Observatories and Research Facilities for EUropean Seismology
Volume 2, no 1 March 2000 Orfeus Newsletter

Surface-wave Centroid Moment Tensors in the Mediterranean region: the MEDNET-Harvard project

A. Morelli1, G. Ekström2, S. Mazza1, S. Pondrelli1, E. Boschi1, A. M. Dziewonski2
1 Istituto Nazionale di Geofisica (ING), Via Vigna Murata 605, 00143 Roma, Italy
2 Department of Earth and Planetary Sciences, Harvard University, 20 Oxford St. Cambridge, MA02138, USA

Introduction - Method and data - Recent Mediterranean seismicity - Central Italy sequence - Pre-1977 earthquakes - Conclusion - References


The Alpine Mediterranean region is characterized by a complex active tectonic environment, connected to the convergent motion between Africa and Eurasia, but also to complicated interaction of several microplates, producing a wide range of tectonic regimes. Seismicity is rather diffuse, and characterized by mostly moderate energy release. Earthquake focal mechanisms can contribute considerably to the understanding of the active tectonics. The Centroid Moment Tensor (CMT) method (Dziewonski et al., 1981; Dziewonski and Woodhouse, 1983;) has shown to be one of the most robust and reliable ways for computing focal mechanisms. Being based on long period seismograms, it reconstructs the average characters of the entire fracturing process of an earthquake, and the centroid of moment release. The CMT Catalog - spanning the period from 1977 to the present - is routinely updated by the Harvard group for earthquakes distributed globally, and with magnitude approximately above 5.2. The uniqueness and reliability of the Harvard Catalog is testified to by its wide use for tectonic studies (stress maps, global plate motion models and cumulative moment tensors studies; e.g. Pondrelli et al., 1995) for which it constitutes a reference.

In the Mediterranean area, however, moderate-energy (4.5 < Mw < 5.5) seismicity is particularly important because it is widely spread, and more common than the relatively infrequent larger-magnitude events. Small or moderate earthquakes are impossible to model at teleseismic distance with the classical CMT method, owing to the low signal to noise ratio of the long period body waves used. We resort, then, to modelling surface waves, which exhibit much higher amplitudes, in a modified CMT algorithm. Besides showing prominently on seismograms, surface waves can also be modelled at closer distance, thereby further decreasing the magnitude threshold of the analysis when seismographs are appropriately available at local and regional distance. The method uses detailed surface wave phase velocity maps (Ekström et al., 1997) and is described in Arvidsson and Ekström (1998). Here we briefly review some aspects of the implementation, show some recent applications, and discuss our plans for future activity.

Method and data

Regional CMTs (RCMTs) are computed with a modification of the standard CMT algorithm to deal with smaller magnitude events (Arvidsson and Ekström, 1998; Ekström et al., 1998). This can be accomplished by using intermediate period surface waves recorded at shorter epicentral distance. While the Harvard centroid moment tensor method fits seismograms in two frequency bands (long period body waves, T>45s, and, for large earthquakes, mantle waves with T>135s) we model Love and Rayleigh surface waves after low-pass filtering with a cut-off at 35 to 45 seconds. At close distance, the seismogram is not yet dispersed (Figure 1) and is dominated by the fundamental mode of surface waves. Fundamental mode synthetic seismograms are computed by excitation in PREM and propagation through the phase velocity maps by Ekström et al. (1997). Overtones are calculated by normal mode summation in a 3D mantle model (S20U7L5, Ekström and Dziewonski, 1995). Figure 1 shows the fit between observed and synthetic seismograms at different distances, ranging from local to teleseismic.

Click to enlarge

Figure 1. Waveform fit for the surface wave Centroid Moment Tensors. Left panel shows local and regional seismograms (epicenter to station distance ranging from 0.8 to 14.4 degrees) for the smallest of the events analyzed (Mw=4.2). Right panel shows seismograms at larger distances (13.4 to 16.4 degrees) where dispersion of surface waves allows to discriminate between the dominant fundamental mode and overtones (from Morelli et al., 1999). 

We use data from available stations at local, regional, and teleseismic distance. Earthquakes with magnitude 5.5 and above are most conveniently recorded and modelled at a global scale, with the standard CMT technique, and are routinely analyzed at Harvard. We are instead normally interested in events smaller than 5.5. Our magnitude threshold may reach 4.2 in the best instrumented areas, but varies depending on data availability.

