Every year, the monsoon turns a large part of Bangladesh into a shallow inland sea, and every year, the country depends on it. The gentle flood, the barsha, spreads silt across the rice fields and carries the fish on which millions rely. The destructive flood, the bonna, is the same water arriving too fast, too deep or for too long. BWDB records put the average inundation at about 31,000 square kilometres a year, roughly a fifth of the country, while in the worst years the water has covered more than two-thirds of it.
This case study traces how that exposure works: the geography that makes flooding inevitable, the four kinds of flood Bangladesh faces, the landmark floods from 1988 to 2026, the embankment debate, the forecasting revolution, and the remarkable, if incomplete, success in keeping people alive. Throughout, tolls and damage figures are given as their sources reported them: in Bangladesh's flood history the numbers themselves are often part of the argument.
Bangladesh sits at the bottom of one of the largest drainage systems on Earth. The Ganges, the Brahmaputra and the Meghna together drain a catchment of about 1.72 million square kilometres across China, Nepal, Bhutan, India and Bangladesh, yet, as the FFWC points out, only about 8 percent of that catchment lies inside Bangladesh itself. More than 90 percent of the water that flows through the country to the Bay of Bengal enters across its borders. By the estimate most commonly cited in the flood literature, roughly 80 percent of the country is floodplain built by these three rivers. The full anatomy of the system is set out in our account of the rivers of Bangladesh.
Three further facts turn drainage into disaster. First, timing: roughly four-fifths of the year's flow arrives in the monsoon months, and when the peak flows of the three great rivers happen to synchronise, as they did in 1988, no channel could carry the combined volume. Second, relief: much of the country lies only a few metres above sea level, so water that arrives spreads wide and drains slowly, held back by the tide and by the silt that continually raises and shifts the riverbeds. Third, rain: Bangladesh's own monsoon rainfall, averaging around 2,300 millimetres a year and far more in the north-east, falls on ground the rivers may already have saturated. The catastrophes occur when all three coincide.
Four Kinds of Flood
Flood professionals in Bangladesh distinguish four types. The monsoon riverine flood is the classic slow-onset event: the great rivers rise over weeks, spill across the central floodplains and recede, dangerous mainly through depth and duration. The flash flood is its violent opposite, water racing off the hills of Meghalaya, Assam and Tripura into the north-eastern haor basin and the south-eastern hill districts within hours, often before the monsoon proper, the sworn enemy of the haor's single boro rice crop, as our article on the Haor Basin describes. The third type is urban waterlogging: intense rain on paved, poorly drained cities such as Dhaka and Chittagong, where filled and encroached canals leave storm water nowhere to go. The fourth is coastal flooding, where high tides, and in cyclone years storm surge, push the sea into the southern polders. Most bad years mix several types at once.
The Landmark Floods: 1987 to 2007
The Water Development Board's table of major floods records the 1987 flood at about 57,300 square kilometres inundated and 1,657 deaths. The 1988 flood was worse: the peaks of all three great rivers arrived within days of one another, Dhaka itself was badly hit, and the Board's table puts the inundated area at 77,700 square kilometres, 52 percent of the country, though later accounts often round the figure to about 60 percent, with 2,379 deaths, the highest flood toll in the country's recorded history.
The 1998 flood remains the benchmark. Heavy rain, synchronised river peaks and a strong backwater effect from the sea combined to inundate 100,000 square kilometres, 68 percent of Bangladesh, the share most commonly cited, including half of Dhaka. The water stayed for more than two months. Some 30 million people were made homeless, yet the recorded death toll, 1,050 in the Board's table, was less than half of 1988's, an early sign that preparedness was changing the arithmetic of survival even as the water grew worse.
In 2004, government figures recorded about 35,000 square kilometres inundated at the peak and 36 million people affected across 39 of 64 districts; a joint Asian Development Bank–World Bank assessment put the damage at about US$2.2 billion, while the official death toll stayed under 800, which the government's own post-flood assessment credited to disaster response and emergency health work. In 2007 the floods came twice, in July and again in September: the Red Cross reported about 10.6 million people affected, health authorities had counted 838 deaths by mid-September, most of them by drowning, and later Red Cross reporting put the season's toll at roughly 1,100.
