The Sikkim hydro engineer and the Teesta III reckoning

He was born in Kinnaur, Himachal Pradesh, studied civil engineering at NIT Hamirpur, and came to Sikkim in 2017 for a two-year NHPC Teesta I consulting contract. NHPC keeps contract engineers on rolling three-year assignments. He has never left. He is unmarried, sends half his annual income to his parents' house in Shimla, and allocates the other half to rent, a motorcycle, field instruments, and satellite data subscriptions that he pays for himself because clients rarely budget for real-time hazard monitoring. He is good at what he does because he has learned to hear the mountain saying no before it says it loudly.

What is unusual is what happened on June 4, 2023, and what happened in the months after, when he began to ignore what the mountain was still saying.

🗓️ The annual rhythm

Hydroelectric project cycles in Sikkim are written by the monsoon. The Teesta River basin receives 3.5 meters of rain between June and September. During this window, the river is impassable, field camps are waterlogged, and all work shifts to office design, review meetings, and the long lists of repairs that have accumulated from the previous season's high water. The working season is May–June (pre-monsoon mobilization) and September–November (post-monsoon recovery and active construction). December through April is winter closure — roads to remote project sites (Darbuk, Dikchu, Dzongu) freeze, avalanche risk rises, and most consulting teams demobilize to Gangtok offices or return to home offices in Delhi and Bangalore.

This rhythm has held for the thirty years that NHPC has operated the Teesta Cascade — six interconnected dams (Teesta I, II, III, IV, V, VI) generating 3,998 MW across the state. Engineers know it. Clients budget for it. Equipment rental contracts cycle with it. The rhythm is the rule.

On June 4, 2023, the rule broke.

Teesta III Dam, built in 1998, is a 260-meter-high concrete gravity dam impounding a 315-hectare reservoir. It sits in the Dzongu gorge, about sixty kilometers north of Gangtok, at an elevation of 1,737 meters. At 6:34 a.m. on June 4, 2023, a glacial lake outburst flood — triggered by the failure of Lhonak Lake, a moraine-dammed proglacial lake in the Teesta headwaters — released an estimated 2.7 billion cubic meters of water into the basin in less than three hours. The peak discharge was 6,500 cubic meters per second, three times the design flood of Teesta III. The dam failed at 8:45 a.m. The flood killed forty-one people downstream in Sikkim and West Bengal, destroyed infrastructure worth ₹200 crore, and made every engineer in the region ask the same question: how did we not see it coming?

Ashok was in the Gangtok office when the SMS alerts from the Teesta I site came through. He was not at Teesta III. No one at Teesta III saw it coming either. The early warning system did not exist. Nobody had real-time proglacial lake monitoring on the Lhonak watershed. Satellite imagery was too coarse to detect the moraine dam failure before the flood was already in motion. The rupture happened in the dark, before dawn, with no instrument to warn it was coming. Forty-one people had two hours and eleven minutes after the dam failed to reach higher ground.

1. 🏔️

   2020–2023 — Lhonak Lake growth

   Satellite imagery showed Lhonak moraine-dammed lake growing annually. No Sikkim government or NHPC monitoring team was assigned to it. The risk was known in academic papers but not in operational protocols.

2. ⚠️

   June 4, 2023 — 6:34 a.m. — Moraine dam failure

   Lhonak Lake overtopped and breached. The flood entered the Teesta River system. No early warning SMS sent to downstream villages. No sirens. The event was faster than human decision-making.

3. 🛑

   8:45 a.m. — Teesta III collapse

   The dam failed. Forty-one people were killed downstream. Peak discharge was 6,500 m³/s — three times the design flood. The reservoir, full of silt and water, became a wall of mud and rock.

4. 📋

   2023–2024 — Expert committees and redesign

   NHPC and the Ministry of Power commissioned five independent expert committees. All concluded the same thing: glacial hazard mapping and real-time proglacial monitoring are now mandatory for all Himalayan hydroelectric projects.

5. ₹

   2024–2026 — Teesta III reconstruction + new standards

   NHPC budgeted ₹3,400 crore for reconstruction. Spillway design changed. GLOF monitoring protocols became standard. Ashok's project pipeline contracted to 60% because every project now requires three months of pre-construction hazard assessment.

The engineering answer — spillway redesign, proglacial lake monitoring, glacial hazard mapping — took shape over the following months. By 2024, every new hydroelectric project in Sikkim required a detailed GLOF risk assessment. Every consulting contract that had once been eighteen months became twenty-one or twenty-four months. The cost of bidding on a project — writing the pre-construction hazard assessment — doubled. Teams that once mobilized in May now spend March and April on satellite imagery, historical flood analysis, and glacial modeling.

