Ž .Physics of the Earth and Planetary Interiors 113 1999 75–88

First application of ASDP software: a case study at Mt. Etnaž /volcano and in the Acri region Southern Italy

Domenico Patane a,), Ferruccio Ferrari b, Fabrizio Ferrucci c`a Istituto Internazionale di Vulcanologia, C.N.R., P.zza Roma, 2, 95123 Catania, Italy

b Soften, S.A., Li Battiati, Catania, Italyc Dpt. Scienze della Terra, UniÕersita della Calabria, Rende, Cosenza, Italy`

Received 19 March 1998; received in revised form 6 July 1998; accepted 10 August 1998

Abstract

Ž .The PC-based automatic seismic data processing ASDP software module uses a multi-algorithm approach and a newŽ .procedure MSA for signal detection, phase grouping and event identification and location. It is designed for an efficient

and accurate processing of local earthquakes records provided by single-site and array stations. Results from ASDPprocessing of two different data sets are analysed to evaluate its performance. By comparing of ASDP pickings with thoserevised manually the detection and subsequently the location capabilities of this software are assessed. The first data set is

Ž .composed of 330 local microearthquakes 1.7(M(3.8 recorded in the Mt. Etna volcano area by a telemetered analogseismic network. Digital data conversion of data is performed at the network centre operated by the Istituto Internazionale di

Ž .Vulcanologia of C.N.R. The second data set comprises 40 ultra-microearthquakes 0.5(M(2.0 recorded by a temporaryŽ .array of four digital three-component stations, deployed in the Acri region southern Calabria . For the Etnean earthquakes, a

comparison of the automatic results with the manual picks indicates that the ASDP module can accurately pick 80% of theŽ . Ž .P-waves differences are within 0.15 s and 65% of the S-waves differences are within 0.3 s . A peculiarity for these events

is that the majority of the 3 three-component station records do not exhibit distinct S-waves. This statistics includes recordsŽ .with a strong noise ground tremors andror system electronic disturbances , but does not include false alarms due to

non-seismic signals. Conversely, if we consider high-dynamic recordings of seismic digital data in a low-noise environmentŽ . ŽAcri temporary array the ASDP module can accurately pick 93% of the P-waves differences with manual estimations are

. Ž .less than 0.1 s and 87% of the S-waves differences with manual estimations are less than 0.2 s . For this type of data with afavourable signal-to-noise ratio the automatic back-azimuths estimates are also extremely accurate in the 87% of the casesŽ .differences estimations with those revised by an analyst and deriving from hypocenter solutions are generally within 108 . Ingeneral, our results indicate that both automatic ASDP and manual hypocenter locations are comparable within the estimatederrors bounds. Furthermore, the comparison of ASDP automatic locations with those performed by XRTP-IASPEI software,routinely used for automatic locations at the Mt. Etna seismic network, indicates a better performance of the ASDP module.q 1999 Elsevier Science B.V. All rights reserved.

Keywords: Automatic seismic data processing; Mt. Etna; Acri region; Italy

) Corresponding author. Tel.: q39-95-448084; fax: q39-95-435801; e-mail: patane@iiv.ct.cnr.it

0031-9201r99r$ - see front matter q 1999 Elsevier Science B.V. All rights reserved.Ž .PII: S0031-9201 99 00031-X

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88`76

1. Introduction

Regional and local telemetered seismic networksfor earthquake monitoring have been in existence formany decades. Recent rapid technological advancesand the advent of microcomputers offer even im-proved flexibility in seismic systems design andrecording capabilities. The capability of the micro-computers to transfer instructions to external hard-ware components has shifted the emphasis in systemdesign from hardware to software. Modern digitalseismograph networks encompass a wide range ofdisciplines from fine mechanics to RF and satellite

Ž .telemetry e.g., Calderoni et al., 1997 to real-timeseismic processing software and advanced computernetworking. Obviously, the availability of high-dy-namic digital data has renewed the seismologist ef-forts to study the seismic process of earthquakes.

