Flood Early Warning in Nepal: River Danger Levels and DHM Alerts

Flood forecasting in Nepal is the legal mandate of the Department of Hydrology and Meteorology (DHM), a department of the Ministry of Energy, Water Resources and Irrigation headquartered at Babarmahal, Kathmandu. Its Flood Forecasting Division runs a round-the-clock operation during the southwest monsoon (roughly mid-June to early October, Asar to Asoj in the Bikram Sambat calendar), when almost all of Nepal's riverine flooding occurs. The division's public portal, hydrology.gov.np, displays near-real-time water levels through its River Watch page and rainfall through Rainfall Watch, alongside daily flood bulletins issued throughout the monsoon.

The system rests on a nationwide observation network of hydrological (river-gauging) and meteorological stations. Many gauges are still read manually by trained local gauge readers, but a growing share are automatic telemetry stations that transmit water-level readings to the DHM server as frequently as every 15 minutes. Forecasters combine these observations with rainfall data, numerical weather prediction and simple statistical gauge-to-gauge models that relate the level at an upstream station to the level expected downstream hours later.

DHM's Standard Operating Procedure for the Flood Early Warning System describes three escalating stages of public information: flood watch (be prepared to act), flood warning (act on the advisories issued) and severe flood warning (act together with emergency services). Warnings are pushed out through the website, electronic display boards, sirens at district emergency operation centres, social media, FM radio and television, and bulk SMS delivered in cooperation with Nepal Telecom (NTC) and Ncell.

Every DHM flood-forecasting station has two preset thresholds. The warning level corresponds to the condition when water reaches the top of the riverbanks at the downstream sites at risk; the danger level corresponds to water overflowing the banks and entering downstream communities and floodplains. These definitions come from the community-based systems pioneered in the Karnali basin and are now applied nationally: when a river crosses its warning level, alerts go out and people should prepare to move; when it crosses the danger level, flooding of settlements is expected and evacuation advisories follow.

The thresholds are calibrated station by station, using records of past floods, flood-hazard mapping carried out by DHM and the lived experience of downstream communities. At Chisapani, where the Karnali River leaves the hills, the warning level is 10.0 metres and the danger level 10.8 metres. Studies of the West Rapti basin report a warning level of 5.0 m and danger level of 5.4 m at the Kusum station, while older DHM-linked literature gives roughly 5.5 m and 6.1 m for the Babai at Chepang; because riverbeds shift and gauges are re-rated, DHM periodically revises thresholds, so the live River Watch page is always the authoritative figure.

Crossing a threshold triggers automated actions as well as human ones. District-level flood monitoring display boards sound sirens automatically when a station passes its warning level, and the telemetry system fires automatic SMS messages to chief district officers, DHM basin offices and the security forces, who in turn alert police posts, army posts and local FM stations for wider dissemination.

DHM issues basin-level flood forecasts for Nepal's major southward-draining rivers. In its verification of the September 2024 floods the department listed forecast stations for ten basins: the Saptakoshi, Kankai, Kamala, Bagmati, East Rapti, Narayani, West Rapti, Babai, Karnali and Mahakali. Community-based early-warning schemes extend the coverage, including a system on the Seti River at Pokhara set up in 2013 after the deadly 2012 Seti flash flood, which killed more than 70 people.

The most-watched stations, and the ones behind common searches such as 'koshi water level' or 'bagmati water level', are listed below with verified reference points. Historic gauge heights give a sense of scale but are not the operational thresholds, which are shown live on River Watch.

Nepal's flood early-warning system is celebrated internationally because it pairs national telemetry with a 'last-mile' community model. Trained gauge readers, part-time DHM staff recruited from the local community, watch the river at upstream stations and, once water approaches the warning level, telephone and SMS downstream contacts using pre-agreed communication charts that are updated before every monsoon. The message cascades through Community Disaster Management Committees (CDMCs), district disaster relief committees, security forces and local FM radio, and is amplified locally with sirens, megaphones and hand-held alarms.

