Vol49_2_2006 801 ANNALS OF GEOPHYSICS, VOL. 49, N. 2/3, April/June 2006 Key words ocean bottom seismometer – integrated seismic networks – Southern Tyrrhenian Sea – Aeo- lian Islands 1. Introduction Ocean Bottom Seismometer Networks (OB- SN) are used in different regions of the world to improve the locations of earthquakes occurring off-shore and in coastal areas and to complete da- ta sets from the land-based networks. In the Northern Japan Trench hundreds of seismic events, recorded between July-August 1992, have been relocated by Hino et al. (1996) using both OBS and land network data, improving focal depth errors to less than 5 km. Lawton et al. (1982) studied the seismicity in the area of the Kodiak continental shelf (Alaska) and reported a mean dislocation of about 12 km of the hypocen- ters by integrating the OBS readings in their land network data set. Earthquake relocations, record- ed in the Central Juan de Fuca plate (west coast of An improved seismicity picture of the Southern Tyrrhenian area by the use of OBS and land-based networks: the TYDE experiment Tiziana Sgroi (1), Thomas Braun (2), Torsten Dahm (3) and Francesco Frugoni (1) (1) Istituto Nazionale di Geofisica e Vulcanologia, Roma, Italy (2) Istituto Nazionale di Geofisica e Vulcanologia, Sede di Arezzo, Italy (3) Institut für Geophysik, Universität Hamburg, Hamburg, Germany Abstract The problem of large location uncertainties for seismicity occurring in the Southern Tyrrhenian Sea have been par- tially overcame during the implementation of the long-term scientific mission of the TYrrhenian Deep-sea Exper- iment (TYDE), which allowed the installation of 14 wide-band Ocean Bottom Seismometers (OBS) and Hy- drophones (OBH) in the period December 2000-May 2001 on the seafloor around the Aeolian and Ustica Islands. Local events recorded at land-stations have also been observed on the seismograms of the Ocean Bottom Seismic Network (OBSN). Moreover, some hundreds of low magnitude events, undetected from the land networks, have been recorded. We combined the readings of body wave arrival times from OBS-OBH with those from land-sta- tions to locate seismic events. We focused our study on three clusters of events representative of the seismic activ- ity of the area: i) «deep» events, ii) Ustica, iii) NE-Sicily. The analysis of the integrated data set of the seismicity off-shore and on-shore, obtained from the combined land-OBS seismic network (Ustica sequence and Deep events), has improved locations in terms of RMS residuals, azimuthal gap, epicentral and hypocentral errors. More- over, further classes of events have been analysed: the first includes some local events that could be located only by integrating single trigger readings from the few available land-stations with the OBSN-data; the second com- prises local events that have been detected only by the OBS-OBH stations. In particular, the last cluster underlines the importance of an OBSN in the Tyrrhenian deep basin to reveal its unknown intense micro-seismicity, permit- ting to better understand both the tectonic and geodynamic picture of the area. Mailing address: Dr. Tiziana Sgroi, Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata 605, 00143 Roma, Italy; e-mail: sgroi@ingv.it 802 Tiziana Sgroi, Thomas Braun, Torsten Dahm and Francesco Frugoni North America) by a joint land and OBS network, provided a more accurate evaluation of the seis- micity of the intraplate region (Bebel et al., 1992). Also the use of OBSN allowed other aspects of seismological research to be developed, as an ex- ample in tomographic studies performed in the coastal region of Northern Chile, improving both the hypocenter locations and three-dimensional vp and vp /vs structure (Husen et al., 2000). The precise location of off-shore seismicity performed by land-based seismic stations is often difficult. Especially in the case of an insular or peninsular seismic network, the observations of local earthquakes occurring in coastal areas show large azimuthal gaps, resulting in a large location error. In Italy this problem particularly concerns the Southern Tyrrhenian Basin, one of the most interesting regions of the Mediterranean Sea con- sidering its intense geodynamic activity linked to tectonic and volcanic processes. Up to now only two short-term OBSs experiments have been per- formed in the Southern Tyrrhenian Sea: in 1987, a small OBS network operated for 11 days close to the Calabrian coast, recording microseismicity not detected by permanent and temporary land- based seismic networks (Soloviev, 1990); in 1996, a three week passive seismic campaign was realised by a joint Italian-Japanese collaboration around the Aeolian Islands and Ustica Island (Aoyagi et al., 1997). The attempt to integrate seismic data from land-based and OBS-stations and to improve the epicentre determinations re- vealed the following problems: i) The use of short-period seismometers (4.5 Hz), mainly used for seismic reflection sur- veys, turned out to be not appropriate for stud- ies of local seismicity in the area. ii) The three week long experiment took place in a seismic quiescent period. iii) The spacing of the OBSN (about 40 km) proved to be too large for the low magnitude lo- cal seismicity location, therefore only a few seis- mic events were simultaneously detected by more than 2 OBS-stations (Beranzoli et al., 1999). Fig. 1. Location of OBS-OBH (diamonds) and land-based seismic stations (triangles and squares) used in the present work. Each ocean bottom station was equipped with an OBH (1 