Page  1 ï~~"UNDERGROUND SOUNDS" EXPERIMENTING WITH THE SOUNDS OF EARTHQUAKES Anna Saranti Institute of Electronic Music and Acoustics (IEM), University of Music and Dramatic Arts Graz Inffeldg. 10/3, 8010 Graz, Austria anna.saranti@student.kug.ac.at ABSTRACT The composition "underground sounds" - an interdisciplinary project including a concert piece as its artistic element - deals with the phenomenon of the constantly moving, therefore resonating earth and is based on data taken from an earthquake which reached 7.8 on the Richter scale and triggered a tsunami on April 1st, 2007 close to the Solomon Islands in the Southwestern Pacific. The data from several related seismic events was provided via a real-time data server belonging to the GEOFON network of seismic stations and converted to audio data using programs specifically developed for that purpose. 1. INTRODUCTION Audification of earthquake data has been practised and explored since the 1960's mainly for the distinction of "natural" earthquakes and atomic explotions and therefore is not a new field of investigation. In particular the work and publications of Ch. Hayward, "Listening to the earth sing", presented at the ICAD (International Conference on Auditory Display) in 1992 and F. Dombois [1], interested in the connection of arts and science concerning exploration of earthquake data, were studied. Even so, this project's starting point cannot be seen as the problem of audification but of earthquake prediction. The possibility of deriving and connecting global earthquake data and efficient networking, as well as the immense amount of available data, were the reasons for our interest in the topic and the extension of our artistic work, which ranged from playful and spontaneous experiments in perception to the close examination of data streams in the attempt to reveal and communicate complex connections and information. Besides dealing with the question of earthquake prediction, further intentions of the composition were to explore and examine very specific sounds and Manuela Meier Institute of Electronic Music and Acoustics (IEM), University of Music and Dramatic Arts Graz Inffeldg. 10/3, 8010 Graz, Austria manuela.meier@mur.at organizations appearing during earthquakes processes. Concentration focused mainly on following characteristics: the sound of earthquake events and the earth's movements (trying to derive audio data as pure as possible e.g. though filtering of interfering signals), the earth itself as resonance corpus, the measuring instruments and their sound's harmonic components of their sound during the recording process and the influences on triggering earthquakes, in the case of this project the sun's electromagnetic field. Another field related to the artistic implementation is that of human perception and emotions during an earthquake. The possibility and/or expedience of drawing a parallel between the prediction of an earthquake event and the culmination of a musical composition was questioned. For this examination and the need to observe several parameters during the course of an earthquake, long-term recordings were required. A direct conversion of the data, in this case audification, and a direct mapping of parameters in spatialisation was not intended. Figure 1. GEOFON Stations

Page  2 ï~~2. "UNDERGROUND SOUNDS" other sources in its sounding radius. "underground sounds"sonic explorations with earthquake data (electroacoustic composition in four parts) 1 _tremor 2_trigger 3_wake 4_vacillation_1_2_3 The data used to compose "underground sounds" includes records of the earth's activity beginning two days before the earthquake event on April 1st, 2007 and continuing several days after the tremor. The four parts of the composition concentrate on different characteristics of seismic events including sounds of the same seismic event recorded by different stations, the filtered harmonic sounds of the measuring instruments and the output of the separation of the earthquake's impulse-like components from the earth's constant movements, each used as separate instruments in the composition. Spatialisation was used as a function for displaying the artistic implementation of the field of human perception during an earthquake event. The first performance took place in the Institute of Electronic Music and Acoustics (IEM) concert room Cube [2], equipped with a hemisphere consisting of 24 loudspeakers and allowed reproduction of three-dimensional soundfields following ambisonic principles [3]. Figure 2. Spatialisation of 1 tremor, Detail from Cubemixer GUI [4] Figure 1 shows a detail of spatialisation from the first part of "underground sounds" and is a two-dimensional projection of the three-dimensional sound fields of the Cube [2]. The center of the circle corresponds to the highest point of the hemisphere. Each number signifies.............................