Bangladesh is a country being built and dismantled by the same rivers. The Padma, Jamuna and Meghna carry down the silt that made the delta, and every monsoon they take some of the finished land back — a field here, a homestead there, sometimes a whole village in a single season. Riverbank erosion rarely makes the dramatic pictures that cyclones do. It works quietly, a few metres of bank at a time, but its cumulative toll is among the largest of any hazard in the country: land that took generations to settle can vanish in one flood season, and its families join a population of erosion-displaced people that Bangladesh has never counted in full.
This article sets out what is actually known about the scale of that loss, and what is not. The satellite record assembled by the Centre for Environmental and Geographic Information Services (CEGIS) now measures five decades of erosion and accretion along the three great rivers. The human figures, by contrast, rest on rules of thumb, partial tallies and competing estimates that differ by an order of magnitude — a gap this article keeps visible rather than smoothing over.
Why the Banks Fall: A Delta That Will Not Sit Still
Erosion in Bangladesh begins with sediment. The Ganges–Brahmaputra–Meghna system carries on the order of a billion tonnes of silt a year in recent estimates — among the largest sediment loads of any river system on Earth — through channels whose flow swings violently between dry season and monsoon. The Jamuna, the Brahmaputra's course through Bangladesh, is a braided river: a shifting weave of channels separated by sandbars, in a belt more than ten kilometres wide at high water. Where the current strikes a bank of loose sand and silt it undercuts it, and the overhanging block slumps into the river in slabs. Morphologists describe the Jamuna's exceptional instability as the product of that sediment supply — swollen, many argue, by the debris of the great Assam earthquake of 1950 — meeting a river still adjusting to the course it seized in the floods of the late eighteenth century.
The Padma erodes by the same undercutting and slumping along its meandering bends, and the Lower Meghna adds tidal scour as it widens into the estuary, where island districts such as Bhola, Hatiya and Sandwip are eaten on one shore and built up on another. Human works can move the problem rather than solve it: embankments and bridge guide bunds that pin one bank concentrate the current against another, a charge levelled for years at the Jamuna Bridge's training works. Illegal sand mining from riverbeds, reported along the Teesta and other northern rivers, deepens channels and steepens banks further. None of this is primarily a climate phenomenon — the rivers were devouring banks long before the climate changed — though sharper monsoon bursts are expected to load still more energy into the system.
Five Decades in the Satellite Record
What satellites can measure precisely is land. CEGIS, a trust under the Ministry of Water Resources, has tracked the three great rivers by satellite imagery since 1973, and its long-run balance sheet, published in 2026, is stark. Between 1973 and 2025 the Jamuna eroded about 94,220 hectares and gained back 14,955 hectares of new char; the Ganges eroded 30,290 hectares and regained 23,192; the Padma eroded 33,843 hectares and regained 5,112. Together the three rivers swallowed 158,353 hectares — roughly 1,585 square kilometres — while newly risen chars returned about 433 square kilometres. The net loss, around 1,153 square kilometres, is close to four times Dhaka's area; the Jamuna alone accounts for about three-fifths of everything lost.
Two qualifications matter. First, the CEGIS study covers the three great rivers, not the hundreds of smaller ones — the Teesta, Dharla and their neighbours erode banks too, outside this ledger. Second, the annual rate has fallen sharply: CEGIS's figures show combined erosion running at roughly 9,500 hectares a year in the mid-1990s but down to 1,034 hectares in 2025 — 593 hectares along the Ganges, 348 along the Jamuna and 93 along the Padma — a decline engineers attribute to protection works, the rivers' own shifting geometry, and better forecasting. Older figures still circulate widely: estimates of 8,700 to 10,000 hectares lost a year, drawn from earlier CEGIS reporting and repeated in Mohammad Zaman and Mostafa Alam's 2020 study of char dwellers, describe the river of the 1990s and 2000s more than the river of 2025. Both sets of numbers are "CEGIS numbers"; they answer questions about different periods, and quoting the larger one as today's rate misleads.
