What we recommended after the 2015 earthquake, and what 26 August 2026 showed 

Bottom Line 

• Only 3 of 120 Nepal Telecom sites were flooded. On the worst day, 83 were down. The network failed on power and access, not water. Assessing that risk was Goal I of our 2016 report. 

• Day two, 40% restored, same pattern from the 2015 earthquake and the 2026 flood. Restoration started on the same curve, eleven years apart. The last few sites took weeks, because they could only be reached by helicopter. 

• The right kit arrived, as donations, days later. RREACT, Goal II, asked for restoration capacity that earns its keep every day, can be flown in and switched on within hours, and is written into operators’ service commitments. 

• Emergency services still worked independently. In 2026, fifteen coordination teams worked without one common picture. Goal III asked for systems used every day by police, ambulance and fire, so that they scale on the day of the crisis. 

• Ten years on, the 2026 regulator’s lessons are our 2016 goals. Knowing was never the gap. Deciding was. 

This is the second part of a three-part blog series examining Nepal’s Bhotekoshi disaster.

Part 2 of 3: why it happened. Part 1, “Nepal’s Bhotekoshi Disaster: Every Link Existed, Nothing Joined Them”, set out what happened on 26 August. Next, Part 3, “Degrade by Design, or Collapse by Default: A Governance Choice”, sets out how to design for the next crisis. 

In April 2016, LIRNEasia and ISOC-Nepal published an evaluation of Nepal’s emergency communication after the 2015 earthquake (LIRNEasia, 2016). We call it the 2016 report throughout. It set three goals. The Bhotekoshi disaster of 26 August 2026 tested all three. This post grades each one against the 2026 record. 

How We Graded 

Each recommendation is graded Done, Partly done, Not done, or Open (evidence not yet in), against the record of 26 August 2026 and the weeks after. We compare by function, not by hazard. The 2015 earthquake and the 2026 ice-rock avalanche were different hazards. The communication functions that failed were the same. 

The three goals of the 2016 report are also the three circles of the Nepal Internet Forum (NIF) 2026 keynote, shown in Figure 1: 

• Goal I. A telecommunications survivability and availability assessment, using risk mapping and step-wise refinement (RASTER), to support business continuity and disaster recovery plans (BC-DRP). 

• Goal II. A Rapid Restore of Access to Telecoms (RREACT) programme: interconnection, access points and backhaul in dark areas. 

• Goal III. Multi-agency situational awareness (MASA) and incident command platforms for all-hazard, all-media warning and response. 

The Report Card 

2016 report recommendation	Grade and 2026 evidence
Goal I (a). Assess network survivability	Not done. 80 of 83 Nepal Telecom sites down on the worst day were not flooded; they failed on power and transmission links.
Goal I (b). Define availability service levels	Not done. No emergency service levels defined; operator licences carry no emergency clause.
Goal I (c). Rehearse BC-DRPs, not just write them	Partly done. Operators told to prepare BC-DRPs in August 2026, with no targets or drills.
Goal I (d). Build sensors into survivability analysis	Not done. Upstream gauges never saw the flood; nobody was monitoring the slope.
Goal II (a). RREACT: rapidly deployable restoration capacity	Partly done. Satellite phones and terminals arrived by donation and helicopter, days later.
Goal II (b). Infrastructure sharing in a declared disaster	Partly done. Operators and the electricity utility coordinated; no mandate found.
Goal II (c). VHF/HF as a guaranteed fallback	Open. No 2026 evidence found.
Goal II (d). Design in the community layer	Not done. Police, a head teacher, volunteers, monasteries and schools stepped in unplanned.
Goal III (a). All-hazard, all-media alerting, including cell broadcast	Not done. One channel, SMS; cell broadcast still under discussion in July 2026.
Goal III (b). Register of alerting authorities (WMO OID) and CAP	Not done. One registered authority for Nepal; those who acted were unregistered.
Goal III (c). Always-on multi-agency situational awareness	Not done. Fifteen coordination teams, no common picture.
Goal III (d). Community radio continuity plans and an emergency band	Open. No 2026 evidence found.

Goal I: Survivability and Availability 

The network failed on power. The Nepal Telecommunications Authority’s figures for the affected area show it plainly (Karki, NTA, NIF 2026, slide 2). Nepal Telecom had 120 sites. Only three were flooded: Timure, Chilime Khola and Trishuli 3B. Yet 40 sites were down on 26 August, and 83 the next day. Ncell had 78 sites, one flooded and 27 down. NTA names the major issues as power for Nepal Telecom, and transmission links and power for Ncell. The count doubled on day two, as sites the water never reached ran out of power. The 2016 report recorded the same failure after the 2015 earthquake: tower batteries were sized for about a day. 

