Sikkim and the 1% Flood

Just after 10 pm on 3 October 2023, a slab of frozen mountainside — roughly 14.7 million cubic metres of lateral moraine — sheared off and crashed into South Lhonak Lake, a glacial lake perched below Kangchenjunga in far north Sikkim. The impact threw up a wave about 20 metres high that overtopped and tore open the lake’s natural dam. In minutes, some 50 million cubic metres of water were loose in the Teesta gorge, building to a peak discharge of about 48,500 cubic metres per second — a wall of water and sediment funnelling down one of the steepest river valleys on Earth.

By 12:30 am on 4 October, it reached Chungthang and destroyed the 1,200 MW Teesta-III dam — the largest hydropower project in the state — as if it were not there. The flood scoured 385 km downstream, ripping out more than 30 bridges, army camps, roads and whole riverside settlements, and carried debris out of the mountains into Jalpaiguri district in north West Bengal. The Government of Sikkim recorded at least 46 dead, more than 77 missing, and about 88,400 people affected; roughly 25,900 buildings were damaged or destroyed. It is the defining Himalayan flood disaster of the modern era — and it explains almost everything about how to read Sikkim in this atlas.

Sikkim’s flooding is not the plains’ flooding. There is no broad floodplain here to inundate. The entire state is essentially the Teesta gorge and its tributaries, dropping thousands of metres from the Kangchenjunga massif to the Bengal border. Its water hazard is sudden: glacial-lake outburst floods (GLOFs), cloudburst flash floods, and the landslides that come with them. This piece uses the India Flood Atlas to show where that risk concentrates — and is candid that in the steepest state in the country, the atlas’s modeled depths are the least reliable numbers it produces.

The scale

Sikkim is a small state, and its flood footprint is small in absolute terms but not in share:

• Of the state’s 357,000 building footprints, 24,588 — 6.9% — sit in the modeled 1% annual-chance (100-year) floodplain. That is roughly one building in fourteen, a meaningful share for a state where almost every settlement clings to a valley side above a river.
• About 798 km of road — 8.8% of the network — lies in the modeled floodplain. Note one gap: the new Sevoke–Rangpo railway, Sikkim’s first, is not yet in the exposure data, so rail risk is not captured here.
• The atlas returns a modeled 29 m mean depth over exposed buildings — the highest in the atlas — but this is not a usable depth. It reflects a roughly 30 m DEM struggling to represent the near-vertical Teesta gorge, not 29 metres of water. On the same distorted basis, 93% of exposed buildings model as being under more than 3 m of water — which is a sign of the terrain problem, not a reading of reality. (More on this below; it is the single most important caveat in this article.)

So the reliable headline is the extent: 6.9% of buildings and 8.8% of roads fall inside the corridor a severe flood would follow. Where is trustworthy. How deep is not.

Where it concentrates

Risk pools in the Teesta and its tributary corridors — the valley floors where roads, bridges, towns and hydropower all compete for the same narrow ribbon of flat ground. By share of buildings exposed across Sikkim’s four districts:

District	% of buildings exposed	Exposed buildings	Modeled depth
North Sikkim	19.0%	6,143	DEM-uncertain
East Sikkim	6.9%	7,592	DEM-uncertain
South Sikkim	5.7%	6,394	DEM-uncertain
West Sikkim	4.4%	4,459	DEM-uncertain

North Sikkim stands out at 19.0% — nearly one building in five. That is exactly where the hazard is greatest: North Sikkim holds the high glacial lakes, including South Lhonak, and the upper Teesta corridor through Chungthang and Lachen–Lachung that the 2023 GLOF devastated. East Sikkim carries the largest count of exposed buildings (7,592), because that is where the population concentrates — Gangtok, Singtam and the lower Teesta towns. The depth column is deliberately left as “DEM-uncertain” for every district; in this terrain, a district-level mean depth would be a false precision.

Why Sikkim floods

1. Glacial-lake outburst floods (GLOFs). Sikkim’s mountains hold hundreds of glacial lakes — inventories range from around 230 sizeable lakes to over 1,100 water bodies depending on the threshold used. As glaciers retreat under warming, many of these lakes are growing behind fragile moraine dams. When one bursts — as South Lhonak did — it releases tens of millions of cubic metres in minutes. This is the state’s signature, and deadliest, flood.
2. Extreme vertical relief. Sikkim falls from over 8,500 m to under 300 m across a few tens of kilometres. That gradient turns any water release into a fast, erosive torrent, and leaves settlements no room to retreat — the only flat land is the valley floor the flood claims.
3. Cloudburst flash floods. Intense, localised monsoon downpours on near-vertical slopes generate flash floods and debris flows with almost no warning, independent of any glacial lake.
4. Landslide-dammed and landslide-driven floods. The 2011 Mw 6.9 Sikkim earthquake killed at least 111 people and triggered nearly 1,200 landslides — a reminder that in this terrain the ground itself moves, blocking rivers and then releasing them, and that seismic and flood hazard are entangled.
5. A hydropower cascade in the flood path. The Teesta is lined with a chain of dams. A GLOF or flash flood hitting one — as at Teesta-III — can cascade downstream, and critics argue the projects were cleared despite the GLOF risk being flagged in the basin’s own carrying-capacity studies.
6. Climate change is the accelerant behind both the swelling glacial lakes and the intensifying cloudbursts.

