Are India's heat deaths being counted?

Blue years have vanished, replaced by deep red stripes over the last two decades.

This chart strips away numbers and shows each year since 1901 as a coloured stripe: blue for cooler than the 1991-2020 average, red for warmer. The early 20th century is mostly blue, the mid-century mixed, and then from the 1990s onward, a relentless shift to dark red. It is the same data as the line chart, but rendered as a visual story that needs no explanation. The stripes make it impossible to miss that the baseline has shifted, and recent years are outside the historical range.

The Himalaya belt has warmed fastest in this state-level ERA5 cut, with Ladakh approaching 1.8°C above the 1951-1980 baseline.

This choropleth map colours each Indian state by its warming between the 1951-1980 baseline and the 2015-2024 period. The darkest reds appear in the northern mountains: Ladakh, Himachal Pradesh, Uttarakhand. The rest of the country shows warming of 1-1.5°C. The map highlights that a national average of roughly 0.7-0.8°C warming hides dramatic regional differences. While the mountains warm fastest in degrees, the most dangerous heat stress occurs in the humid, populous plains. The map uses ERA5 reanalysis data.

Dangerous heat-index days tripled from 5 to 14 per year by 2014, and under high emissions could reach 133 by 2100.

This chart combines observed data (1950-2014) with three climate model projections (SSP1-2.6, SSP2-4.5, SSP5-8.5) to show the annual number of days when the heat index exceeds 39°C. The observed line climbs from 5.22 in 1950 to 13.87 in 2014. Projections diverge sharply: under the low-emission scenario, days rise to 32 by 2100; middle-road, 57; high-emission, 133. These are not predictions of what will happen, but what could happen under each emissions pathway. The gap between the scenarios shows how much of the future is still a choice.

Since 2015, India has averaged 70-85 warm nights and 10-30 hot days each year, showing exposure didn’t pause.

This chart uses ERA5 reanalysis to fill the observation gap from 2015 to 2025, plotting both the number of hot days (max ≥ 40°C) and warm nights (min ≥ 26°C). Hot days range from 19.8 (2015) to 13.1 (2025), while warm nights stay high: 73.2 (2015) to 70.3 (2025). The warm night count dwarfs the hot day count, underlining the relentless night-time heat. The data show that the recent decade, post-model era, matches or exceeds the exposure trend earlier. This is the actual lived experience that any death count must reflect.

Interior north and central cities now see a long season of very hot days, while Bengaluru and Srinagar remain near the floor.

Each row compares a city’s average number of very hot days in the 1940s with its 2016–2025 average. The chart is sorted by the latest-decade burden, so the hottest city climates rise to the top. Jodhpur, Delhi, Lucknow, Varanasi, Nagpur, Patna and Ahmedabad sit high because they now see roughly a season of days at or above 35°C. The change column makes trend visible without asking the reader to decode 18 tiny lines. Coastal cities can look lower on this dry-temperature threshold because humidity, not just maximum temperature, drives their danger.

Hot nights are rising in several city climates, with Chennai near 99 nights in 2025 and Patna, Lucknow and Bhubaneswar showing the regional burden beyond the metros.

This two-panel chart tracks annual hot nights, defined as nights when the minimum temperature stays at or above 28°C. The first panel shows major metros: Delhi, Mumbai, Kolkata, Chennai and Bengaluru. The second shows predefined city points from vulnerability-context states: Patna, Lucknow, Bhubaneswar, Hyderabad, Raipur and Ranchi. Both panels use the same scale. Chennai is the sharpest metro signal, rising from 22 hot nights in 1940 to 99 in 2025, while Patna, Lucknow and Bhubaneswar show sustained night-time heat beyond the megacity frame. Hyderabad and Ranchi are kept in because the selection rule is regional, not value-hunting.

Under high emissions, large parts of the Gangetic plain, west coast and east coast could see 150-200 dangerous heat-index days per year by the 2080-2099 period.

