1. Volcano seismograms sonification application
Domenico Vicinanza, CERN
EELA Conference, Santiago, September 04-05, 2006

2. Volcano eruption forecasting
Currently no definitive method to predict the eruption of a volcano has been discovered or implemented (yet).
Scientists monitor
seismic waves
number of earthquakes and the intensity of a specific type of quake (harmonic tremors) in the run up to eruptions.
changes in the shape of the volcano or concentrations of gases emitted from the cone.
Santiago, Chile, EELA Conference,

3. The hope: music for forecasting
By correlating spectra and melodies with precise stages of volcanic activity we hope to discover a sort of “signature tune” of an imminent eruption or earthquake.
By identifying musical patterns that warn of an eruption it would be possible to implement civil protection measures, days or even hours before the event
Santiago, Chile, EELA Conference,

4. Data Audification
Data audification can be considered as the acoustic counterpart of data graphic visualization, a mathematical mapping of information from data sets to sounds.
Data audification is currently used in several fields, for different purposes:
science and engineering,
education and training,
mainly as data analysis and interpretation tool.
Although most data analysis techniques are exclusively visual in nature, data presentation and exploration systems could benefit greatly from the addition of sonification capabilities.

5. Motivations
Sonic representations are particularly useful when dealing with complex, high-dimensional data, or in data monitoring tasks where it is practically impossible to use the visual inspection, or for pattern detection.
Research has shown that people are quite more confident in recognizing patterns audibly rather then visually
Music theorists and researchers have carried out in centuries of history lots of techniques and methods to detect, study and classify musical phrases
Main idea: Music as a language and music analysis as a tool to inspect scientific data

6. Advantages of hearing Hearing is non-directional
It is surely easier to recognize a change in a sound with respect to a modification in something which has to be looked at.
It is quite impossible to distinguish a blinking light flashing 100 times a second from another one flashing at 200, 1000, or times a second, while anyone can recognize periodic signals from 20 Hz to (almost) Hz
Santiago, Chile, EELA Conference,

7. Sonification on the GRID network
First experiments involving sound production with INFN-GRID facilities started during the last months of 2003.
In September 2003, it was installed CSound, a free and cross-platform acoustic compiler, on a GRID test site, the Catania INFN-GRID computer farm
The compiler was tested within the new environment and since its beginning, the test phase produced interesting results: efficient use of the calculus resources, customizable quality of the audio files.
Santiago, Chile, EELA Conference,

8. First Phase: CSound
After a four-five months test phase, CSound has been distributed to other GRID farms in Europe,
Data sonification, sound synthesis and algorithmic composition have been investigated using couple of Csound files and Python (or sometimes Perl) scripts.
Python scripts have been used to prepare the right data files during the tests.

9. Second Phase: Java
Second test phase: development of a sound production suite based on Java (equipped with the standard audio and math libraries), more flexible and easy to manage.
All the results presented in this website have been carried on using this last approach: sample computation, audio rendering, DFT computing were obtained with the Java sonification program on the GRID
Executable = "/bin/sh"; StdOutput = "sonification.out"; StdError = "sonification.err"; InputSandbox = {"sonification.sh", "Sonification.java", "etna.dat"}; OutputSandbox = {"Sonification.aiff", "Sound.dat","Spectrum.dat", "sonification.out","sonification.err", "logfile"}; RetryCount = 7; Arguments = "sonification.sh";

10. Tungurahua (Picture: M. Monzier IRD/IG-EPN)
Sound form volcanoes
Sonified data were geophysical data collected by digital seismographs placed on the Etna volcano in Catania (Italy) and on Tungurahua volcano in Ecuador.
We carried out two sonifications:
seismogram straight sonification (tranformation into an audible waveform)
seismogram melodisation (tranformation into a melody)
Etna Volcano
Tungurahua  (Picture: M. Monzier IRD/IG-EPN)

11. About seismograms audification
In both the cases, structural properties of the seismographic information would be straightly mapped into sound or melody properties
In the first case, regularities in the seismograms will be reflected by the existence of spectral lines in the sonified signal
In the second case, regularities in the seismograms will be transferred into regularities in the melody (such as a repeated set of data will become a repeated musical phrase)
Santiago, Chile, EELA Conference,

12. Seismograms sonification
Seismographic data have been recorded onto the surface of volcanoes, by digital seismographs at a sampling frequency of about 100 Hz ( Hz).

