|Observatories and Research Facilities for EUropean Seismology|
|Volume 2, no 1||March 2000||Orfeus Newsletter|
Regional Moment-Tensor Inversion in the European-Mediterranean AreaJ. Braunmiller, U. Kradolfer, M. Baer, D. Giardini
At the Swiss Seismological Service, we incorporate these new data to our routine analysis procedures. Our goal is two fold. First, we rapidly determine earthquake source parameters of prominent local and regional events for dissemination to scientists and to the public. Second, we want to provide a regional moment tensor catalog utilizing near real-time and additional, later available data. In this letter, we illustrate our efforts towards rapid moment tensor determination and towards building a moment tensor catalog.
The method uses the entire three component waveforms (body and surface waves) and inverts for the earthquake moment tensor by minimizing the misfit between observed and synthetic seismograms in a least-squares sense. Synthetic seismograms are calculated with a frequency-wavenumber algorithm (Bouchon, 1982).
Data sources for rapid moment tensor determination are the Swiss Digital Seismograph Network (currently 20 stations are running in Switzerland), the ORFEUS and IRIS data centers (where event based data are available for larger earthquakes), data available via AutoDRM (e.g., data from stations in the area of the former Soviet Union available from the USGS), and data from station TRI (Trieste, Italy) available via telnet. Data from the German Regional Seismic Network and from some stations of the Geofon network become available within one day after an event.
Figure 1. Star: Epicentral region of the Izmit and Duzce earthquakes. Circles: Seismic stations used for analysis of the seven Turkey events (see Figure 3); red and black: used for more than, respectively, less than half of the events.
For the Duzce earthquake, broadband data from the Swiss stations were available immediately after the earthquake and additional data from several broadband stations in the European-Mediterranean region could be accessed through the ORFEUS data center within a few hours after the event. We extracted and processed the data; processing consists of windowing, filtering and removal of the instrument response. At the same time, we calculated synthetic seismograms for a source depth of 12 km and a PREM crust-mantle velocity-depth model. We then inverted the displacement seismograms in the 40-125 s pass-band for the source parameters. Our solution was distributed locally and posted on our web site about four hours after the earthquake. Figure 2 shows the waveform fit and the source mechanism of the quick moment tensor solution. We repeated the analysis when additional data became available through ORFEUS; the resulting source parameters are almost identical (compare the fault plane solutions of the Duzce main shock in Figures 2 and 3) illustrating the stability of the waveform inversion procedure.
Figure 2. Observed (solid) and synthetic (dashed) seismograms for the Duzce main shock quick moment tensor solution. Seismogram amplitudes are normalized to 100 km epicentral distance assuming cylindrical geometric spreading. Stations are listed in azimuthal order; numbers beneath station codes are event-station azimuth and distance. Z, R, and T are the vertical, radial, and transverse components. Triangles on the fault plane solution (lower hemisphere projection) depict the station coverage.
Similar analyses were performed for the Izmit main shock and for five large aftershocks. Figure 3 shows a map of the epicentral region with the source mechanisms and the aftershock activity (obtained from the USGS). Color-coding of the mechanisms and of the epicenters illustrates the spatio-temporal behaviour of the earthquake sequence. Waveform fits and details of the the source mechanisms can be found at the ETH info web page.
Figure 3. Epicentral region of the Izmit and Duzce earthquakes. Fault plane solutions are from this study and epicenters are from the USGS; size is proportional to magnitude.
A strong limiting factor for rapid source parameter determination is data availability. Figure 1 shows the stations used for one or several of the Turkey earthquakes. Almost all stations are located far more than 1000 km from the epicentral area, thus limiting the analysis to larger events (about magnitude 5 and larger). In addition, the station distribution is uneven: a few, relatively close stations to the south and east, and most stations, at greater distances to the northwest. For the smallest event analyzed (Mw = 5.0), neither ORFEUS nor IRIS had collected any data. We thus had to wait until data from the Geofon network started to become available one day after the earthquake. Another limiting factor for source parameter retrieval are structural complexities along the long event-station travel paths and the heterogeneity of the paths. The simple one dimensional crust-mantle structure (PREM) used for all stations is an oversimplification. To obtain proper phase alignment and an adequate fit between observed and synthetic seismograms, we had to invert the data at relatively long periods (T > 40 s). Resorting to long period data also limits analysis to larger events. Additional quickly available data from close-by, azimuthally well distributed stations would lower the magnitude threshold for quick moment tensor analysis considerably and would improve source parameter resolution.
Figure 4. Map of the broadband Swiss Digital Seismograph Network and fault plane solutions for the February, 14 1999 Fribourg (red), the May, 20 1999 Boltigen, and the December, 29 and 31 1999 Bormio earthquakes. Stations in operation during the Fribourg event are shown as pink squares; additional stations which became operational meanwhile are shown in black.
All available stations are located less than 200 km from the epicenter. Despite the complex Alpine crustal structure, we could model the data at relatively high frequencies (0.03-0.1 Hz). The waveform fit is shown in Figure 5. The fault plane solution from the moment tensor analysis agrees very well with the first motion data (Deichmann pers. comm., 1999). Compared to the lower frequency data used for the Turkish earthquakes, the higher frequency data are more sensitive to depth variations; performing a grid search we found a best fitting source depth of 4 km.
Figure 5. Observed (solid) and synthetic (dashed) seismograms for the Fribourg earthquake. See Figure 2 for further details.
