This new satellite data is a remarkable advance. We are seeing for the first time the detailed behaviour of the ice streams that are draining the Greenland ice sheet. They show that Greenland seems to be losing at least 200 cubic kilometres of ice a year. It is different from even two years ago, when people still said the ice sheet was in balance.

Hundreds of cubic kilometres sounds like a lot of ice. But this is just the beginning. Once a sheet starts to disintegrate, it can reach a tipping point beyond which break-up is explosively rapid. The issue is how close we are getting to that tipping point. The summer of 2005 broke all records for melting in Greenland. So we may be on the edge.

Our understanding of what is going on is very new. Today's forecasts of sea-level rise use climate models of the ice sheets that say they can only disintegrate over a thousand years or more. But we can now see that the models are almost worthless. They treat the ice sheets like a single block of ice that will slowly melt. But what is happening is much more dynamic.

Once the ice starts to melt at the surface, it forms lakes that empty down crevasses to the bottom of the ice. You get rivers of water underneath the ice. And the ice slides towards the ocean.

Our Nasa scientists have measured this in Greenland. And once these ice streams start moving, their influence stretches right to the interior of the ice sheet. Building an ice sheet takes a long time, because it is limited by snowfall. But destroying it can be explosively rapid.

How fast can this go? Right now, I think our best measure is what happened in the past. We know that, for instance, 14,000 years ago sea levels rose by 20m in 400 years - that is five metres in a century. This was towards the end of the last ice age, so there was more ice around. But, on the other hand, temperatures were not warming as fast as today.

How far can it go? The last time the world was three degrees warmer than today - which is what we expect later this century - sea levels were 25m higher. So that is what we can look forward to if we don't act soon. None of the current climate and ice models predict this. But I prefer the evidence from the Earth's history and my own eyes. I think sea-level rise is going to be the big issue soon, more even than warming itself.

It's hard to say what the world will be like if this happens. It would be another planet. You could imagine great armadas of icebergs breaking off Greenland and melting as they float south. And, of course, huge areas being flooded.

How long have we got? We have to stabilise emissions of carbon dioxide within a decade, or temperatures will warm by more than one degree. That will be warmer than it has been for half a million years, and many things could become unstoppable. If we are to stop that, we cannot wait for new technologies like capturing emissions from burning coal. We have to act with what we have. This decade, that means focusing on energy efficiency and renewable sources of energy that do not burn carbon. We don't have much time left.

Jim Hansen, the director of the Nasa Goddard Institute for Space Studies in New York, is President George Bush's top climate modeller. He was speaking to Fred Pearce

"The emissions that caused this past episode of global warming probably lasted 10,000 years," Professor Zachos told the American Association for the Advancement of Science at a meeting in St Louis. "By burning fossil fuels, we are likely to emit the same amount over the next three centuries."

He warned that studies of global warming events in the geological past indicate the Earth's climate passes a threshold beyond which climate change accelerates with the help of positive feedbacks - vicious circles of warming.

Professor Zachos is a leading authority on the episode of global warming known as the palaeocene-eocene thermal maximum, when average global temperatures increased by up to 5C due to a massive release of carbon dioxide and methane.

His research into the deep ocean sediments suggests at this time about 4.5 billion tons of carbon entered the atmosphere over 10,000 years. This will be the similar amount of carbon released into the atmosphere from cars and industrial emissions over the next 300 years if present trends continue, he said.

Although carbon can be released suddenly and naturally into the atmosphere from volcanic activity, it takes many thousand of years for it to be removed permanently by natural processes. The ocean is capable of removing carbon, and quickly, but this natural capacity can be quickly overwhelmed, which is probably what happened 55 million years ago.

"It will take tens of thousands of years before atmospheric carbon dioxide comes down to pre-industrial levels," the professor said. "Even after humans stop burning fossil fuels, the effects will be long-lasting."

Satellite measurements of the entire land mass of Greenland show that the speed at which the glaciers are moving to the sea has increased significantly over the past 10 years with some glaciers moving three times faster than in the mid-1990s.

