# News this Week

Science  28 Mar 2003:
Vol. 299, Issue 5615, pp. 251
1. RESEARCH FUNDING

# U.K.'s Biomedical Agency Gets a Parliamentary Tongue-Lashing

1. Richard Stone,
2. Gretchen Vogel

5. ASTRONOMY

# Distant Quasars Give Astronomers a Massive Puzzle

1. Dennis Normile

TOKYO—The race to unravel the early history of the universe heated up last week, as the Japanese Subaru Deep Field (SDF) Project announced first results from the 3-year-old Hawaiian telescope: an in-depth look at two of the three most distant galaxies ever sighted. The findings, reported in the April issue of the Publications of the Astronomical Society of Japan, give the earliest glimpse yet of how vigorously galaxies created stars in the first billion years after the big bang and may help astronomers nail down that and other key cosmological parameters.

“The [star formation rate] is a number we all want to know, and this work is making progress toward that,” says Esther Hu, an astronomer at the University of Hawaii, Manoa, who leads another team hunting for deep-sky objects. Studying how the birth rate of stars has changed over time, she explains, gives cosmologists crucial clues to the evolution of galaxies.

The SDF survey uses the 8.2-meter Subaru telescope on Mauna Kea, Hawaii, to probe a full moon-sized patch of sky in the constellation Coma Berenices. Its quarries are young galaxies at a redshift of 6.5, or about 13 billion light-years away. “It's one of Subaru's biggest missions,” says Keiichi Kodaira, first author of the paper and a former director of the National Astronomical Observatory of Japan, which operates Subaru. Hu's group, meanwhile, scans a different part of the heavens using the 10-meter Keck telescopes, also on Mauna Kea. But whereas Hu often relies on a technique called gravitational lensing, in which an intervening cluster of galaxies amplifies light from a more distant source, the Subaru group peers into a blank field containing no lensing galaxies.

Lensing allows astronomers to see fainter objects than the blank-field approach does, a big help in estimating rates of star formation. To offset that advantage, Subaru sports Suprime-Cam, a mammoth 80-million-pixel charge-coupled device camera that can image 20 times as much sky as Keck's cameras can. As a result, Subaru can collect a greater number of distant objects more quickly, says Yoshiaki Taniguchi, an astronomer at Tohoku University in Sendai and a co-author of the paper.

In 2001 and 2002, Kodaira's team found 73 candidate galaxies. Follow-up observations with a spectrograph confirmed one of them to be at redshift of 6.58 and one at 6.54. (Hu sighted the only other object yet observed beyond redshift 6, at 6.56.) Using empirical relationships linking star formation to luminosity and other parameters established by studying nearby galaxies, Kodaira and his team estimated that the two galaxies were forming several solar masses' worth of stars a year. Combining their observations of the 73 candidate galaxies and making a number of assumptions, they derived a star formation rate density of 0.001 solar mass per year per cubic parsec, an order of magnitude less than the current rate density in the universe around Earth.

Taniguchi says the small number of samples and large number of assumptions led them to calculate very conservative lower limits for star formation. But over the next year, they hope to confirm that 30 to 40 galaxies in their deep field are beyond redshift 6. That number would enable them “to narrow the error bars” on their star-formation rates.

It would also shed light on a landmark cosmic event, called reionization, that occurred when ultraviolet light from the first stars broke hydrogen atoms—then the dominant form of matter—back down into the protons and electrons from which they had formed as the universe cooled after the big bang. “Trying to understand exactly when and how the universe was reionized is the frontier [issue of cosmology] right now,” says Avi Loeb, a theorist at Harvard University. Recent results from the Wilkinson Microwave Anisotropy Probe suggest that the process started as early as 200 million years after the big bang (Science, 14 February, p. 991). Taniguchi hopes that analyzing the spectra of light from the galaxies in their deep field will yield estimates of how quickly reionization progressed.

6. ASTRONOMY

# Rare Outburst Lights a Dusty Shell

1. Robert Irion

An intense burst of light from an otherwise nondescript star has lit up a cocoon of dust in a rare “light echo,” shown in these Hubble Space Telescope (HST) images. The star, named V838 Monocerotis, suddenly inflated into a cool supergiant in January and February 2002, growing 10,000 times brighter and becoming the most luminous star in our galaxy for about 40 days. Light from the flash is now racing outward through matter the star shed in previous outbursts. As the light scatters off dust farther and farther from the star, HST sees thin cross sections of the nebula in a sort of tomographic scan, says astronomer Howard Bond of the Space Telescope Science Institute in Baltimore, Maryland. “The nebula itself isn't growing on this time scale,” says Bond, lead author of a report on the light echoes in this week's issue of Nature. “The only thing moving is the light.”

Bond's team doesn't yet know what triggered the strange flare, although spectra suggest that the star is a binary pair. Matter falling from a companion star onto a white dwarf might have induced a thermonuclear runaway that forced the dwarf to expand radically without exploding into a more common nova, Bond notes. The illuminated wisps are far more intricate than those seen in any light echoes to date, says astronomer Benjamin Sugerman of Columbia University in New York City. The stunning details, he says, will lead to a three- dimensional map of the dust around V838 Monocerotis as the echoes grow—promising unprecedented insights into the past tantrums of an active star.

7. INFECTIOUS DISEASES

# A Second Suspect in the Global Mystery Outbreak

1. Martin Enserink*
1. With reporting by Gretchen Vogel in Berlin.

As a new respiratory disease that originated in Southeast Asia spread this week and raised alarms, especially in Hong Kong, the global hunt for the mysterious pathogen behind the outbreak took an unexpected turn. Just as evidence seemed to point to a paramyxovirus, researchers fingered a candidate from a very different virus family, the coronaviruses. The World Health Organization (WHO) says both may be involved.

By Tuesday, health officials had counted 487 probable cases of severe acute respiratory syndrome (SARS) worldwide and 17 deaths. Over the past 2 weeks, researchers from 11 labs around the planet have collaborated under WHO guidance in a frenzied search for the pathogen, exchanging samples and reagents and sharing information on a secure Web site. But their work has produced varying, and at times confusing, results.

Last week, German researchers said electron micrographs suggested that the virus belonged to the Paramyxoviridae, a broad group that includes agents causing common diseases such as parainfluenza, mumps, and measles—as well as Nipah and Hendra, two zoonotic agents that created outbreaks in Asia and Australia in recent years. The search seemed to narrow on 21 March, when scientists at the National Microbiology Laboratory in Winnipeg reported finding one particular paramyxovirus, called the human metapneumovirus, in tissue samples from six out of eight patients in Canada. The metapneumovirus was identified in 2001 by Ab Osterhaus and his colleagues at Erasmus University in Rotterdam, the Netherlands, as a cause of respiratory tract infections in young children and has since been found in many parts of the world.

Researchers in Hong Kong were the first to identify a coronavirus as the possible culprit, says Klaus Stöhr, WHO's lead virologist on the case. Coronaviruses are known to cause the common cold in humans and more serious diseases—including pneumonia—in animals. On Monday, Julie Gerberding, director of the U.S. Centers for Disease Control and Prevention (CDC) in Atlanta, announced that, although her agency is “very respectful” of the paramyxovirus findings, it considers the coronavirus etiology its “leading hypothesis.” CDC scientists have cultured a coronavirus from tissue samples from two patients, Gerberding said; they have also found it to occur specifically in lung and kidney tissues affected by the disease. And in an important clue that gives CDC confidence, researchers have found that one patient started producing antibodies to the virus as the infection progressed.

Osterhaus says the data look strong, but to clinch the case, researchers will have to identify the virus in more patients and possibly confirm the findings with animal studies. From the start, the metapneumovirus seemed a less likely candidate, Osterhaus says, because the syndrome it produces is not quite like SARS. But it's still possible that a paramyxovirus is involved, he said; SARS might require an infection with multiple viruses.

WHO officials kept all options open during a press conference on Tuesday, including the two-agent scenario and the possibility that something entirely different is to blame. It could be that both a paramyxovirus and a coronavirus are innocent bystanders, WHO's executive director for communicable diseases David Heymann said.

Members of the coronavirus family—so called because they appear to have a crown, or halo, in microscopic images—can infect mammals and birds. Although the viruses are known to cause severe disease in some animals, those infecting humans are known primarily for causing the common cold. They fall into two groups, called 229E-like and OC43-like. But based on the sequence of one of the virus's genes, obtained by a group at the University of California, San Francisco, CDC believes SARS may be caused by a distinct, as-yet-unknown coronavirus with a genetic signature different from that of known types. In animals, researchers have found that different members of the group can recombine their RNA relatively easily, giving rise to new viruses, says Susan Baker, who studies coronavirus replication at Loyola University Chicago Stritch School of Medicine in Illinois—but it's unclear whether this could have given rise to SARS, she says.

8. ASTROPHYSICS

# Competition Heats Up for Underground U.S. Lab

1. David Malakoff

Not all that glitters is, indeed, gold. Just ask Marvin Marshak of the University of Minnesota, Twin Cities. The physicist, a central player in a controversial bid to convert the Homestake gold mine in South Dakota into the world's deepest underground laboratory, now says a retired Minnesota iron mine might be a better bet. He'll soon ask the National Science Foundation (NSF) for up to $200 million to expand the Soudan mine to house cutting-edge detectors to study neutrinos and other natural phenomena. “The Homestake project is stuck, and we need to get unstuck,” Marshak told Science, adding that he doesn't view his move as “a defection” from the Homestake campaign. But Marshak's request will put him in the unusual position of backing two competing proposals. Homestake backers, meanwhile, also reject the idea of a split, preferring to see Marshak's proposal as another vote for a world-class buried laboratory somewhere in the United States. Marshak is one of five researchers who 2 years ago submitted a$281 million proposal to NSF to turn parts of the 2300-meter-deep Homestake mine into a lab before its owners—Canadian mining giant Barrick Gold Corp.—abandoned and flooded the tunnels (Science, 15 June 2001, p. 1979). Astrophysicists covet the mine because it is deep enough to shelter sensitive detectors from cosmic radiation.

NSF is still reviewing the proposal. But Barrick executives say they won't hand over the mine to scientists unless the government protects the company from legal liability for safety, environmental, or other problems that may show up later. Congress passed such “indemnity” legislation in 2001, but Barrick rejected the deal amid election-year charges that it could saddle taxpayers with millions of dollars in cleanup costs. A proposal for South Dakota to take over the mine has also languished. Several scientific panels, meanwhile, have endorsed the concept of an underground laboratory but remained silent about its location.

The Homestake gridlock convinced Marshak that physicists need to look elsewhere. That led him to the Soudan facility, a 710-meter-deep shaft near Tower where he and other researchers have conducted experiments for more than 20 years. It would be faster and cheaper to expand and deepen the Soudan mine (eventually to about 2500 meters) than to start from scratch, he says.

Soudan isn't the only alternative to Homestake. Scientists have touted plans for doing underground science beneath the granite San Jacinto Mountains in California and in salt domes under New Mexico. This month, they toured a working Colorado mine. But a 2001 study led by astrophysicist John Bahcall of the Institute for Advanced Study in Princeton, New Jersey, deemed Homestake the best bet.

Congress will have the final say on whether and where to build a lab. In the meantime, NSF officials say they are happy to help the physics community set priorities. Researchers are likely to unite behind Homestake if it gains momentum, predicts Wick Haxton, the University of Washington, Seattle, physicist who is the lead investigator on the proposal. And Marshak says there's consensus that “we need an underground laboratory. We just disagree on the timing and tactics.”

9. CLINICAL RESEARCH

# AIDS Vaccine Results Draw Investor Lawsuits

1. Jon Cohen

VaxGen, the Brisbane, California, biotech that came under scientific fire last month when it revealed results of the first-ever real-world trial of an AIDS vaccine, is now dodging legal bullets. On 17 March, the San Diego branch of the law firm Milberg Weiss filed a class-action suit on behalf of investors against VaxGen, a publicly traded company, and two of its officials. The lawsuit charges that the company and its officials “deceived” investors with a “fraudulent scheme” that “concealed” negative results from the trial. Within days, four other law firms filed what appear to be copycat suits.

VaxGen says the lawsuits have no merit. “These are nuisance lawsuits,” says company spokesperson Lance Ignon. “We'll obviously defend ourselves vigorously. We have every confidence that we'll prevail, and they'll never make it to court.” Milberg Weiss attorney Mary Blasy declined to discuss the allegations. Attorneys from the other firms did not reply to requests for interviews.

VaxGen has staged two placebo- controlled efficacy tests of its AIDS vaccine in various countries. Both trials are double-blinded, meaning that no one knows whether participants received vaccine or placebo until the data are decoded at the end of the trial. On 24 February, the company revealed data from the largest trial that officials said they had decoded 1 week earlier (Science, 28 February, p. 1290). The data showed that, overall, the vaccine had failed to protect against infection. But the company reported that responses in blacks, Asians, and people of mixed race showed that the vaccine had had at least a 67% efficacy in these racial groups. VaxGen at first claimed that these results were statistically significant but later acknowledged that statistical “penalties” for breaking the data into subgroups had not been factored in (Science, 7 March, p. 1495).

The Milberg Weiss suit alleges that by 6 August 2002, VaxGen CEO Lance Gordon and president Donald Francis knew that the infection rate among those vaccinated differed little from that in the general population, leading to a “statistically irrelevant” efficacy. Yet the pair continued to make “materially false and misleading” statements about the vaccine's prospects, the complaint claims. The complaint offers no evidence to explain how Gordon and Francis could have known the results before the data were decoded, however. The complaint further alleges that VaxGen misstated the significance of the subgroup analyses in order to have a “corrective effect” on the company's stock price, which declined by more than 50% following the report of overall failure.

VaxGen's Ignon says the lawsuits are no surprise. “If you're high profile and your stock has a sudden and precipitous drop, chances are someone will file a lawsuit,” he says. “It's the way of the world these days.”

10. WAR IN IRAQ

# Bracing for Gulf War Syndrome II

1. Martin Enserink

Unexplained medical problems among war veterans may be inevitable—but this time, the military says it's much better prepared

Long before U.S. and British troops swept into Iraq last week, military doctors and epidemiologists were planning for the war's aftermath. They are anticipating a repeat of Gulf War syndrome, the vexing series of unexplained medical problems that affected tens of thousands of veterans after the 1991 war. This time, they say they are much better prepared to prevent, recognize, and treat such problems than they were 12 years ago.

In the Iraqi desert, continuous environmental monitoring and improved record-keeping should help avoid health hazards and prevent the widespread confusion over who was exposed to what that erupted after the last war. Back home, there's a plan in place to take better care of patients. “We have learned our lessons,” says Michael Kilpatrick, deputy director of the U.S. Department of Defense's (DOD's) Deployment Health Support Directorate. “I'm confident that we're well prepared to welcome our veterans home.”

That confidence reflects a profound change in attitude. After the first Gulf War, DOD was chastised by veterans, Congress, and the media for ignoring the problems or even branding patients as malingerers. Today, a new philosophy has taken hold among military leaders that accepts that the panoply of medical problems is real—indeed, likely to occur after every major deployment—and that science cannot pinpoint the causes. Acknowledging the problems and addressing them early on, instead of shuffling veterans through the medical system in a fruitless search for answers, may in itself prevent a “snowball” effect of illness, worries, and suspicions, says Charles Engel, an influential Gulf War researcher and clinician at Walter Reed Army Medical Center in Washington, D.C.

## Battlefield epidemiology

Science's failure to get a handle on what ails Gulf War vets isn't for want of trying. Over the past dozen years, researchers have spent almost a quarter of a billion dollars on hundreds of research projects. They have zoomed in on potential causes such as chemical and biological weapons, pyridostigmine bromide (PB, a drug used to counter nerve gases), depleted uranium in ammunition, vaccine side effects, smoke from burning oil wells, infectious diseases, and pesticides. But despite vast epidemiological studies and an array of animal experiments, most researchers and a dozen expert panels agree that almost none of these factors can be conclusively linked to any of the medical complaints, which include fatigue, muscle and joint pains, asthma, dizziness, and rashes, as well as memory, attention, and concentration problems (Science, 2 February 2001, p. 812).

