Sikkim Ranked Its Glacial Lakes: Using the Register Before Underwriters Do in the Teesta Corridor

Sikkim has identified 40 high-hazard glacial lakes in the state and sorted them into a risk order. Sixteen sit in Category A, the highest-risk band. Two are Category B, nine are Category C, and 13 remain unclassified. The classification was presented at a state GLOF risk-mitigation workshop in early September 2026, alongside the method behind it: continuous monitoring using remote sensing and satellite imagery, backed by field investigation that includes electrical resistivity tomography, discharge and water-level monitoring, and DGPS surveys.

For the state, this is a disaster-management input. For anyone placing property or engineering cover in the Teesta corridor, it is something different. Until now, glacial lake outburst flood exposure in North Sikkim was argued in generalities: the valley is glaciated, the 2023 South Lhonak event happened, therefore the peril exists. An underwriter could neither price that nor decline it with confidence, so the market defaulted to blunt instruments, mostly low sublimits and wide flood deductibles.

A published, named, ranked list changes the conversation. The peril now has an inventory. Sixteen specific water bodies carry a state-assigned top-tier hazard label, each with a traceable downstream channel. That is exactly the shape of data a reinsurer asks for when it decides whether a treaty will absorb Himalayan hydro and transmission risk or whether it wants the exposure carved out.

The register is a risk-management asset before it is an underwriting weapon

Risk managers who read the register first, and read it properly, control how it enters their renewal file. Risk managers who wait for a broker to forward an underwriter's question about Category A lakes are answering someone else's framing of their own exposure.

The register landed weeks after the market received a fresh number for what a Himalayan glacial collapse costs. On 26 August 2026, a glacial collapse on the Trishuli in Nepal triggered flooding that commercial insurers estimate could cost more than NPR 20 billion, roughly USD 132 million, with hydropower the largest single share of that figure. Across the wider flood event, the Nepal Insurance Authority had received 583 flood-related claims worth NPR 25.87 billion, about USD 171 million, as of 31 August 2026.

Two features of that loss matter for Sikkim placements.

First, the concentration. Hydropower dominating the insured share is not an accident of that particular valley. Run-of-river schemes are built where the gradient is steep and the channel is narrow, which is where an outburst wave carries the most destructive energy. Barrage, intake, desilting chamber, headrace and powerhouse sit strung along a single watercourse, so one event touches most of the asset base at once. Conventional sum insured allocation across locations gives no diversification benefit when the locations share a river.

Second, the speed of the claims count. Five hundred and eighty-three claims logged by 31 August tell you the event footprint extends well past the generating assets into road, bridge, contractor plant, transit cargo and commercial property in the valley towns. Corporates that only stress-test their own generating or manufacturing asset miss the contingent layer sitting in their supply corridor.

Expect the register to work through Indian property and engineering placements in three ways over the 2026 to 2027 renewal cycles.

1. Question sets get specific. Proposal forms and underwriter queries move from "is the risk exposed to flood" to "state the distance and vertical drop between the insured location and the nearest Category A lake, and the channel it drains into". A vague answer will be read as a poor answer.
2. Sublimits get keyed to the register. A GLOF sublimit that currently sits as a flat figure will start to vary by category exposure. Locations downstream of a Category A lake should expect a tighter sublimit, a higher deductible, or both, unless the buyer brings mitigation evidence.
3. Exclusion language gets sharper. The risk is a drafting shift from a general GLOF sublimit toward a named-source exclusion that references state-classified high-hazard lakes. That wording, if it enters a schedule unchallenged, can strip cover for the one scenario the buyer most needs it for.

The precedent for how badly a mismatched sublimit ages is already in the market. The 2023 South Lhonak outburst hit the Teesta-III hydro scheme, where the GLOF cover available was a small fraction of the declared sum insured. The gap was structural, not a claims dispute. Our earlier analysis of GLOF exposure in Himalayan hydropower, transmission and project cargo sets out how those wordings fail in practice.

The register makes that structural gap harder to defend in either direction. A buyer can no longer say the peril was unforeseeable. An insurer can no longer say the exposure was unmodellable.

The work a risk team should do before its next renewal is a mapping exercise, not a modelling exercise. It takes days, not months.

Build an asset schedule for every location in the Teesta and its tributary corridors and record, for each:

• Distance along the watercourse from the nearest classified lake, measured along the channel rather than straight line.
• Vertical elevation above the active channel bed and above the 2023 flood mark where one is observable.
• Category of the upstream lake, and whether any upstream lake is in the unclassified 13.
• Whether the location's only road access crosses the same channel.
• Whether critical utility feeds, transmission line towers, or fibre routes cross the same channel.

That last pair of rows is where most schedules are thin. An asset can survive the wave and still stop earning for months because the highway crossing, the substation, or the evacuation line went with it. That is a business interruption loss with no material damage at the insured location, and whether it responds depends entirely on how the denial-of-access and public-utilities extensions are drafted and sublimited.