The surface wave regional centroid moment tensor calculation is very fast. Its speed allows rapid calculation of source mechanisms, a feature of great importance for scientific and relief operations following an earthquake. For this reason, we also analyze strong earthquakes in a rapid manner, depending on the quasi-real time availability of data from a number of seismographic stations. For the determination of rapid RCMTs we rely on data recorded at MedNet (Mediterranean Network) stations accessible by telephone dial-up or the Internet. Long period seismograms are automatically extracted in nearly-real time by the MUSCLES system (Mednet Unmanned Stations CalLer for Extraction of  Seismograms), by calling MedNet stations in the occurrence of a seismic event (Mazza et al., 1998). MUSCLES is launched by an e-mail reporting an earthquake, and it is based on five unix shell scripts running independently from each other every minute. When available, we also use data from other seismographic stations, reachable through the ORFEUS or IRIS Spyder® systems.; We estimate the availability of local and regional data generally sufficient to grant approximate completeness for events with magnitude equal or greater than 4.5.

Recent Mediterranean seismicity

Figure 2 presents moment tensor solutions that we obtained for the years 1997 and 1998. Solutions for 1999 will be presented elsewhere, together with a more complete analysis and discussion of the method. We show 73 solutions for the period 1997-1998. The map also includes solutions for some events that occurred between 1977 and 1995, plotted in blue. The benefit of the RCMT technique for smaller magnitudes is clear considering that, for this time period, the number of moment tensors is more than tripled with respect to what was achieved by the standard, global analysis.

Click to enlarge

Figure 2: Regional moment tensor solutions for years 1997 and 1998 (Pondrelli et al., 1998). The map also shows, in blue, RCMTs for some older events. The map shows 73 solutions, in red, in the 2-year period, with magnitude between 4.5 and 5.5, which had not been previously analyzed on a global scale.

The Central Italy sequence of 1997-98: a case study

The 1997-98 Central Italy earthquake sequence provided a particularly important test case for our RCMT technique. The automatic data retrieval system MUSCLES was operational, and could rely on several seismographic stations conveniently located in Italy and surrounding areas. During the crisis, we routinely computed RCMTs in a rapid fashion, usually after one or few hours of event occurrence, for 20 events of the sequence with moment magnitude ranging from 4.2 to 6.0 (Figure 3). Provided that they cover a fair azimuthal range, as few as 3 stations at distances of the order of 100's of kilometers proved to be sufficient. The RCMT procedure contributed valuable information for the timely study of the complex process of stress transfer to different fault segments that marked the unusual time evolution of the sequence. Details can be found in Ekström et al, 1998, and Morelli et al., 1999.

Click to enlarge

Figure 3: Regional moment tensor solutions for events of the 1997-98 Central Italy earthquake sequence (Ekstrom et al, 1998; Morelli et al, 1999). Labels indicate month and day of occurrence of events. The radii of the focal spheres are proportional to the moment magnitude. Compressional quandrants are shaded in different colors to distinguish between 1997 and 1998 events.

Being based on a modification of the CMT scheme, our RCMT solutions have similar characters and, for earthquakes whose size allows both to be computed, the two procedures yield compatible and very similar results, as shown in Figure 4. Note that, for each event, the two solutions shown have been computed with different sets of stations, at different distance ranges, and modelling different parts of the seismogram. The level of agreement is also significant as an empirical estimate of stability to possible bias in the data or due to simplifying assumptions in the theory.

Figure 4: Comparison of regional CMTs (top) and standard Harvard CMTs (bottom) for the largest events of the 1997-98 Central Italy earthquake sequence. Full moment tensors are plotted by the red areas, thin lines show best fitting double couple mechanisms. Labels identify event dates (from Morelli et al, 1999).

Pre-1977 earthquakes

1977 marked the beginning of operation of the digital standardized global network GDSN, and the beginning of convenient availability of digital seismograms. Starting from the present, we plan to cover this whole period going backwards in time. However, reliable seismograms exist for earlier times, and their careful analysis can provide valuable information. We plan to examine special cases of older events and sequences of particular interest. Figure 5 shows an important example. HGLP and GDSN data were retrieved to study the 1976-77 Northeastern Italy (Friuli) earthquake sequence (Pondrelli et al., 1999), known for its complexity and characterized by large aftershocks.

