# Budhi Gandaki disaster: satellites map candidate landslide areas in Gorkha **Event:** Landslide and debris flood in the Budhi Gandaki valley, early on 11 October 2026 **Where:** Chumnubri Rural Municipality, Dharche Rural Municipality, Gorkha District, Gandaki Province, Nepal **Event date:** 11 October 2026 **Imagery compared:** Sentinel-2 1–6 October 2026 against 21–26 October 2026; Sentinel-1 radar from two orbit directions **Facts checked:** 2026-11-13 · **Analysis packaged:** 2026-11-13T12:57:33Z **Status:** Satellite-detected candidate landslide areas, not confirmed on the ground. Impact figures are as reported; several are provisional. ## The story in brief In the early hours of 11 October 2026, a mass of rock, ice and debris came down the Ungdilung Khola above Chumchet in Chumnubri Rural Municipality, Gorkha. It blocked the Syar Khola and sent a series of mud-laden surges down the Syar and Budhi Gandaki rivers. Nepal's Department of Mines and Geology timed the landslide at about 04:20 local time. It estimated the largest surge at 20–25 m high when it hit the Budhi Gandaki hydropower camp at Salleri at about 06:30. At Arughat, the river gauge passed its 5 m danger level at 07:50 and peaked at 8.43 m at 08:40. The exact mechanism is still being discussed. Early reports described a landslide dam on the Syar Khola that burst. The Department of Mines and Geology describes a high-altitude avalanche and landslide whose debris briefly ponded the Syar Khola. Landslide researcher Dave Petley reads post-event Sentinel-2 imagery as a rock-slope failure that turned into a rock avalanche and debris flow, without a dam collapse. Rain was not the trigger. Our comparison of satellite images from before and after the disaster picks out **205 satellite-detected candidate landslide areas covering about 8.10 km²** on the slopes of the study area. These are places where green vegetation was stripped to bare ground or debris between early and late October. They are candidates, not confirmed landslides: some may be older scars, rockfall, or changes unrelated to 11 October. Separately, **78 patches of valley-floor change (about 52.7 ha)** lie within 300 m of mapped rivers. These are consistent with ground scoured or buried along the flood and debris path, although that has not been verified on the ground. ## What has been reported The figures below are as published by the sources shown. Where authorities have not confirmed a figure, it is marked as provisional. - **People missing:** 5 (herders at a shed on the Ungdilung Khola, Chumnubri-6); no deaths confirmed — official, as reported on 12–13 October; Nepal Red Cross listed 6 out of contact on 29 October. Source: [Kathmandu Post, citing Gorkha District Administration Office and police; NDRRMA situation report via Nepal News](https://kathmandupost.com/national/2026/10/12/torrent-of-black-mud-and-rocks-engulfs-settlements-along-budhigandaki), 12–13 October 2026. - **People rescued by helicopter:** 95 on the first day; 1,044 between 11 and 15 October, including 70 foreign nationals — official. Source: [NDRRMA situation report via Nepal News (12 October); Gorkha Assistant Chief District Officer via Nepal News (15 October)](https://english.nepalnews.com/s/nation/budhigandaki-floods-more-than-1000-people-rescued-by-helicopter/), 12 and 15 October 2026. - **Houses damaged or destroyed:** about 204 across Chumnubri, Dharche, Arughat (Gorkha) and Tripurasundari (Dhading) — provisional; the split between destroyed and damaged was not confirmed. Source: [NDRRMA situation report via Nepal News](https://english.nepalnews.com/s/nation/budhi-gandaki-flash-flood-authorities-estimate-relief-needs-for-21-days/), 12 October 2026. - **Households affected:** more than 550 — provisional; Nepal Red Cross counted 146 on 29 October, probably with a different scope. Source: [WFP, citing NDRRMA](https://reliefweb.int/disaster/ff-2026-000189-npl), 13 October 2026. - **Bridges:** 2 motorable and 2 suspension bridges lost, including the Arughat bridge linking Gorkha and Dhading — provisional; later Nepal Red Cross figures are higher (2 permanent, 1 Bailey and 6 suspension). Source: [WFP, citing NDRRMA; Nepal Red Cross Bulletin 04](https://website-api.nrcs.org/media/news/attachments/Gorkha_Flash_Flood__Bulletin_4.pdf), 13 and 29 October 2026. - **Hydropower:** 3 projects under construction damaged; the 341 MW Budhi Gandaki project camp at Salleri was swept away — official. Source: [NDRRMA situation report via Nepal News; Department of Mines and Geology report via Telangana Today/IANS](https://english.nepalnews.com/s/nation/budhi-gandaki-flash-flood-authorities-estimate-relief-needs-for-21-days/), 12?14 October 2026. ## What the satellites show - **Candidate landslide areas:** 205 (8.10 km²). - 1 area (0.3 ha) show both optical and radar change: the strongest evidence. - 187 areas (7.86 km²) show optical