The Uneven Heat: District-Scale Warming Across India, 1951–2025

1Introduction

Global-mean temperature is a single number that hides almost everything a person actually experiences. A farmer in Palakkad, a commuter in Patna and a herder in Bikaner do not live in the global average; they live in the trend of their own district — its nights that no longer cool, its monsoon that arrives differently, its pre-monsoon heat that starts earlier. India’s national warming has been documented at roughly +0.7 °C over 1901–2018 by the Ministry of Earth Sciences,¹ but a national figure cannot tell a district what is happening to it.

This note takes the opposite view — bottom-up rather than top-down. Using the per-district climatology that powers the VayuMet District Climatology map, we compute the linear temperature trend for every Indian district from 1951 to 2025 and ask three questions. How fast is each district warming? Where are the extremes — the fastest-heating districts and the places that have barely moved? And do the spatial patterns line up with what climate science attributes to human activity — greenhouse-gas forcing, aerosol pollution, irrigation and ocean heat?

The headline is not that India is warming; that is settled. The revelation is how unevenly it is warming, and how legibly that unevenness maps onto human fingerprints on the regional climate.

2aEarlier studies

The modern baseline for Indian climate change is the Ministry of Earth Sciences’ 2020 Assessment of Climate Change over the Indian Region (Krishnan et al.), the first national report to attribute regional change explicitly to human-induced forcing. It reports a rise of about 0.7 °C in surface air temperature over 1901–2018 and finds the warming significant and accelerating in recent decades.¹ Our district results are consistent with, and refine, that national picture.

A second strand concerns the ocean. Roxy and colleagues have shown that the tropical Indian Ocean — and the western basin in particular — has warmed faster than any other part of the tropical oceans, its sea-surface temperature rising about 1 °C over 1951–2015 against a global-mean 0.7 °C.² Because peninsular India is nearly surrounded by this rapidly heating water, the ocean signal is a natural candidate for the south-coastal warming maximum we find below. The same body of work links Indian Ocean warming to a weakening land–sea thermal gradient and to shifts in monsoon rainfall.³

A third strand is the most surprising, and the most relevant to attribution: the Indian warming hole. Multiple studies have documented that parts of the Indo-Gangetic Plain warmed little, or even cooled, through the late twentieth century — a regional exception to global warming.⁴ The debate over its cause is essentially a debate about which human intervention dominates: anthropogenic aerosols (which dim and cool the surface) versus the spread of intensive irrigation (which cools through evaporation). Recent work increasingly favours aerosols as the primary driver, finding that irrigation’s cooling effect has been substantially over-estimated in models.^5,6 Crucially, both mechanisms are expected to weaken as air-quality policy cuts aerosols, implying the hole may fill in and warming there could accelerate.⁷

2bData and methodology

The analysis rests on three gridded observational products, all sampled at each district’s representative point (the centroid of its largest polygon):

• Temperature — IMD Pune gridded monthly mean temperature (1° resolution), 1951–2025.
• Rainfall — IMD Pune gridded monthly rainfall (0.25° resolution), 1951–2025.
• Cloud cover — NCEP total cloud cover via NOAA PSL, 1979–present (context only).

For each district and variable we aggregate to complete calendar years — annual mean for temperature, annual total for rainfall — and fit an ordinary least-squares straight line against year. The reported trend is the fitted slope expressed per decade (slope × 10); districts with fewer than ten valid years are dropped. The temperature series is rebased onto each district’s own 1991–2020 climatological mean so that absolute levels agree with the finer-resolution normals; because a constant offset does not change a slope, the warming trend is unaffected by this step.

3Results and discussion

3.1  Warming is near-universal, but modest in the mean

Across 755 districts the warming-rate distribution is tight and almost entirely positive (Figure 1). The mean is +0.072 °C/decade and the median +0.070; the interquartile range runs from +0.04 to +0.10. Only four districts — Kaimur, Bhojpur and Buxar in Bihar and Shrawasti in eastern Uttar Pradesh — show a slight negative slope, and none exceeds −0.01 °C/decade. Over the full 75-year record the district-mean rate implies roughly half a degree of warming, the lower edge of the national 0.7 °C because district-averaging gives equal weight to the many slow-warming plains districts.

3.2  The south and the coasts lead; a warming hole sits on the plains

The mean hides a strong geography. Grouping districts into broad regions (Figure 2) exposes a near-fivefold spread: the deep south below 15°N warms at +0.127 °C/decade while the Indo-Gangetic Plain crawls at +0.030. The Northeast hills and the arid Gujarat–Rajasthan coast and desert also run hot (≈+0.10), whereas the northern and Himalayan-fringe districts and central India sit in the middle.

