Settlements at risk along the Trishuli — Rasuwagadhi to Galchhi

Bhote Koshi / Trishuli flood of 26 August 2026 · Copernicus EMS activation EMSR927

This is a list of every village and town close enough to the Trishuli, and low enough above it, to be in danger from the river. Two things put a place on the list: it sits within 2 km of the river, and its ground is no more than 50 m higher than the water. It covers the 80.0 km of valley from Rasuwagadhi, at the Tibet border, down through Timure, Syabrubesi and Betrawati to Galchhi.

Height above the water is the measure that matters most. A village 2 km away but 300 m up a hillside is safe; a village 100 m away and level with the river is not. Because the Trishuli drops 1,387 m between these two towns, every place is measured against the water level at its own point on the river, not against a single number for the whole valley.

These are pre-flood counts. The houses were mapped before 26 August, and inside the surveyed stretches they are the buildings the survey found damaged or destroyed. So the figure splits two ways: 10,579 people lived in the 2,421 houses of the 18 settlements now destroyed — a displacement caseload, not an evacuation target — and 21,833 are in settlements still standing, which is the number an evacuation plan is sized against.

Two ways of counting, and they agree. The headline figure is the 6,278 houses counted inside the danger zone times 4.37, the national average household size from the 2021 census. An independent modelled population grid puts 33,176 people in the same zone — 16% apart, which is as close as two methods of this kind get. Both are shown in the table, and the second is a cross-check, not a second answer.

What P1, P2 and P3 mean

Two things decide the class, and only these two: how close the settlement is to the river, and how high its ground sits above the water.

	Within 500 m of the river	More than 500 m from the river
Less than 25 m above the water	P1Evacuate first Close and low — on the valley floor, where the water went.	P2Warn and prepare Low, but set back from the channel.
25 to 50 m above the water	P2Warn and prepare Right beside the river, but higher up the slope.	P3Monitor Neither close nor low, but still inside the zone.

P1 27 settlements  ·  P2 24  ·  P3 6  ·  every one of the 57 is in some danger; the class says how urgent.

Population and damage are deliberately not part of the class, so a large village and a small one in the same position get the same one. Use the columns in the table to weigh size on top of it.

The class describes the ground, not the buildings standing on it. 15 settlements on this list, 9 of them P1, are already destroyed; the condition column and the Still standing filter separate them.

Where they are

The valley in profile — how far each settlement sits above the river

The blue line is the river falling from Rasuwagadhi on the left to Galchhi on the right. The pink strip above it is the 50 m limit. Every dot is a settlement, drawn at its own height: dots inside the pink strip are on the list.

The 57 settlements

The corridor is cut into five stretches of 8 to 10 km, at Betrawati where the road from Kathmandu meets the river, at Bidur the Nuwakot district headquarters, at the Devighat powerhouse, and at Tarukabesi. Each holds 8 to 13 settlements, so the map stays readable when you filter to one — and it zooms to the stretch you pick. All four filters combine, so you can ask for P1 settlements still standing in Nuwakot between Devighat and Bidur.

Totals by stretch of valley, downstream first

Stretch	Settlements	of which P1	Houses	People
Galchhi - Tarukabesi	13	6	1,191	5,205
Tarukabesi - Devighat	12	5	1,501	6,559
Devighat - Bidur	13	5	1,621	7,082
Bidur - Betrawati	8	6	839	3,667
Betrawati - Rasuwagadhi	11	5	1,126	4,922

Totals by palika (local council)

District	Palika	Houses	People
Nuwakot	Bidur	3,002	13,118
Nuwakot	Belkotgadhi	1,363	5,957
Dhading	Galchhi	627	2,740
Rasuwa	Uttargaya	438	1,915
Nuwakot	Tarakeshwor	434	1,896
Rasuwa	Gosaikunda	265	1,158
Nuwakot	Kispang	149	651

Second tier — just above the line, but partly inside it

A village is not a single point. These 15 places sit above 50 m if you measure the middle of them, so they miss the cut, but part of each one lies lower than that — together 1,139 houses and 3,658 people. They carry the same measurements as the main table.

Timure is one of them, and it is already destroyed — 99% of its known buildings were graded destroyed. Measured at its centre it stands 55 m above the water, its lowest ground 38 m, and 47% of the ground around it is inside the zone. 3 of these 15 are destroyed.

How the list was made

In short: trace the river from Rasuwagadhi to Galchhi, work out how high the water sits at every point along it, then measure every named village in the valley against that — how far to the river, and how far above it. Anything within 2 km and 50 m goes on the list. Houses come from building maps and the emergency survey; people come from a population grid built from those same buildings. The detail below is for anyone who needs to check the work.

River centreline. HydroRIVERS v1.0 supplies the downstream topology from Rasuwagadhi to Galchhi; that chain then selects which OpenStreetMap waterway=river ways form the main channel, and a shortest-path search over them yields one continuous 80 km centreline carrying only Bhote Koshi and Trishuli segments. The reach is clipped to the Rasuwagadhi and Galchhi anchor points, and the whole analysis is clipped to Nepal, so the band does not run on into Tibet past the border crossing.

Water level. Copernicus GLO-30 elevation is sampled every 50 m along the centreline, taking the minimum in a 150 m box to find the channel, then a running minimum downstream to strip bridges and surface-model spikes. The result falls 1,387 m over the reach, steeply in the upper gorge (25–40 m/km) and gently below Betrawati (2–5 m/km).

Selection. Distance is horizontal to that centreline. Height is the settlement's ground elevation — the median of a 150 m box — minus the water level at its nearest point on the centreline, so the 50 m band follows the river down. The lowest ground in the same box and the share of the 250 m neighbourhood inside the band are given alongside, because a single elevation sample on this terrain is worth ±10–20 m.

