Observing Volcanoes, Satellite Thinks for Itself

by Michon Scott • design by Robert Simmon • December 6, 2007


Nyamuragira Volcano (also known as Nyamulagira), in eastern Democratic Republic of the Congo, ranks among Africa’s most active volcanoes. After a month of unrest, Nyamuragira erupted on November 27, 2006, and scientists at the nearby Goma Volcano Observatory were concerned that the local population could be in danger.

  Photograph of Nyamuragira (Nyamulagira) Volcano, Democratic Republic of the Congo

They had good reason to worry; in January 2002, neighboring Nyiragongo Volcano had erupted, sending a river of molten rock through the town of Goma and into nearby Lake Kivu. According to some estimates, that disaster had killed more than a hundred people and displaced a quarter million.


Nyamuragira Volcano rises up from the Western Rift Valley near the border of Democratic Republic of the Congo and Rwanda. The 3,053-meter volcano erupted most recently in November 2006. (Photograph ©2006 Tom Pfeiffer, Volcano Discovery.)

  Photograph of a hardened lava flow in the town of Goma, Democratic Republic of the Congo

At the onset of the November 2006 eruption, volcanologists watching the distant glow in the nighttime sky knew that Nyamuragira produced fast-flowing lava. If it flowed very far, it might reach the nearby town of Sake, perhaps mimicking the 2002 nightmare. Years of civil war, however, had left the country with lingering strife and few disaster-response resources. To reach the volcano, scientists would have had to risk not just lava but sniper fire.


Nyamuragira’s neighbor, Nyiragongo, erupted in 2002, sending lava through the streets of Goma. The eruption killed roughly 50 people, and thousands remained displaced years later. (Photograph ©2006 Tom Schaul, EAUasis.)

  Satellite image of Nyamuragira's summit Caldera acquired July 14, 2005.

On December 1, 2006, the international science community received an appeal for assistance from the scientists at the Goma Observatory asking for any satellite imagery that could help them safely monitor the situation. One of the scientists who fielded this appeal was Ashley Davies, a research scientist at NASA’s Jet Propulsion Laboratory (JPL).

Davies is part of NASA’s Volcano Sensor Web, an experimental artificial intelligence (AI) project that links orbiting and ground-based volcanic sensors with NASA’s Earth Observing-1 satellite. When the satellite receives an alert of a volcanic trouble spot, it can change its data collection and transmission plans accordingly, without waiting for a human command. The Sensor Web is an offshoot of the lab’s Autonomous Sciencecraft Experiment, a broader effort to test advanced software and AI technology that could network space exploration assets—satellites, orbiters, and rovers—and make them more self-directed.

In 2004, the Earth Observing-1 (EO-1) satellite had first demonstrated its abilities by detecting and self-directing observations of an eruption of remote Erebus Volcano in Antarctica and delivering data to scientists within hours. “It worked perfectly,” Davies recalls, “although human lives weren’t at stake in that situation.”


Nyamuragira is among the world’s most active volcanoes, erupting every few years for at least the past century. Recent lava flows (within a decade or two) spilled out of the volcano’s caldera to the north and southeast, replacing green vegetation with barren black and brown rock. The November–December 2006 eruption occurred on the volcano’s lower slopes, southeast of the summit. This image from the Ikonos satellite was taken on July 14, 2005. (Image ©2005 Geoeye.)

  Visible and Thermal-infrared Satellite Image Composite of Mt. Erebus

In contrast, when Nyamuragira erupted in late 2006, Davies understood that human lives could be in jeopardy. “I immediately picked up the phone and started to do the human thing and call the person who manually handles spacecraft ops for EO-1 to re-task it,” he says. “And then I noticed this email from the sensor web system. The planning software at JPL had picked up an alert originating from the Volcanic Ash Advisory Center in Toulouse [France], and had already re-tasked the spacecraft to get the data.”

A step ahead of him, the sensor web had given EO-1 new orders to observe Nyamuragira Volcano at the next available opportunity, on December 4, 2006. The system transmitted a preview of the data to JPL on December 4, showing the intense thermal emission characteristic of ongoing volcanic activity. Complete high-resolution images and thermal data pinpointing the location of vent and lava flows arrived on December 5. By this time, EO-1 had re-tasked itself to obtain more data on December 7.


On May 7, 2004, the EO-1 satellite detected heat emission from the Erebus Volcano lava lake, alerted scientists on the ground, and automatically rescheduled the satellite to obtain more observations. This image combines nighttime thermal data (yellow and orange) of Erebus' lava lake with a photo-like image of the volcano. (Image courtesy NASA/JPL.)

  Satellite Image of the November 2006 eruption of Nyamuragira

The December 4 data led scientists to revise their prediction of where the lava would flow. Based on the best estimate of the vent location from staff at the Goma Observatory, models run by geologist Paolo Papale and his colleagues at the L’Istituto Nazionale di Geofisica e Volcanologia (INGV) in Italy initially predicted that lava would flow both to the east and to the southwest, toward the town of Sake. The estimate for the vent location, however, was off by about 2 kilometers.