Counting the Displaced: The Missing Ledger
If the land loss can be measured from space, the people cannot. Bangladesh has no comprehensive count of those displaced by river erosion — a point CEGIS's own specialists make openly. What exists instead is a rule of thumb: Sudipta Kumar Hore, a senior morphology specialist at CEGIS, has put it at roughly 1,000 people displaced for every square kilometre eroded, which on recent rates implies a yearly toll in the low thousands of families along the great rivers. Official tallies are fragments. The Department of Disaster Management's most recent compilation, from 2019, recorded 24,328 families as erosion victims across nine districts, with damage put at Tk 366 crore. In September 2026 the planning minister told parliament, citing Bureau of Statistics reports, that erosion displaces an average of over 15,000 people — some 3,500 families — a year. Older government reporting cited by the World Bank spoke of around 50,000 households made homeless in an average year, and estimates of 100,000 or more people displaced annually, derived by multiplying eroded area by assumed settlement density, remain common in the literature. These figures cannot all be true at once; they count different things, in different periods, by different methods. The honest summary is that the true annual number is unknown but certainly runs to thousands of households.
The local record shows what those abstractions mean. In Shahjadpur upazila of Sirajganj, more than 200 houses in two Jamuna-side villages were destroyed in a fortnight in late August and September 2026. Beside the Padma Bridge in Shariatpur, a char that rose about three decades ago and grew into a union of six villages lost 1,803 families' homes in two years to October 2026, its people dismantling their own houses ahead of the current and camping on bridge land they fear eviction from. Surveys of displaced households in Gaibandha and Jamalpur by the Red Cross and Red Crescent in 2023 found a median of 3.4 displacements per household over twenty years: displacement by erosion is rarely a single event, but a repeating condition. Looking ahead, the climate research organisation Change Initiative has projected that by 2050 one in seven Bangladeshis will have been displaced by climate change, erosion responsible for about 26.6 percent of the total — a projection, not a measurement.
Life on the Chars
Most of the displaced do not go far. They move onto the chars — the sand islands the same rivers are building elsewhere — or crowd onto embankments and relatives' plots nearby; academic studies put the average relocation distance at around 12 kilometres, and erosion-prone households move on average every five years, some three times in a single bad year. A 2022 survey under the Char Development and Settlement Project, cited by the Water Development Board, counted some 1,200 to 1,500 chars covering about 830,000 hectares, roughly 5.7 percent of the country's land. How many people live on them is, again, contested: BWDB officials have estimated 1.8 to 2 million, other estimates run to five or six million, a 2025 Nature Communications study worked with about 6.5 million, some 65 percent of them on the Jamuna chars, and the char researcher Mohammad Zaman has argued for 15 to 20 million once coastal chars are included. The spread reflects definitions — which islands count, and whether attached mainland chars are included — as much as error.
What is not in dispute is the condition of char life. New chars are raw sand; only after years of silt deposition, grass and settlement do they carry the chilli, groundnut, maize and sesame crops that char farmers have made pay — yet a char can be a generation old and still vanish in two monsoons. Services arrive late or never: many chars have no permanent school, clinic or road. Land rights are the deepest insecurity. Under the law of alluvion and diluvion, land lost to a river is lost, and land that later re-emerges becomes khas — state — land, to be settled under policies such as the Agricultural Khas Land Settlement Policy of 1997, which reserves it in principle for the landless. In practice, re-emerged chars are fought over, sometimes violently, between returning families with old deeds, new occupants and local strongmen who seize and lease out land they do not own. Those who give up on the river altogether swell the slums of the cities — Dhaka's Bholar Basti, home to more than 30,000 people, is named for the Bhola district most of its residents fled.
Prediction: Forecasting Where the River Will Bite
Bangladesh's most distinctive answer to erosion is not concrete but information. Since 2004 CEGIS has issued an annual erosion forecast for the Jamuna, Ganges and Padma, built from pre-monsoon satellite imagery and channel modelling. Its forecast for 2026 put about 915 hectares at risk across eleven locations in Gaibandha, Jamalpur, Kurigram, Pabna, Kushtia and Faridpur, including 73 established sites — among them four schools, three mosques and a health clinic. The 2025 forecast broadly matched what the rivers then did, and CEGIS aims to extend the horizon to two years. A forecast, of course, saves nothing by itself: its value depends on whether administrations, schools and families act on it — moving a school building, harvesting early, relocating a homestead before rather than after the bank goes — and turning prediction into timely relocation support remains the weak link.
