wet heat stops the body cooling itself. 57% of all districts, home to 76% of Indians, sit in the high or very high band. This is the synthesis: the deadliest heat is humid, and it lands where the most people live, many of them poor and unable to cool down.

Where heat is hardest to survive

The most vulnerable places are not simply the hottest. They are the places where high heat risk meets weak protection. This scatter combines CEEW’s state share of districts at high or very high heat risk with NSS cooling ownership, NITI’s multidimensional povertyMultidimensional povertyPoverty is not just about low income. Multidimensional poverty captures deprivations in health, education and living standards at the same time. Someone might be poor in money terms but own a house; someone else might have an income but no sanitation or clean cooking fuel. The MPI headcount ratio tells you the share of people who suffer multiple overlapping deprivations. In India, about 15% are multidimensionally poor. This is a strong lens for vulnerability because it identifies those who cannot afford a buffer against shocks like extreme heat.The poverty map shows that the regions with the highest heat risk also have the deepest poverty. It explains why a heatwave turns deadly: the people exposed are the ones least able to afford cooling, decent housing or medical care. and HCES consumption context in the tooltip and CSV. States to the right have more districts in the high-risk band. States higher up have more households without a conservative AC/cooler protection proxy.

Bigger bubbles are poorer. That puts states such as Bihar and Uttar Pradesh in a different light: they may not top every humidity map, but they combine large vulnerable populations, limited cooling and high heat-risk exposure. Kerala and Maharashtra show another pattern: very high heat risk, but lower poverty and different cooling profiles. The point is not to rank deaths. It is to show why vulnerability is regional, layered and uneven. Heat kills where exposure, poverty, thin spending room, housing, health and weak cooling meet.

So how many die? It depends entirely who you ask

Here is the heart of the confusion. The viral 3,400 is a modelled estimate of extra deaths on a nationally extreme heat day. The official numbers are head-counts of deaths formally labelled as heatstroke or heatwave. They are not rival versions of one fact. The comparison chart shows seven sources: IMD DWE (460 for 2024), NCRB ADSI (804 for 2023), OWID/EM-DAT (733 for 2024, and 10,398 over 2000–2024), NCDC surveillance (161 confirmed, or 374 by a different cutoff), and the Frontiers model (3,400 for one extreme day, ~30,000 for a five-day heatwave).

The spread is enormous because each source measures a different thing. No single number can be declared the truth.

Most deaths are registered. Most causes are not medically certified.

The missing piece is not only whether a death is registered. It is whether anyone can say, medically, why the person died. On that first test, India now looks strong: in 2023, the Civil Registration System registered 86.6 lakh deaths, a death-registration level of 97.2%.

But the record thins out after that. Only 24.0% of registered deaths occurred in institutions. CRS says 53.4% had no medical attention at the time of death. MCCD, the system that records medically certified causes, covered 19.0 lakh deaths, just 22.0% of registered deaths.

That is the gap heat falls through. Heat often does not arrive on a certificate as 'heat'. It pushes a weak heart, a damaged kidney, a lung condition or a dehydrated body past the edge. Without a doctor certifying the chain, the death may be counted, but the heat in it disappears.

Where causes of death are least visible

The blindness is geographical. MCCD coverage is high in a handful of small, urbanised or better-certified places: Goa reports 100.0% of registered deaths medically certified, Delhi 66.0%, Maharashtra 42.4%, Tamil Nadu 39.1%. Then the floor drops out in several large states where heat vulnerability is already a serious concern: Kerala 11.4%, Jharkhand 11.2%, Madhya Pradesh 10.1%, Uttar Pradesh 5.8%, Assam 5.6%, Bihar 5.5%.

That does not mean these states have fewer heat-related deaths. It means the official record is less likely to know the medical cause of death at all. This matters because heat mortality is rarely a clean label. It is often a trigger hidden inside cardiac, kidney, respiratory or dehydration deaths. Where certification is thin, the undercount is built into the map before any heat estimate even begins.

Where India officially counted its heat deaths in 2024

The IMD’s Disastrous Weather Events report for 2024 lists 460 heatwave deaths. This bar chart shows them by state. Uttar Pradesh alone reported 240, more than half. Bihar added 63, Telangana 46, Jharkhand 35, Maharashtra 28, Odisha 16, Rajasthan 16, Chhattisgarh 7, Madhya Pradesh 5, Kerala 2. The pattern matches the hot plains and the poverty map.