Scientists believe that computer models of how the Greenland ice sheet will react to global warming have seriously underestimated the threat posed by sea levels that could rise far more quickly than envisaged.

The latest study, presented at the American Association for the Advancement of Science, in St Louis, shows that rather than just melting relatively slowly, the ice sheet is showing all the signs of a mechanical break-up as glaciers slip ever faster into the ocean, aided by the "lubricant" of melt water forming at their base.

Eric Rignot, a scientist at the Jet Propulsion Laboratory and the California Institute of Technology in Pasadena, said that computer models used by the UN's International Panel on Climate Change have not adequately taken into account the amount of ice falling into the sea from glacial movements.

Yet the satellite study shows that about two-thirds of the sea-level rise caused by the Greenland ice sheet is due to icebergs breaking off from fast-moving glaciers rather than simply the result of water running off from melting ice.

"In simple terms, the ice sheet is breaking up rather than melting. It's not a surprise in itself but it is a surprise to see the magnitude of the changes. These big glaciers seem to be accelerating, they seem to be going faster and faster to the sea," Dr Rignot said.

"This is not predicted by the current computer models. The fact is the glaciers of Greenland are evolving faster than we thought and the models have to be adjusted to catch up with these observations," he said.

The Greenland ice sheet covers an area of 1.7 million sq km - about the size of Mexico - and, in places, is up to 3km thick. It formed over thousands of years by the gradual accumulation of ice and snow but now its disintegration could occur in decades or centuries.

Over the past 20 years, the air temperature of Greenland has risen by 3C and computer models suggested it would take at least 1,000 years for the ice sheet to melt completely. But the latest study suggests that glaciers moving at an accelerating rate could bring about a much faster change.

"The behaviour of the glaciers that dump ice into the sea is the most important aspect of understanding how an ice sheet will evolve in a changing climate," Dr Rignot said. "It takes a long time to build and melt an ice sheet but glaciers can react quickly to temperature changes. Climate warming can work in different ways but, generally speaking, if you warm up the ice sheet, the glacier will flow faster," he said.

The ice "balance sheet" of Greenland is complex but - in simple terms - it depends on the amount of snow that falls, the amount of ice that melts as run-off and the amount of ice that falls directly into the sea in the form of icebergs "calving" from moving glaciers.

Satellites show that the glaciers in the south of Greenland are now moving much faster than they were 10 years ago. Scientists estimate that, in 1996, glaciers deposited about 50 cubic km of ice into the sea. In 2005 it had risen to 150 cubic km of ice.

Details of the latest study, published in the journal Science, show that Greenland now accounts for an increase in global sea levels of about 0.5 millimetres per year - compared to a total sea level rise of 3mm per year.

When previous studies of the ice balance are taken into account, the researchers calculated that the overall amount of ice dumped into the sea increased from 90 cubic km in 1996 to 224 cubic km in 2005.

Dr Rignot said that there are now signs that the more northerly glaciers of Greenland are beginning to adopt the pattern of movements seen by those in the south. "The southern half of Greenland is reacting to what we think is climate warming. The northern half is waiting, but I don't think it's going to take long," he said.

While deltas are naturally vulnerable because they form in low-lying coastal regions, human activity is adding to the risk. "The subsidence issue is related to population," says Robert Nicholls of the Tyndall Centre at the University of Southampton, UK. It matters because the people live there, but their very presence adds to the risk.

Most deltas are subsiding because dams and canals upstream are preventing sediment from reaching the deltas. To make matters worse, people are also pumping out water and petroleum from beneath the deltas, causing the sediment-starved land to settle further. This is causing the waterline in most deltas to rise faster than the average rise in global sea level of about 1.5-2 millimetres per year.

Ericson and his colleagues found that all 40 of the major deltas they studied were at risk: 68 per cent of them were sinking because they were being starved of sediment, 20 per cent were subsiding, mainly because of the pumping of water and oil, and in 12 per cent of the deltas, rising sea level was the biggest problem.

Five of the six deltas most threatened by rising sea level are also at risk from storms, the only exception being the Nile delta. Storm surges can overwhelm an already low-lying region. In the Bengal delta, for instance, storm surges can go right over the hurricane shelters, says Greg O'Hare of the University of Derby, UK.