Three India maps show the number of dangerous heat-index days per year under SSP1-2.6, SSP2-4.5 and SSP5-8.5 for the 2080-2099 climatology. The colour scale is shared, so darkening from left to right means more danger. Under low emissions, most states see under 50 days. Under high emissions, the plains and coasts turn deep red, with some areas approaching 200 days, more than half the year. These are CMIP6 model projections, not forecasts, and geographic boundaries have coarseness, but the relative pattern is robust. The map is a stark visualization of the emissions choice.

About 43.5% of workers are in agriculture, and another 24.9% in industry, much of it outdoors.

This line chart tracks the share of workers in agriculture, industry and services from 2017-18 to 2023-24, based on the Periodic Labour Force Survey. Agriculture remains the largest sector, rising slightly to 43.5%. Industry holds steady at 24.9%, and services dip to 31.6%. Both agriculture and much of industry (construction, manufacturing) involve outdoor work with direct sun exposure. The chart underscores that for most Indian workers, a heatwave day is not a day off, but a day of heightened physical strain. The informal nature of these jobs compounds the risk.

Almost every Indian lives in a home with electricity, but only 4.9% of households own an air conditioner.

The bars move from access to actual relief. NFHS-6 puts electricity in households at 98.3% of the population, and NFHS-5 puts electric fan ownership at 88.3% of households. But a fan is not cooling when the room is dangerously hot or humid. NFHS-5 reports 23.7% of households with an air conditioner or cooler, and NSS 78 separates that into 14.1% with an air cooler and only 4.9% with an air conditioner. The rural gap is severe: AC ownership is 1.2% rural versus 12.6% urban. The chart makes the vulnerability concrete: the wires arrived much faster than real cooling.

Air-cooler ownership is highest in the dry north (e.g., Punjab ~50%) and lowest in the humid south and east, where they are least effective.

This choropleth shows the percentage of households owning an air cooler by state. The darkest colours are in Punjab, Haryana, Rajasthan, Delhi and Chandigarh, where ownership often exceeds 40-50%. The southern and eastern states barely register, because evaporative coolers work poorly in humid air. Yet the heat risk map (next) shows the humid regions are at highest composite risk. The mismatch means the most vulnerable regions lack even the cheap cooling technology that dry areas use.

Bihar has 34% multidimensionally poor, Jharkhand 29%, Uttar Pradesh 23%, all high-heat-exposure states.

This map shows the share of multidimensionally poor people by state, based on NITI Aayog’s National MPI 2023. The highest poverty rates are in the eastern and central states: Bihar 34%, Jharkhand 29%, Uttar Pradesh 23%, Madhya Pradesh 21%. These states lie in the hot plains and have high population densities. Poverty means inadequate housing, poor health, and no financial cushion to buy cooling or reduce work hours. The map aligns almost perfectly with the regions facing rising heat exposure, creating a deadly synergy.

In Kerala, Goa and Andhra Pradesh, 100% of districts are rated high or very high heat risk by CEEW.

This choropleth shows the share of districts in each state that fall into the high or very high category of the CEEW Heat Risk Index 2025. A value of 100% means all districts in that state are in those two categories; a value of 0% means none are. The darkest states are Kerala, Goa, Maharashtra, Andhra Pradesh and Tamil Nadu, all at or near 100%. In contrast, several Himalayan and northeastern states sit near zero on this district-share measure. The point is the humid-risk geography, not a fine-grained league table: 57% of all districts, home to 76% of Indians, are at high or very high risk.

The worst vulnerability is where high heat risk overlaps with little cooling and high poverty.

This scatter combines four layers: CEEW heat risk, NSS cooling ownership, NITI multidimensional poverty and HCES consumption context in the tooltip and CSV. Right means a larger share of districts is rated high or very high heat risk. Up means a larger share of households lacks the conservative AC/cooler protection proxy. Bigger bubbles mean a higher share of people are multidimensionally poor. The upper-right quadrant is the danger zone: places where heat is widespread, cooling is scarce and poverty limits the ability to adapt. It is a context map, not a death forecast, but it makes clear why vulnerability cannot be read from temperature alone.

Estimates range from 161 confirmed heatstroke deaths (NCDC) to 3,400 modelled excess deaths (Frontiers) for a single extreme-heat scenario.