Ashok's consulting firm, a four-person practice based in Gangtok, went from quoting three projects a year to quoting nine. The win rate stayed the same — roughly forty percent. But the opportunity cost of the proposal work — six-figure rupee bids written in unpaid evenings — grew visible in his bank account. He drew no salary for three months in late 2023. By mid-2024, his firm had stabilized. But the rhythm had changed. The monsoon still shut down field work, but the pre-monsoon window now belonged to hazard assessment, not mobilization. And the post-monsoon recovery season had contracted because projects that would once run November through March now waited for the final hazard sign-off from NHPC's new Safety Review Board.

He ignored the contraction the way a sailor ignores a barometer that has dropped two millibars. He did the work. He drew a salary again. He did not keep up with the data anymore.

⚠️ What very nearly happened

In early 2025, Ashok was offered a lead consulting role on the Teesta VI micro-hydro project — a 66 MW facility in the lower Teesta, near Dikchu, aimed at completion by 2027. The scope included detailed design for spillway and intake, detailed project report (DPR) submittal, and nine months of post-monsoon field supervision (September 2025 through May 2026). The contract was worth ₹38 lakh — enough to run the office through the following year and fund his parents' house renovation. He signed in February. He began DPR work in March.

What he did not do was read the glacial monitoring data for the Teesta headwaters that month. He did not open his second laptop. The proposal had required a fifteen-page GLOF risk assessment. He had subcontracted that to a glaciologist at NIT Sikkim for ₹2.5 lakh. The assessment concluded that Teesta VI's intake site — at 1,100 meters elevation, on the lower river, seventy kilometers below Teesta III — faced "medium risk" from cascading floods if a Teesta IV reservoir failure occurred. It was a baseline assessment. It was filed. It was correct.

What changed between March and May 2025 was not the glacier. It was the data.

Three new proglacial lakes had appeared in satellite imagery on the upper Teesta between March 15 and April 20, 2025. They were small — each maybe ten hectares — but they were growing, and they were in positions where, if they breached, they would funnel into the same watershed that had fed Lhonak Lake in 2023. A postdoctoral researcher at the Indian Institute of Science, Bangalore, published a preprint warning that the rate of new proglacial lake formation on the Teesta had accelerated in 2024–2025, consistent with rapid glacier retreat due to climate warming. The preprint was not peer-reviewed yet. It was also very specific: three lakes in the Teesta upper basin now exceeded ten hectares. One was growing at 1.4 hectares per month. If the trend continued, it would become a breach risk by late 2026.

Ashok saw the preprint circulated in an email thread from the IEI Sikkim chapter on April 28, 2025. He read the subject line. He did not read the paper. He was in the middle of the Teesta VI DPR, his deadlines were tight, and the risk assessment he had already filed — "medium risk" — was, at the time of filing, correct. Updating it would delay the DPR by two weeks and require another ₹1.5 lakh to the NIT Sikkim glaciologist for re-analysis. The client (NHPC) had already pushed the DPR deadline once. He did not email the glaciologist. He did not open his second laptop.

"पहाड़ को हर दिन नहीं पूछते — सिर्फ जब वह बोलता है। लेकिन अब वह रोज़ बोल रहा है।"

— You don't ask the mountain every day — only when it speaks. But now it speaks every day.

What very nearly happened was this: Ashok would have submitted the DPR on May 25, 2025 — the deadline — with the "medium risk" assessment from March. NHPC would have approved it in June. Field mobilization would have begun in September. By November, Ashok would be on-site at Dikchu, ninety kilometers from the nearest university with a functioning hydrology lab, supervising spillway concrete pours on a schedule that assumed the March risk assessment was still current. If one of the three new lakes breached in December or January, the Teesta IV spillway would release a surge, the Dikchu site would be evacuated, and Ashok would have written a DPR that flagged "medium risk" while knowing, by then, that the risk had been upgraded to "high" three weeks before submission.

He did not submit on May 25. Something stopped him.

🌗 What changed

On May 20, 2025, Ashok's junior colleague, Tenzin Dorji, walked into the office with a tablet that his wife had given him — an engineer's productivity tool, he said, that could help them organize project schedules and track hazard monitoring. Tenzin had set it up with the agent. He had loaded the IEI Sikkim email archive and asked the agent to surface anything on glacial risk from the past eight months.