The Recent Decade: 2017, 2020, 2022 and 2024
The floods of the last decade have been regional rather than nationwide, In 2017 the haor country was hit first: a flash flood arriving on 27 March, weeks ahead of the harvest, drowned the standing boro crop, the Centre for Policy Dialogue estimated that about 85 percent of haor farmers lost their rice that year. From August, monsoon flooding spread across more than a third of the country; the humanitarian response plan drawn up that September recorded about 6.9 million people affected and 114 confirmed deaths, with later counts higher.
The 2020 flood was defined by duration. Pre-monsoon soils, saturated by Cyclone Amphan, sent the water up early, and a satellite-based study published in the Proceedings of the National Academy of Sciences identified it as the second-longest flood since 1998, inundating nearly a third of the country. The National Disaster Response Coordination Center put the affected population at 5.5 million; the health ministry reported 257 deaths. At Bahadurabad, the FFWC recorded the Jamuna flowing at 20.79 metres, above the previous record set in 1988.
In June 2022 the north-east suffered what the disaster management minister called the region's worst flood in 122 years. The International Federation of Red Cross and Red Crescent Societies estimated 7.2 million people affected; by its account some 94 percent of Sunamganj district and 84 percent of Sylhet district were submerged, and the FFWC's chief reported the Surma and Kushiyara at their highest levels since records began in the 1970s. The health directorate's final count recorded 136 deaths between mid-May and early August 2022.
The August 2024 flood struck a region with little flood memory. From 21 August, extreme rain over the eastern hills and the catchments upstream sent a sudden surge through eleven districts, with Feni, Noakhali, Cumilla and Lakshmipur hit hardest. The National Disaster Response Coordination Center counted about 5.8 million people affected and 502,501 people sheltering in 3,403 evacuation centres; the Centre for Policy Dialogue estimated damage at about Tk 144 billion (roughly US$1.2 billion). The death toll was reported differently by different counters: the disaster ministry's running total reached 71, while the health directorate's emergency centre, whose figures the World Health Organization summarised, reported 75 flood-related deaths, most by drowning.
The 2024 flood also produced a bitter argument about cause. In Bangladesh, officials and much public commentary blamed a sudden release of water from upstream in India, centred on the Dumbur dam on the Gumti River in Tripura. Tripura's power minister responded that no floodgates had been opened, water had escaped automatically over the spillway once the reservoir passed its full level. India's Ministry of External Affairs stated on 22 August 2024 that the claim the flood was caused by the dam's opening was "factually not correct", arguing that the Gumti catchment had received the year's heaviest rain and the flood was "primarily due to waters from these large catchments downstream of the dam", a structure more than 120 kilometres upstream of the border and about 30 metres high. On 54 shared rivers, a flood in Bangladesh very often begins as weather, or as a decision, in another country.
The Embankment Era and the Flood Action Plan Debate
The 1964 Master Plan prepared for East Pakistan proposed 58 flood control and drainage projects, and by the end of the 1980s the country had accumulated some 7,555 kilometres of embankments, 3,674 kilometres of them coastal, and nearly 7,907 hydraulic structures. They demonstrably protect crops and homesteads from moderate floods. But they carry costs: embanked land, starved of silt, sinks relative to the rivers outside; drainage channels clog; and water that would once have spread shallowly across a floodplain stands deep against the embankment or ponds behind it, the waterlogging that afflicts parts of the south-west. Embankments also interact with the delta's other great hazard, river erosion, since a channel confined on one bank often attacks the other.
After the 1988 catastrophe, the argument went global. A French proposal, championed by Danielle Mitterrand, envisaged embanking the major rivers along both banks and dividing the protected land into compartments, a scheme costed in studies of the time at between US$5.4 billion and US$10.2 billion. At the July 1989 G7 summit the donors instead commissioned a World Bank-coordinated Flood Action Plan: 26 studies and pilot projects, run from 1990 to 1995. American and Japanese studies were sceptical that embankments could control a once-in-a-century flood at all, and recommended preparedness and forecasting instead. Opposition inside and outside Bangladesh, char dwellers who stood to lose their land between embankment and river, fisheries specialists, environmental groups who took the argument to a European Parliament debate in 1993, charged that the scheme would displace millions (one World Bank-cited study put about five million people in the zone a set-back embankment line would leave unprotected) and enrich consultants more than farmers.
In the event, the grand embankment scheme was never built. What survived the Flood Action Plan was its quieter half: studies, pilot projects and a lasting policy turn from "flood control" to "flood management", forecasting, warning, flood shelters, raised homesteads: learning to live with floods.