The deltas in greatest danger lie in the arc from China to India, where populations are high, and powerful typhoons and tropical cyclones are common. Geologically active mountains and heavy rains make south-east Asia the source of half the world's sediment supply, making large deltas commonplace in the region, says Janok Bhattacharya of the University of Texas in Dallas. Intense development has robbed some deltas of water as well as sediment. For instance, extensive damming meant that no water flowed from the mouth of the Yangtze on 230 days of the year in 1998, compared with just 20 days in 1992, he says.

Ericson's study, which will appear in the journal Global and Planetary Change, calls for an increased surveillance of the river deltas using technology such as remote sensing to prevent unprecedented displacement of people. "Greater awareness of potential threats to such systems is a precursor to the design of responses that maximise the protection of life, infrastructure and economic development," the researchers say.

Bengal delta, Bangladesh: 3,430,000 at risk

Residents of this densely populated delta are pumping out so much groundwater that in some areas the land is subsiding by almost 25 millimetres a year. Submarine geology amplifies storm surges at the northern end of the Bay of Bengal, some of which have exceeded 10 metres. Such problems and a lack of warning made the Bengal delta the site of the two deadliest storms of the last century: a storm killed about 300,000 people in 1970 and 138,000 were killed in 1991.

Mekong delta, Vietnam: 1,910,000 at risk

Also densely populated, this sediment-starved delta is on the southern fringe of the Asian typhoon zone and on average is hit once a decade by a typhoon. These storms can be devastating. Typhoon Linda, which developed rapidly over the South China Sea in 1997, was blamed for more than 3000 deaths as it passed along the southern edge of the Mekong delta. The storm flattened villages, wiped out crops, and left tens of thousands of people homeless.

Nile delta, Egypt: 1,300,000 at risk

Though it is not at risk from storms, the Nile delta suffers heavily from human activity. The completion of the Aswan High dam in 1964 has had even more devastating consequences. The dam stopped the delivery of silt carried by the river's seasonal floods and the silt now lies trapped in the dam's reservoir. And beyond the dam a dense network of drainage and irrigation canals is preventing much of the remaining sediment from reaching the coast.

Yangtze delta, China: 484,000 at risk

Shanghai, a coastal city on the delta, has records of typhoons dating back to the 7th century, though their frequency increased rapidly over the 20th century. "Shanghai's defences are as good as London's - one of the best in the world," says Robert Nicholls of the University of Southampton, UK. They are also better maintained than the defences that failed in New Orleans. This extensive protection does not extend into rural parts of the deltas, however.

Mississippi delta, USA: 480,000 at risk

Hurricane Katrina showed just how vulnerable this low-lying region is. The delta also scores high on another measure of risk: 20 per cent of the delta could be lost to subsidence and rising sea levels by 2050. Tens of square kilometres of wetlands to the south of New Orleans were turning to open water each year even before Katrina hit, as petroleum pumping combined with sediment starvation was causing the land to sink and subsidence within New Orleans has left large areas of the city below sea level.

Godavari delta, India: 453,000 at risk

This delta on the east coast of India could lose an even larger fraction of land than the Mississippi delta - about 23 per cent by 2050. Groundwater pumping is a major cause of subsidence, and storms are a constant threat to a region that has few coastal defences. More than a thousand people died when a tropical cyclone hit the delta in November 1996. In the Mahanadi delta, further north along the coast, some 10,000 people died in 1999 when hit by a "super cyclone".

Note: populations at risk by 2050

This could happen if the North Atlantic current system collapsed, denying Britain the warming effect of the Gulf Stream. Ocean surface temperatures would fall by 3C (5.4F), but as the Arctic sea ice melted, they would rise again by 8C (14.4F) in an abrupt turnabout over a period of no more than about 20 years.

Climate Change on the Millennial Timescale is the first study to examine comprehensively the impacts of global warming beyond the end of this century. It calls for continued efforts to cut the emission of global-warming gases to prevent the changes from getting out of control.