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................. a single sound source (channel), which differs from the Earthquake sounds typically oscillate between two poles: the normal sounds of the earth's resonation and the sounds of the earth excited by an earthquake's tremor. The undulatory sound of the normally resonating earth, varying slightly, contains noise as well as harmonic components. Although the overall impression of the sound is unitary, it shows clearly distinguishable differing densities and volumes in particular frequency ranges of its rich spectra, influenced by the choice of specific sampling rates. The use of different sampling rates also enables both differing gradations of sound color and the highlighting of unique textures which proceed without end and without any sign of a regular or predictable pattern. In contrast, the brief, impulse-like tremor appears unexpectedly, its sudden intensity drawing attention to itself. It is of high amplitude and even broader frequency range than the resonating earth. When composing "underground sounds" and "1 _tremor" in particular, one of the concepts pursued was the exploration of our own perception. In this case we chose extracts from earthquake recordings to work out specific questions (see Introduction). The extracts became musical motives, which were combined to construct musical patterns. The variations were executed using several signal-processing effects. However, the focus lay on creating manipulated versions of, and patterns within, a continuing overall sound. 2.1. 1_tremor In the first part of the composition, human perspective during the process of an earthquake, an inherently unpredictable event is examined. Questions concerning the parallels between a listener's expectations of earthquake data and of a musical composition arose. The earthquake's development is examined and recreated artistically, using material from various stations. Signal alterations of the audio data include various playback speeds and the same channel type from different stations. In terms of spatialisation, the sound in the first part of underground sounds comes from above and behind from the audiences perspective, underlining the unpredictability and indirectness of the process. 2.2. 2_trigger The envelope of a specific earthquake's audio file is decisive for the form of the second part of the composition, triggering time windows where another layer of sound can be heard. The envelope itself is very unique, so even if the second layer of sound is of either harmonic or noisy nature, its structure is still associated with earthquakes. The second layer of sound consists of the filtered harmonics of the measuring instruments, generated by the

Page  3 ï~~excitation of each seismometer's eigenfrequency through the same frequency of seismic ground movement while recording. Switching over to another measuring sample rate does not affect the resonating frequency corresponding in our case to a specific pitch. In thinking about the earth as a resonance corpus, the question of how an extraterrestrial tremor would sound arose. Examples of scientific exploration in this field are mentioned in Till F. Sonnemann's thesis [5], where he evaluates data of lunar seismic events. Bearing in mind that the sun's electromagnetic field exerts a great influence on the earth's magnetic field, which in turn affects seismic activity, audio files of the solar electromagnetic field [6] were used and source filter separations applied as before with data from the earth. The sun's constant movements were reconnected through convolution with the impulse-like part of an earth's tremor. From an artistic point of view this could be interpreted as reexciting the sun's electromagnetic field with the earthquake's impulse. The sound distribution is almost identical to that in "1 _tremor", the sources still positioned mainly in the back. Now, however, single sources also come from the front. The Cube's ambisonic sound system allows sound to emerge at almost ground level, so that the audience is surrounded and covered by sound. Only the front-center and front-left areas of the hemisphere remain silent. 2.3. 