Several software products for seismic data analy-sis include digital seismograph network data acquisi-

tion software, interactive time series analysis fordisplay, manipulation, filtering, spectral analysis and

Žearthquake location e.g., IASPEI Software Library,Tottingham and Lee, 1989; Bache et al., 1990; Bod-varsson et al., 1996; Harris and Young, 1997; SNDA,

.Haikin and Kushnir, 1998; MATSEIS .Ž .We have developed Patane and Ferrari, 1999 an`

Ž .automatic seismic data processing ASDP softwaremodule, for efficient and accurate processing of localearthquake network recordings provided by single-site and array stations. The ASDP exploits recent andadvanced programming PC-based technology for au-tomatic detection and location, interactive waveformand map manipulation tools for analyst review, andsystem management by an integrated network

Ž .database Seismic Network Management, SNMmodule. The input data are record files in PC-SUDSformat and the output includes identified phase de-tection, earthquake locations, original or pre-processing waveform segments, etc. Furthermore,

Ž .Fig. 1. Location map of permanent seismic network Mt. Etna and Aeolian Islands used in this study. The grey square at the top of thefigure indicates the area under special surveillance of the Acri temporary array. The blow up area on the right is a detail of Mt. Etna volcanowith elevation contour lines at intervals of 500 m. One component stations are marked by squares, three-component digital stations bytriangles.

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88` 77

Table 1U UPermanent networks Total One component analog Type of sensors Three-component analog Type of sensorsUU UUand temporary arrays number of stations and digital stations

stationU U UMt. Etna 14 SVN–SML–SCV–PZF Mark L-4C MAN –ECP Lennartz LE-3D,

UU UUCTS–SVN–CZM–CIS EGA –ESP Guralp CMG-4PDN–ESP

UAeolian Islands 14 ACL–FIL–PAN–SAL– Teledyne- VUG–VPL–VLT– Teledyne-Geotech,UŽ .LCV–LLI–STR–NOV Geotech S-13 VCR-PLE all S-13, Guralp CMG-4

UUTF2UUAcri 4 Sd2–So2–Sg2–Co2 Lennartz LE-3D

UUŽ .all

the history of the decision process is also available.Ž .The study of Patane and Ferrari 1999 provides a`

detailed description of the ASDP module architec-Ž .ture. The main features in ASDP are the use of: i a

Ž .new object-based visual interface and ii a modularmulti-algorithm approach for signal detection, phaseidentification and event location.

The current performance and likely future devel-opment of these new approaches for automatic seis-

mic data analysis are evaluated in the present workby the comparison of the automatic ASDP resultswith those obtained by the XRTP-IASPEI softwareand with those revised by an analyst. Other specific

Ž .objectives in this study are: i to assess the opera-tional reliability and performance of ASDP software;Ž .ii to verify the efficiency in the use of an on-line

Ž .database; iii to determine the monitoring capabilityŽ .of the overall system; and iv to explore the advan-

Ž X X .Fig. 2. Example of two Etnean earthquake recordings latitude: 37845 0.38; longitude: 14859 0.21: Depth: 5.22 km . Seismograms areŽ .relative to six stations ECP, CIS, PDN, CTS, CZM and SCV of the Mt. Etna network and to one station of the Aeolian Islands network

Ž .VPL . Automatic P and S pickings are also shown.

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88`78

Fig. 3. Example of an ultra-microearthquake recorded by the Sg2station deployed in the Acri region. The application of the 3-C

Ž .automatic polarization analysis CMD for phase detections is alsoshown. P- and S-phases are clearly recognised for this weak eventŽ .magnitude ca. 0.7 , not detected by all stations of the array. Thet y t computed time is 1.6 s and the related epicentral distanceS P

from Sg2 is ca. 6.0 km.

tages of modular programming for seismic data anal-ysis.

Microearthquake data recorded at Mt. Etna vol-Ž .cano Sicily, Southern Italy and in the Acri region

Ž .Calabrian Arc, Southern Italy have been used totest the performance of the ASDP module.

2. Mt. Etna volcano and Acri region seismicity:data sets, seismic networks and instrumentation

Earthquake activity at Mt. Etna volcano, such asin other volcanic areas, is usually defined as‘volcanic-seismicity’ even though it is often difficult

Žto differentiate between events of volcanic volcano-.tectonic origin and those of tectonic origin. Recent

advances in the quality and quantity of seismologicaldata available on Mt. Etna have allowed to infer thatthe dynamics of the volcano might be tectonically

Žcontrolled Ferrucci and Patane, 1993; Patane et al.,` `.1994 and no apparent difference seems to exist

between short-period microearthquakes recorded inŽthis area with those in no-volcanic regions Patane et`

.al., 1997 . Therefore, we consider that a large classof Etna’s earthquakes, such as those used in thisstudy, are related to shear fracture mechanismsŽ .Patane et al., 1994, 1997 . Other types of seismic`

Žsignal ‘low-frequency’ earthquakes and volcanic.tremor , which are also recorded at the Mt. EtnaŽ .network Chouet et al., 1994 , clearly deviate in their

waveform from earthquake signals and for theseevents a direct participation of magma in the sourceprocesses seems unequivocal.