Most researchers today are pessimistic that clear links will ever emerge. Almost all the symptoms also occur in the population at large—although at lower levels—where they are just as difficult to understand. And similar, baffling complexes of symptoms have appeared after almost every major war, suggesting a more general cause, such as the stress and horror of warfare. “These things are part and parcel of going to war,” says Simon Wessely of the Gulf War Illness Research Unit at Guy's, King's and St Thomas' Hospital in London. “It would be naive to think you can prevent them completely.”

But Wessely thinks they can be moderated—which is exactly what U.S. and British forces are trying to do. The strategy includes avoiding real hazards and collecting better data. This time, for instance, special environmental teams are traveling along with the troops to take air, water, and soil samples and test them for a range of contaminants, Kilpatrick says. Each unit's geographical position during the course of the war is being recorded, which should further help tease out potential exposures. Medical record-keeping is automated and is as accurate and complete as possible—a far cry from the sloppy practices of 12 years ago, when it wasn't even clear who had gotten anthrax shots and PB pills in the Gulf. Further helping reduce both actual threats and anxiety is a new generation of chemical and biological warfare detection systems.

The chances of accidental exposure are also much smaller this time, says Kilpatrick. During the last war, U.S. forces blew up what they thought was a conventional ammunition depot near Khamisiyah in Iraq, which later turned out to contain rockets filled with sarin and cyclosarin. Although there were no direct casualties, about 100,000 troops were later estimated to have been exposed to trace amounts of gas as it wafted across the desert. This time around, any weapons caches will be approached with much more caution, Kilpatrick says—if only because Saddam Hussein's alleged weapons of mass destruction served as the casus belli, and the U.S. has a keen interest in investigating and exposing them.

If all goes according to plan, all medical and exposure data will be stored in the Defense Medical Surveillance System, a database that already contains the medical records of 7 million military men and women, linked to a blood serum repository currently containing 27 million samples. This vast resource allows researchers to answer all kinds of questions quickly, says Mark Rubertone, head of the U.S. Army Medical Surveillance Activity. For instance, some scientists have speculated that infections with a microbe called Mycoplasma fermentans contributed to Gulf War syndrome. But a comparison of blood samples taken before and after deployment, published in 2000, showed no higher incidence of mycoplasma infections among those suffering from medical problems.

Gulf War veterans and activists give the military credit for some of its preparations but criticize others. To allow better studies, for instance, a law passed in 1998 requires that each soldier have a serum sample stored before and after every deployment. But rather than taking special samples, DOD relies on samples taken annually during mandatory HIV tests, says Stephen Robinson, director of the National Gulf War Resource Center. That means that up to a year could elapse before returning veterans have their blood sampled—a time during which traces of exposures may disappear, Robinson argues. (A House subcommittee had scheduled a hearing to discuss that issue this week.)

## Better treatment

Things will also be much different for soldiers after they come home, the Pentagon asserts. Engel and others have crafted a so-called clinical guideline for dealing with postdeployment health problems. It aims to avoid the fruitless search for a cause that patients with unexplained illnesses often go through. If no diagnosable disease can be found, therapy should focus on rehabilitation and “physical reactivation.” Patients should be fully informed about their exposures and what science can tell about the associated risks.

For the most severe cases, the Health Clinical Center that Engel directs at Walter Reed has pioneered an intensive 3-week treatment program, modeled after therapies for other elusive chronic illnesses such as chronic fatigue syndrome, that includes so-called cognitive behavioral therapy (CBT) and exercise. A key factor, says center psychologist Roy Clymer, is that patients let go of their “find-it-and-fix-it” expectations of medicine and start taking steps themselves to deal with their problems. The treatment team, for its part, does not question the reality of veterans' illnesses.

Critics like Robinson say such therapies miss the point because they don't address the disease itself; they just teach people how to live with it. But patients at the Health Clinical Center praise the new approach. “The big difference is, here they believe you,” says David McNease, one of three veterans who completed the program last week.

Glowing testimonials aside, the long-term success of this type of therapy is unclear. Just last week, the Journal of the American Medical Association published the results of a trial in which its key components—CBT and exercise therapy—were tested among 1092 veterans in 20 Veterans Affairs and DOD medical centers. Cognitive therapy had only a very modest effect on overall physical functioning, and exercise did not seem to help at all, the study found.

Engel, who co-chaired the panel overseeing the study, says that therapists at some of the study sites were inexperienced and veterans' adherence to the protocol was poor; a highly focused program like his own might yield higher success rates, he says.

His biggest hope, however, is that the preparations now in place will mean that fewer veterans will need therapy in the first place.

11. WAR IN IRAQ

# U.N. Inspections Find Wisps of Smoke But No Smoking Guns

1. Richard Stone

Scientific teams were withdrawn on the eve of the war after spending 3 months looking for weapons of mass destruction; in interviews, several inspectors relate what they found

CAMBRIDGE, U.K.—Even before the first shots were fired, the war in Iraq claimed a prominent casualty: the United Nations (U.N.) and its efforts to disarm Saddam Hussein's regime through diplomacy and on-the-ground inspections. That process came to an abrupt halt last week when the U.N. withdrew its weapons inspectors from Baghdad on the eve of the war. As the endgame in Iraq commenced, Science spoke with a range of experts, including U.N. inspectors just back from Baghdad, to get a sense of what had been accomplished during more than 3 months of efforts to ferret out weapons of mass destruction (WMD)—and what should be done to engage Iraqi scientists after the war ends.

The inspectors uncovered some wisps of smoke but no smoking guns: no proof that the Iraqi regime, in its dying days, maintained stocks of biological or chemical weapons. Nor did evidence turn up to back claims that Iraq had resurrected a nuclear weapons R&D program snuffed out by the U.N. in the early 1990s. But although Iraqi officials this time around allowed speedy access to institutes and labs, scientists detailed to the U.N. Monitoring, Verification, and Inspection Commission (UNMOVIC) say their efforts to penetrate this patina of cooperation often were met with obfuscation. “The Iraqis could not have been expected to confess their sins,” says UNMOVIC inspector Heinz-Ulrich Gläser, a WMD expert at the Federal Office of Defense Technology and Procurement in Koblenz, Germany. Researchers quizzed at the sites often failed to answer questions or did so deceitfully. Other clues fed suspicions that the regime was hiding something. “Our work merely proved that Hussein hadn't changed his goals; he just got better at hiding it,” asserts one veteran bio weapons inspector who served on UNMOVIC and earlier U.N. teams.

Much of the fog should soon lift. As Science went to press, military planners were girding for the grim possibility that Hussein might unleash chemical or even biological weapons, if he possessed them. Even if he were to refrain, stocks of such weapons could come to light during the war or its aftermath.

The war's end should also bring a new dawn for Iraqi scientists. With Hussein's police gone, they are likely to provide critical observations and perhaps documents on Iraq's WMD programs. Analysts are now debating the merits of a program to fund peaceful research for former Iraqi weapons scientists styled after a decade-long initiative to support nonproliferation activities in the former Soviet Union.

## Drawing blanks

Saddam Hussein's development and use of WMD is not in doubt. His regime is notorious for having gassed the Kurdish town of Halabjah in northern Iraq in March 1988, killing hundreds, and the U.N. Security Council condemned Iraq for its repeated use of mustard gas during the Iran-Iraq War. But when U.N. inspectors began their investigations into Iraq's WMD activities after the Gulf War in 1991, the Iraqi regime insisted that it had abandoned its chemical weapons program and claimed that it had pursued only a limited R&D effort on defenses against biological weapons. It took 4 years of dogged sleuthing, along with key revelations from a defector, before Iraq divulged that it had filled munitions with thousands of liters of anthrax, Clostridium botulinum toxin, and aflatoxin and had cooked up large quantities of C. perfringens toxin, which causes gas gangrene, and wheat smut (Science, 16 August 2002, p. 1110). By the time the Iraqi government threw them out of the country in November 1998, U.N. inspectors had overseen the destruction of the known chemical and biological materiel and had tagged equipment used in R&D at dozens of facilities.

When inspectors prepared to return to Iraq after a 4-year hiatus, they knew they faced a daunting assignment. The majority of the members of UNMOVIC's 350-strong roster had never been in Iraq. The organization's predecessor, UNSCOM, had been tarred with allegations toward the end of its tenure that it was a tool of the United States. In part to avoid this perception, UNMOVIC chief Hans Blix built his team mainly from scratch. “It was a very different animal from UNSCOM,” the U.N. teams of the 1990s, says Graham Pearson, former director-general of the U.K. Ministry of Defence's Chemical and Biological Defence Establishment. “My general impression is that almost everyone with UNSCOM who knew anything was sidelined.”

When the new teams began work last November, they devoted part of their time to retracing the steps of their predecessors, visiting sites known to have been involved in past weapons work and probing for signs of fresh illicit activity. But they also cast a wide net, sweeping in potential dual-use facilities that the Iraqi government had been forced to declare last autumn: vaccine plants, breweries, or other outfits that could be converted readily to military use or could be masquerading as civilian operations. UNMOVIC officials in New York selected targeted sites; the personnel in Baghdad had little advance warning. “I did not know the nature of a site earlier than late afternoon of the day before an inspection,” says Gläser. The teams, often composed of a dozen or more scientists, kept up an unrelenting pace; each inspector averaged two site visits every 3 days. They took hundreds of samples for analysis at their field office in Baghdad.

About 25% of the 600-odd inspections took place at sites that the Iraqi government had not declared. Many visits were based on tips from the CIA and other national intelligence agencies. But these turned out to be wild WMD chases, inspectors say. The intelligence reports were “worthless to the point of embarrassment,” asserts the veteran inspector. For example, on several occasions the U.N. teams were ordered to investigate what appeared on the surface to be chicken farms. These turned out to be, well, farms for chickens. One inspector had T-shirts made that lampooned their mission as “Ballistic Chicken Farm Inspectors.”

Far more inscrutable were the Iraqi scientists encountered on inspections. “They would be very evasive on even the most basic questions,” says Rocco Casagrande, an inspector with expertise in weapons aimed at crops and livestock. For instance, in a laboratory at one institute, he asked a researcher leading the tour whether the lab worked on viruses. “He said ‘No, no one does any work with viruses.’” But when Casagrande saw viral cultures in a lab down the hall, he asked about the discrepancy. The researcher's response: Those cultures belonged to someone else; he did not know what research went on there. “After getting these kinds of responses everywhere you went,” says Casagrande, “there are only two conclusions you can draw: Either the people are very nervous” or, much less likely, he says, “every facility you go to is involved in a covert biological weapons program.” The researchers had good reason to be nervous: Government minders from Iraq's National Monitoring Directorate (NMD) were on hand for the inspections. Also present on most inspections were Special Security Organization officers who reported directly to Hussein.

Even more disappointing were the private interviews that UNMOVIC sought with scientists presumed to have been key players in Iraq's past WMD programs. Of approximately 30 bioweapons experts initially invited for an interview, half declined. “All the scientists were very scared, since all were driven there and picked up by NMD minders,” says the veteran inspector, who participated in the interviews. Even more crippling was a Blix directive, which drew flak from some inspectors, forbidding tape recordings by either party. (Nuclear inspectors, on the other hand, were allowed to tape interviews.) A dozen scientists who did show up had asked to make a recording for their own protection and declined to proceed when that request was denied. By the end of February, just three scientists had acquiesced—but it was hardly worth the effort. “They obviously only had knowledge of the old programs,” the veteran inspector says; the same was true for the NMD minders, many of whom were holdovers from the UNSCOM days. The pace picked up earlier this month: Eleven more scientists had been interviewed before UNMOVIC staff were pulled out, but the U.N. had not released details before Science went to press about what it had learned.

One of the more dramatic allegations that U.N. inspectors were asked to investigate was that Iraq had mobile bioweapons labs that could be moved from site to site on rails or by truck. “We took it very seriously,” says Casagrande. He and other inspectors looked for evidence such as signs that large trucks had been present at a biological site. But it was like chasing phantoms: Apart from a handful of foreign catalogs depicting mobile platforms, they found nothing to indicate that the movable beasts existed, the veteran inspector says. Even if there were mobile labs, Gläser doubts that they could produce substantial quantities of bioweapons. “Otherwise,” he says, “Iraq would be the leading nation in biotechnology.”

Although in the end they came up empty, many inspectors remain doubtful that they came away with a true picture. After all, “Iraq still hadn't provided any evidence that they actually decommissioned their bioweapons program,” Casagrande says. “If all this valuable equipment and all these valuable reagents are being destroyed, the person who's doing it is damn well going to have some paperwork to back himself up just in case someone asks him why he did this down the line. It's a relatively momentous decision in their military history. Someone has to be holding some paperwork somewhere.”

## A new beginning

After the war, Iraqi scientists will presumably find themselves working for a new government that makes a clean break with the country's WMD past. “It's crucial to get the government of Iraq to turn around completely, like South Africa did” when the postapartheid government relinquished its nuclear arsenal in the early 1990s, says Pearson. Few expect that Iraqi scientists will pose a grave proliferation threat after the war. “We did not see big facilities with huge numbers of scientists working to develop WMD,” says Gläser, who doubts that there will be “an exodus of scientists supporting Osama bin Laden or rogue states.”

Nevertheless, some observers argue that Western nations should move quickly to establish a program that steers key individuals into civilian research. “Iraqi weapons scientists who find themselves unemployed and impoverished in the aftermath of an invasion could be tempted to sell their expertise to the highest bidder,” warns nonproliferation expert Jonathan Tucker, a senior fellow at the United States Institute of Peace in Washington, D.C. One model he and others are touting is the International Science and Technology Center, an organization funded largely by the United States, the European Union, and Japan that has handed out peer-reviewed grants to thousands of scientists in Russia and other former Soviet nations. Establishing a similar center for Iraq, argues Tucker, who served on a U.N. bioweapons inspection team in 1995, would “make it easier to track the movements of Iraqi weapons experts.”

Others say Iraq may not need such a program. A postwar government may be able to provide sufficient funding for science from its oil revenues, Casagrande argues. “There will be an explosion of legitimate science after the war,” he predicts. “Iraq has very talented Western-trained scientists who could do outstandingly well with regime change.” And for the first time in years, many of the Western-educated scientists will have the opportunity to travel abroad.

Most UNMOVIC inspectors, meanwhile, have now returned to their old jobs or, like Casagrande, taken new ones. (He has just taken a position as director of homeland defense at Abt Associates Inc., in Cambridge, Massachusetts.) It's unlikely that any of them will be on hand if the illicit weapons they were hunting are indeed flushed out. That job, the Pentagon says, will fall to scientists attached to specialist U.S. military units.

Although UNMOVIC's work failed to prevent a war, many feel it was worthwhile. The 14 weeks of grueling inspections, Casagrande says, identified facilities that should continue to be monitored after the war, providing “a road map of how to basically clean up once things settle down.” And whatever else Iraq had managed to hide should soon see the light of day.

12. PLANETARY SCIENCE

# Can NASA's Promethean Vision Bring Back Heavenly Data?

1. Andrew Lawler

Researchers applaud NASA's new plan for solar system exploration, but a nuclear component, major technical challenges, and a hefty price tag will make it a tough sell

Prometheus stole fire from the gods to benefit humanity, according to Greek mythology. Now NASA has borrowed his name for an ambitious plan to develop nuclear power and propulsion systems to explore the heavens. Officials hope this gift—just one part of a new road map for exploring Earth's neighbors—will provide planetary scientists with an abundant source of power to speed spacecraft bristling with sophisticated instruments to the far reaches of the solar system.