Lead time is set by hydrology. From Chisapani, Karnali floodwaters take about two to three hours to reach the vulnerable settlements downstream, so the roughly 74 flood-prone villages of the lower Karnali get a two-to-three-hour head start; the Chisapani gauge reader's calls and the automated SMS chain help protect an estimated 60,000 people between the gorge and the border, about 20 km downstream. On the East Rapti, the Rajaiya station gives Chitwan communities at least two hours of advance notice. That window is enough to move people, documents and livestock to raised shelters, though researchers note it is tight for protecting property and for safe night-time evacuation.

The model proved itself in the August 2014 floods, when warnings from Chisapani and Kusum allowed downstream communities in the Karnali and West Rapti basins to evacuate before the peak arrived. Since then the approach has been replicated eastward, and annual mock-flood drills before the monsoon keep response plans fresh.

Anyone in Nepal can consult the same data DHM forecasters use. The quickest routes are the River Watch page, the daily flood bulletin and the toll-free hotline. Bulk warning SMS does not require registration: under DHM's arrangements with Nepal Telecom and Ncell, mass text alerts are pushed to mobile subscribers in districts at risk when a river crosses its thresholds. During the 27-29 September 2024 emergency alone, the Flood Forecasting Division issued 49 SMS alert events totalling about 4.09 million messages, alongside bulletins, e-mails and social-media posts.

For residents of flood-prone wards, the most reliable habit is layered: know your nearest upstream station and its danger level, check River Watch when heavy rain is forecast, keep the 1155 hotline saved, and follow your municipality's disaster committee. The BIPAD portal of the National Disaster Risk Reduction and Management Authority (NDRRMA) also aggregates hazard alerts and incident reports nationwide.

Nepal ranks among the world's most flood-exposed countries; floods and landslides account for roughly three-quarters of its disasters. The 18 August 2008 (Bhadra 2065 BS) breach of the Koshi embankment at Kusaha displaced about 70,000 people in Nepal and showed how catastrophic embankment failure can be even without record river flows. The 19-24 August 2017 (Bhadra 2074 BS) Tarai floods affected some 1.7 million people, killed 143 and caused around NPR 61 billion in losses — about three percent of GDP — inundating an estimated 80 percent of the Tarai belt.

The most severe recent test came on 27-29 September 2024 (Asoj 11-13, 2081 BS), when a late-monsoon depression dumped record rain — 25 stations in 14 districts broke their 24-hour precipitation records on 28 September. The Saptakoshi at Chatara, Narayani at Devghat, Trishuli at Kali Khola and Sunkoshi at Hampachuwar all exceeded their historic maximum flood levels, and more than 200 people died, many in the Kathmandu Valley. DHM's verification found the event was forecast well in advance: a special flood bulletin issued at 5 pm two days before the peak correctly flagged extremely high flood risk (above danger level) for the Narayani, Bagmati, West Rapti and Babai basins.

That episode illustrated both the system's strength — timely bulletins, tens of lakhs of SMS, accurate basin-level risk calls — and its central weakness: warnings only save lives when local governments and households act on them quickly, especially for floods that peak at night.

The biggest gap is flash floods on small, ungauged rivers. Streams draining the Siwalik hills — such as those that devastated Susta Rural Municipality in July 2017 while the Narayani itself stayed below warning level at Narayangadh — can rise within minutes, faster than any gauge-to-gauge model can warn. For these catchments DHM's SOP proposes rainfall-based triggers, where three-hour rainfall totals or an antecedent precipitation index, rather than water level, set off flood watches. ICIMOD's satellite-aided flash-flood prediction tools complement this for thousands of river segments.

Researchers are also working to stretch lead times. A pilot probabilistic model for the Karnali showed that forecasting the Chisapani level five hours ahead, added to the existing two-to-three-hour travel time, could give downstream communities seven to eight hours of warning. Internationally supported anticipatory-action frameworks now use three-day and seven-day forecasts (including the Global Flood Awareness System, GloFAS) to release humanitarian funds before a flood peak, with DHM alerts as a required trigger. Other known weaknesses being addressed include sirens at district centres that sound for threshold crossings anywhere in Nepal rather than only the local basin, power and telecom failures during severe floods, and the need for real-time data exchange with India on cross-border rivers such as the Mahakali, Karnali and Koshi.