Bhote Koshi Flash Flood Impact and Exposure Atlas

Errata & CorrectionsKnown Issues With This Dataset

• corrected 2026-08-27
  Casualty and damage figures revised upward. Published 26 August with 157 confirmed dead and "more than 1,400" reported missing. Nepal Police put the toll at 177 on 27 August (41 women, 69 men, 10 girls, 7 boys; 50 unidentified). Missing counts have been restated by source rather than aggregated: 85 security personnel (44 Nepal Army, 28 Nepal Police, 13 Armed Police Force), 60 hydropower workers, and either 484 tourists (Prime Minister's Office) or 579 travellers (NDRRMA). Rescue figures updated to 336 airlifted from Timure, Dhunche and Mailung. Stat chips, overview and limitations on this page have been updated; the GeoPackages, CSVs and the Zenodo deposit are unchanged, because no casualty figure is carried in the spatial data.
• corrected 2026-08-27
  Downstream scope limitation superseded. The page originally stated that Dhading, Gorkha and Chitwan received warnings but no reported damage, and justified ending the corridor at Devighat on that basis. Bodies have since been recovered in Dhading, Gorkha, Chitwan, Tanahun and East Nawalparasi. The 62.4 km corridor still reflects the reported structural damage reach, but it is no longer a proxy for where the flood caused harm, and the limitation has been rewritten to say so.
• open 2026-08-27
  Source location disputed between two authorities. This page places the source at the USGS seismic origin for event us7000tbwb (28.2710 N, 85.5150 E) - 13.5 km from Rasuwagadhi on a bearing of 094 degrees. NDRRMA describes the trigger as a glacier and ice-rock collapse approximately 20 km northeast of the Rasuwagadhi border point, which resolves to roughly 28.407 N, 85.522 E - 15.1 km from the USGS point. The seismic location for a surface mass movement carries large horizontal uncertainty, so the discrepancy is expected rather than an error in either source. The damage register retains the USGS point, flagged location_confidence=low. This will be resolved when imagery-based source mapping is published.
• open 2026-08-27
  Official infrastructure damage counts exceed what this analysis found. NDRRMA reports 35 motorable bridges, 45 suspension bridges and 40 km of paved road damaged or destroyed. This dataset counted 38 OSM-mapped main-stem bridge crossings and 28.5 km of the Pasang Lhamu Highway inside the 80 m envelope. The gap is expected - OSM does not map most suspension footbridges in this corridor, and the official road figure spans tributaries the corridor analysis excludes. The official figures should be preferred for any damage total; this dataset's value is in where those assets sit relative to the river, not in counting them.

Keystone GIS documents known data issues here rather than silently revising published datasets.

OverviewWhat Happened, and What This Records

At 02:52:10 UTC on 26 August 2026 (08:37 Nepal time) an ice and rock avalanche fell into the Lhende Khola, a Bhote Koshi tributary rising in Gyirong County, Tibet. The United States Geological Survey catalogued the seismic signal as event us7000tbwb and, after long-period waveform analysis, classified it not as an earthquake but explicitly as event-type 'landslide' — the catalogue entry is titled 'M 5.2 Landslide — 55 km NW of Kodari, Nepal', at 0 km depth. Nepali officials had initially attributed the flood to a magnitude 4.4 earthquake.

The flood entered Nepal at the Rasuwagadhi border crossing and ran 62.4 km down the Bhote Koshi and Trishuli to Devighat, falling 1,339 m at a mean gradient of 21.5 m per km. There had been no rain and no warning. As of 27 August, 177 people are confirmed dead in Nepal and 3 in Tibet. Nepal Police record the dead as 41 women, 69 men, 10 girls and 7 boys, with 50 of the 177 still unidentified. Bodies have been recovered not only in Rasuwa and Nuwakot but downstream in Dhading, Gorkha, Chitwan, Tanahun and East Nawalparasi - far beyond the 62 km reach where structural damage was reported. 336 people have been airlifted out of Timure, Dhunche and Mailung, 45 of them foreign nationals.

This atlas pairs a source-cited register of reported damage with a Copernicus-DEM height-above-channel envelope, so every claim about what was hit can be checked against where it sits relative to the river. Sixteen damage records, 13 exposed settlements, 38 main-stem bridge crossings, 28.5 km of the Pasang Lhamu Highway and five of the six damaged Nepal Electricity Authority facilities are placed on the same 62 km line.

Why This DatasetThe Problem It Solves

In the first days after a Himalayan flood, the damage picture arrives as a list of place names in news copy and agency statements. It is rarely georeferenced, almost never height-referenced, and impossible to reconcile with what the terrain would allow. That makes it hard to answer the questions responders actually have: how far above the river was this settlement, how much of the valley floor could plausibly have been reached, and how many people lived on it.