The gradient is monotonic enough to show up as a clean correlation with latitude (r = −0.45): warming decays northward from about +0.14 °C/decade in the 8–11°N coastal south to a minimum in the 20–32°N plains belt (Figure 3). This is exactly the footprint expected if the surrounding ocean is doing much of the work. The tropical Indian Ocean is the fastest-warming tropical basin on record,² and the districts pressed against it — Kerala, coastal Tamil Nadu, the Konkan, and the Saurashtra coast of Gujarat — are precisely the ones heating fastest.

3.3  The extremes

At the top of the table, warming saturates near +0.15 °C/decade across a belt of Kerala, coastal Gujarat and the offshore islands; Lakshadweep leads at +0.170, a small ocean-surrounded territory where the SST signal is undiluted by land. At the bottom sit the Bihar–eastern-UP districts of the middle Ganga plain, several with flat or marginally negative trends. The contrast between, say, Thrissur and neighbouring-latitude Kaimur is not noise — it is the warming hole in two rows of a table.

Ranked by state, the same story holds: Kerala (+0.148) and Tamil Nadu (+0.130) head the list, the Northeast hill states and the Gujarat–Rajasthan arid zone follow, and the great plains states — Uttar Pradesh (+0.019), Jharkhand, Bihar, Punjab — sit at the bottom (Figure 4).

3.4  Rainfall is reorganising, not simply rising

Rainfall trends are far noisier than temperature and split almost evenly — 359 districts wetter, 387 drier — so there is no single national direction. The structure, however, is telling. The Indo-Gangetic Plain is drying (−20 mm/decade on average), consistent with the weakening land–sea thermal gradient that a rapidly warming ocean produces.³ The wettest-trending districts cluster in the Meghalaya–Mizoram hills and the Andamans, and the driest in the Siang and Dibang valleys of Arunachal — but the Northeast’s sparse rain-gauge network makes these magnitudes uncertain, and we flag them as indicative. The robust rainfall message is redistribution: a drying breadbasket and a shifting monsoon, layered on top of the temperature trend.

Nearly every district is warmer than it was in 1951 — but the country is not warming as one body. It is warming as a coast, a desert, a hill range and a river plain, each on its own clock.

3.5  Reading the map as human fingerprints

The spatial pattern is not random, and each piece has a human-attributable reading in the literature:

• The peninsular–coastal maximum tracks the tropical Indian Ocean, which has warmed ~1 °C since 1951 — well above the global ocean — under greenhouse forcing, with the western basin the single largest contributor to global-mean SST rise.² Land districts ringed by that water inherit its heat.
• The Indo-Gangetic warming hole is the clearest regional human fingerprint of all — but of a masking kind. Dense anthropogenic aerosol pollution over the plains dims incoming sunlight and cools the surface, and intensive irrigation adds evaporative cooling; together they have offset much of the greenhouse warming there for decades.^4,5 The uncomfortable corollary, now widely noted, is that as India cleans its air the mask lifts — the hole is expected to fill and warming on the plains to accelerate.⁷
• The arid-northwest warmth reflects dry-land amplification: with little soil moisture to evaporate, added energy goes straight into heating the air.

In other words, the map is legible. Where the ocean is warmest, the land warms fastest; where human pollution and irrigation are thickest, warming is suppressed — for now.

4Conclusion

Seven decades of district-level records deliver a clear and slightly counter-intuitive verdict. India’s warming is near-universal but deeply uneven: 751 of 755 districts have warmed, yet the fastest region heats almost five times faster than the slowest. The gradient is not arbitrary — it descends with latitude, peaks along the ocean-hugging south and coasts, and collapses into a warming hole over the Indo-Gangetic Plain. Read against the attribution literature, the pattern resolves into human causes: an exceptionally fast-warming Indian Ocean pulling up the peninsula, and aerosol pollution plus irrigation temporarily holding the plains down.

Two implications follow. First, adaptation is a district problem, not a national one — a uniform policy calibrated to the +0.7 °C national figure will badly under-serve a Kerala warming at twice that pace and mis-read a Bihar that has, so far, barely moved. Second, and more sobering, the plains’ calm is probably borrowed. The very interventions suppressing warming over India’s most populous, most food-critical belt — dirty air and heavy irrigation — are the ones expected to fade, and when they do the warming hole is likely to close. The districts that look safest in this record may have the most abrupt adjustment ahead.

The per-district series behind this note are open in the VayuMet District Climatology map; the “Warming Trend” layer renders the same numbers analysed here, district by district.