Names. OSM place nodes plus the GeoNames Nepal gazetteer, deduplicated on transliteration-tolerant name similarity within 700 m; where both hold a place the OSM position is kept, since GeoNames coordinates can sit a few hundred metres up the valley side. 336 distinct named places lie within 2 km of the reach; 57 also pass the height test.

Houses. OSM building footprints whose centre falls in the hazard band, excluding tagged non-residential structures. Inside the three Copernicus EMS survey windows the graded-building census is larger than OSM's coverage, and the larger of the two is used, recorded against the settlement it came from.

People. Houses times 4.37, the national average household size in the 2021 census. The count is pre-flood on both sides: OSM footprints were mapped before the event, and the EMS layer grades buildings that were hit — Syabrubesi's 253 are all damaged or destroyed structures, so its 1,106 people are who lived there on 25 August, and any undamaged building there is missing from the count entirely. This is tied to buildings actually counted in the band and never returns zero where buildings stand, which is what an evacuation figure has to do. It assumes one household per building — high in a bazaar where several share a roof, low where a building is a shed — and it inherits any gap in the building maps.

The cross-check. WorldPop's 2020 UN-adjusted building-constrained 100 m grid gives a second, independent estimate, each cell credited by the share of it inside the band and attributed to the nearest named settlement within 1 km. Over the corridor the two agree to 16%. Per settlement they do not: the grid attributes by catchment, so a village among sparsely mapped houses absorbs a whole neighbourhood — Majhitar reads 3 houses and 607 people — while a village beside a larger neighbour can be left with none, as Bhainse is with 81 graded buildings and zero people. That is why the grid is the check and not the answer. It is shown in the second half of the People column so any row can be interrogated.

Administrative address. District, palika and pcode from the OCHA COD-AB Nepal boundaries (v02, valid 2024-03-14); ward from OSM admin_level=9 relations. All 57 resolved to a ward.

Does this test actually find the damage?

The Copernicus emergency survey inspected satellite imagery of three stretches of this valley and marked 2,521 buildings as destroyed. Testing the two rules against those buildings: 97.6% were within 2 km of the river, and 89.6% met both rules. So nearly nine out of ten buildings actually destroyed would have been picked up by this list in advance.

It works almost perfectly around Bidur and Betrawati, where 99.7% of the destroyed buildings fall inside the zone, and reasonably at Syabrubesi (63.5%). It works worst at Timure (37.1%): the village and the dry port sit on a high terrace and the damage there reached up to 3 km from the channel. That is exactly why the second-tier list matters, and why Timure is on it.

What this cannot tell you

• This is not a flood forecast. Nobody has modelled how much water would come down or where it would go — there is no river gauge record behind this. A settlement listed at 45 m above the water is not predicted to flood. It is inside the zone where the damage from this event mostly happened.
• The heights are satellite measurements, not survey. The elevation data measures the top of whatever it hits — treetops and roofs included — so heights run a few metres high, and on steep valley sides any single figure is worth give or take 10–20 m. Read everything between 40 and 60 m as one group rather than ranking those places against each other.
• House counts are lower bounds. The building maps are volunteer-made and uneven: around Syabrubesi they hold 37 buildings where the emergency survey found 202. Outside the three surveyed stretches, expect the real number of houses to be higher than shown.
• Every population figure is pre-flood. It is the houses counted in the zone times the census household size, and it takes no account of who has already been killed, displaced or evacuated since 26 August. For the 18 destroyed settlements it is a record of who lived there, not of who is there now.
• 279 named places inside 2 km sit above the band and are held separately so the exclusion can be audited. Four unnamed OSM place nodes also fall in the band and carry no name to report.
• Safe-site distances come from the existing relief-workstream safe-camp survey, which only covers the three EMS windows. A 9 km figure in the lower valley means no site has been surveyed nearby, not that evacuation requires a 9 km walk.

Terms used on this page

Water level

The height of the river surface itself, in metres above sea level, worked out from satellite elevation data along the whole channel. It falls steadily downstream, so each settlement is compared with the water beside it.

The zone (or band)

The strip of ground that is both within 2 km of the river and no more than 50 m above its surface — 34.5 km² in total along this valley.

Palika and ward

Nepal's local government units. A palika is the rural or urban municipality; a ward is the numbered subdivision inside it. Together they are the address relief coordination works from.

km down

Distance measured along the river from Rasuwagadhi, so places can be found in order going downstream rather than by map coordinates.

People

The houses counted inside the zone times 4.37, the national average household size from Nepal's 2021 census. The smaller figure beside it is an independent modelled population grid, kept as a cross-check.

With neighbours

The same count for this settlement plus every other one within a kilometre. A row is a settlement point — in a bazaar, a tole — so this is the figure to quote when someone asks about a town by name.

Copernicus EMS

The European Union's emergency mapping service. After the flood it examined satellite imagery of three stretches of this valley and graded each building as destroyed, damaged or possibly damaged. The rest of the valley was not surveyed.

Stretch of valley

Five working sections, each 8 to 10 km long and holding 8 to 13 settlements, divided at Betrawati, Bidur, the Devighat powerhouse and Tarukabesi. The long stretch above Betrawati is a gorge and holds few settlements despite its length.

Condition

What the Copernicus survey found on the ground. Destroyed means 60% or more of the settlement's known buildings were graded destroyed; heavily damaged means 25% or more. Not surveyed means the settlement lies outside the three surveyed stretches.

Second tier

Settlements whose centre is above the 50 m line but which have lower ground inside the zone. They fail the test on a technicality and should be treated as exposed.

Safe site

Ground already checked by the relief team as high enough and flat enough to camp on. Only three stretches of the valley have been checked, so a large distance here means none has been surveyed nearby — not that people must walk that far.