Depth here is a warning, not a measurement

This is the caveat that governs how to use the atlas for Sikkim, and it deserves to be stated plainly.

The atlas models flood depth on a digital elevation model of roughly 30 m resolution. On a flat delta, that is fine — the ground barely changes across a 30 m cell. But Sikkim is near-vertical: within a single 30 m pixel, the real ground can drop tens of metres from a valley wall to the river bed. When the model tries to fill that gorge with a 1% flood, the arithmetic produces enormous “depths” — the 29 m mean, the 93% of buildings over 3 m — that are artifacts of the terrain colliding with the resolution, not measurements of water. They are the least reliable depth figures in the entire atlas. They should never be quoted as literal depths, and we do not.

What the atlas can say reliably is different, and still valuable:

• Where water concentrates — the Teesta trunk and its tributary corridors — is robust. The floodplain footprint (6.9% of buildings, 8.8% of roads, 19% in North Sikkim) does not depend on resolving exact depth.
• Sikkim’s real risk is sudden GLOF and flash-flood events, where what saves lives is detection, early warning and evacuation time — not a modeled depth number. A GLOF gives minutes to hours of warning if a monitoring and alert system exists; South Lhonak had been flagged for years as high-risk, and the tragedy was in part a warning that was not acted on in time.

Read the atlas here as a map of the danger corridor, not a depth gauge.

A strategic approach

Sikkim’s strategy is not about walls and floodplain zoning across a broad plain — it is about watching the mountains and buying time:

Pillar 1 — Monitor and drain the dangerous lakes. Instrument the high-risk glacial lakes (South Lhonak’s level was already being lowered with siphons before 2023; the effort was too limited). Continuous satellite and in-situ monitoring of the highest-hazard lakes is the foundation.

Pillar 2 — Real-time GLOF and flash-flood early warning. Automated sensors on lakes and upper-catchment rivers, tied to sirens and mobile alerts in the valley towns below. In a GLOF, minutes of lead time are the difference between evacuation and catastrophe.

Pillar 3 — Rethink the hydropower cascade. Design dams for GLOF-scale discharges, or reconsider the most exposed projects. A dam that fails becomes part of the flood, as Teesta-III did.

Pillar 4 — Keep development out of the danger corridor. The valley floor is the flood’s path. Restrict critical infrastructure, camps and dense settlement in the Teesta and tributary corridors the atlas maps.

Pillar 5 — Prepare communities to move. Rehearsed evacuation routes to high ground, resilient bridges and lifelines, and disaster drills in the exposed towns. In a state where the flood arrives in minutes, preparedness beats any wall.

The honest bottom line: Sikkim is the steepest state in the atlas, and its flooding does not behave like the plains’. Its danger is the sudden flood — the glacial lake that bursts, the cloudburst that becomes a torrent — funnelled down the Teesta gorge. The atlas tells you reliably where that corridor runs: 6.9% of buildings, 8.8% of roads, and nearly a fifth of North Sikkim sit inside it. It cannot reliably tell you how deep, and its record-high modeled depths are a DEM artifact of the vertical terrain, not real water. October 2023 already showed what the real number looks like — 46 dead, 77 missing, a 1,200 MW dam gone. Here, resilience is not a depth map. It is a monitored lake, a working siren, and a rehearsed route uphill.

Sources

• 2023 Sikkim GLOF — Wikipedia: 2023 Sikkim flash floods · Science: drivers, causes and impacts of a multihazard cascade · Eos: The 3–4 October 2023 Sikkim GLOF · SANDRP: Glacial lake flood destroys Teesta-3 dam
• Glacial-lake inventory and GLOF risk — ScienceDirect: increasing GLOF risk in Sikkim under warming · India Water Portal: GLOFs, Sikkim’s ticking time bomb
• Teesta hydropower risk — The Wire: ecological destruction behind Teesta damage · Deccan Herald: why were alerts on Lhonak lake ignored
• 2011 Sikkim earthquake — Wikipedia: 2011 Sikkim earthquake
• Exposure statistics — India Flood Atlas (modeled 100-year / 1% annual-chance event), NEER.

Exposure figures are modeled estimates (~30 m) for a 1% annual-chance flood, intended for awareness and prioritisation — not parcel-level certainty, and least reliable in the steepest terrain.