But Prometheus also brought down on humans the wrath of the gods, who retaliated by dispatching a host of worldly woes in Pandora's box. To avoid the same fate, NASA must solve some tricky technical problems, calm any public anxieties about nuclear-powered probes, and persuade Congress to cough up as much as $9 billion for the new program, which includes a mission using the new technologies to study Jupiter's moons. To pay for Prometheus as well as several other planetary initiatives, including a flight to Pluto, the White House wants to double solar system spending to more than$2 billion by 2008.

Researchers regard the initiatives as a strong vote of confidence for their field. But they also know that the program's high cost and technical challenges make it vulnerable in a time of recession, war, and growing budget deficits. “There's no jubilation in the streets yet,” says Mark Sykes, an astronomer at the University of Arizona in Tucson. “This could be a mirage.”

## All together now

Only 18 months ago, NASA's planetary program was in shambles. The White House nearly aborted both a Pluto and a Europa mission (Science, 4 January 2002, p. 32) in an angry response to freelance lobbying by Pluto advocates and to Europa cost overruns. So NASA ordered up a National Research Council (NRC) study, which ranked Europa as the best large-mission target and put Pluto atop a list of five medium-class missions (Science, 19 July 2002, p. 317). The panel also called for more intensive technology efforts, such as nuclear power system development. It left Mars exploration for another group.

The academy report became a rallying point for the research community, and it impressed NASA and White House officials. “Out of this chaotic, factious kind of feeling, we got some order,” says NASA space science chief Ed Weiler. “And what came together was a real vision for the next decade.” That vision received a boost from NASA chief Sean O'Keefe, a former Navy secretary with great enthusiasm for nuclear systems. “It's amazingly exciting; we're opening up the whole solar system,” says Colleen Hartman, NASA's solar system exploration manager.

Weiler admits that Prometheus so far is more vision than reality. Its most difficult element is nuclear-electric propulsion, which would provide spacecraft with long-duration thrusts by using fission to convert heat to power. NASA hopes to have a working system ready for the 2012 launch of the Jupiter Icy Moons Orbiter, a mission expected to cost up to $4 billion. It's a far cry from the 1990s philosophy of smaller and cheaper spacecraft championed by O'Keefe's predecessor, Daniel Goldin. Replacing tons of chemical propellant with a small reactor would allow probes to travel faster, carry more powerful instruments, and defy the constraints of gravity. Instead of a brief trip around Jupiter's intriguing moon Europa, for example, the proposed spacecraft could spend 6 months or more orbiting Europa and also visit Callisto and Ganymede. The prospect of a new generation of power-hungry instruments intrigues researchers. “What if you are no longer limited to the volts in this office?” says Hartman. She estimates that Prometheus could provide 1000 watts for science, allowing high-powered radars that would penetrate the thick ice of the three moons, lasers that would comb the surfaces, and multiple landers. By comparison, the agency's most sophisticated deep-space probe, the Cassini mission to Saturn, provides only 290 watts for its instruments. Hartman predicts that data rates from the spacecraft to Earth could expand as much as 1000-fold. Prometheus also would provide funding for a new generation of radioisotope thermal generators developed by the Department of Energy. The generators would replace conventional power systems that can't draw enough power from the sun to visit the cold and dark of the outer solar system. A$500 million probe to Pluto (which the Administration had opposed until the NRC report gave it high priority) has dibs on the last generator from the current series. The new generators, which like the current ones produce heat and then electricity through the decay of plutonium, could also allow Mars landers to operate at any latitude and last 10 times longer than those with solar arrays. The 2009 Mars Mobile Science Laboratory would be its first customer, providing coverage that Hartman describes as “all of the surface, all of the time.”

## Billions and billions

Such power wouldn't come cheap. Weiler says the entire Prometheus effort, with includes the Jupiter mission, would cost between $8 billion and$9 billion through 2012. Two years of in-depth study are needed to pin down the cost, he says, which is not out of line with previous major missions such as the Hubble Space Telescope. And the technology would be available for a whole generation of future missions. “It's like building the first B-2 bomber,” says Weiler. “The second and third ones are cheaper.”

But such huge numbers worry many planetary scientists and congressional staffers. “I feel uneasy,” says one former member of the NRC panel, who predicts that “the numbers will go up and up and up—and what we get will go down and down and down.” Trying to develop new space propulsion systems along with new instruments worries others. “Doing R&D in parallel with a mission is kind of scary,” says one congressional staffer, who says the failure of one piece could halt the entire project. “Without a usable space nuclear power system, those very powerful instruments would be worthless.”

Prometheus isn't the only item on NASA's wish list. This spring, it will begin a competition for new missions costing up to $650 million. Dubbed New Frontiers, the contestants will be four possible missions strongly backed by the decadal survey: a lunar sample return effort, a Jupiter polar orbiter, a flight to Venus, and a comet rendezvous. The winner would be named by fall 2004. NASA is already under the gun to launch the Pluto mission by early 2006 so that it can use a Jupiter gravity assist to avoid a lengthy delay in reaching the distant planet. The problem is a lengthy approval process—Cassini took 6 years—that NASA must navigate. Cassini drew legal challenges as well as street protests. Environmentalists were concerned about the risk of putting nuclear material on rockets as well as the danger of a later spacecraft flyby that supplies a gravitational boost from Earth. Expected opposition to the Pluto mission could extend to Prometheus as well. NASA officials play down the environmental issues, insisting that safety will be a priority. But the recent Columbia tragedy could lead to second thoughts among lawmakers as well as the public. Legislators began looking into the new planetary road map this week as House and Senate appropriators took up NASA's 2004 budget request. Weiler already faces financial pressure from the Gravity Probe B effort (Science, 21 March, p. 1827), the James Webb Space Telescope, and several other complex spacecraft under design or in development. Congress must also find the money to pay for the Iraq war, while tax revenues fall and the deficit soars past$300 billion. The new plan, worry some planetary researchers, could succumb to a Pandora's box of technical and budgetary problems.

“There may be a morning after,” warns Arizona's Sykes. But for now, he and other planetary scientists are hoping that Prometheus can fire up the imagination of U.S. politicians and the public—and bring home the heavenly goods for researchers.

13. COLUMBIA INVESTIGATION

# NASA's Hypersonic Lab Studies Factors Leading to Breakup

1. Charles Seife

Langley Research Center in Virginia is one of the few places with the facilities to shed light on what happens when a vehicle travels at Mach 18

Even to an aeronautical engineer, the 30-centimeter-long ceramic model of the space shuttle Columbia might seem no match for a jet of gas traveling at more than 7000 kilometers an hour. But each time, the model holds up bravely for several seconds before being pulled out of the flow, which travels much faster than even the speediest rifle bullet. After measuring its flash-heated surface, engineers reinsert the battered model into the tunnel for another blast of wind at six times the speed of sound (Mach 6).

Engineers and physicists at NASA's Langley Research Center in Hampton, Virginia, have begun conducting these and other tests for the federal commission investigating the 1 February disaster over Texas that killed the seven-member Columbia crew. Their efforts could reveal what sort of wing damage would have caused Columbia to break up during its hypersonic journey back to Earth. “Nobody, nowhere, no other vehicle does that,” says Charles Miller, head of aerothermodynamics research at Langley, referring to the shuttle's unique ability to travel at speeds that can exceed Mach 18. Langley is one of the few places on Earth with the facilities and know-how to investigate flight under such extreme conditions.

The tests Miller and his team are conducting for the Columbia investigation build on decades of work here and elsewhere on the peculiar rigors of hypersonic flight. Hypersonic aerodynamics aren't just a faster version of supersonic aerodynamics: The rules of flight begin to change above about five times the speed of sound. The shock wave, the abrupt jump in air pressure that surrounds the speeding aircraft, moves closer and closer to the body of the vessel. The enormous velocity of the air rushing past the wings, nose, and fuselage heats up the craft to thousands of degrees, temperatures that would melt ordinary metals such as aluminum. Even the air itself begins to break down, with oxygen and nitrogen molecules torn apart into atoms and their electrons ripped free.

Even the definition of “hypersonic” is a bit elusive. “Mach 5 is not a hard cutoff,” says Dean Kontinos, an aerothermodynamicist and branch chief at NASA's Ames Research Center at Moffett Field in California. “There's a gradual change as certain physical processes become important.” One key shift is in the position of the cone-shaped shock wave created by a supersonic body. “When the flow is hypersonic, the shock wave moves much closer to the body and almost, but not quite, embraces the body,” says Ramesh Agarwal, an aerospace scientist at Washington University in St. Louis, Missouri. The relatively quiescent “boundary layer” of air that coats the aircraft as it slides through the viscous atmosphere may be as thin as 25 micrometers before it gives way to the much more violent flow of faster moving and more turbulent streams of gas.

Conditions get even more complicated when the boundary layer changes from a smooth, laminar flow to a swirling, turbulent flow. That always happens at some point during the shuttle's reentry as the speed, atmospheric conditions, and angle of attack change. “If it's smooth and laminar, then the drag on the vehicle is greatly reduced, and it has a good effect on cooling,” says Dennis Jenkins, a Florida-based space historian who has written a comprehensive history of the space shuttle. “If it's turbulent, then the drag on the vehicle increases dramatically, the heating increases dramatically, and controllability issues compound rapidly.”

That loss of smooth flow may have contributed to the shuttle disaster, say experts. In fact, Columbia, whose wing design is slightly different from that of the other active shuttles, had experienced earlier-than-expected transitions to boundary-layer turbulence in two previous flights in 1989 and 1995. “Nobody understands why Columbia has a greater magnitude of variance than other vehicles,” says Jenkins.

Yet hypersonic flight is very hard to study. Good computer simulations need to be grounded in actual data from flights or from wind tunnels, and traditional wind tunnels can't generate an airflow much faster than a few Mach numbers—well short of hypersonic flight. Fortunately, Langley has a few special-purpose tunnels that can be used to study air flows at speeds relevant to the shuttle investigation, although none is large enough to accommodate the shuttles themselves.

One, a 50-centimeter tunnel that pumps high-pressure, heated air over a ceramic model, can achieve speeds up to Mach 6. Thanks to an evacuated 30-meter-diameter sphere that helps maintain the pressure differential between the source and the exhaust, a test run can last 10 minutes. Yet another wind tunnel blows Mach-6 carbon tetrafluoride gas over a metal model, a test that simulates air conditions during Mach-18 flight. Before the accident, these tunnels were used for studying hypersonic flight and for designing the next generation of spacecraft and airplane engines.

Langley researchers declined to go into detail about what they were testing for the commission. They would say only that they were simulating the effects of changes to the surfaces of the shuttle's outer mold line. Presumably, they are looking at how subtle changes, such as a dent or a gouge, could affect the heating and airflow around the shuttle's wings and fuselage. Such gouges could have been caused by the foam that struck the shuttle's wing upon launch or by the impact of a micrometeorite during orbit.

The answers may hold the key to the future of the shuttle program, including any modifications needed to return the fleet to orbit. Whatever the design of the next generation of shuttles, however, one thing is certain: There's no avoiding the dangers of hypersonic flight upon reentry. “You have a lot of energy to bleed off,” says Kontinos about what happens when the shuttle decelerates through the upper atmosphere. “There's no cheating nature.”

14. MEDICINE

# Building Better Mouse Models for Studying Cancer

1. Jean Marx

Researchers are creating new strains of mice with tumors that more closely mimic human cancers, for use in both basic cancer studies and drug screening

The lab mouse, long a favorite of researchers seeking to understand the mysteries of cancer, is being remodeled. The small, whiskered mammal has served on every front in the war on cancer: providing insights into chemical carcinogenesis, helping the pharmaceutical industry screen for potential new anticancer drugs, and yielding a better understanding of the functions of the numerous cancer-causing oncogenes and tumor-suppressor genes discovered over the past 25 years. But, crucial though these animals have been, they have serious shortcomings. Now, researchers are building more sophisticated models: new breeds of tumor-prone animals that should reflect human disease much more faithfully, both for basic studies of tumor biology and for drug screening.

Many of the newer mouse strains were on display last month at a meeting on “Mouse Models of Cancer,”* held appropriately at a kingdom built by a mouse: a Disney resort in Lake Buena Vista, Florida. These animals have been designed to mimic more closely the gene changes that foster human cancer development. For example, researchers have often used gene-transfer technology to put active oncogenes into mice. Such animals are cancer prone, but the oncogene may be expressed at high levels in more than one cell type and from early in the animal's life—a far cry from the situation in humans where the gene becomes mutated, usually later in life, only in the specific cell that gives rise to a cancer.

The flaws are just as severe in mouse models currently used for drug screening. Typically, such experiments try to block a tumor that has been created by injecting human cancer cells from a lab culture under the animals' skin. Mouse model expert Tyler Jacks of Massachusetts Institute of Technology (MIT) in Cambridge describes these so-called xenograft models as “not attractive.” Cells grown in culture, he notes, likely have acquired many additional genetic alterations not found in the original tumor cells. In addition, tumors growing under the skin miss environmental influences, such as interactions with normal tissue cells, that affect tumor growth. As a result, the growth properties—and drug responses—of the transplanted tumors may be far different from those of naturally occurring tumors.

To create more lifelike models, researchers are employing a panoply of sophisticated gene-transfer techniques that allow them to turn genes on or off in specific cell types at specific times. Often, they combine two or more cancer-promoting genetic alterations within the same cell type, taking their cues from the gene mutations identified in human cancer cells. “By putting these [gene] modifications precisely in the mouse, you see [cancerous] lesions that more resemble the human condition,” says Anton Berns of the Netherlands Cancer Institute in Amsterdam, who co-organized the Buena Vista meeting with Terry Van Dyke of the University of North Carolina (UNC), Chapel Hill, and Ronald DePinho of the Dana-Farber Cancer Institute in Boston.

Mouse developers hope that these models will provide more accurate tests for identifying therapies, although much more work will be needed to prove that. Meanwhile, researchers are using the models to gain insight into how tumors develop. Among other things, new evidence points to the possibility that at least some tumors arise in stem cells, the immature cells that produce the various cell types that form tissues. If so, that would counter previous ideas that tumor cells look immature because they have “dedifferentiated”—that is, gone backward developmentally—although, again, more work is needed.

## Tunable mice

Researchers in labs around the world have been using a variety of genetic engineering tricks to create their mouse models, but a common theme is that all allow oncogene and tumor-suppressor gene expression to be controlled much more precisely than in earlier models. As a result, researchers can follow the full course of tumor development, from initiation to growth and progression and ultimately to metastasis.

Fine control is critical, for example, in mice developed by Jacks and his colleagues. Like many other researchers, they are using a variation of the so-called bacteriophage Cre-lox system that was adapted for genetic engineering about 15 years ago by scientists at E. I. du Pont de Nemours and Co. in Wilmington, Delaware. In this system, researchers flank a piece of DNA that they want excised from a cell's genome with lox sequences, which are segments of DNA obtained from a bacteriophage, a virus that infects bacteria. When a bacteriophage enzyme called Cre encounters those sequences, it clips out the DNA between them. The Cre-lox scissors are often used to remove or inactivate genes, including tumor suppressors, but the MIT team used it to turn on the Ras oncogene specifically in the lung cells of mice.

Mutations in Ras have been linked to many human cancers, including those of the lung and colon. In previous work, when Jacks and his colleagues introduced one of the mutant Ras versions into mice, they found that the animals developed lung tumors. Indeed, the experiment was almost too successful. “The animals die so quickly with so many tumors that none progress to the late stages,” Jacks says. He set out to create mice that develop only one or a few tumors that could be followed from beginning to end.