This product does the reconciliation. Every reported impact is georeferenced from a named open feature — not from the news text — and then intersected against a terrain surface. Where the two disagree, the disagreement is published rather than smoothed over: Dhunche's reported flooding turns out to be tributary flooding 733 m above the main stem, and the Trishuli 3B hub substation sits 20 km from the river, where direct inundation is implausible.

MethodologyHow It Was Built

Flood path. The main stem was assembled from OpenStreetMap waterway=river ways named Bhote Koshi, Trishuli River, Trishuli Ganga River and Trisuli, linemerged, and the longest continuous component taken: 62.42 km from 28.2781 N, 85.3770 E at the border to 27.8666 N, 85.1127 E at Devighat.

Copernicus GLO-30 tiles N27_00_E085 and N28_00_E085 were mosaicked to the corridor at roughly 29 m pixels. The channel was rasterized, a Euclidean distance transform with index return gave every pixel the elevation of its nearest channel cell, and subtraction produced the height-above-channel surface. The 80 m band is not arbitrary: reporting records floodwaters reaching settlements about 80 m (270 ft) above river level, and the chief customs officer at Timure described a wave “rising like a wave nearly 100 metres high”.

Population. WorldPop 2020 UN-adjusted unconstrained counts were reprojected onto the DEM grid by density-preserving bilinear resampling. The constrained BSGM raster was tested first and rejected: it sums to 25.3 million against a UN 2020 total of 29.1 million, has only 272,253 valid pixels nationally, and places 35,449 people in a single 92 m cell.

Damage register. Each of the 16 records is georeferenced independently of the damage claim, from OSM, Wikidata, Wikipedia coordinates or Wikipedia GeoSearch, and carries that source, a location_confidence flag and an evidence URL. Nothing is inferred: where a coordinate could not be established from a named open feature, that is recorded rather than estimated.

Key FindingsWhat the Data Shows

• USGS reclassified the trigger from earthquake to landslide. Event us7000tbwb carries event-type: landslide, Ms 5.2 and a fixed 0 km depth — the signature of a surface mass movement. The distinction is not academic: an earthquake-triggered avalanche and a spontaneous glacier collapse imply different monitoring regimes.
• The valley floor is thin, and almost everything is on it. Only 39.4 km² of the 62 km corridor lies within 80 m of the river, and 8.8 km² within 10 m. Into that ribbon fall 13 mapped settlements, 28.5 km of the Pasang Lhamu Highway, 38 main-stem bridge crossings and five of the six damaged NEA facilities.
• 17,167 residents in the 80 m envelope — a floor, not a ceiling. WorldPop under-counts these hill districts against the 2021 census by 27 percent in Nuwakot and 37 percent in Rasuwa; the district-ratio adjustment gives roughly 23,700. Exposure is concentrated: 15,850 of the 17,167 are in Nuwakot and 14,191 in Bidur alone.
• The road and the river are the same corridor. The Department of Roads reported the entire 42 km Betrawati–Rasuwagadhi stretch destroyed. Measured along the river that reach is 45.5 km — within 9 percent, which corroborates the figure. The 16 km Syaphrubesi–border section, rebuilt to two lanes with a Chinese grant after the 2025 floods, was swept away again.
• 231 MW of hydropower on one glaciated headwater. Rasuwagadhi (111 MW, river km 5.0), Chilime (22 MW, powerhouse at km 12.4 and 7 m above the channel), Upper Trishuli 3A (60 MW, km 39.0, on the channel), Trishuli (24 MW, km 53.9) and Devighat (14.1 MW, km 58.9) all sit inside the envelope. Cascade siting concentrated the risk the flood realised.
• The steepest reach is where people lived closest to the water. Between Rasuwagadhi and Syabrubesi the channel loses 348 m in 14.7 km, and Timure and Syabrubesi sit just 40 m and 37 m above it on valley floors a few hundred metres wide. Timure was hit first, about three river kilometres from the border, and much of it was gone within moments.

Damage RegisterWhat Was Reported Destroyed or Damaged

* outside the 80 m main-stem envelope. Nine bank branches along the corridor were also reported swept away with 26 staff out of contact. The full register, including the USGS seismic source record and per-record location sources, confidence flags and evidence URLs, ships in the data package.