change only. - 17 areas (24.4 ha) are radar-only detections where cloud or shadow hid the ground; treat these with the most caution. - **Valley-floor change beside rivers:** 78 areas (52.7 ha). - **Roads and paths:** about 27.7 km of OpenStreetMap roads, tracks and trails lie within 30 m of a candidate landslide area. - **Buildings:** 126 OpenStreetMap building footprints lie within 50 m of a candidate area. This shows exposure, not damage. - **Coverage:** a usable optical comparison was possible for 83% of the study area; radar covered 100%. ### Largest candidate areas | ID | Area (ha) | Evidence | Mean slope | Nearest mapped place | Roads or paths within 30 m | |---|---:|---|---:|---|---| | LS001 | 576.0 | optical | 43° | Chumling (3.0 km) | Tsum Valley Path, path | | LS002 | 31.5 | optical | 41° | Chumling (4.8 km) | — | | LS003 | 27.2 | optical | 44° | Chumling (2.3 km) | Tsum Valley Path | | LS004 | 12.6 | optical | 39° | Lokpha (0.6 km) | — | | LS005 | 6.1 | optical | 30° | — | — | | LS006 | 5.9 | optical | 44° | Chumchet (0.7 km) | — | | LS007 | 5.2 | optical | 26° | Philim (0.3 km) | Manaslu Circuit, path | | LS008 | 4.8 | optical | 22° | Philim (0.5 km) | Manaslu Circuit, path, unclassified | **LS001, the largest candidate.** The largest candidate zone covers forested valley sides below the reported source, where trees turned grey or were stripped between the before and after images. Its sharp edges fit the air blast, dust and debris of a rock avalanche better than a single landslide scar. Treat it as a zone of disturbance that needs field or very-high-resolution confirmation, not as one slide. ### Roads and paths closest to candidate areas - Manaslu Circuit: about 13.66 km within 30 m of candidate areas - Unnamed path: about 6.17 km within 30 m of candidate areas - Unnamed minor road: about 5.46 km within 30 m of candidate areas - Tsum Valley Path: about 2.27 km within 30 m of candidate areas - Tatopani Foot-Bridge: about 52 m within 30 m of candidate areas - Njak Phedi Bridge: about 51 m within 30 m of candidate areas - Unnamed track: about 40 m within 30 m of candidate areas ## How we did it 1. **Optical (Sentinel-2 L2A).** We averaged cloud-free pixels from 3 dates before (1–6 October 2026) and 3 dates after (21–26 October 2026), using the scene classification layer to remove cloud, cloud shadow and snow. A pixel was flagged when its vegetation index (NDVI) fell by at least 0.20 more than the typical fall for its altitude, and ended at or below 0.25, which is typical of fresh bare soil, rock or debris. The typical fall is the median for each 200 m band of altitude, which removes autumn browning of high meadows. Above 4,000 m, where browning, fresh snow and shadow still mimic scars, optical change counts only if radar agrees. 2. **Radar (Sentinel-1).** Radar sees through cloud. We compared terrain-corrected VH backscatter from descending and ascending passes, averaging two dates on each side and smoothing over 5 × 5 pixels to reduce speckle. A change of 3 dB or more supports an optical detection. Where the optical view was blocked, a change of at least 4 dB over at least 0.5 ha is reported as a lower-confidence, radar-only candidate. 3. **Masks.** We removed: - slopes gentler than 15°, using Copernicus DEM GLO-30; - water; - cloud, cloud shadow and terrain shadow; - snow and ice; - cropland, built-up land and wetland, using ESA WorldCover 2021, so that the October rice harvest on terraces is not mistaken for slides; - OpenStreetMap buildings and residential land. 4. **Clean-up.** Groups smaller than 0.2 ha were discarded. The remaining pixels were turned into polygons and attributed with OpenStreetMap roads, paths, buildings and the nearest mapped place. ## Limits - **Candidates, not confirmed slides.** Vegetation loss can also come from rockfall, burning, landslides earlier in the monsoon that were hidden by cloud in the "before" images, or seasonal browning. Field reports or very-high-resolution imagery are needed to confirm any individual slide. - **Dates.** The before and after images are about three weeks apart. Change between them cannot be pinned to 11 October alone. - **Steep terrain.** Radar suffers layover and shadow on steep Himalayan slopes, and those areas have no radar evidence. Deep shadow also hides some north-facing slopes in October optical images. - **Resolution.** Ten-metre pixels miss narrow slides and debris channels, and areas are planimetric, meaning they are not corrected for slope. - **Coherence.** Interferometric coherence was not used. It needs Sentinel-1 SLC processing that was not practical for this story, so radar evidence is backscatter change only. - **OpenStreetMap completeness.