13. Original data Example of ASCII files processed
(sample from Etna data):
Starting time: 15/06/ :03: Frequency: Hz Samples:

14. First Sonification: From data to waveform
Scaling procedure to properly arrange the samples in the [-1,1] interval, according to their sampling frequency.
Users can specify in the sonification program a certain resample frequency (pitch shift).
Setting resample factors in the Java code greater than 1 won't preserve the original pitch, allowing a frequency shift...
....making audible regular phenomena happening at very low frequencies.
In this way it is possible to observe and study periodical patterns, regular behaviors, long-range correlations, which can happen at different time scales.
Santiago, Chile, EELA Conference,

15. Sample and hold and linear interpolation
It have been implemented two interpolation methods to generate the waveform:
linear interpolation and sample-and-hold (SH) interpolation.
Linear interpolation returns particularly "smoother" sounds.
Nevertheless, for some analysis application, it could be useful to prefer the SH audio file, where no external values are added to the original (scaled) data set.
Santiago, Chile, EELA Conference,

16. Creating the waveform: Linear interp.

17. Creating the waveform: Sample & Hold

18. Quasi-regular phenomena
The waveform coded in the audio file will have exactly the same regularities, also recognizable thanks to the presence of some higher lines in the spectrum.
The order of magnitude of the frequency of quasi-regular phenomena is in the range 0-50 Hz, with a spectral envelope centered around Hz.
20x resampled Etna seismogram

19. Etna waveform and sonogram

20. Tungurahua Waveform
Santiago, Chile, EELA Conference,

21. Tungurahua Spectrum Spectral lines = Regular patterns
Santiago, Chile, EELA Conference,

22. Oscillation pattern variations are clearly visible in the pattern
Tungurahua Sonogram
Time evolution of the spectrum. Each vertical slice is the spectrum at a certain time
Oscillation pattern variations are clearly visible in the pattern
Santiago, Chile, EELA Conference,

23. Etna sonification within GILDA

29. Listening to sonification.aiff file
Etna sonogram

30. Data melodization
The whole data set interval is mapped on the (equally tempered) piano keyboard
The min value of the seismographic data will correspond to the lowest playable note on the piano keyboard
The max value to the highest playable note
Santiago, Chile, EELA Conference,

31. Main advantages
Already available tools to manage MIDI files and analyze them
Tracking the evolution of the musical intervals, the dynamics of their patterns, it is possible to detect, with an high level and in a customizable way, any kind of modifications in the shape of seismogram.
Two example of MIDI analysis (free) software:
Rubato (
MIDI Toolbox (
Santiago, Chile, EELA Conference,

32. MIDI Toolbox
Santiago, Chile, EELA Conference,

33. Dynamic evolution of the tonalities (related to couples of adjacent values)
Santiago, Chile, EELA Conference,

34. Example: Sinusoidal behavior
Original data
Graphical representation

35. Sinus melodization Music representation (of the same set of data)
Data are periodic, so the melody is periodic, with the same period
Music representation
(of the same set of data)

36. Melodization: a pictorial view
Pictorially we can say that the melodization works by overlaying seismograms with music notes
To create the volcanic score, we take a seismogram and trace the shape on to blank music bars.
Then we overlay the contours with musical notes.
Santiago, Chile, EELA Conference,

37. Seismograms Melodisation

38. Seismograms Melodisation

39. Seismograms Melodisation

40. Seismograms Melodisation
… have you ever heard a volcano playing a piano ?

41. Melody follows the shape of the oscillation
Santiago, Chile, EELA Conference,

42. First Etna-Tungurahua duet
Players:
Mt Etna: Piano
Mt Tungurahua: Guitar
Santiago, Chile, EELA Conference,

43. Reference sites:
Etna Sonification website:
Tungurahua Sonification web repository:
MIDI Toolbox manual:
Santiago, Chile, EELA Conference,

44. Thanks! …Questions
Santiago, Chile, EELA Conference,