3_wake A phase of comparative calm follows the earthquake's activity. The acoustical waveform seems to return to its moderately moving median. Impulse-like earthquake activity and the sounds of the earth's resonance are separated and constitute sound-supporting components in the third part of the composition, implemented by source filter separation. Random playback speeds are also utilized. The idea of placing different source signal separated audio files in linear succession through convolution arose. This describes a concentrated audio file combining several earthly resonances with the result of a "hyper-earth". The change in sound distribution from the second to the third part of the composition displays a noticable switch from ground level to the ceiling. The distancing of the sound from the audience corresponds to the distancing of a person from past events. 2.4. 4_vacillationl_2_3 In three parts differing sounds of the earthquake's aftershock are presented. Besides source signal seperation, different playback speeds were used. 4_vacillation closes the parenthesis opened with "l_tremor" at the beginning of "underground sounds". It reopens the possibility of an unpredictable event, building and releasing musical and psychological tension in a quick, dense succession without any preparation.................................................................................................................................................................. I F i --------------- -------------------------------------------- y x 3............................................................................................................................................................................................................................................................:!:!!::!:!!:::!:!!::!:!!::!:!!!::!:!!::!:!!:::!:!!::!:!!!!!!!::::::::::::::.......................................................................................................................................................................................................................................................................................................................................................................................................................................................................................................... Figure 3. Details of audio tracks from the stations EIL, CSS and APE 3. DERIVING SEISMIC DATA 3.1. Data formats The seismic network GEOFON [7] is just one of several networks which provide earthquake event data in different formats, such as SEED (Standard for the Exchange of Earthquake Data), SAC (Seismic Analysis Code) and SEG-Y (file format developed by Society of Exploration Geophysicists) and has to be converted to audio data afterwards. The main reasons for using SEED [8], an international standard format for the exchange of digital seismological data, were that it is one of the newest formats, is well-supported and documented and shows a great deal of detailed information, such as glitches and sample rates. miniSEED [9] (as one of the two parts of SEED) data records are without any of the associated control header information but contain the seismic waveform. Information about specific earthquake events and stations was preselected by the composers and therefore available for querying the appropriate waveform stored as miniSEED. 3.2. Data retrieval programs During the course of the project several programs, either GUI-oriented or command line-based, were tested as request tools for data retrieval. SeismiQuery and jweed are some examples for GUI-oriented software, whereas jrdseed, verseed and SeedLink are command line-based. There are also programs which operate seismic waveform processing, such as SAC, interesting for analysis and filtering of seismic signals. The decision to use the miniSEED format and slinktool (SeedLink protocol client) for the project was made because slinktool is a non-GUI command line tool. The user is able to derive waveform data streams of a

Page  4 ï~~station for a user-defined period of time. SeedLink is a data transmission protocol intended for use on the Internet or private circuits that support TCP/IP. Requested data streams may be limited to specific stations, locations and/or channels. All data packets are 512-byte Mini-SEED records. The most common implementation is the SeedLink [10] server within the SeisComP (Seismological Communication Processor) package developed by GEOFON. slinktool, used in version 3.8 for the project, is a command line SeedLink protocol client and part of SeisComp package as well. The number of samples from data blockettes, though equal in time span, can vary due to data gaps caused by timing [11], recording and transmission errors [12]. This must be considered when audifying earthquake data. 3.3. Conversion to audio data