Holding the Bank: Embankments, Geo-Bags and Concrete
The engineering repertoire has been rebuilt around a Bangladeshi invention of necessity. The country has almost no natural rock, so from the late 1990s, under the ADB-supported Jamuna–Meghna River Erosion Mitigation Project, the Water Development Board developed revetments of sand-filled geotextile bags — geo-bags — placed to form a flexible apron that launches into the scour hole as the bed deepens. A decade of experience put geo-bag protection at around US$2 million per kilometre against roughly US$5 million for cement-concrete block revetment, and geo-bags are now the default armour of Bangladeshi riverbanks, supplemented by concrete blocks at high-value sites such as towns and bridge guide bunds. Alongside revetments stand the older embankments — the Brahmaputra Right Embankment above all — which keep floodwater off farmland but do not by themselves stop a channel migrating into them.
The record is mixed, and the criticism is specific. Much BWDB work remains reactive: emergency geo-bag dumping at a threatened point, often begun when homes are already at the edge, rather than planned protection of forecast reaches — a pattern protested again through the 2026 monsoon, when residents of eroding Teesta villages in Sundarganj demanded permanent works instead of yearly bag-dumping, and when an 18-kilometre Gaibandha embankment built in 2023 under a Tk 996 crore protection project cracked and subsided within three years, drawing local allegations of sand used where soil was specified. The strategic response now taking shape is corridor-scale: the World Bank-supported Jamuna River Economic Corridor Development Program, a planned series of projects costing some US$2.8 billion over about fifteen years, whose first phase (US$135 million, 2023–2028) pilots protection and navigation works and whose second phase is to extend bank protection along some 100 kilometres of the Jamuna from 2025. Even its planners do not propose to armour every bank: the arithmetic forces triage — protect towns, infrastructure and the densest settlements, forecast and assist movement elsewhere.
The Policy Debate: Whose Loss, Whose Land, Whose Duty?
Running beneath the engineering is an argument about obligation. Erosion victims occupy an odd place in policy: the country has disaster law, a national strategy on managing internal displacement adopted in 2021, and repeated commitments to enumerate and rehabilitate the displaced — yet commentators noted in 2025 that the displacement action plan's promised coordinating committee had still not been formed, there is still no national database of the erosion-displaced, and relief remains the ad hoc distribution that follows a bank collapse rather than an entitlement that precedes it. Editorials in 2026, following CEGIS's fifty-year reckoning, returned to the same demand: count the displaced, by name and district, so that housing, social protection and livelihood support can reach them.
Three positions contend. The protection-first view holds that a state that can forecast erosion a season ahead has a duty to defend settled land wherever defence is feasible, and points to the falling erosion rates where revetments have been built. The adaptation view answers that the delta's rivers cannot be frozen in place at affordable cost, that hard works export erosion downstream, and that money is better spent on secure tenure for char settlers, raised homestead platforms, portable housing, movable schools and clinics, and planned relocation with livelihood support. The third argument is about land itself: whether re-emerged char land should return to the families whose deeds the river swallowed, go to the landless as the 1997 policy intends, or remain a prize for whoever can hold it — a question the courts and land offices have never answered cleanly. These debates now sit inside the larger frame of climate change in Bangladesh: erosion is not caused by global warming, but a hotter monsoon, a rising sea backing up the estuary, and the arrival of loss-and-damage finance have made the erosion-displaced part of the climate story, and of Bangladesh's claim that those least responsible should not bear its costs alone.
What the satellite record settles is narrower but real: the great rivers are still taking land, more slowly than they did in the 1990s, and the difference is partly the result of works and warnings Bangladesh built for itself. What remains unsettled — how many people the rivers have moved, where they went, and what the country owes them — is the unfinished business of a delta learning, reach by reach and char by char, how to live on land that moves.