But the total, 460, is dwarfed by any excess-death model. This is the official count, from the meteorological authority, and it is the figure most often cited in policy. It is almost certainly an undercount, but it does tell us where the counting system is at least recording some deaths.

Heat kills in a tight, predictable window

The official deaths cluster in May and June, the pre-monsoon furnace. The IMD’s monthly breakdown for 2024 shows 1 death in March, 39 in April, 185 in May, 235 in June, and zero the rest of the year. This predictable cycle is why heat-related deaths are among the most preventable disasters. If the death window is known, the interventions, public alerts, cool shelters, water stations, hospital readiness, can be timed precisely. The fact that deaths still occur in that tight window, despite heat action plans, suggests the current measures are not reaching the most vulnerable.

In 2015 India counted 2,000 heat deaths. In 2021, zero.

This single line is the strongest proof that India does not really know its heat toll. IMD’s annual reported heatwave deaths from 2013 to 2024 swing wildly: 1,400 (2013), 500 (2014), over 2,000 (2015), then a collapse to just 27 in 2020, exactly zero in 2021, and 460 in 2024. The sun did not take a year off. What changed was attention and bookkeeping. After the catastrophic 2015 heatwave, states launched Heat Action Plans that genuinely saved lives, but the recorded number also collapsed because systematic counting was not maintained.

The official figure measures how hard India is looking, not how many die. It can fall even in years when the heat is rising.

Why the disaster databases always look too small

International disaster databases, like EM-DAT compiled by OWID, catch the big, named heatwaves and miss the slow burn. For India, reported extreme-temperature disaster deaths since 2000 show spikes in 2015 (2,248) and 1998 (2,541), with many years of zero or near-zero. The cumulative total is 10,398 over 2000–2024. This spiky pattern is typical: the database relies on media reports, government submissions and disaster declarations, not daily mortality surveillance. It is a record of recognised events, not a measure of the true toll, and it almost certainly undercounts heat’s real impact, especially in years without a headline-grabbing heatwave.

How a single hot day can plausibly reach four figures

This is the test of the viral 3,400 number. India’s 2024 population is about 1.45 billion, and the crude death rateCrude death rateFor every 1,000 people alive, the crude death rate tells you how many die in a year. In India, it is now about 6.6 per 1,000. If you had a village of 1,000, that would mean roughly 6–7 deaths a year from all causes combined. It is ‘crude’ because it doesn’t adjust for age; a country with many old people will have a higher rate even if healthcare is excellent. Here it is used as the baseline: on a normal day, across India’s 1.45 billion people, around 26,000 people die. A heatwave nudges that number up.The sensitivity check that shows thousands of excess deaths is plausible starts from this crude death rate. Without this denominator, it is impossible to gauge whether 3,400 extra deaths is huge or tiny relative to the normal deaths each day. is 6.6 per 1,000, yielding roughly 26,236 deaths on a normal day. The sensitivity chart asks: if 20% of the population is exposed to extreme heat, and the death rate among that exposed group rises temporarily by just 3%, that yields 157 excess deathsExcess deathsExcess deaths are the number of deaths above what would normally be expected in a given place and season. Think of a town where about 10 people die every day in June. If one brutal heatwave day sees 15 deaths, the excess is 5. It is a modelled population estimate, not a pile of death certificates. It does not tell you the cause – it might include heart attacks triggered by heat, but also other things. It is fundamentally different from a reported heatstroke death, which requires a specific diagnosis.The viral 3,400 number is an excess-death estimate. Understanding what that means, and how it differs from official reported deaths, is the core of this page. Without this distinction, the numbers appear contradictory when they are actually measuring different things. in a day. Scale up the exposure share to 75% and the mortality lift to 15%, and you get about 2,952 excess deaths. The Frontiers model’s implied combination, a 70% exposed share and an 18.5% lift, reproduces the 3,400 figure.

The arithmetic shows thousands is plausible, not proven. A real attribution would need daily all-cause deaths by district, age and season, which India does not publish.

The number any honest estimate has to start from

Every excess-death claim begins with the baseline: how many people would have died anyway? India’s crude death rate has fallen from 19.4 per 1,000 in 1960 to 6.6 in 2024, a massive public health success. But a low death rate with a huge population still produces enormous absolute numbers.

This line chart tracks that rate over time and is the denominator for the sensitivity check. Get this baseline wrong, and the whole heat-mortality number is wrong. The data comes from the World Bank, and while it is not a precise daily series, it is the best available national anchor. The caveat: a crude death rate hides age and seasonal patterns that are essential for accurate excess-death modelling.