Baroness Young of Old Scone, the chief executive of the Environment Agency, said: “We are running out of road on decision-making. Unless we dramatically change the use of fossil fuels then we will be committing future generations to the most severe impacts of climate change.”

By the year 3000, the report says:

# Global warming could have more than quadrupled, with temperature rises of as much as 15C, if we continue burning fossil fuels

# Sea levels will still be rising at the end of this millennium and the total increase could reach 11.4 metres. This dwarfs estimates made by the Intergovernmental Panel on Climate Change that sea levels will rise by between 16cm and 69cm by the 2080s

# Anything more than a two-metre rise would flood large areas of Bangladesh, Florida and many low-lying cities, and displace hundreds of millions of people

# Abrupt climate changes are possible even after emissions cease because changes may be set in motion that cannot be stopped

# The acidity of the oceans will fall significantly, posing a threat to marine organisms such as corals and plankton. That, in turn, would affect the whole marine ecosystem

# The changes could be even greater than this if the climate turns out to be more sensitive to greenhouse gas emissions than the study assumes.

The solution, the team says, is to reduce emissions to zero by 2200.

Tim Lenton, lead author of the study, said: “While most studies stop at year 2100 with temperatures and sea level rising, we explored where they are heading into the next millennium. Only by starting to reduce carbon dioxide emissions now can we avoid dangerous climate change.”

The message is that the world can afford to burn only about a quarter of its known reserves of fossil fuels. This implies a small increase in global emissions up to 2025, and reducing and eliminating them by 2200.

“If we follow business-asusual then we will commit future generations to dangerous climate change,” Dr Lenton said. The risk was that of returning the Earth to a hot state it had not been in since 55 million years ago.

Baroness Young said: “Tough decisions are needed soon. Many of our coastal towns could be in jeopardy and immediate action needs to be taken if we are to avoid many of these impacts.

“We need to get tough on energy efficiency. This means much tighter standards for buildings and government providing proper incentives for businesses from the transport sector if we are to meet our 2010 target of 20 per cent carbon reduction and tackle more strenuous targets for 2020.”

To add to the confusion, there are actually three types of wobble. Two of them were discovered and measured in the 19th century, though only explained in the last decade: the Chandler wobble takes 14 months to complete a cycle, while the annual wobble, as the name suggests, takes exactly a year. Both are driven largely by weather and ocean currents.

In 1960, William Markowitz of the US Naval Observatory in Washington DC noticed what appeared to be another, subtler form of polar wander, which moved the poles by about a metre on what seemed to be a 24-year cycle. Nobody could explain it.

Now a young geologist at the UK's University of Leeds says he has found an explanation for the Markowitz wobble. If he's right, studying the wobble could reveal some of the secrets of the most mysterious part of the Earth - its inscrutable inner core.

The first clear evidence that the Earth's poles were moving came in the 1890s, when astronomers at the International Latitude Service began to monitor the position of six small observatories around the Earth, by watching the movements of the stars overhead. They found that the latitude of all these stations was changing very slightly, by just a few metres a year. The only way to explain it was if the entire crust of the Earth was slipping relative to its axis, moving some observatories to higher latitudes and others lower. The result was the discovery of the Chandler and annual wobbles and, much later, the wobble identified by Markowitz and named after him.

From the 1970s onwards, the evidence for the Markowitz wobble became less and less convincing. Advances in technology allowed the Earth's rotation to be measured in relation to quasars billions of light years away and in relation to artificial satellites, and when GPS with its fleet of satellites came along it made measuring ground position relatively simple. All the new measurements conflicted with the old ones, and many researchers wrote off the Markowitz wobble as an illusion caused by unreliable methods of measuring the Earth's spin. "It was kind of dismissed 10 to 15 years ago," says David Crossley, a geophysicist at Saint Louis University in Missouri.

Even though the Markowitz wobble as he described it was later shown to be a mistake, as further space-based readings came in, something very interesting emerged. There is indeed some kind of long-term movement of the pole - not the periodic wobble that Markowitz saw, but something less easy to pin down, some kind of motion that may go back and forth, or may change speed on timescales of a few decades.