To achieve this, Jacks and his colleagues attached an activated Ras oncogene to a “stop” signal that was flanked by lox elements and then introduced this gene into mice. As long as the stop signal was present, the gene was inactive. To eliminate it, Erica Jackson, a grad student in the Jacks lab, used an aerosol to introduce an adenovirus carrying the Cre gene into the lungs of the mice. When the virus infected the lung cells, Cre went to work excising the stop element, thus allowing the mutant Ras to be turned on only in the infected lung cells. By keeping the adenovirus content of the aerosol low, the researchers achieved their goal of getting a small number of tumors per animal.

When Jacks and his group analyzed these tumors, they found that they had developed much as human tumors do, beginning with areas of increased cell growth, which progress to noncancerous growths called adenomas and then to malignant adenocarcinomas. In keeping with another recent trend in mouse model research, Jacks and Todd Golub, also at MIT, have been using DNA microarrays to compare patterns of gene expression in the mouse tumors with those of human lung cancers in order to confirm the resemblance. “We see signatures that mouse tumors do resemble a subset of human tumors,” Jacks says. That's important, he notes: “If they don't look at all like human tumors, why study them or evaluate drugs [in them]?”

## A backward glance

By controlling when tumor development is initiated, researchers can look at the very earliest events in the process. That's not possible with most human tumors, which have progressed to more advanced stages by the time of diagnosis. For example, by examining the cells in the premalignant lung lesions of their mice, Jacks and colleagues identified cells that carry markers for two different types of lung cells: Clara cells that form the epithelial lining of the bronchi and the type II cells found in the alveoli, the tiny sacs from which oxygen passes into the blood.

This might be because the gene changes involved in cancer formation caused the cells to revert to a more immature form. That would agree with what many researchers have thought for years. However, Jacks thinks it's more likely that the cancer-initiating mutations arose in a lung stem cell. He points out that stem cells are a naturally proliferating population. If their differentiation is somehow blocked, Jacks says, he would expect them to form a tumorous mass.

Other researchers, including Luis Parada of the University of Texas Southwestern Medical Center in Dallas, are also beginning to suspect that cancers arise in stem cells. At the meeting, Parada described work by his team on a very different type of tumor: those afflicting patients with a hereditary disease called neurofibromatosis 1 (NF1). This disease is caused by a mutation that inactivates the NF1 tumor-suppressor gene, which makes a protein needed to turn off the Ras protein. As a result of abnormal Ras activity, NF1 patients get numerous neurofibroma tumors associated with their peripheral nerves. Although benign, these tumors can cause severe disfigurement, including enormous moles and swelling of the face and limbs, and ultimately some neurofibromas develop into malignant cancers.

Neurofibromas contain several different cell types; researchers didn't know which type actually gave rise to the tumor. But Parada and his colleagues, using Cre-lox technology, made a mouse in which the NF1 gene was inactivated only in the stem cells that give rise to Schwann cells, which produce the membranous sheath encasing some neurons. The mice developed tumors that closely resemble human neurofibromas. “You can recapitulate the neurofibromatosis tumors in a mouse model by mutating NF1 in Schwann cells,” Parada says. Because a mutation in a Schwann cell precursor led to neurofibromas, Parada is now using similar technology to investigate whether stem cell mutations give rise to astrocytomas, an incurable type of brain tumor. Preliminary evidence suggests that they do, he told the meeting participants.

The influence of the environment on tumor development comes through in the Parada team's work as well. Human NF1 patients inherit one inactive copy of the NF1 gene and develop neurofibromas only when a mutation inactivates the second copy, presumably in their Schwann cells. But in the mice, Parada and his colleagues find that neurofibromas develop only in the presence of mast cells, a type of immune cell, that have one active and one inactive NF1 gene copy.

How those particular mast cells influence tumor growth isn't clear, although Parada notes that they infiltrate the mouse nerves long before the tumors develop. In addition, they secrete a variety of growth-stimulating proteins. But whatever the mast cells are doing, the results suggest that it may be possible to inhibit tumor formation in NF1 patients by blocking mast cell activity.

## Ras turn-off

While the new mouse models are helping researchers trace the early stages of tumor development, they are also helping unravel the later events that underlie tumor growth and progression. A central question is whether the gene mutations that initiate the formation of cancers are also required for tumor maintenance. Recent results show that for some oncogenes, including Ras, the answer is yes, making the gene products potential targets for anticancer drugs.

To address this question, researchers attach an activated oncogene to a regulatory sequence that allows it to be turned on by some agent, such as an antibiotic, and then introduce the gene into mice. Investigators give the animals the antibiotic to turn the gene on, and once the resulting tumors are growing well, they withdraw the antibiotic to cut off the gene expression.

DePinho's group found that melanomas, virulent skin cancers, shrink dramatically once the Ras gene is turned off. Similarly, a team led by Harold Varmus, former director of the National Institutes of Health and now head of Memorial Sloan-Kettering Cancer Center in New York City, showed that continued Ras activity is needed to maintain lung tumors. The challenge now will be find compounds that can inhibit the oncogenic form of the Ras protein without inhibiting normal Ras, which is needed to regulate cell growth.

Researchers have long known that full cancer development requires mutations in two or more genes; because they can now combine mutations within the same cells, they can also use the new mouse models to explore these gene interactions. UNC's Van Dyke has been focusing on the Rb gene, a key tumor suppressor whose inactivation contributes to the formation of many human cancers.

Van Dyke notes that just knocking out the Rb gene in mice doesn't have much effect in most mouse cells because it has two relatives, called p107 and p130, that can take over its functions. As a result, two or three of the proteins have to be inactivated to get tumors. She and her colleagues do this by introducing into mice the DNA encoding a small fragment of the T (for tumor) antigen of simian virus 40 (SV40). This fragment binds to Rb and its relatives, taking the proteins out of commission. The researchers control where the T antigen fragment is expressed by attaching it to regulatory sequences that cause it to be turned on only in certain cell types.

In their first set of experiments, the UNC team expressed the fragment in cells of the choroid plexus, an epithelial tissue in the ventricles of the brain. This caused the choroid plexus cells to divide, but the tumors grew slowly, Van Dyke says, because the cells were also dying from apoptosis, a form of cell suicide. However, when the researchers did the same gene transfer in mice lacking p53, another tumor suppressor that is known to help trigger apoptosis, the tumors grew very quickly, killing the animals in just 3 to 4 weeks.

In addition, Van Dyke and her colleagues saw another key indicator of tumor progression—formation of new blood vessels—once p53 was out of the picture. “Clearly, apoptosis is modulating tumor growth,” Van Dyke says, “but p53 inactivation has more impact than just loss of apoptosis.”

Since then, Van Dyke and her colleagues have used similar methods to inactivate Rb and its relatives in other tumor-prone cells, including breast epithelia and astrocytes, neuronal companion cells that can produce astrocytomas. “In the mammary epithelium, everything we learned recapitulates what we learned in the brain epithelium,” says Van Dyke. But although the behavior of the astrocytomas looked similar to that of the choroid plexus cells, the event that fostered malignancy was not loss of p53, but of a different tumor suppressor called PTEN.

Malignant tumors don't just grow uncontrollably, however. They also spread to new sites in the body. Indeed, this metastatic spread is the major reason why people die of cancer (Science, 14 February, p. 1002). Gerhard Christofori of the University of Basel, Switzerland, leads one of the efforts to use mouse models to study metastasis.

Christofori and his colleagues work with the so-called RIP1-Tag2 transgenic mouse originally developed by Douglas Hanahan and his colleagues at the University of California, San Francisco (UCSF). The UCSF team created this model by introducing into mice the SV40 T antigen gene attached to the rat insulin promoter (RIP). RIP causes the T antigen to be expressed in the insulin-producing β cells of the pancreas, which produce numerous tumors. The tumors don't metastasize on their own, but the Christofori group found that they can generate metastatic pancreatic tumors by inactivating certain other genes, including those for cell-adhesion molecules. Absence of such molecules may foster tumor cell movement in the body by disrupting normal cell adhesion.

## Difficult to package

Although there's no doubt that mouse models can shed a lot of light on tumor biology, a big question remains: Will they help researchers improve cancer therapy? That is, as Scott Lowe of Cold Spring Harbor Laboratory in New York asks, “is this just an interesting exercise, or can you take it to a higher level and directly affect patient care?”

Lowe, for one, hopes to answer that question in the affirmative. In particular, cancer therapists know all too well that some tumors respond well to drug treatment, whereas others, even though they look the same to the pathologist, fail to respond. Using a variety of mouse model strategies, Lowe's team has pinpointed a number of gene changes that are involved in blunting the response.

For example, in a mouse model of the blood cancer lymphoma, the researchers found that increased expression of the Bcl-2 oncogene, which produces an apoptosis-inhibiting protein, greatly decreases responses to chemotherapy. When the same lymphoma cells were maintained in lab culture, rather than in a mouse, however, the resistance-promoting effects of Bcl-2 were much reduced. That points up problems with the current drug assays, Lowe notes, which use either cultured cells directly or cultured cells transplanted into mice.

In other work, the Lowe team found that loss of the p53 tumor-suppressor gene also increases drug resistance. “Clearly, we get very different responses, depending on the genotype of the tumor,” Lowe concludes.

Although gene variations may explain the differing drug responses of tumors, researchers still need to learn how to overcome the problem. It's also too early at this point to tell whether the new mouse models will help in drug screening. “We need to show that drugs that work in mice work in humans, and that those that don't, also don't in humans,” says Jacks.

But even if that can be done, widespread adoption of the newer models, at least for drug screening by the pharmaceutical industry, is still in doubt. DuPont holds a patent on Cre-lox technology that some researchers say hinders use of animals created with the technology (Science, 4 July 1997, p. 24). So far, though, academic researchers seem to be coping. “With the tricks you can do with the mouse, the field is just exploding,” Lowe says.

• *The meeting, held from 19 to 23 February, was sponsored by the American Association for Cancer Research.

15. # High-Tech Fingers on Earth's Erratic Pulse

1. Richard A. Kerr

Volcanologists are throwing new sensors and instruments at a problem of immense societal importance: predicting when a volcano near an inhabited area may erupt. The results are encouraging but far from consistent

After 2 decades of slumber, Alaska's Redoubt volcano was growing restless. Earthquakes rattled the mountain as magma welled up through the crust. But would the magma rend the surface or fizzle out? Thomas Miller, chief scientist of the then-new Alaska Volcano Observatory (AVO), was “very concerned” that the mountain would blow, but he was not ready to translate the tremblings into a prediction of an imminent eruption. However, an hour after Redoubt began spouting ash and gas on 14 December 1989, Miller got a phone call from Bernard Chouet, a colleague at the U.S. Geological Survey (USGS) in Menlo Park, California, who wanted to chat about the volcano. Oh, it's erupting? “I expected it was going to erupt,” said Chouet. Miller was perplexed. “If this guy knew it was going to erupt,” he thought, “why didn't I know it?”

Chouet's edge was a peculiar squiggle that he as well as AVO scientists had seen on seismic recordings from the volcano, but only Chouet had taken them to be sure harbingers of eruption. More than a decade after Redoubt's reawakening, this sort of “hum,” called a long-period (LP) seismic event, has become one of several once-inscrutable inklings of doom that scientists now routinely eavesdrop on. The hope is to anticipate a volcano's behavior well enough to move people out of harm's way. A number of successful evacuations across the globe in recent years suggest that scientists are becoming more and more adept at deciphering the cryptic signals of rising magma and, most important, deciding when to brace for a volcano to explode.

“Really, we do pretty well” at forecasting eruptions, asserts volcanologist John Eichelberger of AVO and the University of Alaska, Fairbanks. And he and other volcanologists at listening posts around the world are getting better all the time, he says. “The picture is analogous to weather forecasting: Some mistakes, but we provide vital information about a gathering storm.”

The success stories are due in large part to leaps in the sophistication of the technology used to probe the hearts of volcanoes. Using instruments ranging from satellite-borne radars to strainmeters buried in boreholes, researchers can now track magma from the time it first rises and pools in the upper crust, pushes on into the volcano, and then, sometimes with deadly consequences, breaks through the surface. A major concern, however, is that “most volcanoes are not well monitored,” says Dan Dzurisin of USGS's Cascades Volcano Observatory (CVO) in Vancouver, Washington, including ones in populous parts of Central and South America, Africa, and the southwest Pacific.

But even well-monitored volcanoes can get cranky and confound their watchers. Soufrière Hills on Montserrat fussed and fumed for 2 years after its July 1995 reawakening, triggering one false alarm before really letting loose in 1997. Thus some volcanologists view the prediction glass as half-empty. “We're quite a ways from [consistently] forecasting an eruption before the event,” says C. Dan Miller of CVO. “We can do that part of the time.”

Miller recalls the two dozen overseas missions in the past 18 years of USGS's Volcano Disaster Assistance Program (VDAP) team he now heads and the sometimes maddeningly erratic behavior of rising magma. For example, he says, in September 1999 “all indications were that [Ecuador's] Tungurahua was about to suffer a major explosive event.” On the advice of scientists, officials evacuated 20,000 people from the volcano's flanks. Nothing much happened. Three years later, Tungurahua sputters on without a major eruption. “Despite what some scientists think and despite excellent monitoring,” says Miller, “we have been unable to predict when many of these explosive events are going to happen in the course of years-long eruptions. There are huge challenges for scientists.”

## From the bottom up

The first challenge for volcanologists is deciding which of the planet's 600 or so active volcanoes to monitor closely enough to allow any hope of predicting an eruption. Geologists teasing out the history of a dormant volcano from layer upon layer of its ash, lava, and mud might warn of the possibility of an eruption in coming decades, as they did before Mount St. Helens blew in 1980. They continue to sound warnings on Italy's Mount Vesuvius, which according to the intervals between past eruptions is due for a major eruption (see p. 2020). But for a more immediate wake-up call, volcanologists are coming to rely on the first signs that fresh magma, which will fuel an eruption, has moved into chambers in the upper crust—roughly 5 to 10 kilometers below the surface.

The best way to track such magma movements is to detect minute deformations of the surface. Since the late 1990s, satellite-borne radar monitoring of ground swelling has been providing early alerts years before an eruption. In the Cascades of central Oregon, radar measurements revealed a 15-kilometer-wide, 10-centimeter-high bulge that had formed between 1996 and 2000 on the flanks of the Three Sisters volcanoes, which have slept for 2000 years (Science, 11 January 2002, p. 260). The size and shape of the bulge suggest that about 21 million cubic meters of magma have risen into a chamber located 6 or 7 kilometers underground. The upwelling of hot, gas-laden magma can tip a volcano toward an eruption. Such magma is more buoyant than magma lingering from a previous eruption, so it pushes upward, pressurizing the plumbing leading to the surface.

Geophysicist Charles Wicks and his colleagues at USGS in Menlo Park measured ground swelling around the Three Sisters by combining two sets of radar signals bounced off the surface, one in 1996 and one in 2000. The resulting interference fringes formed a bull's-eye that gauged the amount of uplift. Thus alerted, USGS researchers deployed another recent tool in their arsenal—a Global Positioning System (GPS) instrument that measures precise latitude, longitude, and elevation. GPS has confirmed a continuing rise. The longer the land swells, researchers figure, the more likely that the magma will break out of its chamber and start heading for the surface.

USGS also installed a seismometer on the bulge. Earthquakes—the squeals of rock breaking as magma pushes upward—have long been the primary warning sign of an impending eruption. But volcanologists have come to recognize that bursts of a peculiar, resonating hum from deep in a volcano's plumbing system can also alert watchers to the resupply of a deep magma chamber. These LP events are occasionally detected at midcrustal depths of 5 to 10 kilometers, but they are not the sounds of rock breaking, which come at periods as short as a few hundredths of a second. Rather, LP events have periods of 0.2 to 2 seconds.

