TRENDS & INNOVATIONS
Singing Songs of Volcanoes
If a picture is worth a thousand words, how much for a musical score? Artists and scientists alike have long recognized that a visual image, whether a painting or a graph, has the power to instantly convey subtleties and shadings that would take many words to describe.
But visual images are only one way to express complex data — music, too, has its charms. Domenico Vicinanza, a physicist at CERN in Geneva, Switzerland, is exploring the power of sound, making music out of the seismic rumblings of active volcanoes. This process of converting data into music and other types of sounds audible to human ears, known as sonification, could help researchers detect otherwise hidden patterns buried within complex sets of geological data. And recognizing the telltale auditory pattern, or musical hook, of an impending eruption could save lives, Vicinanza says.
Humans can instinctively detect changes in a sound, such as its tempo, pitch or amplitude, says Bruce Walker, a psychologist and computer scientist at the Georgia Institute of Technology. “Speech is exceedingly complex; it changes all the time in time and frequency,” Walker says. “We still don’t really understand how humans understand speech. But…we are capable of extracting a great deal of meaning from auditory signals.”
That ability includes not only recognizing patterns that, when represented visually, may appear chaotic, but also being flexible “in terms of attention,” he says. “If you’re in the audience of an orchestra, and have some practice, you can listen to the whole orchestra, holistically, or you can choose to listen just to the violins. And if you’re very good, you can listen to just the first violin. It’s a very sophisticated data-analysis tool that we have between our ears.”
Sonification uses humans’ innate pattern-recognizing ability as a way to examine complex datasets and sort out truly random data from recurring events. Finding ways to convert prohibitively large amounts of data into a relatively user-friendly format has multiple applications in the earth sciences, Walker says. At the heart of any geological data analysis task, whether it’s examining seismic data, well log data from a drill site or marine buoy data to measure tides and ocean patterns, “there’s a person trying to figure it out,” he says. “Should I drill here or not, is there an impending earthquake or not, is the volcano going to go or not? Ultimately, it’s a decision made by a person.”
Scientists have been “listening” to earthquake seismograms for decades, Walker says. Seismic vibrations translate well into sounds, because the shapes and characteristics of the waves traveling through the planet are not unlike acoustic waves traveling through air. During the Cold War era, for example, researchers translated seismic signals into audible sounds, hoping to hear patterns that would signal whether the data were the result of natural earthquakes or underground nuclear tests, an idea first proposed by Sheridan Speeth of Bell Telephone Laboratories in 1961.
Other scientists, such as geophysicist Florian Dombois (now at Bern University of the Arts in Switzerland), have used sonification techniques to distinguish between different types of plate tectonic movements, such as subduction zones, spreading ridges and slips along transform faults. Even rock musician Thomas Dolby is experimenting with sonification, transforming the tides from the December 2004 Sumatra tsunami into music.
Scientists have also listened to volcanoes, although Vicinanza’s team is the first to give volcanoes a singing voice. Although he began his work alone, the current team combines researchers from the Sound Laboratory of the University of Salerno in Italy, CERN and the Istituto Nazionale di Fisica Nucleare of Catania, Italy.
Using seismic data from the Italian Institute of Geophysics and Volcanology of Catania, Italy, collected for 15 minutes at a time over three years from Mount Etna in Sicily, Italy, Vicinanza and his team converted the seismogram data in two different ways, creating not only sound waves that preserve the shape and spectrum of the seismic data, but also volcanic musical scores. Since 2006, Vicinanza’s team has also turned its attention to volcanoes in South America, specifically Mount Tungurahua in Ecuador.
Changing the seismic waves into sound waves is “like putting our ear on a volcano slope, listening directly to its voice,” Vicinanza says. One second of volcano oscillation equates to 100 numbers in the data, which the team converts into a sound or melody using specialized sonification software. To make the sounds audible to human ears, the researchers also shift the pitch, a useful tool because it “preserves the spectrum” of the voice, Vicinanza says. By using this process, the team found that different volcanoes have different voices: Tungurahua’s voice “has a different character” than Etna’s voice, he says.
To create the volcanoes’ songs, the team traced the peaks and troughs in the seismogram’s wave-like data onto blank musical bars. The waves’ contours became the musical notes, forming unusual and sometimes unsettling musical scores. Played through a computer, one of Etna’s songs, for example, sounds like a frenetic jazz solo played by a hyperactive piano player.
Hidden within the barrage of tinkly notes, however, may be musical “hooks,” or patterns of recurring sound, that could translate into recurring, and therefore potentially predictable, movements within the volcano, Vicinanza says. The first step, currently under way, is to sort out the different sounds, and determine which patterns require closer attention. “We are trying to understand which changes happen when an eruption approaches,” he says. “But the real challenge would be to use music as a description language and music analysis techniques as a tool to study scientific data.” Ultimately, he says, he hopes to try to correlate the stages of volcanic activity with musical properties, to learn “the signature tune or sound of an imminent earthquake or eruption.”
Music is a convenient tool for humans to use when interpreting such data, because “the musical notes provide an already linear scale as far as human listeners are concerned,” Walker says. Still, he adds, the musical scale does have limitations, particularly if a dataset is too big and too variable. A piano, for example, has only 88 keys, he says, but if a dataset has values ranging from zero to 10,000 “you have a huge data reduction problem,” he says.
Making audible sounds from a volcano’s rumblings is not an entirely new idea, says Steve McNutt, a volcano seismologist at the University of Alaska at Fairbanks. Such techniques are useful, but “generally only for special occasions, when certain highlighting is needed,” McNutt says. “Our regular analyses catch the same features but just display them in a different way.”
Converting seismic data to music, however, does have something unique to offer, by making the sounds more accessible and pleasing to human ears, and in a format that is instinctively recognizable to people, who are the ultimate interpreters, Vicinanza says. Currently, he is working with volcanologists and other earth scientists who “are interested in studying correlations between spectral properties of the sounds, and musical properties of the melodies, and some characteristics of the volcano,” he says.
Ultimately, he and his collaborators hope that this work will help scientists better understand just what a volcano might have to sing about.