Forecasting: Buying Days
The institution at the centre of the new approach is the Flood Forecasting and Warning Centre (FFWC), established in 1972 under the Water Development Board. When it began, warnings ran a few hours ahead of the water. Today its routine public products are five-day deterministic forecasts and ten-day probabilistic outlooks for the major river stations; satellite-altimetry-based forecasts developed with the University of Washington, the Institute of Water Modeling and ICIMOD extend the horizon to about eight days at selected locations; and a 15-day probabilistic system built with the regional agency RIMES now underpins forecast-based financing, which releases cash and supplies to threatened districts before the peak arrives. Evaluations of the SUFAL anticipatory-action project after the 2020 flood found that warned, supported households protected measurably more of their assets.
The FFWC receives gauge readings only once a river has entered Bangladesh; upstream flows must be estimated from satellite data and partial data-sharing with India, under which stations such as Amarpur on the Gumti transmit real-time readings. Flash floods remain the hardest problem: experimental guidance looks about three days ahead, and a newer forecasting tool for the upper Meghna basin, developed with ICIMOD, works about 54 hours out, a different world from the riverine flood's stately advance.
Why Fewer People Die
The recorded tolls tell the story: 2,379 deaths in 1988; 1,050 in the larger flood of 1998; under 800 in 2004; 257 in 2020; 136 in 2022; and, in 2024, a toll in the low seventies by either official count. Counting methods differ across the decades, and the figures are not perfectly comparable, but no plausible adjustment reverses the direction. Assessments and researchers attribute the fall to several reinforcing changes: warnings that now reach villages days ahead by mobile phone and voice message; a network of flood shelters and the raised earthen plinths on which homesteads increasingly stand; embankments that, whatever their long-term costs, keep moderate floods off the homestead; emergency health services that have cut the diarrhoeal deaths which once followed every flood; and, most recently, anticipatory cash delivered on a forecast rather than relief delivered after a disaster.
The achievement has a clear boundary. What has fallen is mortality, not loss. Damage estimates have not declined with the death toll, about US$2.2 billion in 2004 by the joint ADB–World Bank assessment, about US$1.2 billion in 2024 by the Centre for Policy Dialogue's, and drowning, especially of children, remains the dominant cause of death in the health directorate's counts. Bangladesh has learned to keep people alive through floods it cannot prevent; protecting what they own is the unfinished half of the lesson.
A Warming Climate, and the 2026 Season
Climate change presses on every term of this system, as our overview of climate change in Bangladesh sets out. The IPCC's Sixth Assessment Report projects increases in mean and extreme monsoon precipitation over South Asia. A CMIP6 modelling study published in 2024 found that flood-causing extreme monsoon rainfall events over western Meghalaya and north-east Bangladesh, and over the coastal south-east, have already quadrupled in frequency since 1950–1979, with further increases projected by mid-century. Sea-level rise compounds the problem from the other end, slowing drainage and lifting coastal flood levels; the World Bank's 2022 Country Climate and Development Report estimated that in a severe flood year, flooding could cut Bangladesh's GDP by as much as 9 percent.
The 2026 monsoon supplied a live demonstration, though of the flash-flood kind, not the great-river kind. After an unusually dry June, the season turned in July: from 5 July, persistent heavy rain and runoff from the hill catchments triggered flash floods, waterlogging and landslides across the south-east. The UN's Inter-Cluster Coordination Group reported on 23 July 2026 that an estimated 1.28 million people had been affected across seven districts, Chittagong, Cox's Bazar, Rangamati, Bandarban and Khagrachhari the worst hit, with 59 deaths, about 436,000 people in need of humanitarian assistance, 74,479 houses damaged or destroyed and some 27,007 hectares of cropland flooded; UNICEF's end-of-September situation report carried the same season figures. No nationwide Ganges–Brahmaputra–Meghna flood on the 1988 or 1998 pattern occurred in 2026; the year's catastrophe arrived where forecasting tools are shortest, off the hills, in hours.
The country cannot opt out of its geography: the water of a subcontinent will keep arriving, in a climate loading the monsoon with more of it. What five decades of flood history show is a nation that stopped trying to abolish the flood, and got steadily better at surviving it, first with earth, then with information, and now with money moved ahead of the water rather than after it.