In this study, the first data set analysed consists ofŽ .330 local Etnean microearthquakes 1.7(M(3.8

recorded at 14 seismic stations of the Mt. EtnaŽ .permanent network Fig. 1 and Table 1 . The data

recorded by this network are integrated with dataŽacquired by the Aeolian Islands network Fig. 1 and

.Table 1 . Both networks are run by the InstitutoŽ .Internazionale of Vulcanologia I.I.V. , C.N.R.,

Ž .Catania, Italy Privitera et al., 1989, 1993 . All ana-log and digital seismic signals were transmitted bywire andror radio links to I.I.V. data acquisitioncenter in Catania, where an automatic processing

Fig. 4. For the 261 Etnean earthquakes automatically located byŽ .ASDP 79% of the overall data set , histograms reporting the

differences between the ASDP automatic and manual picking forŽ . Ž .the P-waves a and for the S-waves b . To identify S-waves we

Ž .applied a characteristic function see the text for further details inŽ .which the application of the polarization analysis CMD present a

more greater weight. Therefore, the positive bias in the S-phasedetection will depend on the choice of the fixed and moving

Ž .windows used in the CMD analysis Patane and Ferrari, 1997 .`

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88` 79

system provided the conversion of the analogic sig-nals in digital format and automatic computation of

Žmain focal parameters were performed routinely by.the XRTP-IASPEI software; Tottingham, 1994 . The

Žautomatic analysis of analog digitised with a 12 bit.ArD converter and digital signals are performed in

Ž .continuous sampling rate of 100 Hz and in triggerŽ .mode sampling rate of 200 Hz using a PC-based

Ž .Fig. 5. Histograms relative to the Acri region earthquakes recordings are related to the four stations Sg2, So2, Sd2 and Co2 reporting theŽ .differences between the pickings obtained by the two automatic procedures STArLTA and CMD implemented in ASDP and pickings

Ž . Ž .reviewed by the analyst, both for P- a and S-phases b . The differences between the azimuths obtained using the automatic polarizationŽ . Ž .filter CMD and those revised by an analyst are also reported c .

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88`80

computer. Seismic signals acquired by digital sta-tions are sampled at 125 Hz. In Fig. 2 an example oftwo Etnean earthquakes recorded at six stations ofthe Mt. Etna network and at one station of theAeolian network is reported.

The availability of seismic data in digital formalso from 3 three-component stations provides ac-ceptable surveillance for characterisation of seismicactivity in the Mt. Etna area. The crust has now been

Žmapped in detail beneath the volcano e.g., Hirn et.al., 1991 . Currently, the major seismological prob-

Ž .lems at Mt. Etna are related to: i the abundance ofŽ .events with emergent phase onsets; ii the high level

of noise due to tremor sources which affect record-ings at the stations near the summit of the volcano;

Ž .and iii the distortion in the signals caused byanalog transmission of data to the network centre bythe use of radio andror dedicated telephone lines. Itis noteworthy that the focal depths of Etnean seis-micity are usually comprised between 1 and 25 km.The overall geometry of the Mt. Etna network inte-grated with the Aeolian Islands one and the type ofequipment used allow us to put high-quality con-straints on seismic activity at Mt. Etna occurring atall depths, with the exception in some cases for thevery shallow earthquakes. In fact, for these kinds ofevents it is worth emphasising the need to collectseismic data as near as possible to epicentral areas.The average distance between seismic stations at Mt.Etna area is ca. 10 km. Theoretical studies andexperimental evidence prove that the above require-ment is as important as the availability of unambigu-ous S-wave times for earthquake location.

Seismic activity in Calabria region is usuallymonitored by a regional short-period seismic net-

Ž .work Guerra, 1991 integrated by some stations ofthe National Institute of Geophysics. However,knowledge of seismicity in the Calabrian Arc is very

Ž .poor due to: i the low level of seismic activity inŽ .recent times and ii the lack of an adequate local

seismic network with a sufficient number of stationsto monitor the region. The space distribution of themicroearthquakes recorded however reflects the main

Žgeological feature of the region Guerra and Corea,.1989; Guerra, 1991 . The operation of a temporary

short period network, deployed during 1996 in orderto monitor some seismogenetic structures in Cal-abria, allows for the observation of the local seismic-ity in the Acri region. We used 40 ultra-microearth-

Ž .quakes 0.5(M(2.0 , recorded by this temporaryarray, in order to evaluate the quality of automaticlocations by the ASDP module, also when an arrayoperates with few stations. The temporary Acri arrayŽ .Fig. 1 , covering an area of about 10=10 km, wascomposed of four high-dynamic three-componentdigital stations run in trigger mode for event detec-tion. Seismic signals were sampled at 250 Hz. Digiti-sation is performed by the use of a 16 bit ArDconverter and the station offered the possibility tostore about three megabytes of seismic signal on twoindustry-standard 3.5 in. floppy disks, which wereperiodically changed. Fig. 3 shows an example of aultra-microearthquake recording by one station of thearray.