Analytical UsesWhat You Can Do With This

• Search and relief sequencing — the register orders every reported impact by river chainage, so the order in which the wave arrived is explicit and the envelope shows which ground between the named places is low enough to have been reached.
• Reconstruction siting — height above channel is the single most useful number for deciding what to rebuild where. Timure sat 40 m above the channel and was still destroyed; that is the design constraint.
• Hydropower cascade risk review — five NEA schemes on one glaciated headwater, all inside the envelope, is a portfolio-level exposure question rather than five engineering ones.
• Transboundary early warning — source, tributary and first settlement lie in three jurisdictions within 25 km. The corridor geometry quantifies how much warning time a Lhende Khola monitoring station would buy downstream.
• Method transfer — the height-above-channel workflow needs only a public DEM and a river centreline, so it can be run for any Himalayan valley before an event rather than after one.

LimitationsKnown Caveats

• This is an exposure envelope, not an observed inundation extent. The envelope says what ground lies below a given height above the river; it does not say what was wetted. Copernicus EMS has since been activated as EMSR927 for flood extent and damage assessment, and monitoring stations recorded a 7 m water level; when EMS delineation products are released they should supersede this surface. Planet Labs has published post-event imagery of the corridor, but under a CC-BY-NC licence, so it is not usable in this openly licensed dataset.
• Casualty and missing figures are still moving. Reported deaths ran 8 → 22 → 95 → 157 → 177 across the first 36 hours, and missing totals have been reported inconsistently: the Prime Minister's Office gives 484 tourists out of contact (391 foreign, 93 Nepali) while NDRRMA gives 579 travellers (466 foreign, 113 Nepali). Both are reproduced here with their source rather than averaged. Every figure is attributed and time-stamped; none is final.
• WorldPop under-counts these districts by 37 percent in Rasuwa and 27 percent in Nuwakot against the 2021 census. The 17,167 figure is a lower bound; the district-ratio adjustment gives about 23,700.
• The DEM is a surface model and predates the event. GLO-30 is a DSM, so canopy and buildings raise valley-floor heights in forested gorges, and it records the pre-flood channel — the bed has since been reworked by debris.
• Two hydropower coordinates are inferred from capacity and position. Rasuwagadhi and Upper Trishuli 3A are not named in OSM, have no Wikipedia coordinate and postdate the 2021 WRI power plant database; both were matched to unlabelled Wikidata items and flagged location_confidence=medium.
• OSM completeness here is uneven. Bridges are almost never named, the Rasuwagadhi customs post carries no border-control tag at all, and transliterations differ from those used in reporting. Absence from OSM is not evidence of absence on the ground.
• Scope stops at Devighat, and the event did not. The corridor ends at river km 62.4 because that is the reach where structural damage was reported. Bodies have since been recovered in Dhading, Gorkha, Chitwan, Tanahun and East Nawalparasi, hundreds of kilometres further down the Trishuli and Narayani. The exposure envelope therefore covers the damage reach, not the full extent of the flood's effects, and downstream fatality locations are outside every count on this page.

Data DictionaryFields in the GeoPackages

Jurisdiction BreakdownResults by Jurisdiction

Sources & CitationsAll Data is Openly Licensed

• [1]U.S. Geological Survey. ComCat event us7000tbwb, “M 5.2 Landslide - 55 km NW of Kodari, Nepal”, 2026-08-26 02:52:10 UTC, event-type: landslide, depth 0 km. earthquake.usgs.gov
• [2]ESA / Airbus. Copernicus DEM GLO-30, tiles N27_00_E085 and N28_00_E085.
• [3]WorldPop. Nepal 2020 UN-adjusted unconstrained population count, 100 m (CC BY 4.0).
• [4]geoBoundaries gbOpen. Nepal ADM2 and ADM3.
• [5]OpenStreetMap contributors. Waterways, places, roads, bridges, power and route relation 4662089, via Overpass (ODbL).
• [6]Wikidata. Nepal power stations with P625 coordinates and P2109 installed capacity (CC0).
• [7]Wikipedia. 2026 Nepal floods; Wikipedia API coordinates and GeoSearch (CC BY-SA 4.0).
• [8]The Kathmandu Post. “Nepal flash flood kills at least 157, leaves nearly 400 missing - including foreign tourists”, 26 August 2026.
• [9]The Kathmandu Post. “Flood washes away 42-km road linking Nuwakot to China border”, 26 August 2026.
• [10]The Kathmandu Post. “Flash flood hits Rasuwa, damages hydropower projects and markets”, 26 August 2026.
• [11]OnlineKhabar English. “Rasuwa flood damages hydropower plants and 220 kV substation, power supply disrupted”, 26 August 2026.
• [12]NPR. “At least 98 killed and hundreds missing after flash floods in Nepal and China”, 26 August 2026.
• [13]Nepal Central Bureau of Statistics. National Population and Housing Census 2021, district totals.