This horizontal bar chart, on a log scale, displays seven different death counts from various sources and methods: IMD DWE 2024 (460), NCRB ADSI 2023 (804), OWID/EM-DAT 2024 (733), NCDC surveillance (161), NCDC parliamentary report (374), Frontiers model single day (3,400), Frontiers model five-day (30,000). The bars are coloured by evidence type: reported, modelled, surveillance. The huge spread occurs because each counts something different, cause-specific reported deaths, disaster deaths, or statistical excess. They cannot be ranked against each other; they must be read as complementary but not interchangeable pieces of evidence.

India now registers nearly all deaths, but only about one in five registered deaths gets a medically certified cause.

The bars separate two things that often get blurred: recording that someone died, and knowing medically why they died. CRS 2023 says India registered 86.6 lakh deaths, equal to 97.2% death registration. But only 24.0% of registered deaths occurred in institutions, and 53.4% had no medical attention at the time of death. MCCD then records medically certified causes for 19.0 lakh deaths, or 22.0% of registered deaths. For heat, that distinction is everything. A death can be registered and still lose the heat signal if no doctor certifies how heat interacted with the heart, kidney, lungs or dehydration.

Cause-of-death visibility falls to about 6% of registered deaths in Uttar Pradesh, Assam and Bihar.

This map shows why the undercount is not just national; it is local. MCCD coverage is strong in a few places, including Goa at 100.0%, Delhi at 66.0% and Maharashtra at 42.4%. But in several large states the certified-cause record is extremely thin: Uttar Pradesh is 5.8%, Assam 5.6% and Bihar 5.5%. A low value does not mean fewer heat-related deaths. It means fewer deaths have a medically certified cause that could carry a heat label or reveal heat as a trigger. The map therefore shows where official heat-death counts are most likely to be blind before the counting even starts.

Uttar Pradesh reported 240 deaths, more than half the national total of 460.

A bar chart of IMD’s 2024 heatwave deaths by state. Uttar Pradesh tops at 240, followed by Bihar (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 poverty map. This is the most official count, but it is almost certainly an undercount because it relies on reporting from states and excludes deaths not directly attributed to heatwave by meteorological criteria. Yet it shows where the counting system is at least recording some deaths.

IMD’s annual heatwave deaths collapsed from over 2,000 in 2015 to zero in 2021, exposing the counting system’s flaws.

A line chart of IMD-reported heatwave deaths from 2013 to 2024: 1,400 (2013), 549 (2014), >2,000 (2015), then a sharp decline to near-zero in 2020 and exactly zero in 2021, before a rise to 460 in 2024. Heat did not take a break; the counting did. After the 2015 heatwave disaster, states launched Heat Action Plans that saved lives, but the recorded number also fell because systematic reporting was not maintained. The number swings because it measures attention and bookkeeping, not actual deaths. This is the clearest evidence that India does not know its true heat toll.

EM-DAT records 10,398 extreme-temperature deaths in India since 2000, but most years show zero or near-zero outside major events.

A line chart of reported extreme-temperature disaster deaths from OWID/EM-DAT (1900-2025). The series is spiky, with large events in 1998 (2,541), 2015 (2,248) and then many years with few or no entries. The database relies on disaster declarations and media reports; it misses the slow, accumulating toll of heat. The total of 10,398 over 25 years averages about 415 per year, but that average masks the fact that most years have no data. This is a record of recognised events, not a mortality surveillance system.

A 3% mortality lift among 20% of India’s population yields 157 excess deaths; scale up to 75% exposed and 15% lift, it reaches 2,952.

This horizontal bar chart shows excess deaths per day under five illustrative scenarios, from ‘very low’ (157) to ‘high’ (2,952). The assumptions are: exposed population share and mortality lift among exposed baseline deaths. India’s daily all-cause deaths are about 26,236. Even a small temporary increase in death rate for a slice of the population adds up quickly because the base is huge. The Frontiers model’s 3,400 can be reproduced with 70% exposed share and an 18.5% lift. The bars are not proving 3,400 is correct; they simply show the arithmetic is plausible. Real attribution needs better data.