Figure 5: Centroid moment tensor solutions for earthquakes of the 1976-1977 Northern Italy seismic sequence (after Pondrelli et al., 1999). PO is the Periadriatic overthrust, SF the Sequals fault, and TF the Tricesimo thrust fault. Event locations from Piromallo and Morelli (1998). The size of focal mechanisms is proportional to moment magnitude.


One of our goals is to provide moment tensor solutions quickly after significant earthquakes of the Mediterranean region. The moment tensor calculation is not automatic, and needs control by an operator. Besides prompt availability of an operator, two other intrinsic requirements limit the promptness by which a solution can be computed after the occurrence of an earthquake: an initial estimate of the epicenter, and about 30 minutes of long period seismograms. Independent detection and location of an earthquake is, of course, a prerequisite also to start the data downloading process. Large (M>5.5) events generally are not our targets, as their focal mechanisms are routinely computed at Harvard. For strong earthquakes, our RCMT implementation is in fact equivalent - both in terms of time required, and results -  to the Harvard CMTs, also calculated in a quick-response routine. For moderate magnitudes, instead, our solutions extend the more established CMTs. In occasions holding special interest - such as the Central Italy earthquake sequence, or other seismic crises - we are committed to providing rapid information to the civil defense and scientific communities with response times of one or a few hours.

All RCMT determinations, once revised and improved by modelling of off-line data retrieved through ORFEUS or the IRIS DMC, are collected in a regional catalog of moment tensors. The catalog contains considerably more solutions, as nearly-real time data are only available for larger magnitude events and for selected seismographic stations. The catalog is being organized for publication, and will be regularly updated. The Mediterranean regional catalog of seismic moment tensors will shortly be also hosted on a web site, now under construction.


Arvidsson, R., and Ekström, G., 1998, Global CMT Analysis of  Moderate Earthquakes Mw>4.5, using intermediate period surface waves, Bull. Seism. Soc. Am., 88.
Dziewonski, A. M., Chou, T.-A., and Woodhouse, J. H., 1981,Determination of  earthquake source parameters from waveform data for studies of global and  regional seismicity, J. Geophys. Res., 86, 2825-2852.
Dziewonski, A. M. and Woodhouse, J. H., 1983, An experiment in the systematic study of global seismicity; Centroid moment tensor solutions for 201 moderate and large earthquakes of 1981, J. Geophys. Res., 88, 3247-3271.
Ekström, G., Tromp, J. and Larson, E. W. F., 1997, Measurements and global models of surface wave propagation, J. Geophys. Res., 102, 8137-8158.
Ekström, G. ; Morelli, A. ; Boschi, E. ; Dziewonski, A. M., 1998, Moment tensor analysis of the central Italy earthquake sequence of September-October 1997, Geophys. Res. Lett., 25, 1971-1974
Mazza, S., Morelli, A., Boschi E., 1998, Near real-time data collection and processing at MEDNET, EOS Trans. Am. Geophys. U., 79, 569.
Morelli, A., Ekström, G., and Olivieri M., 1999, Source properties of the 1997-98 Central Italy earthquake sequence from inversion of  long-period and broad-band seismograms, Journal of Seismology, in press.
Piromallo, C., and A. Morelli, 1998, P-wave propagation heterogeneity and event location in the Mediterranean region, Geophys. J. Int., 135, 232-254.
Pondrelli S., Morelli A., and Boschi E., 1995. Seismic deformation in the Mediterranean area estimated by moment tensor summation. Geophys. J. Int., 122, 938-952.
Pondrelli, S., E. Boschi, A. M. Dziewonski, G. Ekström, S. Mazza, A. Morelli, and C. Piromallo, 1998, Regional Centroid Moment Tensors of the Mediterranean area and their tectonic implications, IUGG99 Abtract Book, A, 169.
Pondrelli, S., Ekström, G., and Morelli, A., 1999, Seismotectonic re-evaluation of the 1976 Friuli, Italy, seismic sequence, Journal of Seismology, in press.

page 4
Copyright © 2000. Orfeus. All rights reserved.