3. Analysis of data

We used the ASDP automated module to processand analyse the two data sets above mentioned. Thedata processing by ASDP included automatic signaldetection, event identification, recognition of P- andS-phase if 3-C data are available and phase groupingfor earthquake location.

We also compared results obtained by ASDP withthose obtained by an analyst to ensure proper phaseidentification, reliable timing and more accurate lo-cations. Finally, a comparison with the results de-rived from the XRTP-IASPEI software, routinelyused for automatic locations of earthquakes recordedat the two seismic arrays of Mt. Etna and of AeolianIslands, is also performed.

3.1. Automatic detection statistics

The accuracy of P and S onset-time estimation isan important issue and critical in the automatic anal-

Ž . Ž .Fig. 6. Comparison between the automatic ASDP gray circles and the manual black squares locations for the 261 Etnean earthquakes.Ž . Ž . ŽMaps and cross-sections are related to the period January–June 1997 a and July–December 1997 b . Maps also report some stations with

.squares and triangles of the Mt. Etna network.

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88` 81

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88`82

Ž .Fig. 7. a Comparison between the number of the automatic andmanual locations as a function of the magnitude for the 261

Ž .Etnean earthquakes. b–d Differences between the automatic andŽ .manual location results latitude, longitude and depth as a func-

tion of the magnitude.

ysis for earthquake location. To analyse the signaldetection capability of the ASDP module, we com-pare automatic signal onset time picking with man-ual picking both for Etnean and Acri earthquakes. AtMt. Etna accurate determinations of micro-earthquakes pickings are generally problematicmainly due to the high noise. In the histogram of

Fig. 4a the P-waves picking time differences be-tween ASDP and those revised manually, d t sTP PŽ . Ž .AUTO yT MAN are reported. After the manualP

revision we clearly recognise that these time differ-ences decrease with increasing of event magnitudesŽ .for MG2.5 . It is found that about 65% of the d t ’sP

are within 0.05 s while in 80% of the cases the d t ’sP

are within 0.15 s. The availability of three-compo-nent data improves the detection of S-arrivals on thebasis of the differences in polarization characteristic,spectral content and a new estimation of theSTArLTA in the P-coda region. Therefore, for S-waves detection, we adopt a similar procedure to that

Ž .described by Cichowicz 1993 , even though wecompute the characteristic function in a different

Fig. 8. Histograms reporting the differences between the auto-Žmatic and manual location parameters latitude, longitude and

.depth for the 261 Etnean earthquakes.

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88` 83

Ž .manner see Patane and Ferrari, 1999 . In the his-`togram of Fig. 4b the S-waves picking time differ-ences between ASDP and those revised manually,

Ž . Ž .d t sT AUTO yT MAN are reported. On theS S SŽ .overall set of automatic S-picks 129 performed by

ASDP about 40% of d t ’s are within 0.1 s while inSŽ .65% of the cases the d t ’s differences d t areS

within 0.3 s. It is noteworthy that the majority of theŽrecords on the 3 three-component stations ESP,

.EGA and ECP available at Mt. Etna do not havedistinct S-waves. If we consider that the bulk ofseismicity used in this study occurred beneath thecentral-western parts of Mt. Etna and that the three-

Ž .component stations EGA and ESP are located inthe south-eastern sector of the volcano, the presenceof magma at depth in the central-southern part of thevolcano may be responsible for the S-wave radiationattenuation. This effect is clearly observed at stationECP where a strong attenuation of seismic signalsoccurred due to the presence of magma in the plumb-ing systems of the volcano. Rarely at this stationgood recordings for events with magnitude less than3.0 are present.