"We still call it the Markowitz wobble to honour him," says Richard Gross of NASA's Jet Propulsion Laboratory in Pasadena, California. Crossley, however, is less keen on the name. "Nobody knew what that was in the first place. I'm surprised to see the term come back." Whatever you like to call it, it's a rather puzzling phenomenon. Something must be dragging the planet's crust and mantle around. But what?

Perhaps deep forces are at work, emanating from the Earth's inner core. That's the idea behind a new explanation for the Markowitz wobble, developed by Mathieu Dumberry at the University of Leeds in the UK. Beneath the Earth's crust lies the solid, rocky mantle. Beneath that is a molten metal outer core, and at the centre of the Earth is the inner core: a sphere 2500 kilometres across that is thought to be made of solid iron. Despite the core being isolated from the mantle and crust, and so well lubricated by the liquid outer core that surrounds it, Dumberry thinks he has found a way for it to get a grip on the outer parts of the Earth.
Inner turmoil

Although the mantle and the inner core are solid, their high temperature softens them so that they become more like the consistency of very stiff Silly Putty. If you were to hit mantle rock with a hammer it would crack, but if you squeezed it patiently for thousands of years it would ooze. In fact the mantle is convecting, like a pan of water in ultra-slow motion: over millions of years warm plumes of rock rise from hotspots on the core, and cool slabs of ocean crust fall down from subduction zones on the Earth's surface. It is these plumes and slabs that Dumberry thinks are responsible for the Markowitz wobble.

Cold slabs are denser than their surroundings, so compared to the mantle, they exert a stronger gravitational pull on the core. Assuming the inner core is soft enough to be deformed by this pull, Dumberry says the effect of this gravity is to slowly cause a hump to rise in the surface of the inner core. "It's like a permanent tide, as if the moon were fixed overhead at one place, raising a permanent hump in the ocean," he says. Likewise, where plumes rise from the edge of the outer core, there should be a corresponding dip in the inner core's surface, since the plume exerts less gravity than the surrounding mantle rock.

This effect would have no impact on the poles, except that the inner core is not fixed in place. Seismic studies show that it is spinning slightly faster than the rest of the planet, rotating under our feet by up to half a degree a year. As a result, those bumps and dips should slowly move out of alignment with the slabs and plumes, so that a bump might find itself tugged at in a different direction by a different mantle slab. Some of the extra gravitational attraction would then be acting from the side rather than vertically, exerting a torque on the mantle. The net result of many such mismatches could be a tug on the mantle and crust that is sufficient to change the axis of rotation, and shift the poles.

So far the numbers seem to add up. When Dumberry plugged seismic readings showing the density of different areas of the mantle into his model, they produced about the same amount of tilt as the Markowitz wobble - a maximum polar motion of about a metre.

The timescale of the Markowitz wobble also fits Dumberry's model. The magnetic field deep within the Earth twists slowly back and forth with an irregular beat of a few decades, oscillations that show up in measurements of the field at the Earth's surface. Dumberry thinks that this twisting magnetic field drags the inner core along with it. If he is right, the inner core's rotation should also change on a timescale of decades. And if this rotation in turn drives the Markowitz wobble, then you'd expect the wobble to have a similar timescale, as space-based observations of the past 30 years seem to show.

So, wobble solved? Dumberry is confident. Since he presented his ideas in November 2005 at a meeting in London, reactions from the earth-science community have been positive. "It is probably one of the best explanations we've got," says Gross.

One fly in the ointment is the orientation of the wobble. The measured motion is a gentle side-to-side rocking, and it is aligned along a bearing roughly 30 degrees east/150 degrees west, nodding alternately towards Egypt and Alaska. Dumberry's model does produces a side-to-side motion, but it is closer to 90 degrees east/90 degrees west (Bangladesh and New Orleans). This is clearly a huge discrepancy, but Dumberry says that the orientation depends on exactly where in the mantle the plumes and slabs are found, and that's something seismologists still disagree on. Dumberry plans to plug different published maps of the mantle into his model, to see if any of them provide a better fit.