Exactly how deep LP events are generated remains unclear, but they are widely believed to reflect somehow the gurgling of magma in deep volcanic plumbing. One to 2 weeks before Mount Pinatubo in the Philippines blew in 1991, for example, 400 LP events more than 10 kilometers deep struck each day. It now appears that the cause was fresh magma entering the system, says seismologist Randall White of USGS Menlo Park and the VDAP team. The LP events were saying, “It's coming up, and it's enough to kick things around.” Pinatubo magma that erupted onto the surface contained blobs of the freshly injected magma, suggesting that the resupply was enough to push the volcano over the edge into an eruption. Likewise, concern rose in 1989 when swarms of LP events rippled beneath Mammoth Mountain in east-central California and in late 2000 beneath Mount Fuji. But so far, these appear to have been merely shots across the bow: Neither long-slumbering volcano has awoken as yet.

## On toward the surface

Signs of deep stirrings can flag hot spots, but volcanologists tasked with warning the public of impending danger are more anxious to track magma as it nears the surface. Fumes venting from volcanoes once looked like a promising marker for a coming eruption. Sulfur dioxide and other gases escape ahead of ascending magma. Measured from a safe distance by infrared spectroscopy, sulfur dioxide fumes rising from a volcanic cone have often signaled that magma—not just heat seeping from below—has neared the surface and is driving the venting and ash spouting that can precede a major eruption.

In the past decade, geochemists have added carbon dioxide to their observations because it is more likely than sulfur dioxide to make it through groundwater to the surface. At Mammoth Mountain, 300 metric tons of carbon dioxide—measured at the surface and by flying an analyzer through the plume—ooze up each day from the fresh magma injection 5 to 9 kilometers deep. That amount is down from a more intimidating 1100 tons per day a decade ago, which was enough to kill trees and close parts of the mountain to camping. The slow decline suggests that Mammoth Mountain will not erupt anytime soon.

But some experts argue that sniffing the vapors can easily mislead oracles. “Both an increase in gas emission rates and a decrease in gas emission rates are bad” signs, says William Rose of Michigan Technological University, Houghton. That was made tragically plain in 1993 at Galeras volcano in Colombia (Science, 16 April 1993, p. 289). On 14 January, a party of scientists entered the crater under the assumption that it was relatively safe because its low gas emission meant that magma was not near the surface. But it actually meant that the conduit had plugged itself, allowing gas pressures to build to the breaking point. Seven volcanologists died when the plug shattered in a brief but violent explosion. Gases, says Rose, are “enigmatic without something else” to go on.

In the aftermath of Galeras, the something else looked like it might be LP events—not the kind that occur many kilometers deep, but ones that emanate from just beneath the surface. Shallow LP waves were the precursors that led Chouet to call AVO's Miller as Alaska's Redoubt erupted in 1989. Chouet had deduced that these events a kilometer or so deep reflect an interaction of magma and groundwater that sets cracks filled with magma, water, or gas vibrating like the pipe of an organ. By Chouet's thinking, the more LPs, the more pressurized the shallow plumbing of the volcano is and the sooner it is likely to blow.

The 1989 Redoubt eruption strongly implied that shallow LP waves augur trouble. Redoubt's initial 14 December eruption was followed by a couple of weeks of quiet, but by 1 January 1990 LP events were on the rise again along with a general increase in the number and magnitude of quakes. In consultation with Chouet that day, Miller recalls, AVO released a prediction of “a moderate eruption in the next 12 to 24 hours.” When seismicity began shooting up logarithmically, the warning was changed to a large eruption. That was enough for workers to button up an oil terminal downriver of the volcano and evacuate, just hours before Redoubt's biggest eruption of the series sent mudflows barreling into the terminal. “I became a believer” in LPs, says Miller. “It's a very powerful tool. If we see LP events, they get our attention right away.”

But it took the Galeras tragedy to consolidate opinion that shallow LP events can indeed serve as warning flags. The night before the ill-starred researchers headed to the crater, according to Victoria Bruce in her 2001 book No Apparent Danger, five of the leaders discussed the volcano's state. Fifteen LP events with their constant frequency and descending amplitudes—called tornillos in Spanish for their screwlike appearance—had occurred in recent weeks, as they had in the weeks before the previous Galeras blast in July 1992. Although two of the leaders credited Chouet with predicting that eruption using tornillos, according to Bruce, the field trip went on.

A decade after that wake-up call, researchers regard LP events with a range of feelings. Before Galeras, “what we didn't have in our toolbox was something we could trust to give us reliable magnitude and timing” of an eruption, says retired USGS volcanologist Norman Banks. “Bernard Chouet gave us that ‘missing link.’” Banks was among a number of volcanologists who appeared in a NOVA show aired in the United States last November that portrayed LPs as “the Holy Grail of volcano science.”

Many volcanologists assail that message—one disparages the NOVA show as “absolutely unbelievable”—but concede that shallow LP events are vital to monitor. “Recognition of LP earthquakes is one of the key elements but not the full answer,” says David Hill of USGS Menlo Park. Their absence, for instance, cannot be taken as a clean bill of health. Plenty of seismicity preceded the 1994 Rabaul eruption in Papua New Guinea, Hill notes, but not a single LP event could be discerned. White of the USGS used an LP-type signal himself to predict an eruption of Ecuador's Guagua Pichincha in 1999 within 5 days. But “the biggest problem is that shallow LP seismicity doesn't really begin until the system has opened up a bit,” he says. “It's a late-stage phenomenon. You know [the magma] is coming up the pipe already.” But will it follow through?

## A no-show?

What volcanologists really have to know is whether the magma they can hear or smell, but not yet see, is going to make it out the top. That only happens about half the time. The other half it congeals and stalls. At calderas such as California's Long Valley or Italy's Campi Flegrei—scars of huge ancient eruptions—the success rate is probably more like 1 in 10, says Hill. To force an eruption, magma must push its way through solid rock on the power of its declining buoyancy. If the conduit's walls are sufficiently fractured, for example, rising magma can shed enough gas to lose buoyancy and become far more viscous. Weakened and stiffened, it plugs its own pipe.

Experts typically gauge a volcano's late-stage eruption potential by how big a head of steam it seems to have. After the cataclysmic eruption of Mount St. Helens in 1980, USGS volcanologists were 13 for 13 in predicting subsequent smaller extrusions of lava. When they saw a rapid rise in the intensity of seismic activity and accelerating swelling of the crater floor, they predicted an eruption minutes to weeks ahead. At Pinatubo, unrest escalated so quickly and consistently that it was eventually clear to all that the volcano was going to blow. As White puts it, at many volcanoes “it tends to get noisier and noisier before eruptions.”

Barry Voight of Pennsylvania State University, University Park, has quantified this “noisier and noisier” behavior of soon-to-erupt volcanoes. Drawn from the science of strained materials, his equation essentially says that the volcano's last obstacles to rising magma will fail and the volcano will erupt when exponentially accelerating seismicity or deformation approaches infinite proportions. Applied in hindsight, this method would have predicted a number of eruptions at Mount St. Helens, Redoubt, Izu Shima in Japan, Montserrat, and Galeras, Voight says.

Others have also found exponential acceleration to be a useful predictor. In the September 2002 issue of the Journal of Volcanology and Geothermal Research, John Murray of The Open University in Milton Keynes, U.K., and Juan-José Ramirez Ruiz of the Colima Volcano Observatory in Mexico report that swelling of Colima's upper cone began accelerating almost a year before its 20 November 1998 eruption. Applying Voight's method, successful predictions of the eruption's timing could have been made as early as 25 weeks ahead, they say.

On the other hand, the Rabaul caldera looked to have a good head of steam up when it reawakened in late 1983, says Robert Tilling of USGS Menlo Park. “You would have made a very precise forecast,” he says, “but it would have been wrong.” The populace was put on high alert in October 1984, but things promptly quieted down. It did not erupt until 1994. And in 1997 an intrusion beneath Long Valley drove an alarming crescendo of seismicity without reaching the surface.

Despite some encouraging developments with LP events and runaway activity, many volcanologists remain cautious on prediction. “There's no magic bullet here that will allow us to monitor volcanoes,” says Tilling. “Even if monitoring data is good, it can't [always] distinguish between a large explosive event and a magma intrusion without an eruption.” The rise of magma into a volcano is a complicated phenomenon, researchers say, so prediction will likely remain complicated even as observations and analyses continue to improve. Thus, although volcanologists can take pride in successful predictions that have saved tens of thousands of lives in recent decades, they cannot become complacent: The science and the technology, while impressive, have been tested on only a few volcanoes during short parts of their lives.

“There are so many unknowns,” says Dan Miller. “We can't see the magma chambers or the plumbing system leading up into the volcano.” And characteristics crucial to the successful ascent of magma such as gas content, viscosity, and mechanical resistance to ascent vary from volcano to volcano and even from one intrusion to the next. It makes weather forecasting look like a snap.

16. # Bold Venture Aims to Plumb a Volcano's Fiery Depths

1. Dennis Normile

TOKYO—Most volcanologists monitor a mountain's inner workings with instruments placed on the surface like a doctor's stethoscope. But this summer an international team will try something never done before: drilling into the heart—the magma conduit—of a recently active volcano.

The sleeping giant the Japanese-led team intends to penetrate is Unzen, a volcano in southwestern Japan that erupted in 1995. Last year, the group completed the first phase of the project, punching a pair of exploratory holes in older deposits in Unzen's flanks. The more daring stage, drilling into the conduit, got under way last month. The hope is that the high-risk venture will shed light on the factors that determine whether an eruption is explosive —“one of the fundamental questions of volcanology,” says volcanologist Setsuya Nakada of the University of Tokyo, a leader of the project.

When a volcano starts rumbling, volcanologists try to predict the timing and size of a potential eruption by deciphering the “smoke signals”—seismic waves and ground deformation—it sends out. The Unzen Scientific Drilling Project will be the first time that volcanologists attempt to correlate these data with chemical and physical information from a volcano's interior obtained soon after an eruption. A better understanding of the link between processes in the conduit and the signals monitored at the surface “has the potential to be quite an advance in forecasting eruptions,” says another project leader, John Eichelberger of the University of Alaska, Fairbanks.

Unzen casts a long shadow over the city of Shimabara on Kyushu Island; a 1792 eruption that killed 15,000 people ranks as the fifth deadliest volcanic disaster in history. Its most recent bout of activity, lasting more than 4 years, killed 44, displaced 10,000, and inflicted $2 billion worth of damage as pyroclastic flows combined with rain-driven mud flows to wipe out homes, farms, factories, an elementary school, roads, and a train line. Among the dead were volcanologists Maurice and Katia Krafft of France and Harry Glicken of the United States. When Unzen subsided, scientists quickly laid plans for the drilling. The idea had been floated for other volcanoes, but logistical, environmental, or safety obstacles had always proved insurmountable. At Unzen, however, Nakada and Eichelberger were able to identify a suitable and accessible drilling site outside the boundary of a national park that rings the summit. They put together a team of 60 scientists from five countries for the project, with Japan picking up roughly 90% of the estimated$18 million cost.

In the first phase, cores taken from two boreholes filled in essential details about Unzen's structure and natural history. Researchers knew that over 4 years Unzen—a dacitic volcano in that its magma is largely silica—had effused 200 million cubic meters of magma in roughly 10,000 pyroclastic flows. Such behavior contrasts with that of Pinatubo, a dacitic cousin in the Philippines that in one gigantic explosion in 1991 sent 6 or 7 cubic kilometers of ash into the atmosphere. The team learned from the cores that Unzen's earliest eruptions—about 500,000 years ago—were explosive like Pinatubo's, but that the eruptions grew tamer and tamer, becoming effusive about 200,000 years ago.

The researchers hope that the project's second phase, a 3-year effort launched last month with a Shinto ceremony at the site to pray for safe drilling, will shed more light on the subject. But they are not sure what will happen when they tap into Unzen's conduit, the channel through which magma rises to the surface. Although magma in the conduit has solidified, it is expected to be a bit-melting 600°C and contain highly pressurized water vapor. To cope with such conditions, the drillers will cool the area around the bit to 170°C by circulating a gypsum slurry through the hole. That should allow them to insert a camera and sensors for measuring temperature, density, and permeability. The slurry should also help quell the pressurized water, and a blowout preventer, a valve at the surface, will serve as a backup.

The drillers expect to reach the conduit at a depth of about 1300 meters beneath the summit. Seismic imaging has shown that Unzen's conduit is several hundred meters wide but only 10 to 20 meters thick and slants upward from west to east. The borehole will arc into the conduit, penetrating it horizontally. The first hole is expected to yield rock samples describing the stability of the conduit and its surrounding rock. If all goes well, a second summer of drilling in 2005 will slant out of the same borehole into the conduit at a shallower elevation, taking continuous core samples. That hole would give valuable data on how gas bubble size and the physical characteristics of magma change on the way up the conduit.

Insights into exactly how volcanoes shed gas, says Chris Newhall of the University of Washington, Seattle, “surely will save both lives and unnecessary evacuations at volcanoes.” For the eruption-weary people of Shimabara and the millions of others around the world living in the shadow of volcanoes, better forecasting cannot come too soon.

17. # Living in the Shadow of Vesuvius

1. John Bohannon*
1. John Bohannon is a writer in Lyon, France.

Researchers at the oldest volcanological institute probe the inferno beneath their feet and wrestle with the politics of civil protection

NAPLES, ITALY—From the snow-capped heights of Mount Vesuvius, the city looks like a densely woven blanket wrapped snugly around the volcano's slopes. From up here, with sunlight glinting off the broad crescent of the bay and far from the din of traffic, Naples and its surrounding towns look peaceful. This tranquillity makes it all the more difficult to imagine the worst-case scenario that volcanologists believe Vesuvius could inflict on Neapolitans: a repeat performance of the A.D. 79 eruption that famously destroyed the Roman towns of Pompeii and Herculaneum.

In this scenario, Vesuvius does not ooze lava and throw up a few rocks as Hawaiian volcanoes do. The pressure from below just builds and builds until it goes off like a bomb, sending a superheated column of rock, ash, and gases 20 kilometers into the sky. Within hours, this roiling mixture of gas and dust collapses back and roars down the slopes at up to 240 kilometers per hour toward the 600,000 people living around the base. These pyroclastic flows can blast through stone walls up to 3 meters thick and can flash-fry living matter. Beyond this annihilation zone, ash and volcanic rock fragments called tephra shower down over a wide area, quickly accumulating enough material—up to 500 kilograms per square meter—to collapse roofs, burying people alive.

The good news for modern-day Naples is that, so far, there is no sign of such a head of steam building under the volcano—but just to be sure, it is being watched around the clock by the Vesuvius Observatory. Just shy of 158 years old, the observatory was the first volcanological station ever built. Its recent directors have turned it into an international center of research and a testing ground for urban volcanic risk management. Techniques and models developed by observatory scientists, often with international collaboration, are now used across the globe. But not surprisingly, says Grant Heiken, a volcanologist at Los Alamos National Laboratory in New Mexico, having responsibility for the most heavily populated volcano in the world adds real-world pressures to the science.

Over the past few years, while helping the government draw up an evacuation plan for the area, observatory scientists became involved in a bitter dispute, which split the Italian volcanology community, over how much warning Vesuvius would give of a major eruption. The academic mudslinging has subsided, but observatory scientists are continuing to cajole colleagues in Italy and elsewhere to help them protect the millions of people living in Vesuvius's shadow.

Down in the city, volcanologist Giovanni Orsi chain-smokes behind mountains of paper on his desk at the observatory. “Vesuvius may be the most famous volcano, but it isn't the greatest threat to Naples,” says Orsi. “People are more aware of it because it looks like a classic volcano. But the Campi Flegrei are far more dangerous,” he adds, referring to the western side of the bay that is pockmarked with craters. This was the exit point of the largest eruption in Mediterranean history, 39,000 years ago, which launched at least 200 km3 of magma—the equivalent of 200,000 solid-stone Empire State Buildings. That blast caused the area to collapse into the broad dish-shaped caldera that now cups the western part of Naples and the town of Pozzuoli.

