Read our COVID-19 research and news.

After a 2008 avulsion on the Kosi River, floodwaters overran the Indian state of Bihar, displacing millions of people.

Prashanth Vishwanathan/Bloomberg/Getty Images

Why do rivers leap from their banks? Scientists strive to predict deadly flooding events

Rumors that the Kosi River was about to burst were spreading fast in Kusaha, a Nepalese village on the border with India. The river's levees, towering over the village, were being eroded quickly by the cresting waters. At 2 p.m. on 18 August 2008, the east bank ruptured. People ran for their lives as the breach grew. Soon, the entire river, one of the largest tributaries of the Ganges River, had overrun Kusaha and was spilling into India, drowning farm after farm in search of a faster path to the sea.

Eventually, the rampaging river found a long-abandoned channel, where it flowed for 4 months before engineers wrenched it back to its old course. In the end, the flood destroyed some 800 farming villages, killed more than 400 people, and displaced a further 3 million. At one point, half of the Indian state of Bihar was underwater.

At the time, government officials and scientists blamed the deluge on heavy monsoon rains and poor maintenance of the levees. Both played a role, but this event was triggered by something else, says Rajiv Sinha, a river morphologist at the Indian Institute of Technology, Kanpur, who has studied it in detail.

The Kosi is one of the siltiest rivers in the world, carrying an annual load of some 100 million tons of sediment eroded from its catchment in the Himalayas. Near Kusaha the river hits the flat plain at the foot of the mountains, where it slows, dropping much of its load in the riverbed. As the channel's carrying capacity diminishes over time, the river eventually seeks a new route. For centuries, the Kosi has burst its banks in this region every few decades—a phenomenon known to geographers as an avulsion.

A sediment-laden river in Indonesia is prone to shifting course.

Contains Modified ESA/Copernicus Sentinel Data (2019), Copyright Sam Brooke and Vamsi Ganti (2021)

Avulsions are often called the “earthquakes of rivers,” because they are so sudden and catastrophic. But over the past decade, research has revealed that they are also somewhat predictable. Computer analyses and laboratory models of rivers and deltas have yielded insights into where major rivers tend to avulse—and when. “Ten years ago, we didn't know what was driving avulsions,” says Vamsi Ganti, a geomorphologist at the University of California, Santa Barbara (UCSB). “Now we can model the processes and start to identify hot spots.”

Yet just as researchers are gaining foresight into these rhythmic cataclysms, human activity is undermining it. Upstream deforestation and development is adding silt to rivers in unpredictable ways. Levees and dams are altering flows and paths, sometimes worsening the threat. After the Kosi flood, for example, Bihar began to build even more flood protection levees. But levees force a river to drop silt within a constrained channel, hastening the next avulsion, Sinha says. “If [engineers] don't understand the underlying dynamics,” he says, “they are doomed to fail.”

Climate change is another wild card: Rising seas are shifting avulsion hot spots that occur on coastal deltas, another place where rivers slow down, drop silt, and raise riverbeds. “Humans are now the big instigators of avulsions on rivers,” says Jaia Syvitski, a hydrogeologist at the University of Colorado, Boulder.

For all their destructiveness, avulsions bring benefits to both nature and society. They unleash regular floods that nourish many of the world's great wetlands. For example, the vast Pantanal, in the heart of South America, is kept rich and muddy by the avulsing Taquari River.

And by smearing fresh sediment across flood plains and deltas, avulsions fertilize lands that have nurtured some of the planet's great civilizations. The interlinked flood plains of the Tigris and Euphrates rivers in Iraq bear the imprints of 11 major avulsions over the past 7000 years. Ancient Mesopotamians converted some abandoned channels into irrigation canals; at other times the rivers themselves took over irrigation canals.

It is no surprise that to this day, avulsion zones are among the most densely populated places on Earth. River deltas, where avulsions are frequent, occupy just 0.5% of the world's land, but are home to 4.5% of its population. Those dense populations, drawn by the fertility that avulsions foster, are also in the path of their destructive force. Not for nothing are great avulsing rivers such as the Kosi or the Yellow River (Huang He) in China known as their peoples’ “joy and sorrow.”