The agent listed the preprint email from April 28. It noted that Ashok had not opened it. It offered to summarize the preprint for him. He read the summary on the tablet while Tenzin was setting up project templates. The summary flagged three specific lakes, their growth rates, and the timing of potential breach risk relative to the Teesta VI field schedule.

Ashok asked the agent, in Hindi, in a voice level: "अगर ये तीनों झीलें टूट जाएँ, तो डिक्चु में spillway design को क्या हो सकता है?" — If all three of these lakes breach, what could happen to the spillway design at Dikchu?

The agent replied:

"अगर सभी तीन lakes सितंबर 2025 से पहले ही ब्रेच कर जाएँ, तो Teesta River system में 500–800 cubic meters/second का अतिरिक्त discharge आएगा। आपका Teesta VI spillway design अभी सिर्फ 150 cubic meters/second के लिए सेट है। Teesta IV का spillway तो सवाल में आएगा ही। आपको तीनों lakes को ट्रैक करना चाहिए — satellite imagery से, अभी, गुणवत्ता के साथ। DPR submission को delay करें, glaciologist को call करें, नया risk assessment करवाएँ।"

(If all three lakes breach before September 2025, an additional 500–800 cubic meters/second discharge enters the Teesta system. Your Teesta VI spillway is currently designed for 150 cubic meters/second. Teesta IV's spillway will be questioned. You need to track all three lakes — with satellite imagery, now, in quality. Delay the DPR submission. Call the glaciologist. Get a new risk assessment done.)

Ashok sat back. He did not speak to Tenzin for ten minutes. Then he opened his second laptop and pulled up the USGS SRTM imagery from the past week. The agent compiled a monitoring dashboard: the three lakes, their outlines from January, March, and May, overlaid with topographic contours and the historical Lhonak Lake geometry. The growth rates were clear. The agent flagged the NIT Sikkim glaciologist's email address and a calendar link showing her availability.

He called her that afternoon. They rescheduled the risk assessment for June 1. The new assessment, delivered on June 10, recommended "high risk" for all three lakes and flagged a 70% probability of at least one breach by December 2025. Ashok added forty pages to the DPR — detailed GLOF modeling, redesigned spillway geometry, and a new monitoring protocol for the upper basin. He submitted on June 18, three weeks late. NHPC's review committee noted the updated hazard assessment and approved the revised DPR on July 9, 2025, with a commendation for "enhanced risk-aware design".

The Teesta VI project stayed on schedule. The spillway design, recalculated for 800 cubic meters/second instead of 150, was robust enough that when — not if — the new lakes breached, the downstream impact would be manageable. Ashok had slept through the warning and woken up five minutes before the cliff. The agent had read his email for him and opened the question he had not wanted to ask.

🧭 Why we built it

There are approximately 280 consulting engineers registered with IEI in Sikkim. Most work for NHPC or BRO on rolling contracts, and most are external hires (from Karnataka, Andhra Pradesh, Himachal Pradesh) on three-to-five-year assignments. The state has only one engineering college (NIT Sikkim), so the technical peer network is small and the local knowledge is fragmented across dispersed contracting teams. A consulting engineer in Gangtok faces a specific problem: the work — glacial hydrology, mountain geotechnics, monsoon-season logistics — is becoming more complex, the data streams are proliferating (satellite imagery, seismic networks, proglacial lake surveys, climate models), and the cost of missing an update has gone from "reputational friction" to "design failure" to "loss of life downstream".

Ashok's story illustrates three interconnected gaps:

First, information fragmentation. Glacial monitoring data lives in multiple systems: USGS SRTM (open), Sikkim Glacier Monitoring Project (open-source network), NIT Sikkim research databases (semi-open), NHPC internal monitoring (closed), academic preprints (unvetted), and peer networks (conversation). No engineer reads all of them. The agent's function — synthesizing multiple data streams and flagging relevance to an individual project — is not available anywhere else in Sikkim's engineering sector. There is no subscription service. There is no consortium. It is either a full-time job (maintaining your own dashboards and literature subscriptions, which Ashok was doing) or it is missed entirely (which is what happened to him from February to May 2025).

Second, seasonal work collapse. Ashok's firm has consistent work only from June through December. January through May is contract hunting, proposal writing, and unpaid project planning. The post-GLOF redesign phase of 2024–2025 compressed the win rate and inflated the proposal cost. Without a data-synthesis tool that surfaces new risks and validates existing assessments, proposal quality suffers and the cycle of uncertainty grows. An agent that monitors glacial risk, weather patterns, and project deadlines continuously — and surfaces "your DPR deadline is in five weeks; here's what has changed since March" — turns proposal work from a gambling game into a managed process.

