Observatories and Research Facilities for EUropean Seismology  
Volume 5, no 2  September 2003  Orfeus Newsletter 
Moment tensor determination for the IberoMaghrebian regionDaniel Stich, Jose Morales, Flor de Lis Mancilla and Gerardo AlguacílInstituto Andaluz de Geofísica, Universidad de Granada, Campus Universitario de Cartuja s/n, 18071 Granada, Spain Introduction  Moment tensor catalogue  Examples  Summary  Acknowledgements  ReferencesIntroductionThe crustal deformation in the Iberian Peninsula, northwestern Africa and the adjacent offshore areas is controlled by the AfricaEurasia plate collision and extensional processes in the western Mediterranean Basin and Alboran Sea. The regional seismicity is characterized by a diffuse geographical distribution and mainly low to moderate magnitudes. Only in the Atlantic Ocean and northern Algeria, seismicity appears to be focused around a linear plate boundary, and moderate to large earthquakes occur with certain frequency (ISC Online Bulletin). To date, moment tensor solutions for thirtyfour northern Algerian earthquakes are available from routine moment tensor projects on Mediterranean and global scale (like the Harvard CMT project, Dziewonski and Woodhouse, 1983, MedNet regional CMT, Pondrelli et al., 2002, and ETH European Mediterranean moment tensors, Braunmiller et al., 2002). They consistently show reverse to strikeslip faulting style and NWSE oriented Paxes, and can be related directly to the plate convergence between Africa and Iberia.Earthquakes are rarely exceeding magnitude 5 on the Iberian Peninsula, northern Morocco, the westernmost Mediterranean Sea and the Atlantic Ocean south of Portugal. Twentyfour events altogether are included in catalogues of European or global routine moment tensor projects. This small set of moment tensor solutions can not be expected to sample adequately the tectonic deformation over this large area with several geotectonic domains and complex neotectonic evolution. For a more detailed seismotectonic picture in and around Iberia, we apply fullwaveform analysis to obtain the source parameters of the more frequent smalltomoderate events. Analysis of smaller events requires the use of seismograms from the combined Iberian broadband networks and implies performing key parts of the data processing manually. Moment tensor inversion is then performed in a routine manner for all regional earthquakes with local magnitude >= 3.5 for the Iberian Peninsula, Northern Morocco and adjacent offshore areas, and local magnitude >= 4.0 for Northwestern Algeria. To date, our catalogue contains eightyfour moment tensor solutions for the Iberomaghrebian region (Figure 1), ranging in size from moment magnitudes 3.5 to 5.8 (http://www.ugr.es/~iag/tensor/, Stich et al., 2003a).
Figure 1. Moment tensor mechanisms available from the IAG moment tensor catalogue. Solutions are dense in the southeastern part of the Iberian Peninsula and in the Alboran Sea, and sample several other seismic zones over the IberoMaghrebian region as well.
Moment tensor catalogueWe invert for the deviatoric moment tensor by minimizing the least squares misfit between observed and predicted displacement waveforms at regional distances (Langston et al., 1982). Greens functions to predict displacement are obtained with a reflectivity code (Randall, 1994), parameterizing the lithosphere as a plane layered halfspace with average properties for regional travelpaths in the study area. Three different models are used for wave propagation predominately offshore, within Alpine domains, or within Hercynian domains (Stich et al., 2003a). Prior to inversion, the observed waveforms and Greens functions are filtered in an intermediatetolong period pass band, to correct adequately for propagation effects by restricting the inversion to wavelengths that average out smallscale lateral heterogeneities. Typical filter bands are 50 to 20s for events with Mw > 4, and 35 to 15s for events Mw <= 4. Absolute travel times for Pwaves are calibrated by aligning seismograms and synthetics at the first arrival. A unit step source time function is assumed when working with the regional long period waveforms of the small and moderate events (Mw < 6).Moment tensor inversion for small events in a geologically heterogeneous environment is not a trivial problem. The basic requirement for inversion is an appropriate set of broadband waveforms, possibly with good azimuthal coverage, and possibly including stations within nearregional distances and tectonically uniform travelpaths. Also, the inversion procedure cannot be fully automated. A crucial manual task is refining the natural distancedependent amplitude weighting until obtaining a stable inversion result and adecuate waveform matches. This is generally an iterative procedure. Criteria to downweight or exclude waveforms are high noise level or propagation along complex paths (where the Greens functions for plane layered media do not provide an appropriate correction). Traces may also be weighted to balance the station coverage over the focal sphere. Moment tensor estimates are usually sensitive to focal depth, and we combine the linear moment tensor inversion with a grid search over a range of depths to evaluate this nonlinear effect. Based on the L2misfitvs.depth curve and depth dependence of waveform matches, we select a most appropriate combination of depth and mechanism. Once a moment tensor solution is obtained, it is doublechecked by dislocation grid search modeling. This alternative way to invert magnitude and doublecouple focal mechanism serves as a resolution test: We compute waveforms for the full ranges of dislocation source orientations and depths, and then compare them with the observed seismograms to identify the range of valuable mechanisms. The fault angle parameters strike, dip and rake are sampled every 10\272. Those double couple sources that produce significantly larger L2misfits than the global minimum can be excluded (based on waveform matches, we consider 10% difference a conservative estimate). We usually observe equivalence of both, focal mechanism and quality of fit, for inversion and double couple gridsearch, supporting the interpretation in terms of double couple force systems. Nevertheless, the nondouble couple component (CLVD) included in the general deviatoric moment tensor may be helpful to absorb effects of uncorrected propagation, noise and finite