The conversion of miniSEED to audio data works with three libraries: SeisFile [13] (library for reading various seismic file types), SeedCodec [14] (collection of compression and decompression routines for standard seismic data formats), jMusic [15] (for conversion of seismic waveform to acoustic waveform data). Data from different stations can be compressed using different algorithms and must be first decompressed and later audified. One of the most common compression methods is the Steiml compression [16]. A miniSEED 2aiff.java program was developed to open and validate the file, extract parameters and apply an algorithm for the decompression, normalisation and writing of the waveform data as an audio file. 4. STATIONS AND CHANNELS 4.1. Stations The seismic data used for the project was derived from nine stations [17] of the Geofon Network data center provider around the world. Different channels corresponding to different measuring instruments with various sampling rates and directions of the earth movement were chosen. All stations have distinct characteristics according to their position on earth. 4.2. The seismometer's channels The SEED format uses three letters to name seismic channels. The first letter specifies the sampling rate in which the seismic data is recorded, the second describes the family to which the sensor belongs and the third letter specifies the physical configuration of the members of a multiple axis instrument package or other parameters as specified for each instrument [18]. Each station provides several types and a differing number of channels. Examples used for the project [18]: Band Code (first letter): Band Code Band Type Sample Rate(Hz) H High Broad Band > 80 B Broad Band >10 to < 80 L Long Period 1 V Very Long Period 0.1 Instrument Code (second letter): Instrument Type Family of Sensor H High Gain Seismometer N Accelerometer Orientation Code (third letter): Orientation Code Physical Configuration Z Vertical N North-South E East-West The measuring instruments produce high frequency harmonic components according to their eigenfrequency. Code Location Streams APE Apirathos, Naxos, Greece HH/SH/BH/LH/VH CSS Mathiatis, Cyprus HH/BH/LH/VH EIL Eilat, Israel HH/SH/BH/LH/VH GSI Gunungsitoli, Nias, Indonesia HH/SH/BH/LH/VH HL/BL/LL ISP Isparta, Turkey BH/LH/VH MNAI Manna, Sumatra, Indonesia HH/SH/BH/LH/VH HL/BL/LL PUL Pulkovo, Russia HH/BH/LH/VH SFJD Sondre Stromfjord, Greenland 00_BH/LH/VH 10 _HN/LN SNAA Sanae, Antarctica HH/BH/LH/VH BH/LH/VH Table 1. used [7] Location and available streams of the stations 5. AUDIO WAVEFORM ALTERATION Playing the audio data at different speeds allows one to listen to different kinds of sounds and concentrate on distinct phenomena. High- and low-pass filters as well as bandpass filters were used. 5.1. Source filter separation The earth's resonance in cepstral domain is of slow and the tremor signal of fast variation. Therefore the signal recorded by the measuring device corresponds to the output signal (time domain), the source signal to the tremor signal and the impulse response to the earth's resonance between hypocenter and recording station. In the cepstrum (source filter separation) [19] equation, the

Page  5 ï~~earth's resonance is assumed to be linear and time invariant. G. P. Angeleri's article "A statistical approach to the extraction of the seismic propagating wavelet" [20] gives further insight on the topic: a model of the seismic trace is designed as convolution between the propagating wavelet (source) and the reflectivity series of the earth (filter) with white noise added to the trace. With the help of a Matlab program computing the source-filter separation model, the fast, impulsive part (excitation) of an earthquake is separated from the earth's constant movements. This enables the artistic use of the two separated parts described above as independent "instruments". 6. CONCLUSIONS The preoccupation with the topic of earthquake sounds led to the present composition and project description. Several facts read before working on the project proved true but through the process became basic elements of comprehension, very useful for following experiments, such as the measuring instrument's easily perceptible harmonic components or the individual characteristics of each seismic station's sounds. Considering the fact that the earth is constantly moving, the scientific (but not artistic) use of certain tools can be questioned for example the source filter separation: the earth's movements are assumed linear and the general differentiation between the earth's "normal" movement and an earthquake's process from beginning