Avulsions are not just rare catastrophes. An unpublished analysis of satellite images by Sam Brooke of UCSB has identified about 100 abrupt river jumps in the past 50 years: about 40 on coastal deltas and 60 on inland alluvial fans where rivers exit mountain valleys. Many avulsions cluster on rainy, mountainous islands in the tropics that erode quickly, such as Madagascar and Java—something Brooke says the analysis, the first of its kind, drew attention to.

In the aftermath of the Kosi disaster, Douglas Jerolmack, a sedimentologist at the University of Pennsylvania, was one of a handful of researchers to look more closely at the dynamics of avulsions. Often in the past, researchers had claimed avulsions were triggered by earthquakes or extreme floods. But Jerolmack's experiments confirmed a growing suspicion that avulsions happened naturally—and predictably. “Rivers are perturbed all the time but only avulse if they are set up by enough deposition,” he says. Typically, he says, this happens once the riverbed rises above the land outside its banks or levees, so the entire volume of the river is “perched,” as happened with the Kosi.

A 1944 map highlights the site of the present-day Old River Control Structure, where engineers fight the river’s natural urge to jump in an avulsion into the Atchafalaya River—once the Mississippi’s main route.

U.S. Army Corps of Engineers

Jerolmack and his colleagues used a humble laboratory tool for mimicking river avulsions: a stream table, basically a large tank filled with sand and water. At one end of the stream table, the researchers could turn a “river” tap on and off and change the concentrations of sediment in the flow. By varying the steepness of the table, they created miniature mountain streams disgorging into fan-shaped flood plains and bird's-foot deltas.

Geographers had already noticed that rivers tend to avulse at the foot of mountains, where slopes change dramatically and sediment settles out, silting up the bed. The miniature rivers of the stream table allowed the team to quantify how often the avulsions occurred. “The time scale was basically the time it took the channel to fill with sediment,” Jerolmack says.

He and his colleagues also found that the river kept reoccupying a small number of channels. “Old channel paths remained topographically smooth, and so low-resistance pathways to the sea,” Jerolmack says. So if an avulsing river encounters an old channel, he says, it is likely to colonize it.

While Jerolmack was learning about the periodicity and quasi-predictability of avulsions inland, Michael Lamb, a geomorphologist at the California Institute of Technology, wanted to understand the mystery of recurrent avulsions downstream, in river deltas. In deltas, unlike at the foot of mountains, there is almost no slope change to slow a river and cause sediment to build up.

Lamb and his colleagues investigated by setting up not a steep stream table, but a languid laboratory delta, with a gradient of just one in 100,000 and a river spilling into a 3-meter-wide “ocean.” In cycles of floods and gentle flows, they sent water and sediment—low-density crushed walnut shells—into the basin. They again found that the river spontaneously switched pathways to the ocean, and that the avulsions almost always occurred at the head of the delta.

Why that was happening turned out to be more complex than expected. Tides, by slowing the river so that it unloads sediment as it enters the delta, are one factor. But in deltas, scouring from floods is almost as important as deposition in creating avulsion conditions. Floods tend to flush sediment away from the lower part of the delta, Lamb says, preventing avulsions from occurring there. This leads to an avulsion-prone spot one “backwater length” from the mouth of the delta—a distance that can be calculated from basic river parameters.

Why rivers get jumpy

Avulsions occur naturally, when rivers jump from their courses and seek a faster way to the sea, often recolonizing remnant channels. They tend to occur in two hot spots where rivers slow down.