Who stayed dark, and for how long. The flood took about 430 MW of hydropower and solar off the grid, roughly a tenth of national capacity, and destroyed the 220 kV Trishuli 3B hub substation (Spotlight Nepal; OnlineKhabar). Market towns downstream had grid power back within a day. The places nearest the source waited longest. Dhunche waited five days and Bidur’s last wards thirteen (OnlineKhabar); Syabrubesi was still without grid power eleven days on (Kathmandu Post). The Timure tower ran on helicoptered diesel for about a month, until Nepal Telecom installed solar (Nepal News). 

Roads set the pace too. The entire 42 km road from Betrawati to the Rasuwagadhi border was damaged, and 41 motorable bridges were washed away (Kathmandu Post; Kathmandu Post). The road link between Rasuwa and Nuwakot reopened only on 11 October, six weeks later, on a Bailey bridge (Kathmandu Post). Until then, generators, fuel and equipment went in by helicopter, ropeway or on people’s backs (Clickmandu). Power and access are the two common-cause failures a survivability assessment exists to find. 

What Goal I asked for. Survivability means a network withstands shocks: damage, power loss, cut access. Availability means service stays usable through them, measured by how often it fails and how fast it returns (LIRNEasia, 2018). 

In 2018 we set the bar at “near 100% survivability & availability” and a national mean time to repair “less than 08 hours” (LIRNEasia, 2018). Today we would write it as a service level an operator can be held to. Starting from no target at all, a realistic first step is critical sites restored within 8 hours with 80% confidence (P80), set in licences, tested through BC-DRPs and raised as capability grows. “Near 100%” is an aspiration; a P80 target is a contract. On 26 August, availability failed twice: the SMS alert queued for close to an hour (Part 1), and restoration of the last sites ran into weeks. Neither was measured against any target, because none existed. 

What happened instead. By 2024–25, an ITU study found that Nepal’s regulator “has not defined the service levels for emergency situations” and that Nepal reported no redundant-power measures (ITU, 2025). In August 2026, two weeks before the flood, operators were directed to prepare BC-DRPs, without targets, testing or drills (Ratopati). Two warnings from the 2016 report still stand. BC-DRPs are “formulated and never rehearsed.” And operator licences carry no “service level agreement clause in support of emergency communications.” 

On fibre. Burying fibre in Nepal’s rock and steep slopes is often neither affordable nor possible. Aerial cable is far more exposed, with one ESCAP study putting its damage rate at 26 times that of buried cable (Samarajiva and Zuhyle, ESCAP). The answer is not to bury everything, but to assess which links will fail and plan how to bridge each gap before it opens. 

Sensors at the source. The 2016 report called landslides “an inherent characteristic” of Nepal’s mountains, to be built into every survivability and availability analysis (NIF 2026 keynote, slide 3). On 26 August the upstream gauges never saw the flood: their last readings, at 08:40, were normal, and the next never came (Part 1). 

In hindsight the slope gave signs: brown meltwater and a possible crack by 24 August (HiRISK), and radar showing movement over several weeks (ICIMOD). Nobody was watching that slope, and no working cross-border data link existed: the 2019 Nepal–China agreement covers border facilities, not data sharing (ISAS). 

Contrast Blatten, Switzerland. Radar and cameras on a moving glacier led to the evacuation of all 300 residents on 19 May 2025. Nine days later the glacier collapsed and buried about 90% of the village, and one person was missing (Antarctic Glaciers; Geopraevent). The village was lost; its people were not. 

Goal II: RREACT, Restoring Access in Hours 

The same start, eleven years apart. After the 2015 earthquake, 40% of damaged sites were restored by day two, 70% by day five and 100% by day seven (NIF 2026 keynote, slide 8). After the 2026 flood, Nepal Telecom had 40% of its worst-day outage back by day two, 88% by day five and 94% by day seven. 

The two events are measured slightly differently. The 2015-earthquake figures are the share of all damaged sites restored, nationally; the 2026-flood figures are the share of the 83 sites down on the worst day, in the affected area only. Read the match as a strong signal rather than an exact comparison. 

Small damage, long tail. The 2026 curve moved faster mid-week but never closed: three Nepal Telecom sites were still down on day ten, and one Ncell site on day 24. The cause was not the size of the damage. A fraction of the damage still produced a tail of weeks, because the last sites were destroyed and reachable only by air (Kathmandu Post). The last few percent of availability depend on location and access, which is exactly what a survivability assessment maps in advance. 

