Swechha SearchAct
Pine forest below a snow-covered Himalayan peak

Active situation

Bihar flood

Active situation ·Flood

Active situation. Being tracked now. Figures move while you are on this page.

Bihar ·29 Aug 2026

Why this keeps happening

73 Rain, 7 daysmm over the region 08/29 SatelliteNOAA-20 · 28.6% cloud 5 Publishersreporting it

Four readings this page holds, and where each one is explained. Every source behind them.

What is a flood, really

Not too much water. Too little room for it.

Think of the drain on your street. Somebody sized it for the rain they expected. Now give it three times that rain, and half fill it with silt and plastic first.

A river works the same way, only bigger. It has a channel, and beside the channel a floodplain — flat land that is supposed to get wet. That is the river's spare room.

Build on the spare room and the water still arrives. It just arrives somewhere that now has houses on it.

Global heating — heating, not warming — loads the dice: a hotter atmosphere carries more moisture, so when it rains, it rains harder.

Where

Bihar — and the river it runs into. Positions are real coordinates; the extent of the event is not known.

The area the satellite frames cover Bihar Bihar

Region named in the reportingDownstream, in the pathThe satellite frame

Frame 677 km wide. The ring is an AREA, not a point: this page knows the region the reporting names and not where inside it the event was. Positions of the downstream places are true.

  1. Origin Bihar
  2. Direct impact India, downstream

What caused it

Candidate mechanisms, each with the evidence word it has actually earned and, where one exists, the counted thing that makes it plausible. Nothing on this list is established for this event.

Under investigation Extreme rainfall Rain over the catchment beyond what the channel and the drains can carry.
Under investigation Reservoir or barrage release Water released from storage upstream, on top of the rain.
Under investigation Embankment or drain failure A bund, embankment or stormwater drain giving way rather than overtopping.
Under investigation Drainage that could not carry it Built surface and blocked channels turning ordinary rain into a flood.
How would we know?
Extreme rainfall
India Meteorological Department daily rainfall for the district against its own heavy (64.5 mm) and very heavy (115.5 mm) day categories.
Reservoir or barrage release
The operator's own gate-operation log and the Central Water Commission's downstream gauge record for the same hours.
Embankment or drain failure
A state irrigation department breach report, which names the location and the length.
Drainage that could not carry it
Whether the same streets flood at the same rainfall every year, which is a records question rather than a hydrology one.

What the satellite sees

Imagery published here, not linked to. The same public NASA layers a newsroom would use, over the region the reporting names, on the dates either side of the event.

True colour satellite view of the region, 2026-08-22
Before · 2026-08-22 Terra, MODIS, 250 m. 8.4% of the frame is cloud or snow.
True colour, second satellite satellite view of the region, 2026-08-29
After · 2026-08-29 NOAA-20, VIIRS, 375 m. 28.6% of the frame is cloud or snow.

The two frames come from different instruments — true colour before, true colour, second satellite after — because nothing else could see the ground on both dates. They are shown side by side rather than as a slider: sliding between two instruments shows the instrument changing.

Frame 23.5–26.7°N, 83.543–87.077°E. The region reported, not a located site.

Open the live comparison Zoom, re-date, or switch to radar.

NASA Worldview Snapshots. Public domain, courtesy of NASA EOSDIS GIBS.

It has happened before

3 comparable events, with what each one actually cost. Every toll below is an official or peer-reviewed count, not a headline.

2023 CWC changes what it counts 4,472 deaths and missing — not comparable with 2022's 1,673 Official estimate Central Water Commission
2023 Yamuna at Delhi, highest ever recorded 208.66 m against a 1978 record of 207.49 m Confirmed Ministry of Jal Shakti
1977 The deadliest flood year on record 11,305 deaths Official estimate Central Water Commission
What happened in each, and what was learned
CWC changes what it counts · 2023
The reporting basis changed. From this year CWC's column counts deaths AND missing persons on the wider MHA 'floods and heavy rains' definition, and moves bridge, dam and road damage outside the headline total — so 2023 and 2024 cannot be compared with earlier years.
Yamuna at Delhi, highest ever recorded · 13 July 2023
The Yamuna reached 208.66 m at Delhi's Old Railway Bridge, passing the 207.49 m of September 1978 — the highest level ever recorded there.
The deadliest flood year on record · 1977
The deadliest flood year in the seventy-two-year record, and the reason the long-run average sits near 1,700 deaths a year.