Better results are obtained for the Acri earth-quakes even though with a minor statistical signifi-cance, due to the limited number of processed earth-quakes. For this 3-C data set, we applied the twoprincipal ASDP routines for automatic detectionŽ .STArLTA and polarization and compared the re-sults with manual pickings. Fig. 3 illustrates resultsobtained by the application of the automatic polariza-

Ž .tion filter CMD for phase identification. P- andS-phases are clearly recognised for this weak eventŽ .magnitude ca. 0.7 , not detected by all stations ofthe Acri array. This earthquake and the following 23

Žwith very low energy content magnitude less than of.ca. 1.5 are usually recorded only at Sg2 and some-

times at So2. However, back-azimuths and S–P timevalues allow us to associate them to the same focalvolume of the other 16 located events of the anal-ysed sequence. In the histograms of Fig. 5 the com-parison between automatic and manual results, for

Ž .the four stations Co2, Sd2, So2 and Sg2 is re-ported. Fig. 5a and b show, both for the P- and theS-waves, the comparison between the results ob-tained using the STArLTA detector and the polar-

Ž . Ž .Fig. 9. Comparison between the manual black squares and the ASDP grey circles automatic locations for a shallow seismic swarmlocated near the central crater at Mt. Etna area.

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88`84

Ž .Fig. 10. Comparison between the manual gray circles and theŽ .automatic black squares locations. For all the events location

errors are less than 1 km. The grey zone indicates the areadelimited by back-azimuths direction, automatically computed ateach station.

Ž .ization analysis CMD . The statistics between theautomatic and manual back-azimuth computationŽ . Ž .particle motion analysis is also reported Fig. 5c .Results indicate that the ASDP module can accu-

Žrately detect 93% of the P-waves differences with.manual estimations are less than 0.05 s and 87% of

Žthe S-waves differences with manual estimations are.usually less than 0.2 s . For this data set with a good

signal-to-noise ratio automatic back-azimuth esti-mates are extremely accurate in 87% of the casesŽdifferences with estimations revised by an analystand deriving from hypocentral locations are gener-

.ally within 108 .

For the stations Co2 and So2, the presence of anoise source with frequency of about 25 Hz causessome problems, even if the pre-filtering of signalspartially removes them.

3.2. Automatic location statistics

The main task of a seismic array is the accuratedetermination of earthquake hypocenters which de-

Ž .pends on: i the precision of the P and S pickings;Ž . Ž .ii the local velocity structure; and iii the networkgeometry. Conventional earthquake location esti-mates are based on a linear approximation to a set ofnon-linear equations and 1D velocity model. In re-cent years, new methods for hypocentral locationsusing a 3D velocity structure have been publishedŽe.g., Thurber, 1983; Eberhart-Phillips, 1986; Hole,

.1992; Thurber and Atre, 1993 but are not alwaysŽusable in practice e.g., Schwartz and Nelson, 1991;

. ŽHole, 1992 . Two 3D velocity models tomographic

.approaches have been published for Mt. Etna areaŽ . Ž .by Hirn et al. 1991 and Cardaci et al. 1993 ,

respectively. Unfortunately, the model of Hirn et al.Žis restricted to the uppermost crust until to a depth

.of about 6 km beneath the central sector of thevolcano. The model of Cardaci et al. ‘covered’ thewhole crust, but was not considered useful for ourpurpose since it was not possible to trace rays forfarther stations and the model is a little rough.Furthermore, the computation with 3D velocity mod-els is in contrast with our necessity to obtain earth-quake locations within a few minutes of their occur-rence in the automatic analysis for surveillance pur-poses.

Ž .Patane and Privitera 1997 evaluated the velocity`Žmodels available in the Mt. Etna area Sharp et al.,

.1980; Hirn et al., 1991; Cardaci et al., 1993 andŽ .chose the 1D model of Hirn et al. slightly modified .

Ž .Fig. 11. Histograms reporting the comparison of hypocentral parameters Hypo71 results for the 261-Etnean event locations obtained viaŽ .phase manual pickings, via ASDP and via XRTP-IASPEI software. Statistics of the solution quality of hypocenters Quality class , number

Ž . Ž .NO of station readings only P for XRTP-IASPEI and PqS for manual and ASDP , largest azimuthal separation in degrees between theŽ . Ž . Ž .stations GAP , root mean square error of time residuals in seconds RMS , standard error of the epicenters in km ERH and standard error

Ž .of the hypocenters in km ERZ are shown. It is noteworthy that minor RMS values and hypocentral errors obtained by XRTP-IASPEI doŽ .not represent true errors limits, due to the minor number of phase readings without S-phases involved in the calculus and the related bad

azimuthal coverage.

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88` 85

It could also be used for stations several tens ofkilometers away, like those in the Aeolian Islands

Ž .and in Calabria Patane et al., 1997 . The model of`Hirn et al. used in this study is quite similar to the

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88`86

one generally used to locate Calabrian earthquakesŽ .Guerra, 1991 recorded by the Acri array.