channel), those equipped with an addi- tional 3-component seismic sensor (OBS) are represented by underlined numbers. Station no. 07 is not reported on the map as it was not recovered. The label 00 reports the position of the GEOSTAR multi-parametric obser- vatory, hosting an OBH. 803 An improved seismicity picture of the Southern Tyrrhenian area by the use of OBS and land networks The implementation of the long-term scien- tific mission of the multi-parameter Ocean Bot- tom Observatory GEOSTAR (Beranzoli et al., 2000) in a depth around 1900 m near Ustica Is- land (station 00 in fig. 1) was a good opportuni- ty to plan a new OBS experiment. Between December 2000 and May 2001 the Institute of Geophysics of the University of Ham- burg (IGH), the Research Centre for Marine Sci- ence at Kiel (GEOMAR) and the GEOSTAR- group of the Istituto Nazionale di Geofisica e Vul- canologia (INGV) carried out a joint seismologi- cal experiment in the Southern Tyrrhenian Sea, the so called TYyrrhenian Deep-sea Experiment (TYDE). A 14 stations OBSN was installed on the seafloor around the Aeolian Islands in the area comprised between latitude 38.4°N-39.6°N and longitude 14.0°N-15.4°N (stations 01-14 in fig. 1), at water depths ranging between 1500 m and 3500 m. Each station consisted of a wide- band hydrophone (OBH), six of them were addi- tionally equipped with a 3-component seismic broadband sensor (OBS) (Dahm et al., 2002). All the stations except OB07 were recovered. Addi- tionally, some temporary broad-band seismic sta- tions were installed in Sicily and on the Aeolian and Ustica Islands, to increase the density of the permanent seismic network configuration (Braun et al., 2001). The data set collected during the TYDE-experiment overcomes the difficulties en- countered in the past experiments, because: i) we used broad-band instrumentation, able to record both local seismicity and even teleseismic events; ii) the entire experiment lasted 6 months; iii) OB- SN was well-spaced (about 30 km) allowing the simultaneous detection of local seismic events at 3 or more OBSN-stations. The present paper analyses the hypocentral determinations of the local seismicity as deter- mined by the land-based network and the OB- SN. We discuss the difficulties concerning the choice of the crustal velocity model, demon- strating that the integration of observations from land-based stations and from the OBSN is use- ful to locate earthquakes in the Tyrrhenian Basin more accurately. In particular, the integration processes contribute to improve the hypocenter determinations, resulting in less scattered loca- tions associated both to reduced depth and epi- central errors and lower values of azimuthal gap and RMS. Moreover, we show how only the presence of the OBSN placed in the Southern Tyrrhenian Basin allows an intense micro-seis- micity linked to both volcanic and tectonic pro- cesses to be recorded. 2. Structural setting of the Southern Tyrrhenian Sea In addition to high quality seismic data recorded by a suitable network configuration, the accuracy of epicentre locations depends on the reliability of the available crustal velocity model. In the past decades the main structural characteristics of the crust under the Southern Tyrrhenian Sea have been the subject of numer- ous studies, using mainly data from seismic re- fraction experiments. To extrapolate velocity models also for depths or areas outside the stud- ied crustal sections, we briefly review the main results of those former studies on the structural setting of the Southern Tyrrhenian area. The tectonic evolution of the Tyrrhenian Basin must be considered in a more general geo- dynamic process of the collision between Euro- pean and African plates (Dewey, 1989). This col- lision caused the subduction of the ancient te- thyan lithosphere under the European plate, with the fragmentation of the European plate margin and the Alpine orogenic belt and the opening of the Tyrrhenian Basin (e.g., Malinverno and Ryan, 1986; Faccenna et al., 1997; Argnani, 2000), which started in the Miocene (Kastens and Mas- cle, 1990). The Southern Tyrrhenian Basin is charac- terised by the presence of two areas floored by oceanic crust: the Vavilov Basin of Pliocene age and the Marsili Basin, where opening started less then 2 Ma ago (Marani and Trua, 2002). The Marsili Basin is surrounded to the south and to the east by the Aeolian Arc, constituted by volcanic islands and submerged volcanoes of Quaternary age (Barberi et al., 1974). Partly, these volcanoes rest on a crystalline basement of Alpine origin which outcrops extensively on the Calabro-Peloritan sector of the Apenninic thrust belt. A clear deep seismicity is present under the Marsili Basin (Anderson and Jackson, 1987; Gi- 804 Tiziana Sgroi, Thomas Braun, Torsten Dahm and Francesco Frugoni ardini and Velonà, 1991; Selvaggi and Chiarab- ba, 1995), with hypocentral distribution that de- picts a very narrow, continuous slab dipping about 70° to the NW down to a depth of about 500 km. Tomographic images suggest a connec- tion between the subducted slab and the Ionian lithosphere (Piromallo and Morelli, 1997; Cimi- ni, 1999). Seismic catalogues (INGV, 2003) report crustal seismicity in the study area both along the coastal areas and in the submerged sector of the Apenninic thrust belt and in correspondence to the Aeolian Islands. Off-shore Sicily Island, this seismicity has been related by Neri et al. (1996) to two main faults systems crossing the Aeolian Islands archipelago (Sisifo and Vulcano), evi- denced by geological studies (Frazzetta et al., 1982; Ghisetti and Vezzani, 1982) and DDS sur- veys (Finetti and Del Ben, 1986). Crustal seismic events in the Calabrian sector can be related to an extensional tectonic regime. 