** Roads, trails and buildings in this mountain area are incomplete, so proximity counts are minimums. ## Cross-check No Copernicus Emergency Management Service activation was found for this event, and ICIMOD has not published a landslide delineation. The nearest independent product is UNOSAT's preliminary map for the International Charter, made from Sentinel-2 imagery of 11 October. It estimates about 18 km² affected by flood water or mudflow across a 470 km² analysis area and about 33 km of road affected; UNOSAT notes it has not been validated in the field. That map measures the flood and mudflow footprint, whereas ours looks for vegetation loss on slopes, so the two complement rather than duplicate each other. Our valley-floor change class is the closer comparison with UNOSAT's footprint. The main landslide source reported by Dave Petley lies at about 28.534°N, 84.935°E, above 5,000 m on the west side of the valley, south of Chamar. Much of that zone is bare rock, snow or ice, so a vegetation-loss method is not expected to pick it up. Sentinel-1 radar there shows no strong, consistent change: within 1 km of the reported source the median VH change is under 1 dB, and about a tenth of pixels rose by 3 dB or more, which fresh snow or ice alone could cause. The radar therefore neither confirms nor rules out the source. The August 2026 Rasuwa–Langtang disaster (Copernicus EMS activation EMSR927) happened in a different valley and was not used to validate this analysis. ## References 1. UNOSAT / International Charter Space and Major Disasters (via ReliefWeb), “Satellite detected mudflow on 11 October 2026, Gorkha District, Gandaki Province, Nepal (v1)”, 15 October 2026. 2. Sentinel Asia, “Flood in Nepal (activation requested by DHM; escalated to the International Charter; GLIDE FF-2026-000189-NPL)”, 11 October 2026. 3. ICIMOD, “Landslide-dam breach may have triggered Budhi Gandaki flash flood (media advisory)”, 11 October 2026. 4. Nepal Red Cross Society, “Gorkha Flash Flood Situation Bulletin 04”, 29 October 2026. 5. Telangana Today / IANS, summarising the Department of Mines and Geology report released by NDRRMA, “Landslide triggering Budhi Gandaki River floods in Nepal unleashed energy equivalent to 3.5-magnitude quake: Report”, 14 October 2026. 6. Dave Petley, The Landslide Blog (Eos/AGU), “The 11 October 2026 landslide and debris flow/flood on the Budhi Gandaki River in Gorkha district, Nepal”, 12 October 2026. 7. Dave Petley, The Landslide Blog (Eos/AGU), “Sentinel images of the source of the 11 October 2026 landslide and debris flow along the Budhi Gandaki River in Nepal”, 5 November 2026. 8. EU Agency for the Space Programme, “Landslide-triggered flash flood sweeps through Nepal's Himalayan valleys (Copernicus image of the day)”, 14 October 2026. 9. BBC News (Navin Singh Khadka), “Homes and bridges swept away as fresh floods hit Nepal”, 12 October 2026. 10. The Kathmandu Post (Hariram Upreti), “Torrent of black mud and rocks engulfs settlements along Budhigandaki”, 12 October 2026. 11. Nepal News, “Budhi Gandaki flash flood: authorities estimate relief needs for 21 days (NDRRMA situation report)”, 12 October 2026. 12. Nepal News, “Budhigandaki floods: more than 1,000 people rescued by helicopter”, 15 October 2026. ## Data sources and licences - Copernicus Sentinel-2 Level-2A surface reflectance — Copernicus Sentinel data terms (free, full and open). Attribution: “Contains modified Copernicus Sentinel data [2026]”. - Copernicus Sentinel-1 radiometrically terrain-corrected backscatter (from GRD) — Copernicus Sentinel data terms (free, full and open); RTC processing by Microsoft Planetary Computer, CC BY 4.0. Attribution: “Contains modified Copernicus Sentinel data [2026]”. - Copernicus DEM GLO-30 — Copernicus DEM licence (free, worldwide, with attribution). Attribution: “Copernicus DEM GLO-30 © DLR e.V. 2010–2014 and © Airbus Defence and Space GmbH 2014–2018, provided under COPERNICUS by the European Union and ESA”. - ESA WorldCover 10 m 2021 v200 — CC BY 4.0. Attribution: “© ESA WorldCover project 2021 / Contains modified Copernicus Sentinel data (2021) processed by ESA WorldCover consortium”. - OpenStreetMap roads, paths, buildings, places and rivers (Overpass API) — Open Database Licence (ODbL) 1.0. Attribution: “© OpenStreetMap contributors, available under the Open Database Licence (ODbL)”. - EOX Maps WMS (OpenStreetMap locator) — OpenStreetMap data under ODbL; rendering by EOX and MapServer. Attribution: “Locator: data © OpenStreetMap contributors (ODbL); rendering © EOX and MapServer”. Contains modified Copernicus Sentinel data [2026]. OpenStreetMap data © OpenStreetMap contributors, available under the [Open Database Licence](https://opendatacommons.org/licenses/odbl/1-0/). The map and derived candidate polygons are released by TechnicalGIS under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/). OpenStreetMap-derived attributes in the GeoJSON and CSV (road names, place names, proximity counts) remain under the ODbL. Prepared by TechnicalGIS Environmental News Desk. Package revision 2.