To automatically locate the earthquakes in theabove two data sets with ASDP we adopted the

Žstandard HYPO71 routine Lee and Lahr, 1975 re-vised by Lee and Waldes, in IASPEI Software Li-

.brary, 1994 . This choice allows us to compare, forthe Etnean earthquakes, ASDP results with thoseobtained using the XRTP-IASPEI software and withthe locations obtained routinely from the analysts atthe I.I.V. centre. Using the recently developed MSAprocedure for phase grouping and earthquake identi-

Ž .fication see Patane and Ferrari, 1999 , we tried to`reduce the uncertainty from an incorrect associationof time pickings. In Fig. 6a and b the 261 automatic

Žlocations provided by ASDP 79% of the Etnean.data set are compared with those obtained manually

by analysts. Encouragingly, 92% of earthquakes withMG2.4 were well located automatically. For the

Ž .lower magnitude events M-2.4 , 62% of auto-Ž .matic locations were of high quality Fig. 7a . His-

tograms in Fig. 8a–c give the differences betweenŽmanual and automatic location parameters latitude,

.longitude and depth . Regarding latitude, the differ-ences are within 1 and 2 km for 81% and 90% of thecases, respectively. For longitude, the differences arewithin 1 and 2 km for 65% and 81% of the cases,respectively. Finally, for the depth the differencesare within 1 and 2 km in 58% and 71% of the cases,respectively. The slightly better precision of the lati-tude parameter is considered to reflect the geometryof the combined Etnean seismic network and Aeo-lian Islands stations. In the case of focal depth,events with MG2.4 are well-constrained by thefavourable azimuthal coverage of the area due, therelatively higher signal to noise ratio that allows amore accurate time picking. In general, we do notobserve a dependence of location errors on the mag-

Ž .nitude Fig. 7b–d , but earthquakes with major errorsare related to locations computed with an inadequateazimuthal coverage. In Fig. 9 we present a compari-son between automatic and manual locations for aswarm of about 10 events. In general computedautomatic hypocenters are comparable to the manualones within errors bounds. Similar results are ob-

Ž .tained Fig. 10 for the 16 events in the Acri region,even though both the automatic and manual locations

Žinvolved few readings only three or four P-picks

.and usually three S-picks . Finally, in Fig. 11 weŽ .give the location parameters Hypo71 output results

for the Etnean earthquakes obtained from the twoautomatic routines ASDP and XRTP-IASPEI. TheXRTP-IASPEI software is routinely used for auto-matic locations of earthquakes recorded at the twoseismic arrays of Mt. Etna and of Aeolian Islands. Inthe same figure the comparison of the ASDP andXRTP location parameters with those revised manu-ally is reported.

It is noteworthy that minor RMS values andhypocentral errors obtained by XRTP-IASPEI do notrepresent true error limits, due to the minor numberof phase readings involved in the calculus and to therelated worse azimuthal coverage. In fact, theXRTP-IASPEI statistics includes much lower unreli-able RMS values and errors, which are obtained by

Žthe calculus with only three or four 27% of the. Ž .cases and five 46% of the cases readings, without

S-phases. Clearly, we obtained a better performanceof the ASDP module with respect to the XRTPsoftware.

4. Concluding remarks

The ASDP module developed for local earthquakerecordings processing, using as input recorded filesin PC-SUDS format, executes all data analysis neededfor a preliminary seismic bulletin: detection of seis-mic events, identification and parameter estimationof seismic phases, location of event sources andidentification of different types of signal in a vol-canic environment. The program also permits inter-active analysis of intermediate processing resultsduring execution, including: manual picking of seis-mic phases, evaluation of their amplitude, dominantperiod, onset time, duration, etc.

The application of ASDP to two different datasets recorded at Mt. Etna volcano and in the Acriregion, respectively, has provided a first opportunityto examine design features implemented in this soft-ware. A single robust algorithm for on-line signaldetection is a crucial feature of any seismic monitor-ing system. In order to reduce incorrect time pick-ings a multi-algorithm approach for earthquakesidentification has been implemented in ASDP. Fur-

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88` 87

thermore, this program presents an efficient routineŽ .MSA for phase identification and association. Thesedesign features played a significant role in automaticearthquake detection and location.