Moment tensor solutions for three additional earthquakes in or close to Switzerland were obtained during 1999. The black stars in Figure 4 show the locations. The smallest analyzed event had a Mw of 3.5. The moment tensor solutions, in all cases, were stable over a wide, upper crustal depth range and were not very sensitive to the chosen frequency band as long as the band contained significant signal energy.Harvard CMT project or the USGS). The catalog should include future events as well as older events for which waveform data are available from different archives (e.g., IRIS, ORFEUS, Geoscope, Geofon, GRSN). The source mechanisms of the moderate sized events will improve, in particular, our understanding of the tectonics of the central and western part of the Africa- Eurasia plate boundary. There, large earthquakes occur relatively seldom due to the low plate motion rate along the plate boundary's western part (Jackson and McKenzie, 1988) and a detailed seismotectonic analysis requires inclusion of the more frequent moderate sized events.
As an example for a moderate event where no fault plane solution existed before, we show fault plane solutions obtained by regional waveform inversion for the Balearic Sea earthquake (mb = 5.0) of September, 24 1994 (Figure 6 left side). The strike-slip solution obtained with Nabelek and Xia's (1995) method, shown on the left, compares well with the solution obtained in a pilot study by Sicilia (1999) who used the inversion code of Giardini et al. (1993).
Figure 6. Left: Fault plane solutions for the September, 24 1994 Balearic Sea earthquake. Right: Fault plane solutions for the September, 5 1996 Adriatic Sea earthquake. Elevation scale same as for Figure 4. See text for details.
The right hand side of Figure 6 shows fault plane solutions for the September 5, 1996 Adriatic Sea earthquake. The Harvard-CMT solution is shown in gray, while the upper left depicts the solution obtained with the Nabelek-Xia code and the upper right the solution obtained by Sicilia (1999). The differences between the solutions for this thrust event are small.
The examples show that we can reliably determine source mechanisms and indicate that we can analyze much smaller events in the Mediterranean region than possible with teleseismic techniques.ORFEUS starts to extract data for these smaller events or when more national networks provide near real-time access to their broadband data (e.g., via AutoDRM).
Building a moment tensor catalog of the European-Mediterranean region that includes events, which are too small for teleseismic analysis is possible using data from various waveform archives. The solution for the Mw = 4.8 1994 Balearic Sea event was obtained mainly with data recorded at epicentral distances of around 1000 km. We expect similar distances for most other earthquakes along the Africa-Eurasia plate boundary. A question we need to investigate is whether we can analyze even smaller events routinely and, if so, how much smaller we can go.
Several aspects of the regional waveform inversion scheme need to be improved. Currently, the inversion code does not run automatically. We are working on automating the data retrieval and processing procedures. We are also planning to implement several regional moment-tensor inversion codes to compare performances, to estimate parameter uncertainties, and to select the code most suitable for automatic analysis.
Travel paths in the European-Mediterranean region are complex and, unfortunately, also relatively long. Currently we match long period data with synthetics calculated for the PREM Earth model. The mismatches are large and erroneous structure prohibits a reliable estimate of the earthquake centroid depths and probably affects the solution quality. Using lower frequencies also restricts the analysis to larger events. The MIDSEA project at the Swiss Seismological Service (van der Lee et al., 1999) is working on improved crust-mantle models for the Mediterranean region. We are closely cooperating and exchange results with our colleagues involved in the MIDSEA project. With better velocity-depth models we are able to analyze smaller events, can obtain reliable centroid depth estimates, and the source parameters overall will be better resolved.
Currently, near real-time waveform data are available from only a few national data centers in the Mediterranean area via AutoDRM (for an overview of available stations visit the waves4u, Kradolfer, 2000) or from stations directly connected to the internet. However, this situation will probably improve soon. The new "MEREDIAN" project headed by ORFEUS aims at improving data access by installing AutoDRM's at several national data centers. The international data center (IDC) of the Comprehensive Nuclear-Test-Ban Treaty Organization (CTBTO) in Vienna hopefully will become another important source for broadband seismic data in the future; due to the requirements of the CTB Treaty, IDC data should have a high availability and should be available as continuous data.
Braunmiller, J., 1998, Seismotectonics of the Explorer region and of the Blanco transform fault zone, PhD thesis, Oregon State University, Corvallis.
Braunmiller, J., B. Leitner, J. Nabelek, and A. Qamar, 1995, The 1993 Klamath Falls, Oregon, earthquake sequence: Source mechanisms from regional data, Geophys. Res. Lett., 22, 105-108.
Giardini, D., E. Boschi, and B. Palombo, 1993, Moment tensor inversion from MedNet data (2) regional earthquake of the Mediterranean, Geophys. Res. Lett., 20, 273-276.
Jackson, J., and D. McKenzie, The relationship between plate motions and seismic moment tensors, and the rates of active deformation in the Mediterranean and Middle East, Geophys. J., 93, 45-73.
Kradolfer, U., 2000, Waves4U: Waveform Availability Through AutoDRMs, Seis. Res. Lett., 70, 79-82.
Nabelek, J., and G. Xia, 1995, Moment-tensor analysis using regional data: application to the 25 March, 1993, Scotts Mills, Oregon, earthquake, Geophys. Res. Lett., 22, 13-16.
Sicilia, D., 1999, Towards regional moment tensor determination: the moment tensor of the 1994 mb=4.8 event near the Balearic Islands and its uncertainties, Diploma thesis, ETH Zurich, Zurich.
Van der Lee, S., D. Giardini, C. Estabrook, A. Deschamps, and C. Chiarabba, 1999, New temporary broadband stations in the larger Mediterranean region, Orfeus Electronic Newsletter, 1, 5.