What these counts leave out

  • Whatever a state chose not to report. For 2024 Assam filed 'Not Reported' against crop value, house value, public utilities and total damage — while reporting 1.39 million ha affected and 110 deaths. A dozen other states did the same, so the published national total is a floor, not an estimate.
  • Embankment failures and breaches. A 2025 parliamentary answer states this is not maintained centrally at all.
  • Infrastructure, in the headline figure. From 2023 damage to bridges, dams, roads and irrigation sits outside the published total — in 2024 that excluded ₹10,447 crore, more than the entire reported total.
  • Urban flooding as a distinct category. There is no national dataset separating it from riverine flooding.

What happens next

Two horizons. A level only where a published threshold and a number exist to put something against — everything else is named as a watch, not graded.

Next 24–72 hours

Moderate More rain over the region Peak 29.5 mm in one day across the next 3, against IMD's heavy-rainfall threshold of 64.5 mm and very heavy at 115.5 mm.
Watch River levels downstream CWC gauges, and whether any cross a warning or danger mark.
Watch More rain on the same catchment Before the water already there has gone down.
Watch Reservoirs near capacity A release adds to a river that is already high.
Watch Embankments under strain A breach moves the flood somewhere new rather than deeper.
Watch The counts settling Agencies count the displaced differently.

Coming days and weeks

Watch Waterborne disease Counted separately from the flood, and rarely attributed to it.
Watch Crop loss Only becomes a figure at harvest, months later.
Watch Silt changing the channel Where the next flood goes.
Watch Displacement that does not reverse No published series tracks it.

The rainfall level is computed from a forecast model over a representative point — Open-Meteo forecast API, not a gauge at the event. The ungraded items are the mechanisms that follow this hazard. This page does not pretend to know their probability.

Data and sources

Everything this page is built from, counted by kind. Nothing here is quoted in this site’s own voice.

6 official alerts
5 published reports
5 independent publishers
View all 11 sources

Official

Reported

A headline is evidence that something was said. It is never evidence that it is true — which is why every figure at the top of this page carries the outlet that printed it and, where they disagree, all of them.

What this page does not know
  • Casualty and damage figures are not established here. Any number in the headlines below belongs to the outlet that printed it.
  • The mechanism is this page's reading of how the event was first reported, not a finding. Himalayan disasters are often reclassified once fieldwork is done.
  • The extent of the affected area has not been verified against satellite or official mapping.
  • Whether this event is continuing or has passed is not established.

Method: assembled automatically from 5 items across 5 publishers by scripts/detect-climate-events.mjs, which scored this event 17 against a publication threshold of 14. Assembled automatically from published headlines and official alert feeds. Every figure is a count this script performed; no claim about the event is made in its own voice.

Feeds re-read every hour. This page last changed 18:40 IST, 29 Aug 2026 — a re-read that finds nothing new does not move that.

This is one event

The standing picture — how much extreme rain India actually gets, what IMD counts as extreme, who the rain kills, and what is changing across twelve cities — is on the situation page this belongs to. That page moves slowly on purpose. This one moves every half hour.

India’s extreme rain

Every reading, kept

Each reading keeps its own address, with the source that produced it, when it was observed, and the limit it was judged against. Nothing is overwritten when it improves and nothing is quietly restated when it gets worse. An empty day stays empty — a gap in the record is a gap in the record, never a zero.

Cite this page

Reuse freely — CC BY 4.0. Every figure carries its source and its cadence in how the number is made, and every figure carries whether it was counted or modelled on the rule beneath it. If you quote a number from here, quote the kind with it. The grant covers this page; each upstream source keeps its own terms, which is why every figure names one.

This is what we are doing.
It is not enough.

The work against this reading

Mostly the written record, and going to look. Three of the essays on this site were reported out of flood years, and Gram Anubhav takes groups into villages living with this, run with over a hundred grassroots partners in the villages themselves.

And why it is not a reply

Publishing and going to see are not resilience. Swechha does not build drainage, and nothing here reduces the 3,594 deaths this page counts. What the work does is make the count harder to ignore — a smaller claim than it sounds, and the only one the evidence supports.

Explore the workFifteen pages, four kinds Who is SwechhaWorking since 2000, and who runs it Take partGive, volunteer or partner

Once a month, what these numbers did.

One email. What the six readings did that month, what moved and what did not, and what we did about it. Nothing else, ever — no appeals between digests, and your address is never shared with anybody.

Confirm by email before anything is stored. Every message carries a one-click link to stop them.