Aggregation across the corridor

If a group holds several sites in the same valley, the correct unit of accumulation is the watercourse, not the district. Two plants 40 km apart on the same river are a single event exposure. Our note on nat-cat accumulation across multi-location Indian corporates sets out the accumulation grouping logic; the Teesta corridor is the sharpest case of it in the country.

A state-published hazard classification changes the disclosure position. Indian general insurance contracts run on utmost good faith, which requires the proposer to disclose material facts, meaning facts that would influence a prudent underwriter's decision to accept the risk or the terms on which it is accepted.

Before the register, a buyer in North Sikkim could reasonably state that the site was in a glaciated catchment and leave it there. After the register, a proposal that omits a known Category A lake directly upstream of the insured location is a weaker document. It does not automatically void cover, and an insurer would still have to establish materiality and the link to the loss. It does hand a claims department an argument it did not previously have.

The same logic works in the buyer's favour. A proposer who documents that its sites sit above a Category C lake, or outside any classified drainage, has a positive fact to price against rather than an underwriter's worst-case assumption about "Sikkim". Silence gets averaged into the pessimistic end of the range.

Five structural items are worth putting on the renewal agenda for corridor risks.

Separate the GLOF sublimit from the general flood sublimit. They are different perils with different loss shapes. One shared sublimit lets a bad monsoon year erode the capacity meant for the low-frequency total loss.

Size the sublimit against realistic damage, not against comfort. For a run-of-river asset in the inundation path, realistic maximum loss approaches total for civil works and full replacement for electro-mechanical plant. A sublimit set at a small fraction of the property insurance sum insured is a decision to self-insure the peril, and it should be recorded on the risk register as exactly that.

Extend the indemnity period past the reconstruction estimate. Post-event reconstruction in a high Himalayan valley is limited by the working season and by road restoration, not by procurement. An indemnity period sized on plains-India assumptions runs out while the site is still waiting for access.

Price denial of access and public utilities properly. These extensions carry small default sublimits in most Indian schedules. In a single-road valley they are among the most likely sections to be triggered.

Ask what a parametric layer costs. A trigger built on gauged water level or discharge at a defined station pays fast and covers the cash-flow gap while an engineering insurance or property loss is being surveyed. It is not a replacement for indemnity cover, and basis risk needs to be understood before it is bought, but for corridor assets the quote is worth obtaining.

The register's tail deserves attention because both sides will misuse it.

Thirteen lakes are unclassified. Unclassified means the state has not yet completed the assessment, which is a statement about data, not about hazard. An underwriter who reads unclassified as low risk is guessing. A buyer who reads unclassified as low risk and says so in a proposal is making a statement it cannot support.

The practical position is to treat an unclassified upstream lake as unresolved and say so explicitly. Ask whether the field investigation programme covers it, whether electrical resistivity tomography and DGPS survey work is scheduled for that basin, and record the answer. Category assignments will move as survey work completes, and a placement that hard-codes today's category into a multi-year wording will be wrong at some point in its term.

Category B and Category C locations have the opposite problem. Nine Category C lakes and two Category B lakes sit below the Category A tier, and the natural underwriter move is to grant broader terms there. That is defensible for pricing. It should not translate into removing GLOF from the schedule entirely, because the classification ranks the source, not the consequence at your specific site. A lower-hazard lake with a large volume and a narrow downstream gorge can still produce a destructive wave.

One practical ask covers both cases. Get the insurer to state, in the schedule or a side letter, which lake classification vintage the terms were priced against. When the state updates the register, you then have a defined reference point for a mid-term review instead of an argument about what was known when.

A workable order of operations for a risk team with assets in the Teesta corridor, running roughly 8 to 10 weeks before renewal:

1. Pull the state classification and map every insured location against it on the five data points listed earlier. Do this before the broker asks.
2. Separate the schedule into locations with a Category A upstream source, locations with B or C, locations with an unclassified source, and locations outside any classified drainage. These are four different risk conversations.
3. Read the current wording for the GLOF sublimit, the flood deductible, the denial-of-access and public-utilities extensions, and the indemnity period. Write down what each pays on a repeat of the 2023 event.
4. Quantify the gap in rupees at board level. A number gets a mandate; a description of a peril does not.
5. Take the mapping into the market as a disclosure document rather than waiting for a questionnaire. Buyers who present structured hazard data usually get better terms than buyers who answer questions about it.
6. Where the gap cannot be closed on indemnity terms at an acceptable price, decide consciously between retention, a parametric layer, and physical mitigation, and minute the decision.

The broader claims pattern behind all of this is set out in our review of climate and natural catastrophe claims trends in India. The Sikkim register is the first time an Indian state has handed the market a ranked source list for this peril. Whichever side reads it more carefully will set the terms.