Nepal Telecom put its 2026 damage at about Rs 88 million, roughly US$0.6 million (DevelopingTelecoms; Republica); NTA estimated Rs 450 million, about US$2.9 million, across all operators (eKantipur). The 2015 earthquake’s communications damage was US$36 million, nationally (PDNA, 2015). 

The right kit, arriving as donations. NTA routed 30 satellite phones, 20 from ITU and 10 from Ncell, through NDRRMA to the district offices. Ncell and Dish Home supplied 29 low-earth-orbit satellite terminals for the situation room, district offices, police, army and a hospital. Wi-Fi hotspots went up at holding centres, and a 1234 hunting line connected the district emergency operations centres (Karki, NTA, NIF 2026, slides 4–5). 

This is what RREACT described in the 2016 report: interconnection, access points and backhaul for dark areas, including a mobile communications units. The difference is timing, and the business model behind it. Kits that sit idle in a warehouse rarely survive a budget review. Capacity that is portable and in commercial use every day, but can be lifted out, flown in and switched on within hours, does. Writing that capacity into operators’ service levels and BC-DRPs turns a donation in week one into a commitment in hour one. 

The community layer. In 2026 it stepped in unplanned: police on loudspeakers, a head teacher, volunteers, monasteries and schools. The 2016 report asked for that layer to be designed in, registered, trained and equipped (NIF 2026 keynote, slide 7). How to do that is the subject of Part 3. 

Goal III: All-Hazard, All-Media, and One Common Picture 

Still working independently. The 2016 report found that emergency services were working independently and improvising, with no multi-agency situational awareness platform and no integrated incident command system. In 2026, fifteen thematic coordination teams worked without one common picture (NDRRMA SitRep #01). The confirmed death toll more than doubled after the first 72 hours (Part 1). 

What the message lacked. The 2026 alert went out on one channel, SMS, saying “remain on high alert” (NIF 2026 keynote, slide 5). By 09:05 the flood had been observed and confirmed (Part 1), so a Common Alerting Protocol (CAP) message could have carried that judgement in its fixed fields: urgency Immediate, severity Extreme, certainty Observed, response Evacuate, area the Trishuli corridor, onset within the hour. 

The channel not used. SMS is point-to-point: every message is addressed to one phone, so under load the queue grows. Cell broadcast is point-to-multipoint: one message reaches every phone in a zone at once, without congestion (Samarajiva and Waidyanatha, 2009). A GSMA/ITU study cited by Mongabay puts that at roughly 10 seconds per zone, at a setup cost of $1.75–3 million (Mongabay). On 26 August, the SMS queue may have needed close to an hour (Part 1). Cell broadcast was still under discussion five weeks before the flood (The Himalayan Times). 

Who may warn. The World Meteorological Organization keeps a register of alerting authorities, giving each an object identifier (OID) that every CAP message carries, so anyone receiving a warning can trace who issued it. Nepal’s entry, 2.49.0.0.524, lists one registered authority [confirm which agency] (WMO). The 2016 report recommended a register integrating all authorities at all layers, down to the district and the ward. On 26 August, the people who raised the alarm or told others to move were a district office, police on loudspeakers and a head teacher. None was registered to warn (Part 1). 

Why the Cycle Does Not Close 

Progress is real. It stops short of the joins. At NIF 2026, the regulator set out its own lessons from the flood (Karki, NTA, NIF 2026, slide 5): diversify connectivity options; adopt a National Emergency Telecommunication Plan; strengthen coordination. 

Read them against the 2016 report. Diversified connectivity is Goals I and II. The emergency telecommunication plan has been in draft since 2015, and a post-2015 disaster communication plan “has never been implemented” (Kathmandu Post). Stronger coordination is Goal III. 

This is not a criticism of the people who restored the network by hand. It is a finding about the cycle. Recommendations without an owner, a budget and a deadline get written down, admired, and written down again after the next disaster. A system switched on only in a crisis is one nobody has practised. 

Knowing was never the gap. Deciding was. Part 3 sets out what deciding looks like. 

 

By Nuwan Waidyanatha (Senior Research Fellow, LIRNEasia)

---

Technical Appendix 

Appendix A. Calculations 

Inputs are NTA’s daily counts of sites down in the affected area (Karki, NTA, NIF 2026, slide 3, attached). Day 0 = 26 August 2026. 

• Failures not caused by flooding. Nepal Telecom’s worst day was 27 August, with 83 of 120 sites down. Only 3 sites were flooded, so 80 of 83 (96%) failed for other reasons, mainly power. 

• The day-two jump. Sites down rose from 40 on 26 August to 83 on 27 August, as power ran out at sites the water never reached. 