to end cannot really be drawn. Cooperation with seismologists would provide insight on earthquake technical topics as well as enrich the artistic side of the project and allow the artists to go into greater detail. From an artistic point of view the whole project, with the composition as end result, was influenced by both technical and artistic elements. Sound was always a primary consideration. The artistic adaptation of the source materials is of course done with a wink and a nod, so to speak: the usage of source filter separation and the sound of the sun's magnetic field, for example, must not be interpreted as serious in scientific terms. Both the further exploration of earthquake sounds and the connection of different kinds of data for a single event would be of great interest. One example might be the connection of data of the earth's electromagnetic field with earthquake data, though it is questionable whether such a huge amount of data would improve or deepen earthquake prediction or other fields of scientific research. When linking different types of data consideration must also be given to the technical conditions and possibilities, as well as the availability of appropriate material. 7. ACKNOWLEDGEMENTS Thanks a lot for supporting the project: Georg Holzmann, Gerhard Eckel, Johannes Zmolnig, Winfried Ritsch, Markus Noisternig, Thomas Musil, Florian Hollerweger, Robert Hildrich, Alois Sontacchi, Brigitte Bergner, Philip Yaeger 8. REFERENCES [1] Dombois, F. "Using audification in planetary seismology" Proceedings of the International Conference on Auditory Display Espoo, Finland, 2001 [2] Website of Institute of Electronic Music and Acoustics (IEM) on CUBE http://iem.at/services/studios/cube accessed 2 February 2008 [3] Zmolnig, J. M.; Ritsch, W.; Sontacchi, A. "The IEM Cube" Proceedings of the International Conference on Auditory Display, Boston, MA, USA, 2003 [4] Ritsch, W.; Musil, Th.; Zmolnig, J. M.; Holdrich, R. "IEM Report 28/05, 3D Sound Mixer, Implementation eines kombinierten Ambisonic- und Bus-Mixers ftir den Einsatz in 3D Audio Environments" http://iem.kug.ac.at/projekte/publications/iemreport/ report28_05/3D accessed 2 February 2008 [5] Sonnemann, T. F. "Neue Auswertung der Daten des Apollo Lunar Seismic Program zur Suche nach Tiefbeben auf der erdfernen Seite des Mondes" Diploma Thesis, Westfailische Wilhelms-Universitit Mtinster, Germany, 2005 [6] Website of Royal Astronomy Society on Sun's Magnetic Field http://www.ras.org.uk/index.php?option= com_content&task=view&id=1206&Itemid=2 accessed 2 February 2008 [7] Website of Seismic Network GEOFON http://www.gfz-potsdam.de/geofon/ accessed 2 February 2008 [8] Website of IRIS on SEED Manual http://www.iris.edu/manuals/SEEDManuaLV2.4.pdf accessed 2 February 2008 [9] Website of IRIS on SEED Manual http://www.iris.edu/manuals/SEEDManuaLV2.4.pdf Appendix G, Data-Only SEED Volumes (MiniSEED), p.185-188 accessed 2 February 2008

Page  6 ï~~[10] Website of IRIS on SeedLink http://www.iris.edu/data/dmc-seedlink.htm accessed 2 February 2008 [11] Website of GEOFON on SeisComp - Configuration Manual ftp://ftp.gfz-potsdam.de/pub/home/st/GEOFON/ software/SeisComP/2.5/seiscomp-2.5.pdf Chapter 8 - Troubleshooting Acquisition, 8.5 SeedLink's StreamMonitor, p.29 accessed 2 February 2008 [12] Website of IRIS on SEED Manual http://www.iris.edu/manuals/SEEDManuaLV2.4.pdf Chapter 1, Design Goals and Strategies, p.9 accessed 2 February 2008 [13] Website of Lithospheric Seismology Program (Department of Geological Sciences University of South Carolina) on SeisFile Library http://www.seis.sc.edu/software/seisFile/ accessed 2 February 2008 [14] Website of Lithospheric Seismology Program (Department of Geological Sciences University of South Carolina) on SeedCodec Routines http://www.seis.sc.edu/software/SeedCodec/ accessed 2 February 2008 [15] Website of jMusic http://jmusic.ci.qut.edu.au/ accessed 2 February 2008 [16] Website of IRIS on SEED Manual http://www.iris.edu/manuals/SEEDManuaLV2.4.pdf Appendix B, Compression Algorithms: Steiml Compression Scheme, p.131-140 accessed 2 February 2008 [17] Website of GEOFON on Network Stations http://www.gfz-potsdam.de/geofon/new/netabs/ge.html accessed 2 February 2008 [18] Website of IRIS on SEED Manual http://www.iris.edu/manuals/SEEDManuaLV2.4.pdf Appendix A, Channel Naming, p.123-129 accessed 2 February 2008 [19] Zilzer, U. DAFX - Digital Audio Effects Chapter 9.2.3, Cepstrum, p.310-315 John Willey and Sons, West Sussex England, 2005 [20] Angeleri, G.P. "A statistical approach to the extraction of the seismic propagating wavelet" In Geophysical Prospecting Volume 31 Issue 5, p. 726-747, 1983.