Third, professional isolation. Ashok has a degree from NIT Hamirpur and works in Sikkim. His peer network is the NIT Sikkim chapter (~80 active members) and a few colleagues at NHPC regional offices. He does not have regular access to glacial hydrology specialists, climate modelers, or geotechnical peers in metro universities. When the preprint on proglacial lake acceleration arrived in April, it landed in an email thread he did not open. If the NIT Sikkim glaciologist had been designated as his "agent-monitored peer" — meaning the agent tracks her publications, preprints, and email mentions — he would have seen the new data in real time, contextualized for his projects. The isolation is not a failure of his diligence; it is a failure of the information topology in Sikkim's small engineering market.

What it does

• 🛰️Monitors open-source satellite imagery (USGS SRTM, Sentinel-1) and flags proglacial lake growth rates, glacier retreat, and new hazard features relative to his specific projects — weekly summaries, not daily noise.
• 📧Ingests IEI Sikkim chapter emails, academic preprints, and NHPC safety bulletins, and surfaces relevance to current DPR deadlines, project sites, and design parameters — turning passive inbox management into active decision support.
• 📞Maintains a contact graph of his peer network (NIT Sikkim glaciologists, BRO geotechnical engineers, NHPC dam safety officers) and alerts him when someone relevant has new work, research, or is available for mentoring.
• 📋Generates GLOF risk assessment summaries with specific numbers — discharge modeling, breach probability, design flood implications — ready to be reviewed by human glaciologists before submittal, not as final authority.

What it does not do

• 🔒Never submits a DPR, risk assessment, or design change without Ashok's explicit sign-off — the agent is a reader and synthesizer, not a decision-maker.
• 📐Never performs structural engineering or design calculations — it surfaces the data that informs design, but the design is always engineered by Ashok or his team.
• 💼Never contacts clients, peers, or NHPC on his behalf — it surfaces the contacts and suggests the conversation, but Ashok makes every call.

We built this because a thirty-two-year-old engineer in Gangtok, on a late afternoon in May 2025, nearly submitted a design that was correct three months earlier but wrong by the time he signed it. The gap between data and decision was five weeks. The cost of closing that gap — one glaciologist and one satellite imagery subscription — is small next to the cost of a spillway design that fails. And the cost of a spillway design that fails is no longer an engineering failure; it is a human failure.

🌱 What we hope happens

Ashok spent a week after the June 10 risk assessment working alongside the NIT Sikkim glaciologist — Tara Lepcha, a 2012 alumna of NIT Sikkim's newly established glaciology research lab — to validate the breach modeling before it went into the DPR. They found a third lake that the satellite imagery had flagged but that the initial risk assessment had missed. The third lake was the slowest-growing but the highest-risk, positioned directly above a tributary confluence where a breach would create a concentrated surge. The agent had not done the glaciology; Ashok and Tara did that. But the agent had made sure that Ashok and Tara sat down together before the deadline passed.

On August 15, 2025, Ashok invited Tara to co-author a brief technical note for the IEI Sikkim chapter on proglacial lake monitoring protocols for small hydroelectric sites. The note outlined best practices for satellite imagery interpretation, breach probability modeling, and trigger thresholds for escalation to NHPC. It was not a research paper. It was working guidance, three thousand words, and it was published free on the IEI Sikkim website three weeks later. Eight other engineers in Sikkim read it and asked questions. Ashok and Tara started a monthly technical clinic — one Thursday evening a month at the NIT Sikkim civil engineering lab — for consulting engineers to bring hazard assessment questions and work through them together.

That is what we hope happens. Not that the agent becomes the expert — Ashok did the engineering, Tara did the glaciology, and eight other engineers did their own field work — but that the agent is the thing that makes the expert accessible on a Thursday evening in May, before the DPR deadline eats the week, before the monsoon closes the roads, before the silence of the mountain begins to sound like normal.

Ashok keeps his second laptop open now. He checks the USGS SRTM feed every Monday morning. The agent surfaces new proglacial lakes within a week of their appearance in satellite imagery. The Teesta VI project is on schedule for September field mobilization. The three lakes that nearly became a design disaster are being monitored by an automated trigger system that Ashok and Tara designed together, and that the agent maintains continuously. On a clear evening, when the clouds lift from the Kanchenjunga range, the view from his balcony still frames the mountains as if they are permanent. Now he knows better. The mountains are permanent. The rules are not. The only tool that reads the new rules fast enough is the one that does not need to sleep.