As I write this, August has ended, and Delhi-NCR has just come through two of the most punishing monsoon months I can remember. Today ran the pattern the season has followed throughout: a cloudy morning, sun by noon, a hot afternoon, and a humid night that is never pleasant.
It has felt like it barely rained all year and the numbers say something else entirely.
Two notifications from my phone, both from this month. On 8 August at 22:42: severe heat warning, will feel like 46°, light rain. On 27 August at 20:33: severe heat warning, will feel like 43°, mostly cloudy. Severe heat alerts, at night, in the middle of the monsoon – one of them (the title image) issued alongside a forecast of rain.
By the night of that first alert on 8th August, Safdarjung Observatory’s rainfall total had already crossed the whole of August’s normal rainfall, in eight days. The month went on to finish 59% above normal, and the season now stands at 653.9 mm against a long-period average of 506.7 mm.
I am writing this in a study with no air conditioner, and the pull toward the living room (where there is one) has been constant for weeks. That pull, multiplied across a city, is the story of this monsoon in Delhi-NCR.
Severe Heat Warning on my Phone on 27th August
In July 2026, Delhi’s weather was, by every official measure, normal.
Safdarjung logged 258.0 mm of rain against a long-period normal of 209.7 mm, a 23% surplus, unremarkable for a monsoon month. The mean maximum temperature came in at 35.5°C, a tenth of a degree below the July climatological normal. The mean minimum was 27.6°C, two-tenths above. Three headline numbers, all of them normal.
That same month, Delhi recorded its most polluted July in seven years. National peak power demand touched 270.20 GW. Spot electricity on the IEX averaged ₹7 a unit, up from ₹4.2 in July 2025. And anyone who spent the month in the city will tell you it was among the most punishing they can remember. All though every number showed signs of a normal monsoon month, nothing about the month was.
There is a reason the monsoon is behaving this way. The Clausius-Clapeyron relation holds that for every 1°C of warming, the atmosphere can carry roughly 7% more water vapour. A warmer atmosphere doesn’t simply deliver more rain — it holds more, for longer, before releasing it faster.
The output is an all-or-nothing cycle: extended dry gaps between rains, punctuated by rainfalls that discharge the accumulated load in hours. The monthly rainfall hits the total. The rhythm that made it useful does not.
Look back exactly one year. July 2025 delivered 259.3 mm at Safdarjung, within two millimetres of this July.
While the two rainfall figures are almost same, the behaviour of the rain was nothing alike. In 2025 it rained on nearly every day of the month. The near-continuous cloud cover held the mean maximum to 34.5°C, more than a degree below normal, and the mean minimum to 26.1°C. In July 2025, CPCB recorded an average AQI of 79 which was the cleanest July in a decade.
In 2026, the same volume arrived through two heavy-rain days. IMD logs heavy rainfall on 9 and 29 July, and nothing heavier than moderate across the rest of the month. Between 11 and 20 July, the record shows no rain at all. The 9 July event alone put 72.6 mm on Safdarjung and 164 mm on Mayur Vihar in a single morning.
The result: an average AQI of 119, the worst July reading in seven years, against 79 a year earlier. The warmest average day and night temperatures in two years. Near-identical rainfall totals, opposite outcomes on almost every measure that matters.
This is aggregation bias, and it is not a rounding error. When rain arrives compressed rather than distributed, soil cannot absorb it, groundwater does not recharge efficiently, and drainage designed for a distributed load fails against a concentrated one. The same volume, delivered differently, is a different substance.
What sat in the ten-day gap is what I’ve come to call a humidity trap.
In June, Delhi’s heat is dry heat, and the city recovers at night. In mid-July the monsoon entered a break phase: the rain stopped, but the moisture flow didn’t. Southwesterly winds off the Arabian Sea kept feeding humidity into the city without releasing any of it.
The physics of the night changes under those conditions. Water vapour is a powerful absorber of outgoing longwave radiation — the heat a city sheds after dark. Saturate the air and the city’s ability to cool overnight degrades. During the day, the body sheds heat through sweat evaporation, and saturated air disables that mechanism outright. This is why air temperature has stopped being an honest instrument: CEEW estimates that humidity of the levels now common in North India can add 3–5°C to the felt temperature.
The trend underneath is not a Delhi anomaly. CEEW’s 2025 district-level heat risk assessment, covering 734 districts across four decades, found that very warm nights are now rising faster than very hot days: roughly 70% of districts have gained more than five additional very warm nights per summer over the last decade, against only about 28% for very hot days. The same study records relative humidity across the Indo-Gangetic Plain rising by up to 10% — North India has moved from a historical 30–40% band to 40–50%, with Delhi named among the traditionally dry cities making that shift. Delhi’s summer is changing category, from dry heat to a humid public health problem. CEEW ranks Delhi first among Indian states and UTs for overall heat risk.