Mountains Alluvial fan River Remnant channels Delta Ocean 2 1 Flow Riverbed Sediment Indus River New course Second avulsion Remnant channel 31 July 2010 18 August 2010 3 September 2010 21 September 2010 0 100 km Delta Pakistan Arabian Sea Karachi Deadly diversions In 2010, a pair of textbook avulsions unleashed deadly floods along the Indus River in Pakistan, killing thousandsand impacting millions. The events were captured by an instrument on NASA’s Terra satellite. 1 Upland hot spotWhen rivers exitmountains, they slowdown and spreadsediment in alluvialfans. The frequencyof avulsions dependson how silty therivers are. 2 Delta hot spotRivers also slow downand drop silt at the headsof deltas, where theyfeel the effects of tides.Rising seas from climatechange may not onlyshift avulsions upstream,but also increase theirfrequency. Settle downWhen rivers slow down, theydrop sediment that settlesand reduces their channels’carrying capacity. Avulsionsare most likely whenriverbeds are perched abovesurrounding flood plains. First avulsion
(Graphic) V. Altounian/Science; (Photos) NASA Worldview

The theory suggested that on the Mississippi River, an avulsion point should fall some 500 kilometers upstream from its delta mouth. The real world agrees: The Mississippi has avulsed in central Louisiana about once every 1400 years, creating large delta lobes across the Gulf Coast. The researchers also found evidence for repeated avulsions one backwater length up from the mouth of the Yellow, Rhine, Danube, Orinoco, and Nile rivers. “The discovered tie between avulsion location and backwater hydrodynamics has been a game changer,” Jerolmack says.

Some rivers are less predictable, however. Last year, Lamb and colleagues published an investigation into a cluster of recent avulsions on short, steep, and sediment-rich streams in Madagascar. They found avulsions happening as much as 20 times farther upstream than predicted by the backwater length alone. They concluded that scouring from floods penetrates farther upstream on these streams because they deposit unusually fine sediments, which floods can sweep away more easily. “It's a special case, but still conforms to the theory that scour length is determining avulsion location,” Lamb says.

Although avulsions will never be wholly predictable, the day is nearing where engineers can know where and when the dangers are at their greatest. Adrian Hartley, a rivers researcher at the University of Aberdeen, is pleased with how far the field has come. “We are getting much closer to understanding avulsions.” But he warns that just as researchers begin to grasp these natural river cycles, human influences are altering them. “Few river systems are without significant human influence anymore, so simple generic models, however good they are, won't do all the job needed to predict avulsions.”

Some anthropogenic effects are unsurprising. For instance, deforestation and soil erosion can make rivers siltier. That increases the pace at which they fill up their channels and reach the point of avulsion. But other factors are less obvious. Dams can trap sediment, preventing its flow downstream—and so they would seem to reduce the risk of avulsions. But Ganti says dams selectively capture coarse sediment, while allowing finer silt to continue downstream. Paradoxically, the resulting rivers “can carry exceptionally high volumes of sediment,” Ganti says. Those loads, eventually dropped, should increase the frequency of avulsions, he predicts.

Humans are now the big instigators of avulsions on rivers.

Jaia Syvitski, University of Colorado, Boulder

Meanwhile, coastal deltas are facing rising seas. By eroding the coastline and pushing tides farther upstream, they cause rivers to dump more sediment farther inland, shifting the entire backwater reach upstream, along with the avulsion hot spot. Ganti says their frequency increases as well, because the invading tides ensure that less of the sediment reaches. As tides push farther into deltas, they slow down rivers; less of the sediment gets into the ocean and more ends up in the delta where it causes avulsions. “We are turning the knob the wrong way,” Ganti says. “We should brace ourselves for more avulsions.”

If models are to keep up with such a fast-changing environment, they need more and better field data to show when and where tipping points are approaching, Lamb says. “Collecting data on the world's riverbeds is more important than ever.” But that information is difficult to gather remotely and painstaking to collect in person or by boat, Lamb says. “For most of the world's rivers, we lack any data set for river bottom elevations,” he says. “It's actually a major gap in our understanding of the geography of the Earth.”

Ignorant of what is happening underwater, many river engineers continue to make bad situations worse by sticking with conventional methods of trying to hold back floods. For them, ever higher levees are the solution, whereas river scientists say they are the problem.

At the natural avulsion point some 500 kilometers up from the mouth of the Mississippi delta, the U.S. Army Corps of Engineers has for a century fought the river's urge to flow into the Atchafalaya River, which was the Mississippi's main channel until a few thousand years ago. In 1963, the Corps acted to prevent a full-scale avulsion. It completed a floodgate system, known as the Old River Control Structure, that kept most of the river's flow—and sediment—in the main channel. “Without this engineering, the Atchafalaya would have captured the majority of the flow from the Mississippi by now,” Ganti says.