3. Experimental setup For the data analysis described in this paper we created a joint database collecting data re- corded by the following seismic networks: OBSN – The 14 stations of the Ocean Bottom Seismometer Network (OBSN; fig. 1) were e- quipped with wide-band sensors: all the stations included a hydrophone, and six stations were al- so equipped with a 3-component geophone (fig. 2); the 20 bit digitiser recorded seismic data at a sampling rate of 50 Hz. Further details about technical aspects of the entire instrumentation pool can be found in Dahm et al. (2002). More- over, we used data recorded from the hydrophone mounted on the multi-parameter Ocean Bottom Observatory GEOSTAR, deployed near Ustica Island (Gasparoni et al., 2002). RSNC – The INGV operates the National Centralised Seismic Network (RSNC), with more than 100 short-period seismometric sta- tions, mostly vertical-component, installed on the Italian territory. In this study, we use data recorded by about 25 stations, located in areas surrounding the Southern Tyrrhenian Sea (fig. 1). These stations are equipped with a S-13 Teledyne Geotech sensor, acquiring continuous data with a sampling rate of 50 Hz. INGV-CT – In Eastern Sicily operates a local seismic network managed by the INGV section of Catania (INGV-CT), consisting of more than 50 short-period stations. They constitute a local network mainly employed in the seismic moni- toring of Mount Etna volcanic edifice, while the seismic stations placed in the Messina Strait and Iblean area permit to locate tectonic earthquakes occurring near the coastal area. In the present study, we used data from eight (fig. 1) 3-compo- nent stations (Mark L4C), installed in a restrict- ed area surrounding the Messina Strait, with a 160 Hz sampling rate. The time signal for the RSNC and INGV-CT data is set by GPS for all the stations, while the Fig. 2. Organisation of seismic data acquired and used in the framework of the TYDE experiment. GEOMAR, Uni-Hamburg, and GEOSTAR are in charge of the OBS/OBH instruments used during the experiment. Istituto Nazionale di Geofisica e Vulcanologia is responsible for the on land networks: for this study we did not use the MedNet and temporary networks data set. An OBH was mounted on all the off-shore stations; the number of the station is typed in italic style when an OBS sensor was mounted, too. In the figure, the characteristics of seis- mic sensors and the procedures to unify the data format are shown. 805 An improved seismicity picture of the Southern Tyrrhenian area by the use of OBS and land networks timing for the OBSN cannot take advantage of satellite signals when they operate on the sea- floor. Each OBS has an internal clock which was synchronised with the GPS clock signal before the deployment (tstart) and just after its recovery (tend). The time delays (tend − tstart) observed at each single station did not exceed ± 0.6 s during the 6- months campaign. Being the drift of the OBS clock signal linear, the correction for this time delay was realised before the data processing. Figure 2 sketches the instrumental charac- teristic, the seismic data processing, and data format conversion. 4. Preparatory activities for the event locations The earthquake locations based on data from combined land and OBS networks meet several problems not usually encountered by a land network. In the present chapter, we briefly describe all the preparatory work and the prob- lems concerning the integration of different da- ta sets: i) the choice of the location routine; ii) the definition of selection criteria for the local events to be integrated; iii) the choice of velocity models. 4.1. Choice of the earthquake location program Currently, the hypocenters reported in the bulletins of RSNC and INGV-CT are determi- ned by location routines based on Hypo71 (Lee and Lahr, 1972). In order to integrate seismic data recorded at stations deployed on continen- tal crust (RSNC and INGV-CT) and oceanic crust (OBSN), the earthquake location program has to comply with negative station altitudes and with strong crustal inhomogeneities. After balancing carefully the advantages and limits of the different available programs, we de- cided to use a relatively new location routine, HYPOSAT (Schweitzer, 2001). HYPOSAT ac- cepts negative station altitudes, but its main fea- ture of interest for our application is the possibil- ity to use at once different velocity models for the location procedure. Preliminary tests with a small cluster of sam- ple events, demonstrated the adequacy of HY- POSAT compared to other location routines. While the determined epicentres were generally well constrained, the focal depths for most events could be determined with minor uncer- tainties by introducing local velocity models. 4.2. Data selection The choice of the local seismic events suit- able for the relocation analysis was based on the INGV bulletins (RSNC and INGV-CT). The catalogue data were divided into (a) located earthquakes (§5) and (b) non-located seismic events, where only single station triggers were reported (§6). In the first category (a), the waveforms record- ed by seismic land-stations were checked in order to verify the arrival times, the uniformity of the readings, and their relative weights. The final da- ta set consisted of about 800 earthquakes in a magnitude range 1.5