Dumberry's is not the only possibility. Recent polar motions could have been generated by melting ice, which would change the weight distribution of the crust, rebalancing the Earth. That would mean it is not a wobble at all, but more of a glitch. We don't know if ice mass has changed enough to do this, but over the next few years satellite measurements of gravity over Antarctica and other ice-bound areas could settle the question.

For direct confirmation of Dumberry's theory we might have to wait a while longer, to look for a correlation between the wobble and the small variations in Earth's magnetic field over decades that are thought to be linked to inner-core motion. The wobble, magnetic field fluctuation and core rotation should all follow one another; if one were to reverse, so should the other two.

If the figures do add up and Dumberry's model turns out to be correct, it could tell us something profound about the Earth's inner core. Because of its rather inconvenient location, the inner core is the least understood part of the Earth, and although scientists have a fairly good idea that it is mostly made of crystalline iron, they don't know what properties the metal has at the high pressures nearer the centre of the Earth. In fact that is an understatement: calculations of the inner core's viscosity - the extent to which it can be deformed - vary by a factor of a trillion. The stuff could feel like toffee or, well, like iron.

One measure of this is relaxation time: if you were to put a dent in the stuff, would it smooth out in minutes or millions of years? "We don't have a clue," says Dumberry.

Experiments are no help. By squeezing material on an anvil made of diamond, or sending explosive shock waves through it, it is just possible to reproduce the kind of pressure felt in the mantle, but no deeper. The inner core is out of reach.

Dumberry's model only works if the inner core's viscosity is within a rather narrow range. If it is too fluid, the tidal bumps would stay in the same position as the core rotates; too rigid, and the repeated gravitational tugs would make the inner core wobble wildly with a period of about seven years. That seven-year wobble just isn't seen.

So if the Markowitz wobble is indeed generated as Dumberry describes, then the viscosity of the inner core has to be just right, with a relaxation time of about 10 years. That's not exactly rock-hard, but it is far from soft. The question is, will we all find the Earth's core as exciting and mysterious if it turns out to be the consistency of Silly Putty?

"Europe is very well prepared but I hope there will not be a pandemic in the coming months," Zsuzsanna Jakab told Swedish newspaper Svenska Dagbladet on Saturday.

She said the centre was working systematically with each of the 25 European Union members on contingency plans. Six had been completed while the remainder should be finalised within a year.

Confirming its 19th death from bird flu, Indonesia vowed to boost stockpiles of Tamiflu, a general anti-viral medicine seen as the most effective at present, amid a steadily climbing number of cases.

The world's fourth most populous nation has reported eight fatalities from H5N1 this year alone, the highest figure globally as the virus continues its march from Asia into Europe and Africa.

A 23-year-old chicken vendor who died in hospital a week ago had tested positive for the virus at a World Health Organisation-certified laboratory in the United States, the Indonesian health ministry said.

With the virus raging through chicken farms in Nigeria and established in Egypt, the United Nations food safety agency has warned of dramatic consequences for west Africa if the pattern of the Nigerian outbreak is repeated.

"The effects on a region already facing severe malnutrition would be devastating," said the Food and Agriculture Agency (FAO).

"If a poultry epidemic should develop beyond the boundaries of Nigeria the effects would be disastrous for the livelihoods and the food security of millions of people," said veterinary official Joseph Domenech.

After tests revealed a wild duck found dead in a marsh in central France was probably infected by H5N1, the agriculture ministry said Saturday it was examining the carcasses of more than a dozen birds found across the country.

"For some days were have been expecting animals carrying the virus to be found," official Didier Houssin told radio station Europe 1.

"From the moment the hypothesis of the role of migratory birds in carrying the virus great distances was evoked ... we were able to foresee in a way what is happening now."

A three-kilometre (two-mile) safety cordon has has been set up around the spot where the duck was found, near the town of Joyeux, and wildlife surveillance stepped up across a 10-kilometre area.

Britain warned the discovery increased the likelihood bird flu could cross the Channel.

"It is more likely now than it was, but it is not inevitable," Animal Health Minister Ben Bradshaw said on Friday.