In this study, a preliminary evaluation of resultsobtained by the use of a set of 330 local mi-croearthquakes recorded during 1997 at Mt. Etna

Ž .volcano Sicily, Southern Italy has been performed.Furthermore, to investigate the limitations and theperformance of ASDP we used another data setconsisting of about 40 ultra-microearthquakesrecorded by a temporary array composed of onlyfour digital stations, located in the Acri regionŽ .Calabria Arc, Southern Italy . The comparisons ofASDP automatic phase picking with manual onesindicate that for almost the totality of the events themulti-algorithm approach, in the automatic earth-quakes detection, optimises the picking estimationand is better than a data analysis performed using a

Ž .single algorithm e.g., XRTP-IASPEI . If three-com-ponent stations are available and the signal-to-noiseratio is favourable, such as in the Acri region, theautomatic polarization analysis implemented inASDP allows the precise identification of both P-and S-phase. In fact, the comparison between theback-azimuth values estimated by CMD and theazimuth directions obtained by particle motion anal-ysis and location procedure, indicates that the formerprovides very reliable estimates and weaker eventsdo not represent a fundamental limitation in theapplication of a polarization filter on 3-C records. Insome cases, unreliable computation using CMD

Ž . Ž .might depend on: i polarization site effects and iiincorrect orientation of the seismic sensors. Ourresults indicate that both ASDP automatic and man-ual hypocenter locations are comparable within theestimated errors both at Mt. Etna and in the Acriregion.

Therefore, ASDP can be useful for automaticdetection and monitoring of local earthquakes at asingle-station andror at seismic array, and it does sowith a high degree of reliability. Future tests andoptimisations will regard cases of sustained or fre-quently recurring signals, such as those produced byintense seismic swarms or by volcanic tremors. Weforesee that ASDP will be modified and subse-quently integrated with software modules for theanalysis of regional and teleseismic earthquakes. At

present, further developments in the ASDP moduleare in progress.

Acknowledgements

We thank T.B. Larsen and S. Falsaperla and theanonymous reviewer for helpful comments. Financialsupport from Gruppo Nazionale per la Vulcanologia-CNR is acknowledged.

References

Bache, T.C., Bratt, S.R., Wang, J., Fung, R.M., Kobryn, C.,Given, J.W., 1990. The intelligent monitoring system. Bull.Seismol. Soc. Am. 80, 1833–1851.

Bodvarsson, R., Rognvaldsson, S.Th., Jakobsdottir, S.S., Slunga,R., Stefansson, R., 1996. The SIL data acquisition and moni-toring system. Seismol. Res. Lett. 67, 35–46.

Calderoni, G., De Simoni, B., De Simoni, F.M., Merucci, L.,1997. Real-time monitoring of seismic data using satellitetelemetry. Annali di Geofisica 4, 799–810.

Cardaci, C., Coviello, M., Lombardo, G., Patane, G., Scarpa, R.,`1993. Seismic tomography of Etna volcano. J. Volcanol.Geotherm. Res. 56, 357–368.

Chouet, B.A., Dawson, P.B., Falsaperla, S., Privitera, E., 1994. Acharacterization of long-period events recorded during theeruptive activity of Mt. Etna, Italy, in 1992. Acta Vulcanolog-ica, Journal of the National Volcanic Group of Italy 4, 81–86.

Cichowicz, A.C., 1993. An automatic S-phase picker. Bull. Seis-mol. Soc. Am. 83, 180–189.

Eberhart-Phillips, D., 1986. Three-dimensional velocity structurein northern California coast ranges from inversion of localearthquakes. Bull. Seismol. Soc. Am. 76, 1025–1052.

Ferrucci, F., Patane, D., 1993. Seismic activity accompanying the`Ž .outbreak of the 1991–1993 eruption Mt. Etna Italy . J. Vol-

canol. Geotherm. Res. 57, 125–135.Guerra, I., 1991. Recenti osservazioni sismografiche e to-

Ž .pografiche in Calabria. In: Boschi, E., Dragoni, M. Eds. ,Aree Sismogenetiche e Rischio Sismico in Italia, Vol. II.Ettore Majorana Centre for Scientific Culture InternationalSchool of Solid Earth Geophysics, pp. 151–163.

Guerra, I., Corea, I., 1989. Sismicita della Calabria negli anni`1986–1989. Atti dell’88 Convegno Annuale del GruppoNazionale di Geofisica della Terra Solida, Rome, November7–9, pp. 41–52.

Haikin, L.M., Kushnir, A.F., Dainty, A.M., 1998. Combinedautomated and off-line computer processing system for seis-mic monitoring with small aperture arrays. Seismol. Res. Lett.69, 235–247.