The EO-1 satellite viewed Nyamuragira on December 4. The satellite's Advanced Land Imager collected this photo-like image and detected shortwave-infrared energy emitted from the hot lava. The Volcano Sensor Web was a step ahead of Davies in tracking this event, automatically reacting to volcanic ash reports and re-tasking EO-1 to observe the volcano. (NASA image by Robert Simmon, based on data from EO-1.)

  Map of initial and revised lave flow predictions from Nyamuragira.

Two kilometers may not seem like much, but given fast-moving lava and hilly terrain, even a distance that small matters. When EO-1 established a different lava vent position, the modelers revised their predictions, calculating that all of the lava would flow southwest, increasing the probability that it would cut across a major road and reach Sake.

“If the eruption had been bigger, it could have been catastrophic for Sake. In fact, the lava never got that far,” Davies says, “but this was a pretty quick data delivery, and it enabled the ‘boots on the ground’ to plan accordingly. I don’t know if this actually saved lives, but it definitely allowed re-targeting of resources away from the eastern side of the volcano.”


Initial estimates of the vent location led modelers to predict that the lava could pose a threat to areas both east and southwest of Nyamuragira. When EO-1 pinpointed the actual vent location, the updated models predicted the entire flow would head southwest, increasing the chance it would reach the city of Sake. (NASA image by Jesse Allen, based on flow data from L'Istituto Nazionale di Geofisica e Vulcanologia and elevation data from the Shuttle Radar Topography Mission.)


Improving the Sensor Web


The Volcano Sensor Web proved its value in disaster planning, but the process still required some human involvement. After the satellite transmitted the images and thermal data, Davies had to produce maps to send to his colleagues. “It was all done in about 30 minutes,” he recalls, but in the future, even map production and delivery will be automated.

Davies is also working with scientists at the Hawaiian Volcano Observatory and Mt. Erebus Volcano Observatory to build more model data into the sensor web system, allowing it to be triggered not just by ongoing eruptions but also by precursors to eruptions. “When magma moves up from depth in a volcano, the volcano bulges,” he explains. Ground-based tilt meters that measure changes in slope steepness could detect these bulges and trigger a satellite observation.

  • Observing Volcanoes, Satellite Thinks for Itself
  • Introduction
  • Improving the Sensor Web
  Photograph of Tiltmeter on Mount St. Helens

Besides planning future refinements to Earth-based autonomous systems, Davies looks beyond our home planet to envision how they might work elsewhere. “Mars is a good example,” he says. “We’re looking at a time in the future when there are multiple assets [orbiters, rovers], and they can communicate with each other. A smart system can help maximize the resource usage and data return. There’s also the safety aspect. Something in orbit could detect a storm moving in, and it could warn assets on the ground to hunker down for the duration.”

Picturing sensor webs on other planets comes naturally to Davies, who took a circuitous route to watching Nyamuragira. Long before he monitored Earth’s volcanoes, he studied volcanoes a little more remote—over 600 million kilometers away, on Jupiter’s moon Io. Scientists didn’t even know Io’s volcanoes existed before 1979, when NASA’s Voyager 1 spacecraft provided a revolutionary view of the moon.


Tilt meters, like this one installed on Mount St. Helens, can signal impending eruptions by detecting changes in slope caused by magma moving beneath a volcano. The next step for the Sensor Web is to allow it to be triggered not just by signs of active eruption, such as lava or ash, but also by signs of impending eruption, such as slope changes. (Photograph courtesy USGS Cascades Volcano Observatory.)

  Animation of Volcanic Plume Erupting on Io

“One of the great discoveries of planetary science was the detection of these huge, hot volcanoes,” Davies remarks. “Up to that point, it was thought that all the moons of the outer planets were just these little, cold, dead ice balls.” Figuring out how to study this distant moon drew Davies to the Autonomous Sciencecraft Experiment, which uses Earth as an analog for remotely observing other worlds. “When they came to me and asked, ‘What would you like to look at [with the sensor web]?’ I said ‘Volcanoes, of course!’”

His fascination with Io continues, helped by his Earth-based observations. He has just published a book comparing volcanic activity on the two worlds. Yet despite his fundamental interest in distant moons, Davies takes considerable pleasure in his success here on Earth.

“The folks on the ground were delighted with the Nyamuragira data,” he says. He estimates that the Volcano Sensor Web had autonomously tasked itself to take observations hundreds of times before. “But this was the first time where the system demonstrated a vastly superior response to the normal way of doing business. There have been proof-of-concept successes in the past but nothing quite like this.”


Scientists hope to extend automated sensor webs throughout the solar system. These far-flung observatories will be able to catch dynamic events such as Io’s erupting volcanoes. This sequence of images from the New Horizons mission shows the 330-kilometer- (200-mile-) tall plume of Io’s Tvashtar volcano. (NASA animation courtesy Johns Hopkins University Applied Physics Laboratory/Southwest Research Institute.)

  • Page:
  • 1
  • 2