"It's still only in wild birds. It hasn't been found in any poultry so far in EU countries this time," he told the BBC. "But clearly the closer it gets to us the risk grows."

The British government has drawn up plans to set up one-mile (1.6 kilometre) exclusion zones if any wild bird is found in Britain to be infected with H5N1.

Inside the zone, all poultry movements would be halted, and if any poultry was found to be infected the entire flock would face slaughter.

So far in Europe, the H5N1 virus has been detected in poultry and wild birds, mostly swans, in Austria, Bulgaria, Croatia, Germany, Greece, Italy, Romania, Slovenia, Ukraine and Russia.

In Egypt, the virus has been detected in domestic poultry in Cairo and in the Minya region, 250 kilometres (155 miles) south of the capital.

Tests were being carried out on chickens in the southern governorate of Qena, where the popular tourist areas of Luxor and Thebes are located, after the mysterious deaths of some 130 birds in a village, a government spokesman said Friday.

H5N1 kills about half the humans it infects. Since 2003, 93 people, all of whom had come into direct contact with the carcasses of infected fowl, have died, most of them in China and Southeast Asia, but also in Iraq and eastern Turkey, according to WHO figures.

Health experts fear the virus could mutate into a form that can pass between humans, leading to a global pandemic.

A senior military officer, Lt Col Raul Farnacio, said that 19 bodies had been recovered and 57 survivors located. But the rest of the population of 1,857 was unaccounted for, with scant prospects for further recoveries. "We presume that, more or less, 1,800 are feared dead," he told local reporters.

Article continues
Television footage showed only a few tin roofs and bits of mangled debris, some of which had been carried hundreds of metres, visible above the sludge. The village primary school was obliterated, with schoolchildren inside. The mud was 10 metres (30ft) deep in places.

"So many died," said a resident, Eugene Pilo, on local television. "Our village is gone, everything was buried in mud. All the people are gone."

Local relief teams were resuming their grim task early today, with dog teams augmenting the effort. A US warship conducting exercises with the Philippines navy was diverted to help. Relief efforts were hampered by the very soft mud which caused rescuers to sink up to their chests.

"The troops pulled out because big boulders are cascading down the mountain," said Col Farnacio, who is in charge of the military's relief operations.

Captain Edmund Abella, part of the relief operation, said the search was extremely treacherous. Heavy equipment was almost impossible to use. "It's very difficult. We're digging by hand, the place is so vast and the mud is so thick. When we try to walk, we get stuck in the mud."

The disaster is being blamed on two weeks of heavy rain and the replacement of the deep-rooted hillside forests with coconut plantations, which have shallower roots and cannot hold soil.

Illegal logging is rampant across much of the Philippines, but it is unclear whether the clearances in this area were carried out illegally.

Some 3,000 people from other villages spent last night in municipal buildings in the nearby town of St Bernard as heavy rain continued to lash the region.

The provincial governor, Rosette Lerias, said that, as well as the inhabitants living in Guinsaugon, which covers an area of about one square kilometre (250 acres), another 1,150 lived nearby.

"I'm really afraid the death toll is going to be very high," she told the Guardian. "There are no homes left. It is a tragedy, and we can't do much - because even the rescuers' lives are in danger."

Survivors said that they were unable to do anything to stop the disaster, which struck at about 10am. "It sounded like the mountain exploded, and the whole thing crumbled," Dario Libatan told radio station DZMM. "I could not see any house standing any more."

Didita Kamarenta told Reuters: "I felt the earth shake and a strong gust of wind, then I felt mud at my feet. All the children, including my two children, are lost. They might have been buried."

The president of the Philippines, Gloria Arroyo, said in a national television address that she was mobilising "the full resources of the government" to help with the relief effort. "They will come from land, sea and air," she said. "Hopefully you will soon be out of harm's way."

Local people reported an earthquake just before the landslide, but experts said that its magnitude of 2.9 was too small to have triggered the disaster alone.

Deadly landslides are not uncommon on Leyte island. In 1991 more than 5,000 people were killed in floods and landslides triggered by a typhoon.

Some 133 people died in similar circumstances in 1993.