The 1 million people at risk from a Campi Flegrei eruption would be far more difficult to evacuate. Add to that the volcanic island of Ischia just off the western arm of the bay, home to 50,000 and host to thousands of tourists in summer, and it becomes clear that Naples has the potential to produce a disaster on an epic scale. But for the time being, that possibility seems remote, and from a researcher's point of view, says Orsi, the area has long been “a paradise for volcanology.” According to Paolo Gasparini, who was director of the observatory from 1971 to 1983, the first seismometer was invented here. Observatory scientists were also the first to record the swelling and sinking of the Campi Flegrei, now one of the main parameters in volcano surveillance. In 1999, the observatory's scientists moved from the slopes of Vesuvius into a new building downtown, and the following year the old observatory building was transformed into a volcanology museum.

The observatory's new generation of scientists would like it to be not just the oldest but the best institution of its kind. “My main objective is to increase the amount of research,” says Giovanni Macedonio, the director since 2001, “in terms of the number of scientists and their productivity.” Macedonio, a volcano modeler at the observatory since 1998, is promoting it as a nexus for both national and international research collaborations.

Some long-standing partnerships are already bearing fruit. Gasparini and his colleagues are putting the finishing touches on a decade-long Vesuvius project called TOMOVES that has produced the clearest picture yet of the “deep plumbing” of a volcano (Science, 26 November 1999, p. 1685). By using a technique called seismic tomography—setting off small explosions and analyzing the waves as they bounce back off different layers of rock—they are able to locate the limestone “floor” beneath Vesuvius, as well as the location of magma.

The same method is also being used to map the underlying structure of the Campi Flegrei in a project called SERAPIS. Researchers have now gathered the data, says project leader Aldo Zollo, a geophysicist at the University of Naples, and are beginning to analyze them. But some results are already jumping out, he says, such as the buried rim of a caldera likely formed by an eruption 12,000 years ago that encircles the entirety of the Campi Flegrei and loops right beneath the bottom of the bay.

The experiments have challenged some basic assumptions about magma systems. The classic view of a “ball of magma” beneath volcanoes is giving way to a variety of structures, including the broad sheets of magma that seem to exist beneath Vesuvius. “This model makes a lot of sense,” says Jon Davidson, a volcanologist at the University of Durham, U.K. Davidson is coordinating ERUPT, an international project starting up this year that includes observatory scientists, focusing on the geochemical evolution of magma beneath Vesuvius and three other European volcanoes.

Apart from the straight geophysics, the Neapolitan area supports a better “marriage” of archaeology and volcanology than perhaps anywhere else, says observatory volcanologist Sandro de Vita. Orsi agrees: Last year, he authored a provocative paper arguing that the global impact of the Campi Flegrei eruption 39,000 years ago sparked the transition from Neandertal dominance to modern Homo sapiens. De Vita and his colleague Mauro Di Vito are also studying the more recent history of the area, concentrating on the coevolution of the volcanoes with the cultures of their human inhabitants. Evidence of human activities going back 6000 years remains frozen in time between layers of mud and ash from the frequent eruptions. In spite of these periodic catastrophes, the rich mud flats produced by the eruptions keep luring large groups of farming people back to the very same place.

## … And a volcanologist's purgatory

Just west of Naples, ensconced within the dense town of Pozzuoli, lies an enormous yellow-white crater called the Solfatara. Enrica Marotta wrinkles her nose. “Yeah, it always stinks,” she says. The rotten-egg smell of sulfurous gases billowing from cracks is generated by an aquifer interacting with magmatic gases several kilometers beneath the surface. Marotta, an observatory volcanologist, points out two glass bulbs connected to tubes emerging from the ground. These containers are used to measure the amount of carbon dioxide escaping from below. If the composition and rate of gas emissions change, it could presage another eruption.

Data like these, as well as from a vast array of seismometers, tiltmeters, and Global Positioning System devices arranged on and around the three Neapolitan volcanoes—and even on the bottom of the bay—all feed continuously into the observatory's surveillance system. The institution has always had two goals: monitoring the volcanoes for signs of danger while pursuing basic volcanological research. But recently, these parallel threads have been a source of discord.

The trouble began in 1994 when the national government appointed a committee of politicians together with scientists from the observatory and elsewhere to draw up an emergency plan in case Vesuvius erupts. Lucia Civetta, who was director of the observatory from 1993 to 2001 and a member of the committee, says the intention was “to continually update the plan in response to new data and the movement of residents.” But parts of the plan drew storms of criticism, directed at the observatory, that lasted right to the end of Civetta's directorship.

A major source of disagreement was how much warning the volcanoes can be expected to give before erupting, a key parameter for any emergency plan. Civetta and her colleagues maintain that the volcanoes are likely to give clear signals at least weeks in advance of eruption, including ground uplift, gas emissions, and characteristic seismic activity. But Giuseppe Luongo of the University of Naples, Civetta's predecessor as director of the observatory, and Flavio Dobran, an independent volcano modeler who worked with Macedonio in the early 1990s at the University of Pisa, argue that this is far from a safe bet. “[We] disagree with the Vesuvius emergency plan because it is impossible to predict an eruption 2 or 3 weeks before,” asserts Luongo.

With political infighting worthy of Imperial Rome, the Italian volcanological community became polarized between the observatory and its critics. International volcanology meetings, usually friendly, laid-back events, were the scene of shouting matches and the circulation of polemical manifestos denouncing observatory scientists. Tension was increased by two nondestructive but nonetheless worrying earthquakes in the region in 1995 and 1999. Researchers outside Italy watched the struggle with chagrin. “[These conflicts] interfere with research and collaboration,” says Davidson. Clive Oppenheimer of the University of Cambridge, U.K., says such disagreements are “inevitable, given the uncertainties in the Vesuvius problem.”

For the past 2 years, at least in part because of Macedonio's diplomacy, the debate has cooled. In the meantime, Civetta—now at the University of Naples—Gasparini, and others have carried on, revising the Vesuvius emergency plan and incorporating measures for a Campi Flegrei eruption. The city and towns have been zoned in terms of risk from pyroclastic flows, ash, and tephra fallout, as well as mudslides and flooding. The observatory's efforts to educate the public seem to be paying off. After decades of growth, the population in Vesuvius's pyroclastic zone has begun dropping as people voluntarily move off the slopes. The objective, according to Gasparini, is to be able to evacuate these highest-risk areas in 2 to 3 days. Right now, he says, it would require at least 10 days' notice.

In spite of the acrimony, Civetta is sanguine about the Neapolitan volcanology community. “The conflict is part of the unique spirit of Naples,” she says. For his part, Macedonio would like to see the pax Neapolitana last. He has spent a large part of his time consolidating research to cut down on overlap and melding the observatory into a national network for natural-disaster monitoring. “Now we have better data exchange and less conflicts,” says Macedonio, who is tapping the U.S. Geological Survey for exchanges of data and techniques and is organizing a major symposium on Vesuvius at the old observatory in May.

The last time the skies over Naples were lit by eruptions from Vesuvius, the Campi Flegrei, and Ischia were in 1944, 1538, and 1301, respectively. Whether Naples will enjoy another century of peace, or suffer another explosive eruption, only time—and the continuing work at the observatory—will tell.

18. # Dr. Doom's Gruesome House Calls

Peter Baxter's insights into the victims of volcanoes may help protect people from the bodily threat of eruptions as well as the potential long-term health effects of gases and ash

Peter Baxter knows intimately the kind of injuries an angry volcano can inflict. On 14 January 1993, the occupational physician at the University of Cambridge, U.K., was attending a workshop in a town at the foot of Galeras, a volcano in southern Colombia. While he was visiting villages on the flanks, assessing the potential danger from avalanches of hot ash and gas, the volcano erupted. Fearful of getting caught in a pyroclastic flow—and worried about a field party that had been headed for the crater—Baxter rushed back into town. “We knew something terrible had happened,” he recalls.

At the local hospital, Baxter helped treat the victims. “It was awful,” he says. Galeras had claimed a friend of Baxter's, volcanologist Geoff Brown, and several colleagues. Yet, he says, “all the time I was trying to understand what was going on.” That is Baxter's specialty: learning how volcanoes kill people. Most of the bodies were blown to pieces, but he was able to observe one autopsy and later interviewed the survivors. His report on the casualties, published in the Journal of Volcanology and Geothermal Research in 1997, offered those who work on volcanoes some seemingly obvious, but necessary, lessons. For starters: Wear hard hats and heat-resistant clothing.

“When scientists work on an active volcano, it's a very thrilling, adrenaline-pumping act. It's very easy to forget the risks you're running and to take more risks,” says Steve Sparks, a volcanologist at Bristol University, U.K. Thanks in part to Baxter's work, he says, “people have a much better appreciation of just how terrible the deaths can be.” In the wake of the Galeras tragedy, Sparks says, there is “a much greater awareness of risks and sensible precautions.”

Baxter's pioneering work as one of the first physicians to study volcanic hazards, along with his penchant for illustrating talks with slides of corpses, has earned him the nickname “Dr. Doom” among volcanologists. His gruesome repertoire ranges from puzzling out the flesh- searing effects of pyroclastic flows to probing whether chronic ills may be linked to long-term ash exposure. The goal, he says, is to find ways to reduce loss of life during and after eruptions.

Baxter got hooked on volcanoes 2 decades ago, while he was a medical epidemiologist at the U.S. Centers for Disease Control and Prevention in Atlanta. As part of a team that investigated the health implications of disasters, he went to Mount St. Helens soon after it erupted on 18 May 1980. The health consequences of ash exposure, it turned out, were limited to exacerbating asthma and lung disease. And because the ash rapidly entered the soil, any long-term effects were highly unlikely. But it was the 57 blast victims who most intrigued Baxter, who realized that little was known about how pyroclastic surges snuff out life.

Complementing his work on recent fatalities, Baxter has also reached back in time to study the earliest known volcano victims. The eruption of Mount Vesuvius in A.D. 79 destroyed both Pompeii and Herculaneum, but it was the latter, lesser-known Roman town that has proved more suitable for studying the deadly effects of pyroclastic flows. Piecing together a story from the charred remains of 80 skeletons from Herculaneum, Baxter and Italian colleagues discovered that it was the intense heat of the flow—roughly 500°C—rather than suffocation that instantaneously killed ancient Romans sheltering in beach caves. In contrast, accounts of a 1902 eruption on the island of St. Vincent in the West Indies describe ash flows that left some people inside houses gasping for air—but alive. The key may be that less-dense pyroclastic flows can be survivable: “You don't have to have total destruction,” Baxter says. That means civil defense officials must plan to rescue people who may have resisted evacuation or returned home before an eruption.

Baxter is now helping create a computer model to determine the risks of a pyroclastic flow from a present-day eruption of Vesuvius in the hope of better guiding evacuations in the Naples region (see p. 2020). The risk-assessment model is based on one developed on Montserrat, where officials had to decide in 1997 whether to evacuate the 17-kilometer-long island in the West Indies. The Montserrat model estimated the probability of casualties under different scenarios. In the end, the island's governor decided that the north end of the island would be safe—which has turned out to be true, so far (see p. 2027). “It's a success story we're trying to build on,” Baxter says.

In the developing world, volcanoes can create additional problems. In the city of Goma, in the Democratic Republic of the Congo, where Baxter traveled last autumn for the World Health Organization, a large population had been threatened by a spectacular eruption featuring a river of lava that flowed through the downtown area (see p. 2024). Baxter found that Mount Nyiragongo's hazards included disturbance to a shaky infrastructure, with the prospect of disease outbreaks among masses of refugees. “It's a very interesting example of how health hazards like cholera can be more important than the volcano itself,” he says.

Baxter says his next challenge is to build a model that sums up all that's known about the risk of ash on Montserrat, where volcanic activity may continue for years to come. There, the high content of a silica mineral called cristobalite may indeed pose problems: It's extremely fine and easy to breathe deep into the lungs (Science, 19 February 1999, p. 1142), and long-term exposure may cause silicosis, a scarring of the lungs. “People get high exposures just by driving around. So the potential for problems, especially for children, … is quite extensive.” But he emphasizes that the risk is still unknown, as it is unclear how much cristobalite it would take to trigger health effects.

As participants in a new volcano health network (see sidebar), Baxter and other medical sleuths are hoping to solve some of these unknowns. Earlier this month, Baxter headed to Quito, Ecuador, to study the effects of ash and gases from November's eruption of El Reventador—casework, no doubt, that Dr. Doom will relish.

19. # New Network Aims to Explore Hidden Perils

1. John Pickrell*
1. John Pickrell is a science writer in Hertfordshire.

HERTFORDSHIRE, U.K.—Before Mount St. Helens exploded in 1980, most assessments of the risk of living near a volcano centered on in-your-face hazards such as lava streaming down a mountain or magma bombs hurled into the sky during an eruption. But the ash that rained down on vast swaths of the Pacific Northwest for 5 months after the Mount St. Helens eruption got researchers pondering another threat: the potential health effects of long-term exposure to ash and invisible noxious gases. Such risks have been “relatively neglected,” says Clive Oppenheimer, a volcanologist at the University of Cambridge, U.K.