sources. With our magnitude criteria, we selected hundredninety events since autumn 1995, when the permanent broadband networks probably reached a minimum standard for regional time domain inversion for magnitudes Mw >= 4. For seventyseven events, we obtained moment tensor solutions that adequately fit the main characteristics of the regional waveforms and passed the gridsearch resolution test. With seven solutions for events in the 1980s, using data from the temporary NARS experiment, this amounts to a catalogue of eightyfour moment tensors to date. Recent improvements of broadband network coverage contribute to higher success rates. For the 18 months period November 2001 to April 2003, moment tensor solutions were obtained for twentyfour earthquakes, i.e. ~50% of all events with local magnitude >= 3.5. This is noticeable exceeding the amount of eight solutions that provided routine moment tensor projects on the European scale for the same period. Recent moment tensors include multiple event solutions for the seismic sequences in August 2002 in Bulla, Murcia, SESpain (MW<=5.0), January 2003 in Zamora, NWSpain (MW<=4.2) and February 2003 at the Alboran Ridge (MW<=4.8). Currently, sufficient waveform data for the combined Iberian networks become available within weeks after an earthquake. Our inversion results, waveform fits, and gridsearch analysis are posted online at http://www.ugr.es/~iag/tensor/. Examples: Sevilla 2002 and Cordoba 2003Over most of the Iberian Peninsula, the moment tensor focal mechanism map is dominated by normal or strike slip faulting style, and ~NESW orientation of pseudo Taxes. Yet these characteristics do not apply to all mechanisms and several unexpected solutions appear. So far, this is observed mainly in southern Spain, however this may be an effect of better sampling due to higher seismicity. We discuss two selected mechanisms to illustrate heterogeneous faulting and the accuracy of waveform modeling under favorable conditions.Two small to moderate earthquakes (MW=4.1 and 4.2) occurred within 5 months at a distance of ~75 km from each other in Sevilla (September 15th 2002, 020915) and Cordoba (January 24th 2003, 030124) provinces in western Andalusia. For these epicentral locations we have good azimuthal station coverage, and we used a selection of nearregional broadband waveforms that are available through Geofon (MTE, SFS, CART), IRIS (PAB) and from several IAG stations (SELV, ANER, ARAC, SESP). The inverted moment tensor solutions match the waveforms well in a pass band from ~15 to 35s (Figure 2), except for effects of unmodeled local earth structure at SFS (event 020915) and ANER (event 030124). ANER is a nodal station for the Cordoba focal mechanism, as confirmed by small amplitudes in the vertical component, but waves with transverse polarization are projected onto the radial component because earth structure is not plane layered at this part of the coast. Large amplitudes at the horizontal components of SFS may be attributed to local amplification in the sedimentary environment (this has been observed for several events, Stich et al, 2003a). These traces have been excluded from inversion (weight zero), as well as the radial component of CART (event 030124) with high noise level. For both events, moment tensor solutions are well resolved according to dislocation gridsearch modeling.
The mechanisms are fundamentally different from each other: The Sevilla earthquake shows predominately reverse faulting with Paxis azimuth of N17°E, while the Cordoba earthquake is almost pure strikeslip, with Paxis orientation of N283°E, nearly perpendicular to the Sevilla event. Particularly for the Sevilla quake, results are different from the nearest moment tensor mechanisms at distances of about 100km for the Granada Basin, Gulf of Cadiz and Gibraltar regions, and also from the average trend for the Iberian Peninsula. The occurrence of focal mechanisms that are severely rotated relative to the regional reference orientation suggests heterogeneous tectonic stresses on a local scale, and points to fault interaction. SummaryWe apply time domain moment tensor inversion to full regional longperiod waveforms of earthquakes from the IberoMaghrebian region and obtained moment tensor solutions for eightyfour events to date. Seventy solutions for the Iberian Peninsula, westernmost Mediterranean Sea and easternmost Atlantic Ocean give a more complete picture of the tectonic deformation of the Iberian microplate, compared to previous moment tensor studies. Moment tensor mechanisms indicate a predominately normal faulting regime over most of the Iberian Peninsula, predominately strikeslip faulting in the Alboran Sea and predominately reverse faulting in northern Algeria. However, several odd solutions point to heterogeneity of faulting and we are still at the beginning of compiling a moment tensor catalogue that represents adequately IberoMaghrebian seismotectonics. To collect a larger set of solutions, we have to continuously add moment tensors for future small and moderate events, as well as try to recover and process 20th century, intermediateperiod, analogue recordings of moderate earthquakes. An application of time domain moment tensor inversion to the 1910, MW 6.1 earthquake near Adra, southern Spain, was successful (Stich et al, 2003b). Processing of presentday and future seismicity benefits from important recent improvements of several regional broadband networks. The fairly dense distribution of stations corresponds to the availability of sufficient high quality, nearregional broadband seismograms even for most of the smaller events.AcknowledgementsWe appreciate the collection and distribution of high quality broad band waveforms by other institutes and data centers, which are at present Geofon, IRIS, ORFEUS, Instituto Geográfico Nacional, Institut d'Estudis Catalans, Institut Cartográfic de Catalunya, Real Observatorio de la Armada, and Universidad Complutense de Madrid. For several past events we used data from the temporal NARS and MIDSEA stations. We are very grateful to Chuck Ammon for his assistance while getting this project started, to him and George Randall for providing program codes, and to the developers of free software SAC and GMT. We receive financial support by the Spanish DGI project REN200204198C0201, FEDER founds and within the Research Group RNM#104 of Junta de Andalucía.References


page 9  