Harris, M., Young, C., 1997. Matseis: a seismic GUI and tool-boxfor MATLAB. Seismol. Res. Lett. 68, 267–269.

( )D. Patane et al.rPhysics of the Earth and Planetary Interiors 113 1999 75–88`88

Hirn, A., Nercessian, A., Sapin, M., Ferrucci, F., Wittlinger, G.,1991. Seismic heterogeneity of Mt. Etna: structure and activ-ity. Geophys. J. Int. 105, 139–153.

Hole, J.A., 1992. Nonlinear high-resolution three-dimensionalseismic travel time tomography. J. Geophys. Res. 97, 6553–6562.

Ž .Lee, W.H.K., Lahr, J.C., 1975. HYPO71 revised : a computerprogram for determining hypocenter, magnitude and first mo-tion pattern of local earthquakes. US Geol. Surv. Open-FileRep. 75–311, 144.

Patane, D., Ferrari, F., 1997. SeismPol: a visual basic computer`program for interactive and automatic earthquake waveformanalysis. Computers and Geosciences 23, 1005–1012.

Patane, D., Ferrari, F., 1999. ASDP: a PC-based program using a`multi-algorithm approach for automatic detection and locationof local earthquakes. Physics of the Earth and PlanetaryInteriors, this volume.

Patane, D., Privitera, E., 1997. An analysis of fault plane solutions`at Mt. Etna volcano: kinematic implications. 29th GeneralAssembly of the IASPEI, August 18–28, Thessaloniki, Greece,Abstracts, p. 23.

Patane, D., Ferrucci, F., Gresta, S., 1994. Spectral features of`microearthquakes in volcanic areas: attenuation in the crustand amplitude response of the site at Mt. Etna, Italy. Bull.Seismol. Soc. Am. 84, 1842–1860.

Patane, D., Ferrucci, F., Giampiccolo, E., Scaramuzzino, L., 1997.`Source scaling of microearthquakes at Mt. Etna volcano and in

Ž .the Calabrian Arc southern Italy . Geophys. Res. Lett. 24,1879–1882.

Privitera, E., Montalto, A., Neri, G., Patane, D., Pellegrino, A.,`Scarpa, R., Spampinato, S., Torrisi, O., 1989. Il controllostrumentale dell’attivita sismica dell’Etna: la rete permanente`I.I.V. Proceedings of V Convegno Annuale Gruppo Nazionale

per la Vulcanologia, Consiglio Nazionale delle Ricerche, pp.491–508.

Privitera, E., Biviano, A., Cocina, O., Di Prima, S., Ferro, A.,Gambino, S., Marturano, M., Pellegrino, A., Rapisarda, S.,Spampinato, S., Velardita, S., 1993. Progetto di ristrut-turazione della rete sismica permanente delle isole Eolie eprimi risultati di un esperimento condotto tramite una retemobile digitale a Vulcano. Proceedings of XII ConvegnoAnnuale Gruppo Nazionale Geofisica della Terra Solida, pp.903–906.

Schwartz, S.J., Nelson, G.D., 1991. Loma Prieta aftershock relo-cation with S–P travel times: effects of 3-D structure and trueerror estimates. Bull. Seismol. Soc. Am. 81, 1705–1725.

Sharp, A.D.L., Davis, P.M., Gray, F., 1980. A low velocity zonebeneath Mt. Etna and magma storage. Nature 287, 587–591.

Thurber, C.H., 1983. Earthquake locations and three-dimensionalcrustal structure in the Coyote Lake area, central California. J.Geophys. Res. 88, 8226–8236.

Thurber, C.H., Atre, S.R., 1993. Three-dimensional VprVs varia-tions along the Loma Prieta rupture zone. Bull. Seismol. Soc.Am. 83, 717–736.

Ž .Tottingham, D.M., Lee, W.H.K., 1989. In: Lee, W.H.K. Ed. ,User manual of IASPEI Software library, Vols. 1, 2, 3, 4, 5:Toolbox for Seismic Data Acquisition, Processing, and Analy-sis. International Association of Seismology and Physics of theEarth’s Interior and the Seismological Society of America, ElCerrito, CA, pp. 89–118.

Tottingham, D.M., 1994. User manual for XRTP. In: Lee, W.H.K.Ž .Ed. , User manual of IASPEI Software library, Vol. 1, 2ndedn.: Realtime seismic data acquisition and processing. Inter-national Association of Seismology and Physics of the Earth’sInterior and the Seismological Society of America, El Cerrito,CA, pp. 207–254.