Forecasting Volcanic Eruptions

The answer, perhaps not surprisingly, is complicated. In fact, Miller and Rose are not as far apart on this issue as they might sound. John Eichelberger, a volcanologist at the University of Alaska Fairbanks, may have summed up the situation best when I asked him if the field was well on its way to becoming an exact science or was still in its infancy. He chuckled and said, "Well, when it works, it's well on its way. When we have a spectacular failure, it's in its infancy."

When it works, when it fails

Both scenarios have occurred in recent years. Mt. Pinatubo in the Philippines is one of eruption forecasting's great success stories. When it awoke in the spring of 1991, volcanologists rushed in, installed monitoring equipment, and correctly identified the nature of the unrest. They rapidly completed a hazard assessment, based in part on looking at evidence from past eruptions there. And they provided good advice to public officials and the Philippine and U.S. militaries. Just days before the mountain exploded on June 15, the Filipino government successfully evacuated more than 60,000 people from towns and villages (and the U.S. Clark Air Base) that were later partly or wholly destroyed.

Tungurahua volcano in Ecuador, on the other hand, might fall in the "spectacular failure" category of eruption forecasting. (Like meteorologists, most volcanologists tellingly prefer the less precise term "forecasting" to "predicting.") When Tungurahua woke up in 1999, volcanologists from the USGS's Volcano Disaster Assistance Program (VDAP) quickly helped their Ecuadoran colleagues set up monitoring equipment to evaluate the hazards and track the volcano's restlessness. When a magmatic eruption began there, the team smelled trouble: Three well-documented eruptions within the past 300 years that had begun just that way had resulted in explosive eruptions that devastated the volcano's flanks and killed large numbers of people. Based on the 1999 team's urgent recommendation, the Ecuadoran authorities evacuated more than 15,000 people living beneath the volcano.

Five years later, Tungurahua, while gurgling on a daily basis, has yet to produce any major explosive eruptions. "It was an economic and political disaster, and a worse one is in the making, because those 15,000 people returned to their villages and have categorically refused to leave again," says Miller, who directs VDAP and led the 1999 team. Miller fears the mountain could still produce damaging eruptions.

Why the disparity in results between Pinatubo and Tungurahua? The reason is that when a volcano shows signs of disquiet, even one that is thoroughly known and instrumented, volcanologists can still never be certain when it will erupt, in what way, and to what degree—or even if it will erupt at all.

Reason for hope

While that may not sound reassuring, the volcanological community has actually made great strides in eruption forecasting in recent years. Arguably the most significant strides have come in the development of new tools for detecting changes in a volcano's seismicity, ground deformation, and gases—the Big Three to watch at an active volcano.

Seismic networks have improved enormously. Broadband seismic sensors coupled with improved telemetry and computer systems now enable volcanologists, like never before, to receive, analyze, and display information about the earthquakes that typically precede eruptions. When it comes to gauging the slight swellings of ground surface that can signal an impending eruption, volcanologists today use satellite-based Synthetic Aperture Radar systems that can detect even the minutest elevation changes over an entire volcano, as well as ground-based GPS units that can communicate deformation data to volcano observatories in near real time.

So can we forecast eruptions? The best answer, for now, seems to be "sometimes."

As for monitoring gases, volcanologists two decades ago mostly measured releases of sulfur dioxide, which can give an indication of whether potentially explosive magma is approaching the surface. But now they can monitor a suite of gases, primarily sulfur dioxide, carbon dioxide, and hydrogen sulfide. Flying through a plume of gas rising from a volcano, volcanologists can very accurately calculate the flux of these three gases in terms of tons per day.

Researchers are even starting to reach within volcanoes themselves in hopes of getting a better idea of their plumbing. Eichelberger is part of an international team that in July 2004 managed to drill into the main magma conduit that feeds Unzen, a dangerously active volcano in Japan. (An eruption there in 1792 killed more than 15,000 people—see Deadly Volcanoes.) The conduit's now-solidified contents, samples of which the team has collected, hold valuable clues to the magma's composition, temperature, and pressure as it rose to the surface during Unzen's last eruption, which lasted for four years in the early 1990s.

Hurdles

Despite these advances, many obstacles remain to successful forecasting. For starters, it takes a long time to gain a good understanding of even a single volcano's behavior. The more former eruptions to study the better, but if a volcano only erupts on average every several hundred years, that understanding can be elusive.

For some major eruptions, volcanologists have had virtually nothing to work from. The 1912 explosion of Novarupta volcano in Alaska was the most dramatic volcanic event of the 20th century, packing 10 times the force of Mt. St. Helens's May 18, 1980 eruption. But the magnitude of the cataclysm came as a complete surprise. "I doubt even today if we would guess the scale of it, because nothing like that had happened there for millions of years," says Eichelberger, who doubles as Coordinating Scientist for the Alaska Volcano Observatory. "You don't guess complete changes in behavior like that."

Moreover, every volcano has a unique plumbing system, and each has its own supply rate of magma from sources deep underground. That means generalizing from one to the next can be tricky. "In other words, what we've learned about Mt. St. Helens does not tell us much about Mt. Shasta or Mt. Hood," says Miller, who is based in the Cascade Mountains of the Pacific Northwest. Even for the best known and best instrumented volcanoes in the world, volcanologists can never be certain how they will behave when they come alive. Despite numerous well-studied eruptions at Mt. St. Helens since the 1980 catastrophe, for instance, experts were unable to say during its awakening in 2004 whether its eruptions would be effusive or explosive—that is, a burp or a blast.

Other obstacles to forecasting are financial. "The problem is, there are 900 potentially active volcanoes in the world, and only about a tenth of them are instrumented enough to provide forecast capability," Bill Rose says. "We have the technology to do it, but the cost is too high to instrument the whole world." Many threatening volcanoes in Central and South America, Africa, and the southwest Pacific have little or no scientific scrutiny. Cerro Quemado in Guatemala, for example, has tens of thousands of people living around it, yet Miller says he's not aware of any monitoring on that peak, not even a single seismometer. Even the USGS has limited funding for researchers eager to plumb scientific questions about possibly dangerous volcanoes in the U.S. "Basically we're doing the best that we can with the budget we have, using the tools that give us the biggest bang for the buck," Miller says.

So can we forecast eruptions? The best answer, for now, seems to be "sometimes."