After 2 decades of intriguing but inconclusive findings, a disparate group of scientists from 20 institutions and counting has banded together to study the risks of living close enough to an active volcano to breathe its fumes. The International Volcanic Health Hazard Network, launched last month with a 3-year, $207,000 grant from the Leverhulme Trust, a U.K.-based charity, will foster collaborations and equip researchers with data about chronic health risks, information that can be disseminated to governments, volcano observatories, and emergency managers. “Observatories are faced with questions on health effects all the time, and it's important to have a resource to turn to,” says Chris Newhall, a volcanologist at the University of Washington, Seattle. Clarifying the health risks is important: Roughly 500 million people live within 100 kilometers of a volcano that has been active in the historical record. But the several dozen studies on chronic risk launched since Mount St. Helens erupted have yielded equivocal results, says network co-founder Claire Horwell, a volcanologist at Bristol University, U.K. They suggest that people exposed for years to tiny ash particles—often acid-coated silica and other pulverized rock—could be at elevated risk of lung cancer and silicosis. But that conclusion, she says, is derived from studies on the health effects of long-term exposure to silica compounds in mining and other industries; exposure levels to ash, she says, are harder to measure and can be quite variable. Another concern, adds network co-founder Peter Baxter of the University of Cambridge, is gases such as sulfur dioxide, which can cause lung and eye irritation in people living hundreds of kilometers from an active volcano. Two recent eruptions underscore the urgent need for more research. In January 2002, Congo's Nyiragongo volcano belched thousands of tons of sulfur dioxide and sulfate aerosols into Goma, a city of 400,000. Last November, Ecuador's El Reventador volcano sent a plume several kilometers into the sky, blanketing the capital, Quito, with several centimeters of ash. The new network, Oppenheimer says, should help answer how harmful, in the long run, these sorts of experiences could be. 20. # Africa's Davids and Goliaths 1. Kevin Krajick* 1. Kevin Krajick is the author of Barren Lands: An Epic Search for Diamonds in the North American Arctic. For scientists in Cameroon and Congo, the hazards of working on active volcanoes pale next to those of poverty, isolation, and violence EKONA, CAMEROON—In some ways, Aka Festus Tongwa is perfectly equipped for monitoring West Africa's biggest, meanest volcano. He studied geochemistry at top universities and converses easily with foreigners or locals in French, English, Japanese, his ancestral Nweh tongue, or any of three other regional dialects. Whether smashing open rocks on the volcano's slopes or discussing isotope analysis, he is tireless and intense. But he faces handicaps that many volcanologists could not imagine. The dirt track to the Mount Cameroon Volcano Observatory is so bad that it long ago ripped up the underbelly of Tongwa's car; he now walks the last kilometer and a half to work. When he gets there, the building offers little more than a roof and walls—no lab equipment, no library. He can't hook his battered Toshiba laptop to the Internet: There is no telephone line. Although Africa's 130-odd volcanoes raise many intriguing questions, they are largely unstudied because of poverty, warfare, and inaccessibility. Two of the biggest threats—Mount Cameroon and Nyiragongo in the Democratic Republic of the Congo—have observatories to warn nearby populations of eruptions. But these are nations primarily of subsistence farmers with little money for clean water, never mind geology. In Congo, volcanoes do not look so bad next to the genocide, civil wars, and famines that have claimed 3.5 million lives in recent years. There, scientists work in battlefield conditions. “These are brave people, heroes, doing an amazing job with few resources,” says John Lockwood, a former U.S. Geological Survey volcanologist with long experience in Africa. Mount Cameroon is monstrous—50 kilometers across, with over 100 parasitic cones peppering its flanks—but major lava flows, every 15 or 20 years, generally miss settlements, so eruptions have not been disastrous, at least not so far. Observers are nervous because Mount Cameroon's history, underground structure, and very reason for being there remain open questions. It and a line of other volcanic centers across Cameroon cannot be explained by conventional models of plate tectonics or hot spots. Ignorance of the volcano's inner workings and a burgeoning population could be a dangerous combination. The nearby populace has doubled in the last 10 years to 160,000, says structural geologist Emmanuel Suh of the University of Buea, which is located on the mountain's flank. And, he says, the mountain's apparently complex plumbing could burst in unpredictable spots. A 1999 flow overran the main coastal road near an oil refinery; the next year, another site higher up the mountain unexpectedly roared to life, unleashing a 30-meter-high lava wave that ended only a few kilometers from a village. “Our main question is: When and where will it erupt again?” says observatory director Ateba Bekoa. Bekoa's team hardly has the means to tackle that question. The government-run observatory opened in 1989 with six seismic stations, set up with British help, but by 1999 only one station still worked. “We had the instruments but not extra parts or technicians to keep them operating,” laments Bekoa. Just before the 1999 eruption, he knew the ground was shaking, but not much more; at least three stations are needed to pinpoint the sources of tremors. After the 2000 eruption, the scientists lacked air transport, and the only way to inspect the results was to walk up and look. As the team descended into a seemingly safe crater, boulder-sized lava bombs, fire, and smoke suddenly shot out. Everyone ran, and although they lacked fireproof gear or hard hats, incredibly, only one person was seriously hurt. On a hot, humid day last January, Tongwa clambers over sharp-edged brown rock left by the massive 1999 lava flow. He stops near a hole where steam is pouring out and waves his hand cautiously over the fumarole to gauge temperature. If he had a lab to analyze this steam, “it would tell us a lot,” he says. Also on the wish list: tiltmeters, portable gas meters, and satellite imagery. There is some consolation; after the eruptions in back-to-back years, the Institut de Physique du Globe de Paris (IPGP) started installing a new 10-station seismic network, which they plan to help maintain. Like most of his colleagues in Africa, Tongwa got his training abroad. Former colonial powers like France and Britain take on a steady trickle of students. He earned a Ph.D. from Okayama University in Japan, a country with vast experience in volcanoes and the desire and means to help educate Africans. But foreign aid has its limits—and exacts a price. Whenever Tongwa wants to do more than hit rocks with his battered geologist's hammer, he must get a foreign grant to ship samples abroad for analysis or take them out himself. He is away so often that he has seen the birth of only one of his four children; his littlest, 4-year-old Asongbesoh, bawled one recent morning when Tongwa slung on his backpack for a day trip. “He thinks he's not going to see me again,” Tongwa says. And because foreigners front the money, they set the research agendas and often take an unfair share of the credit, claims Celia Nyamweru, a geographer at New York's St. Lawrence University who studies volcanoes in Tanzania. African nations could reduce dependence by sharing instruments or other resources, says Tongwa, but politics and poor communication hinder cooperation. He pays about 25 cents a day to use e-mail at a crowded Internet café in Buea, but connections, especially between African nations, may not work for days, if at all. Last November, a government scientist in neighboring Nigeria claimed on the BBC that Mount Cameroon appeared headed for a major eruption. That shocked the Cameroonians; they had no indication of such an event brewing—nor could they reach the Nigerians to find out more. More insidious factors also squelch collaboration. When Congo's Nyiragongo erupted disastrously in January 2002, Tongwa received an expenses-paid invitation from Japan to join foreign scientists rushing in to assess the situation. He declined; the region is overrun with armed militias on the lookout for suspicious outsiders. “White people stick out there: Everyone knows they're there to help, so actually they're pretty safe,” he says. “But I am a black person. I could easily be taken for a rebel. Before I had a chance to explain, … I'd be gone.” Fighting in Congo and adjoining Rwanda largely destroyed scientists' ability to monitor Nyiragongo, and when it erupted in 2002 it uprooted 400,000 people, killed 150, and leveled much of the Congo-Rwanda border town of Goma. Nyiragongo exudes a peculiar, terrifyingly fast lava that can run down victims with flows reaching 100 kilometers per hour; but for local scientists and others, human marauders are usually more menacing. In the 1980s, Japanese experts began training scientists at the Congo's Center for Research in Natural Sciences, and in 1994 they helped complete an observatory with six seismic stations near Goma. Just then, unrest in adjoining Rwanda exploded into genocide, and within weeks some 800,000 people had been slaughtered. Some 400,000 refugees flooded Goma and the slopes of Nyiragongo, many of them members of armed factions vying for control. In the ensuing carnage and destruction, only one seismometer survived. A rocket-propelled grenade blew a hole in the observatory, which was then thoroughly looted. Phone lines were destroyed and bridges blown up, and the observatory's dozen staffers stopped getting paid. But they kept coming to work. One of them was Celestín Kasereka Mahinda, a young seismologist with six children to feed. The looters did not get Kasereka's Japanese camera, thermometer, or laser binoculars for monitoring changes in the summit's roiling lava lake. “I continued to work with this,” he says. He made the 14- to 16-hour round trip every few weeks. To pay helpers and buy a sleeping bag and other equipment, he sold his television and some of his wife's jewelry. Occasional visiting Japanese gave him money, and his Protestant church, run by missionaries, took up a collection. And in recent years, he has needed protection: armed escorts from the Rwanda-backed rebel Congolese Rally for Democracy, which now essentially runs Goma. Kasereka is not the only one who made sacrifices to stay on the job. Others continued to monitor the seismometer or observe the volcano, surviving by teaching school, selling cigarettes, or, like observatory director Kalendi Sadaka Kavotha, growing vegetables. Although last year's eruption was a disaster for Goma, it was probably the best thing that ever happened to the observatory. Kasereka's flimsy wooden home was flattened, and his family narrowly escaped when lava tore through the heart of Goma on 17 January 2002. But in the aftermath, dozens of foreign volcanologists and tons of equipment flooded in, along with$15 million in aid from the United Nations. Over the past year, a rotating U.N.-organized international volcanology staff has been helping the Congolese study Nyiragongo with the latest instruments, including a new seismic network, and training them in new technology. A U.N. peacekeeping helicopter is available for crater inspections and, courtesy of the U.S. government, for the first time in 7 years staff members are getting paid. “Now it is better than before,” says Kasereka, who has been logging 15- to 18-hour days inspecting fire fountains at the summit and mapping fractures near town.

But the future is far from secure. So far, there is only stopgap funding from donor countries, and the Congolese still need more training. And the volcanoes may not be patient: Nyiragongo and Nyamuragira, an even more active cone nearby, have been rumbling and glowing ominously. Recent data show that Goma sits right on a rift. The new seismic network must be protected by guards with automatic rifles, and payoffs must be made to local chiefs.

Last September, a political intrigue landed Kavotha and three colleagues in jail for several weeks on charges of subversion. There, they were starved, deprived of water, and beaten. Colleagues believe Kasereka would have been hauled in too had he not been on Nyiragongo at the time. The others are now free, nursing injuries. But visiting volcanologist Jean-Christophe Komorowski of IPGP says that armed men visit the observatory every few days demanding donations to buy beer. Grimmer threats aren't far away: In the chaotic countryside beyond Goma, the U.N. has confirmed reports of mass executions, rapes, and torture. Kasereka's new house is filled with relatives who have fled the violence.

All this is too much for some. Jean-Baptiste Katabarwa, a former dean at the National University of Rwanda and once Rwanda's only volcanologist, lost 12 family members in the 1994 genocide, and if he had not been working in Nepal at the time, he probably would be dead too. His wife and two children survived, and he bribed a diplomat to get them out. Now Rwanda—vulnerable to the Congo volcanoes along its border, and having five of its own—has no volcanologist. Katabarwa is a consultant in Sherbrooke, Quebec, currently working on a technique for turning volcanic ash into cheap, strong bricks, a resource that could help rebuild his country. “Maybe I'll go back some day,” he says.

Aka Tongwa is not going anywhere. Yes, life would be easier with a phone or even just a rock crusher—luxuries within reach if only he would emigrate. “I want to work at home, where I can serve my country,” he says. “I was born here. In New York or Tokyo, I can make a lot of money. But at home, the negatives I see, I can help to improve.”

21. # Seeing Volcanoes in a New Light

1. Daniel Bachtold

Two new software packages merge data ranging from seismic patterns to escaping gases into a unified view of a volcano's behavior; the hope is that this bigger picture will improve eruption predictions

CAMBRIDGE, U.K.—For scientists at the Hawaiian Volcano Observatory (HVO), the spectacle of lava gushing from Mount Kilauea in late January—a major eruption in which the summit throbbed like a pair of lungs inflating and deflating—was a familiar thrill, yet one experienced from a totally new vantage point. The researchers had watched this constantly active volcano go full throttle dozens of times in recent years, but the latest flare-up was the first in which a new monitoring system merged data on Kilauea's land heaves, gas emissions, and seismic rumblings into a single picture, in real time. “It made the observation much more dynamic,” says Donald Swanson, scientist-in-charge at HVO. “People were gathering around monitors and watching things happening, and that created a sense of excitement.”

HVO is the first observatory in the world to have wrapped all these kinds of monitoring data together onto one computer screen, in real time. The rationale behind the newly developed Volcano Analysis and Visualization Environment (VALVE) is for scientists to “come up with a kind of unified understanding of what's going on,” says Peter Cervelli, a geophysicist at HVO.

That is an idea that volcano observatories around the world are starting to pick up on. Another major attempt to compile monitoring data into a bigger picture is now under way in Europe. Scientists from Germany, Greece, Italy, and Switzerland are using the Greek island of Nisyros in the southeastern Aegean Sea as a natural laboratory to develop a software package called Geowarn. Much like VALVE, Geowarn unifies various monitoring data in a common time frame. But Geowarn goes a step further, presenting them in a three-dimensional map of the volcano that dominates the tiny island.

Geowarn is meant to serve as an early warning system to catch the first signs of unrest. “We try to grasp a volcanic system in its entirety before it reawakens and threatens to erupt,” says Volker Dietrich, a volcanologist at the Swiss Federal Institute of Technology in Zurich. In this sense, Nisyros is an ideal testing ground. Although it has not had a major eruption in 20,000 years, several small hydrothermal blasts from the central crater in the 19th century, and a spate of earthquakes underneath the island between 1996 and 1998, suggest that the volcano may be slowly emerging from hibernation.

Over the past 2 years, the Geowarn team has been using state-of-the-art monitoring equipment to scrutinize Nisyros from every possible angle. Global Positioning Systems and satellite radar interferometry—a new technique that clocks the travel time of microwave signals beamed from a satellite—have measured ground deformations on the island on the order of a few millimeters, from which scientists can infer alterations in the magma chamber that feeds Nisyros. They glean information on the magma's behavior from the composition of gases wafting from the mountain, the temperature and electrical conductivity of the island's hot springs, and the pattern of seismic activity beneath the volcano. The data suggest that during the latest tremblings, magma welled up but is now shedding heat, so Nisyros is unlikely to erupt any time soon.

Many observatories gather these sorts of data, but what is about to change is the way the data are analyzed. At many places, the routine is for various experts to sit around a table with printouts of their data and discuss what it all means. “That sort of works. But it isn't anywhere near as clever as it could be,” says Christopher Newhall, a volcanologist at the University of Washington, Seattle.

In contrast, “Geowarn brings data together in a unique way,” says Lorenz Hurni, head of the Institute of Cartography at the Swiss Federal Institute of Technology in Zurich. Hurni and his colleagues are the architects of Geowarn's visualization software. His team has combined digital three-dimensional maps with the powerful data-processing tools of Geographic Information Systems used by surveyors and other field specialists to create a template that monitoring data can be fed into and displayed. Experts find the approach intriguing. The system, says Newhall, gives “an unusually good integration of different kinds of data and an ability to look at them together in space and time.” Moreover, Geowarn and VALVE, both of which could be adapted to use on any volcano, promise to fundamentally alter how scientists who study volcanoes interact, says William Rose, a volcanologist at Michigan Technological University, Houghton. “We are all specialists trapped in disciplinary holes unless we … put our observations together,” he says.

One caveat, of course, is that the new systems are blind to instinct. Volcanologists standing on a fuming mountain must have “a gut feeling for how the activity is going,” says geophysicist Paul Segall of Stanford University—something that today's computers, at least, cannot possibly register.

Volcanologists hope that Geowarn or VALVE will help them refine their predictions of when a mountain might erupt, but they would be even more pleased if the new systems could aid in getting the message out to the public. “You could be scientifically superb at predicting something, but unless you can communicate that to the people who might be affected, it doesn't do any good,” says David Hill, scientist-in-charge at the Long Valley Observatory in Menlo Park, California. Geowarn's backers say that scientists outlining hazard zones to elected officials and the public will be able to do so more effectively with three-dimensional, interactive maps. Newhall agrees: “This is absolutely the way to go.” Indeed, plans are already afoot to hook up other volcanoes: VALVE at Long Valley, and Geowarn at Vesuvius.

22. # Bracing for the Big One on Montserrat

1. Richard Stone

The Soufrière Hills volcano destroyed the capital and much of the rest of this island's southern half in the mid-1990s, and it may not be done yet: Its massive and still-growing lava dome has begun to threaten communities in the north, raising the stakes for the scientists who are keeping watch

MONTSERRAT—Some of the hottest pop songs of the 1980s came to life here, in a pale beige ranch-style house perched on a bluff. Elton John, Paul McCartney, and Stevie Wonder, among others, made the pilgrimage to this recording studio on the “Emerald Isle” of the Caribbean. Perhaps they drew inspiration from the shimmering turquoise waters, or maybe their muse was the mountain looming in the southeast: the Soufrière Hills volcano.

The vista is as entrancing as ever, long after the music died. AIR Studios, part of the vast Montserrat estate of Beatles producer George Martin, sits abandoned on a swath of land evacuated by authorities last October. The only sounds on a sunny afternoon last month were the sloshing of a cement mixer and the banter of volcanologist Barry Voight and his colleagues as they yanked yard after yard of plastic piping, conduit for the fresh cement, from a borehole near the studio. If a siren had blared, Voight's team members would have dropped what they were doing and hauled out of the exclusion zone as fast as they could have: The lava dome of Soufrière Hills is bigger than ever since the volcano roared back to life in the mid-1990s, and it could collapse at any time, in any direction.