But half a century on, the Old River Control Structure's effectiveness is dwindling. It allows the Mississippi to send 30% of its water into the Atchafalaya while holding onto most of its sediment. As a result, the riverbed just downstream of Old River Control has risen by about 1.5 meters, according to Bo Wang, a fluvial geomorphologist now at Brown University. The sediment buildup increases the risk that a future flood will back up, overwhelm Old River Control, and surge down the Atchafalaya. Such an avulsion would be catastrophic for trade on the Mississippi River, and could overrun many low-lying regions of the Atchafalaya. But Lamb says the rush of sediment would be an ecological boon for the delta regions of coastal Louisiana now sinking to the ocean. “Many would be nourished straight away if the river went down the Atchafalaya,” he says.

To avoid avulsions, China has tested discharging water from the Yellow River reservoirs to flush sediment from the riverbed downstream and send it to the ocean.

Xinhua/Li An/Getty Images

The stakes may be even higher in northern China, where some 90 million people live in a flood zone downstream of an avulsion hot spot on the Yellow River. But China is taking a more sophisticated approach to mitigating the risk, which it knows well from recent, tragic history.

On the right bank of the river, close to the city of Zhengzhou, is a modest stone plaque. It commemorates events on the morning of 9 June 1938, when the great river burst its banks, flooded the country's breadbasket, and found a new outlet to the sea, 650 kilometers south of the river's former mouth. Imagine the Mississippi heading east to enter the Atlantic Ocean in Georgia. The vast avulsion killed up to 1 million people and made refugees of up to 10 million more. It took engineers 9 years to coerce the river back to its former course.

This was no natural avulsion, however. The river's banks were dynamited on the orders of the country's leader Chiang Kai-shek, to hold back Japanese invaders during the Second Sino-Japanese War. It is reckoned to be humanity's single most deadly act of war, although most of the dead were Chinese, and the Japanese troops were held up for less than a month.

In 1938, China triggered an avulsion on the Yellow River to repel Japanese invaders—thought to be humanity's deadliest act of war.

Bettmann/Contributor/Getty Images

Whatever the military folly, the engineers knew what they were doing. They breached the bank at Huayuankou, a known avulsion hot spot where the river slackens as it emerges from the Loess Plateau and dumps its sediment. For the past 2500 years, the Yellow River has changed course here roughly once a century. If it were to happen again, “it would be a disaster of global importance, by affecting the Chinese economy,” Syvitski warns.

Today, as the river rises as much as 10 meters above the surrounding flood plain, China is testing out new tactics in the vigil to avoid an avulsion. It has reduced sediment loads through soil conservation projects on the Loess Plateau. And it tested using fast flows of sediment-depleted water from dammed reservoirs to flush out sediment from the riverbed downstream and deliver it to the ocean, says Jeff Nittrouer, a hydrologist whose team at Rice University has been studying China's efforts to tame the Yellow River.

Chinese engineers are also intervening at the second avulsion point downstream, at the head of the Yellow River delta, where seven avulsions have occurred between 1855 and 1930. Since then, engineers have stopped trying to prevent channel switches. Instead, they induce them with dynamite and digging. Apart from making avulsions more predictable, the idea is to move sediment deposition to parts of the delta where the created land could be economically useful, such as for aquaculture or exploiting oil and gas reserves, Nittrouer says.

The first major diversion was in 1976. Then in 1996, workers blocked the 20-year-old Qingshuigou channel as it filled with sediment, and forced the river down an old natural channel to enter the ocean 20 kilometers from its former mouth. Nature took the lead again in 2005, when the river cut a crevasse into its left bank, forming a new channel that has now taken over.

“The Yellow River delta is no longer by any stretch a natural delta,” Syvitski says. Even so, the river's urge to shift its route to the ocean never abates. The price of safety on its banks is eternal vigilance.