And this is why the night matters more than the afternoon. A hot day the body survives if the night allows recovery. Compress the recovery window and the exposure compounds.
Discomfort is easy to dismiss as anecdote. Hard numbers like electricity demand and generation are not.
On 16 July, Delhi recorded its highest maximum temperature of the month at 39.2°C. The same day, national peak demand hit 270.20 GW — just short of the all-time high set in May — and spot power on the IEX averaged ₹7 per unit, against ₹4.2 in July 2025. National consumption for the month reached 170.7 billion units, up almost 11% year on year, which analysts attributed directly to humidity pushing felt temperatures above actual and driving cooling loads.
That is the humidity trap expressed as a number, on a month whose average temperature was below normal.
Worth stating plainly, from someone who works in renewables: this is not a straightforward argument for solar. The same thick cloud and haze that suppress daytime generation arrive alongside the humidity driving demand to a record, and weaker hydro output compounded the squeeze in July. Volatility hits both ends at once.
The break ended the way break phases do: the monsoon trough shifted back and a low-pressure system moved in. What made the difference was what it arrived over — a city with no absorptive margin left, in soil, in drainage, or in tolerance.
Delhi took 127 mm in the first seven days of August, its wettest opening week since 2011. By the evening of 8 August, Safdarjung had logged 230.1 mm — crossing the entire month’s normal of 226.8 mm in eight days.
August closed 59% above normal, with a second heavy spell on the 25th putting 99.1 mm on Safdarjung in 24 hours, the heaviest August day in two years. Across the season, Safdarjung stands at 653.9 mm against a long-period average of 506.7 mm.
And then the same signature reappeared: for all that rain, August’s average AQI reached a three-year high, with night-time temperatures staying elevated through the month. Two consecutive wet months, both delivering the opposite of what a wet month is supposed to deliver.
The distribution problem shows up across the city as well as across the calendar. Against seasonal normals, the Ridge is running 63% above and Ayanagar 59%, while Palam — barely twenty kilometres from either — is up only 11%. One season, one city, a five-fold spread in surplus. A single figure for “Delhi rainfall” describes none of them.
The submerged underpasses, the transit paralysis, the flooding across NCR: not a separate weather event from July’s heat. The same system, arriving with the bill.
It’s a fair question, and the answer is that the two aren’t alternatives.
El Niño is strengthening — NOAA has an advisory in effect, and July’s Southern Oscillation Index was the most negative July reading on record. It plausibly explains the timing of the break phase. It does not explain why the break was humid rather than dry, and it does not explain a rise in relative humidity measured across the Indo-Gangetic Plain over decades. One is an oscillation that turns over every two to seven years. The other is a trend line.
If anything, the simple version cuts the wrong way: El Niño years are conventionally associated with weaker Indian monsoon rainfall, and July came in 23% above normal. ENSO shifts when and where the rain falls. A warmer atmosphere changes how it falls when it does.
Not a resilience wishlist. Three measurement changes, because the failure here is one of measurement.
Report distribution alongside totals. A monthly rainfall figure with no accompanying measure of concentration is close to meaningless for planning. Two numbers would fix it: the count of rain days, and the share of the monthly total delivered by the single largest event. July 2026 and July 2025 are indistinguishable on volume and unrecognisable to each other on those two.
Make night-time temperature a heatwave criterion. At present it is not a primary trigger in India’s heatwave declaration protocols, and most city Heat Action Plans are built around daytime thresholds — which is why a month with a below-normal mean maximum can be genuinely dangerous and never register as such. CEEW’s recommendation is the right one: warm nights and humidity stress need to sit inside the declaration criteria, not outside them.
Plan energy systems on variance, not means. Resource adequacy built around an average peak is planning for a month that no longer occurs. The relevant question is not what demand averages, but how long the system runs near its ceiling without a nightly trough to recover in — and what generation is available when cloud cover suppresses solar on exactly those days.
None of this requires new science. It requires reporting numbers we already collect, in a form that reflects the climate we actually have.
July 2026 passed every official test. Rainfall above normal. Mean maximum below normal. Mean minimum barely above. On paper, an ordinary month — and it produced the worst July air in seven years, record cooling demand, and a city that spent four weeks unable to cool down.
The measurements weren’t wrong. They were just answering a question that has stopped mattering.
Averages hide the crisis. Volatility defines it.
Sources: IMD Monthly Weather Report of Delhi, July 2026 (Safdarjung observatory). IMD Delhi August 2026 rainfall records, via PTI, 9 and 29 August 2026. CPCB and CREA air quality data. CEEW, “How Extreme Heat is Impacting India: Assessing District-level Heat Risk,” May 2025. Central Electricity Authority / Ministry of Power peak demand data. IEX spot market prices. ENSO status: NOAA Climate Prediction Center and Australian Bureau of Meteorology.