Today, the volcano is calm, a taupe ash cloud drifting lazily out to sea. The researchers have just about finished installing two devices—a meter-long steel-sheathed seismometer that resembles an artillery rocket and a smaller, more cylindrical tiltmeter—near the bottom of the borehole, one of four drilled around the volcano. The sensors sit atop a strainmeter so sensitive that it can detect motions in the rock of the dimensions of an atom. The recordings from these instruments may not rival the sounds that once filled AIR Studios, but they will be music to the ears of scientists: They are expected to be the most precise readings yet of the mountain's inner stirrings.

“We're hoping to get a feel for what's happening deep in the system,” says Voight, who is from Pennsylvania State University, University Park. “This volcano is a unique natural laboratory.” The mountain's often enigmatic convulsions have been a boon to modelers, who are starting to put together a coherent picture of the relation between magma movements and the varying seismic signals they trigger. “The models are starting to give us some power to explain the phenomena we've seen in this eruption”—insights that can be applied to many volcanoes, says volcanologist Stephen Sparks of the University of Bristol, U.K., who has extensively modeled magma dynamics beneath Montserrat.

Soufrière Hills is also a gathering storm of frightening proportions. After 400 years of relative peace, the mountain erupted in 1995. In an encore 2 years later, it obliterated southern Montserrat, including the capital, Plymouth, and much of the 16-kilometer-long, mountainous island's arable land. To the credit of scientists at the Montserrat Volcano Observatory (MVO), their advice led to timely evacuations that saved all but 19 islanders, who had defied orders to leave an exclusion zone. After 20 months of relative quiet, the volcano kicked into a new phase in 1999 in which lava is piling up at the top, forming a gigantic dome. Last autumn, the size and direction of dome growth prompted authorities to redraw the boundary of the exclusion zone, forcing about 300 more people to relocate. Although some residents still grouse about the evacuation, the volcano's threat is escalating day by day. “The dome is growing, growing, growing,” says MVO director Peter Dunkley, a geologist by training. “It really is quite terrifying.”

Steering a jeep over a buried bridge across a mud-drowned river, Lars Ottemöller, a seismologist at the British Geological Survey (BGS), lurches past villas smeared with wispy brown ash, every last one devoid of life, and a sign saying “Please keep our island clean.” As he nears the top of St. George's Hill, fat drops of ash-laden rain begin to pelt the car, forming a gray slurry on the windshield, and a sulfurous smell in the air grows stronger. A cow matted with ash looks on forlornly as Ottemöller pulls up in front of a dilapidated concrete hut housing a transmitter for a nearby surface seismometer, one of a dozen such stations. He is here for a routine check of the equipment, but the view from St. George's Hill is anything but ordinary. A delta of volcanic mud covers Plymouth, with only a few shells and facades of taller buildings still poking through.

Fortunately, the disaster on Montserrat unfolded over months, reaching a crescendo after the southern half of the island had been safely evacuated. The first tremors of the waking volcano came in 1992 and grew more intense over the next 3 years. The gradual buildup allowed volcanologists at the Seismic Research Unit (SRU) of the University of the West Indies, in St. Augustine, Trinidad, to install better monitoring equipment at Soufrière Hills.

The volcano's first real warning shot came on 18 July 1995, when phreatic, or steam-driven, explosions sent a dense ash cloud over Plymouth. The situation “got progressively worse,” says Lloyd Lynch, an SRU instrumentation specialist. A particularly severe blast on 21 August turned “day into night” in Plymouth, he recalls. “Some people were on their knees praying, thinking this was the end.” Even scientists were nervous: “We didn't understand the volcano,” Lynch says. In November, a lava dome began to grow, ratcheting up the level of concern. Yet, says Voight, who first came to the island in March 1996, officials and residents alike were reluctant to clear out for good, returning after two evacuations. Later that month, the first pyroclastic flows—roiling avalanches of ash, gas, and rock—prompted the third and final evacuation of much of the southern half of the island. “If the activity had built up much faster,” says Voight, “there would have been a ton of deaths in Plymouth.” The volcano convulsed all summer, frequently ashing the ghost town; in September an explosion sent a mortar barrage of incandescent meterwide lava blocks into a recently evacuated village, shattering half the houses and igniting fires.

Thanks to sound management and a bit of luck, Soufrière Hills did not claim a life until 25 June 1997, when 19 people in the exclusion zone died in a massive pyroclastic flow. That morning a pair of young volcanologists, Rob Watts of the University of Bristol and Amanda Clarke of Penn State, had driven out to the volcano to take Global Positioning System (GPS) readings at various sites. At about 12:30 p.m., they radioed back to MVO, asking permission to cut across the volcano's eastern flank. But the seismometers were going wild, so Chief Scientist Willy Aspinall ordered the duo to head for safety at the airport. “It was exactly the right call,” says Gill Norton, a volcanologist based at BGS headquarters in Nottingham, U.K., who later ran MVO. The pyroclastic flow tore through about 10 minutes later. It was the closest any volcanologist has come to perishing on the island.

In early 1998, the mountain inexplicably went quiet. “But it didn't go back to sleep properly,” shuddering and venting periodically, says MVO volcanologist Richard Herd. Nor did it resume its former lifestyle. Following abnormal seismic activity in October 1999, the lava began piling up into a new dome. Inexplicably, the volcano has persisted in a steady dome-building mode, punctuated by two major collapses that were triggered by heavy rains. As Science went to press, the dome held a whopping 200 million cubic meters of lava, about double the volume since the most recent collapse in July 2001. “There doesn't seem to be a limit to dome growth,” Herd says with awe.

The evolving drama, observed using the latest instrumentation, has turned Montserrat into a volcanological wonderland. “It's been a phenomenal scientific opportunity,” says the University of Cambridge's Clive Oppenheimer, one of more than 100 volcanologists who have worked on the island. “A huge array of techniques has been brought to bear on its surveillance,” he says.

For example, at most volcanoes researchers lug bulky gas monitors out into the field for occasional readings of the amount of gas venting from a mountain. In 2001, however, scientists at Montserrat were the first to deploy miniature ultraviolet spectrometers that, during daylight, can take readings of sulfur dioxide flux every few minutes. Such accurate measurements could prove particularly handy at Soufrière Hills, where the size of eruptions, researchers believe, often seems to be linked to the speed and volume of gas escaping from the rising magma.

But so far, some of the most important findings for anticipating the next moves of Soufrière Hills have come from seismometers eavesdropping on the acoustical chatter beneath the mountain. Some of the more intriguing discordant notes are long-period (LP) events, peculiar low-frequency waves that somehow reflect the movements of gases and bubbly magma inside the conduits. Sometimes these LP waves are punctuated with high-frequency squeals of unknown origin and are called “hybrid” events. In an article in press at the Journal of Volcanological and Geothermal Research, Philippe Jousset and Jürgen Neuberg of the University of Leeds, U.K., have suggested that these low-frequency waves come about when the conduit resonates as magma lurches through. The frequency patterns, the researchers propose, could be explained by the amount of gas escaping from the magma to form bubbles.

These LP events are not idle chatter. On Montserrat, when LP or hybrid signals start bunching together in repetitive patterns, trouble is usually afoot: Explosions have occurred within hours after swarms, although other swarms have been linked to magma extruding into the dome.

The volcano's idiosyncrasies could also offer fundamental insights into how magma wells up from deep crustal reservoirs to the surface. A signature feature of Soufrière Hills is its cyclic activity on various time scales. It is clear that the short-term cycles, which last between 3 and 30 hours, stem from pressure buildups that result in the volcano swelling slightly before subsiding with a gaseous burp.

More mysterious are longer-period cycles. Voight and his colleagues were the first to document patterns in which tiltmeters on the crater rim register pressure buildup, after which “the seismicity turns on,” Voight says: Swarms of LP events are unleashed. This is followed by a pulse of magma extrusion and gas release. The strength of these cycles would ebb gradually and then flare every 7 weeks or so with renewed vigor. Finally, increases in seismicity, related to the injection of fresh magma into the volcano's plumbing, have occurred roughly every 30 years at Montserrat since the late 19th century. Voight and others have proposed that these longer-term cycles are linked to changes affecting the magma reservoir more than 5 kilometers below the volcano. The mechanism remains unknown, although it probably involves periodic infusions of hot, water-rich mafic magmas into the crustal reservoir of Montserrat's more viscous, crystal-rich andesitic magma.

That is one hypothesis that the sensors near AIR Studios are supposed to help test. Last month, a team including Voight's group at Penn State, the Carnegie Institution of Washington, Duke University, and the University of Arkansas, with colleagues in Bristol and Leeds, wrapped up the installation of strainmeters, seismometers, and tiltmeters at four boreholes, each about 200 meters deep. Continuous GPS systems will straddle the holes—the first system of its kind on an andesitic volcano. Voight estimates that the new sensors should provide 100 times more resolution on tremors; besides giving a clearer picture of magma movements in the reservoir, the data will feed directly to MVO to aid in monitoring. And with their expected sensitivity, the instruments could detect changes in the volcano's plumbing in reaction to nearby earthquakes in the rough-and-tumble Lesser Antilles islands.

One current mystery concerns the lava dome: Why hasn't the monster collapsed yet? MVO's new observatory, which had its opening ceremony earlier this month, is high above the Belham Valley—high enough, scientists hope, to be out of harm's way. But as the dome just 4 kilometers away grows ever more enormous, it becomes harder to predict the volcano's likely behavior and how far the 500°C flows and surges might travel. If the dome were to topple catastrophically in the observatory's direction, any of the 12 staff members on duty would try to ride it out in a basement corridor designed to withstand the searing heat.

Even with that remote prospect in the back of everyone's minds, it is hard not to become inured to the view from the operations room's picture windows of minor pyroclastic flows tumbling down the volcano's flanks. These events, which happen almost daily, sometimes a dozen or more times a day, set the pens scratching madly on a pair of seismic recording drums in the operations room and often trigger an automated alarm sent to pagers carried by the MVO scientists. Usually there is no need to worry: The flows tend to fall safely toward the northeast and peter out in tens of seconds. So far, the rarer flows to the north or northwest have ended well short of inhabited areas.

Bracing for the Big One, a major dome collapse or an enormous explosion, keeps the researchers on edge. A substantial collapse to the east would probably relieve the tension, at least temporarily removing the live bullet from the loaded chamber. A big collapse to the northwest, however, could be lethal. Based on what is known of the 300,000-year history of eruptions at Soufrière Hills and the mountain's current specifications, the odds are vanishingly low that it would erupt cataclysmically—on the scale of the eruption that destroyed most of Krakatau island in Indonesia in 1883—and render Montserrat uninhabitable. “It's like the possibility of having a hurricane the size of the Atlantic Ocean,” says Bristol's Sparks. Just in case, though, authorities have a plan, “Operation Exodus,” for evacuating the entire island using ships from nearby Antigua and Guadeloupe.

“Montserrat is the wrong size,” laments its governor, Anthony Longrigg. “If it were smaller, the whole island would have been evacuated. If it were bigger, we would have moved everyone farther north and we wouldn't have needed to worry so much about where we draw the exclusion zone. But Montserrat needs every bit of land we can get.” And with people living right on the edge of the safe zone, that places an even heavier burden on the scientists.

## Hanging in the balance

A panel of scientists including the MVO director and a handful of outside experts meets at least twice a year—or more often depending on the volcano's activity—to assess the risk to the community. They provide advice to an emergency committee including the chief minister of the locally elected Montserrat government and Longrigg, an appointee of the U.K.'s Foreign and Commonwealth Office, as the island is an overseas territory of the U.K. Ultimately, an evacuation order is Longrigg's call. It is a fine balance. In 1976, authorities on nearby Guadeloupe evacuated about 70,000 people based on uncertain scientific evidence. After a minor eruption people were allowed back onto the island, but the damage was done. “The volcano didn't kill anybody, but the economy was destroyed,” says Voight.

When the lava dome began bulging toward the Belham Valley last September, the regular meeting of the risk assessment panel could not have come at a better time. “We felt there was a 30% chance of a pyroclastic flow coming into Belham,” says Dunkley, who serves on the panel. Based on that assessment, he and his colleagues estimated that the odds that the volcano, over the course of a year, would kill an individual residing round-the-clock in Belham were about 1 in 100, classified under U.K. medical guidelines as a high risk. “It's a judgment call, but I took the view that we had to evacuate,” Longrigg says. Many volcanologists, however, considered it a no-brainer. “You just can't live with that level of risk,” says Herd.

In consultation with MVO, the authorities redrew the boundary of the exclusion zone, dividing a village. The 10 October evacuation order has proved deeply unpopular, particularly among some foreigners whose villas lie on the wrong side of the line. “People are bitter. It's affected their property values,” says Longrigg. The bitterness may also stem from deflated hopes that the evacuation would be temporary.

The expectation was that a hurricane or sustained downpour during the rainy season last autumn would trigger a dome collapse, presumably away from Belham, which would have allowed the governor to lift the evacuation order. Despite a few serious drenchings, however, the lava dome remained intact, and according to a revised assessment in mid- January, steady dome growth has raised the risk of death to between 1 in 47 and 1 in 75 for a full-time resident in the exclusion zone, guaranteeing that Belham stays closed. Longrigg has tried to relieve the hardship by allowing workers and homeowners into the zone for 5 hours a day, six days a week. “We did that to try to keep the jobs in the zone going,” he says. The most recent calculation puts the odds of death of a person spending the maximum 30 hours a week in the zone at 1 in 575; that is deemed a moderate risk. Daytrippers are expected to be out of the zone within 5 minutes if the siren sounds. The warning could save lives in the event of a minor dome collapse, says Dunkley, “but it won't help you if the big event happens.”

Even as the degree of risk increases, scientists and officials are at loggerheads over the risk-assessment process. The scientists worry that they could be particularly vulnerable to litigation over their advice to Montserrat authorities. According to Voight, the U.K. agencies that have sought the advice have so far refused to assure that they would provide assistance if a scientist were sued. “It's unconscionable,” he says. “They want our advice to protect the public, but up to now they have been willing to leave us holding the bag.”

Discussions are under way between MVO, Montserrat authorities, and the U.K. Foreign Office over a proposal to give the risk panel status as a British scientific advisory committee. Under U.K. guidelines, committee members who have acted “honestly, reasonably, in good faith, and without negligence” would not have to pay civil damages out of their own pockets. But because only a court could determine whether a scientist was negligent, the guidelines do not provide “an absolute indemnity against liability,” says Aspinall, an independent consultant who serves on the risk panel. A way forward was being sought before the panel's next meeting in May.

Montserrat's economic prospects are uncertain as well. The economy, propped up by \$350 million in U.K. assistance since 1995, barely shows a pulse despite a construction boom in the north. And although the population has rebounded a bit, from a low of 2740 in late 1997 to about 4500 today—primarily due to an influx of guest workers from other Caribbean islands—it is still a pale shadow of the pre-eruption community of 12,000. “There is no town center as such, no center of the community,” says BGS's Norton.

Many Montserratians displaced by the volcano have settled in England and on Antigua, and few are inclined to return: A 3-year-old U.K. program to pay the expenses of returnees has so far lured back only a few dozen people. Another brake on growth, Longrigg says, is that “you can't get insurance for investing in the north.” Indeed, 35% of the workforce is employed by the government, and 80% of private sector workers are in the construction industry. The island's big hope may lie in tourism, but for now that is mostly limited to a trickle of people on day trips from Antigua. Those who have chosen to remain on Montserrat, says Herd, “have learned to live with the volcano.”

With Montserratians clustered so close to the danger zone, MVO scientists and outside collaborators know they cannot let down their guard. And they could be in it for a very long haul: One andesitic volcano on Kamchatka, a peninsula in Russia's Far East, has been growing and collapsing a lava dome since 1956. “Presumably Montserrat's magma reservoir is being replenished from a deep source, and if so, then we have no constraints on volume,” says Oppenheimer. The bottom line, says Dunkley, is that “this eruption is going to go on for a long time.” If the past is any guide to the future of the Soufrière Hills volcano, plenty more surprises are in store.