• Restoration against the worst day. Share restored = (83 − sites down) ÷ 83: day 2, 33 ÷ 83 = 40%; day 5, 73 ÷ 83 = 88%; day 7, 78 ÷ 83 = 94%. The 2015-earthquake figures measure all damaged sites restored nationally (see Figure 4). 

• The tail. Nepal Telecom: 3 sites down on day 10. Ncell: 1 site down on day 24. 

• Damage in US dollars. 2026 at about Rs 153 per US dollar (Nepal Rastra Bank, August 2026): Rs 88 million ≈ US$0.58 million (Nepal Telecom; DevelopingTelecoms’ headline figure of US$5.8 million is a conversion error against its own Rs 88 million); Rs 450 million ≈ US$2.9 million (all telecom, NTA). The 2015 PDNA used Rs 100 = US$1: communications damage Rs 3,610 million ≈ US$36 million. 

Note. NTA counts 5 Nepal Telecom sites down on 1 September (115 of 120 up); NDRRMA SitRep #01 reports 110 of 120 the same day. This post uses NTA’s operator-level figures. 

Appendix B. Sources 

LIRNEasia and Sahana 

• 2015 Gorkha Earthquake: Evaluation of Nepal’s Emergency Communication Systems (2016 report): http://lirneasia.net/wp-content/uploads/2016/04/Nepal_Report_on_Emergency_Communications_v1.0.pdf 

• E-Resilience in support of Emergency Communications (2018): https://lirneasia.net/2018/09/unescap-apis-emcomm 

• e-Resilience “contingencies” deck (2018): https://lirneasia.net/wp-content/uploads/2018/09/LIRNEasia-e-Resilience-in-support-of-Emergency-Communication-“contingencies”.pdf 

• Samarajiva and Waidyanatha, Two complementary mobile technologies for disaster warning, info 11(2), 2009: https://emerald.com/insight/content/doi/10.1108/14636690910941885/full/html 

• Samarajiva and Zuhyle, The Resilience of ICT Infrastructure and its Role During Disasters (ESCAP): https://www.unescap.org/sites/default/d8files/knowledge-products/The%20resilience%20of%20ICT%20Infrastructures.pdf 

• Situational awareness can save people with disabilities in crises (2020): https://lirneasia.net/2020/08/situational-awareness-PWD-crises/ 

• SAMBRO usability and acceptability, CPRsouth 2017: https://sahanafoundation.org/cprsouth2017/ 

• Extending SAMBRO for hydromet warnings in Nepal (2017): https://sahanafoundation.org/extending-sambro-for-hydromet-warnings-in-nepal/ 

• Waidyanatha, The Last-Mile is the First Priority, NIF 2026 keynote (slides attached) 

• Karki, K., When Connectivity Breaks? Nepal’s Experience from the 2026 Flash Flood, NTA, NIF 2026 (slides attached): slide 2, site distribution; slide 3, restoration counts; slides 4–5, satellite kit and lessons 

• NDRRMA Situation Report #01 (1 September 2026): https://ndrrma.gov.np/mediafiles/rasuwa/Rasuwa_Flood_SitRep_Temp_ENG_01_01092026.pdf 

• Kathmandu Post, Bhotekoshi floods expose gaps in Nepal’s emergency communications: https://kathmandupost.com/national/2026/09/03/bhotekoshi-floods-expose-gaps-in-nepal-s-emergency-communications 

• DevelopingTelecoms, Nepal Telecom damage from extreme floods (Rs 88 million; headline USD figure misconverted): https://developingtelecoms.com/telecom-business/operator-news/20913-nepal-telecom-suffers-5-8m-damage-from-extreme-floods.html 

• Republica, Nepal Telecom suffers Rs 88.51 million in damages: https://myrepublica.nagariknetwork.com/news/nepal-telecom-suffers-rs-8851-million-in-damages-from-august-26-flood-60-38.html 

• ITU, Connect2Recover resilience study (2025): https://www.itu.int/en/ITU-D/Regional-Presence/AsiaPacific/Documents/Events/2025/Sendai%202025/ResilienceStudy_Final_08102025.pdf 

• Nepal PDNA 2015, Volume A: https://www.worldbank.org/content/dam/Worldbank/document/SAR/nepal/PDNA%20Volume%20A%20Final.pdf 

 

Access the presentations, “When Connectivity Breaks? Nepal’s Experience from the 2026 Flash Flood” by Kailash Karki (Assistant Director, Nepal Telecommunications Authority) and “The Last-Mile is the First Priority: Safeguarding a Resilient Internet in Times of Crisis” by Nuwan Waidyanatha (Senior Research Fellow, LIRNEasia) below

  Download PDF   Email  Download PDF   Email