Microseismic feasibility study : detection of small magnitude events ( M L < 0 . 0 ) for mapping active faults in the Betic Cordillera ( Spain )

We present the results of the first application of the newly developed concept «Nanoseismic Monitoring» on active faults in the region close to Murcia, Spain. The aim of this microseismic feasibility study is to test if it is possible to record small magnitude events (ML<0.0) within a short period of time with surface installations and to investigate if these events are related to the regional catalog in terms of amount of events. The seismic monitoring was performed with one small array called the Seismic Navigating System. It consists of one central three component and three one component seismometers arranged tripartitely around the central station. In the measurement period of two nights at two different sites we were able to detect 19 microearthquakes down to ML = -2.6. The results correlate well with the frequency-magnitude distribution of the regional bulletin. This in turn will allow for estimation of monitoring rates before actual field measurements just from bulletin data. Given an activity rate of 5 to 10 events per night one may map active fault zones within just a few weeks of field campaign. Mailing address: Dr. Martin Häge, Institute for Geophysics, Universität Stuttgart, Azenbergstrasse 16, 70174 Stuttgart, Germany; e-mail: haege@geophys.uni-stuttgart.de


Introduction
Characterizing recent seismicity and mapping active fault segments must be based on the compilation of seismological bulletins.The fundamental data collection by semi-permanent seismic networks is a time-consuming and costly task.Only a few studies deal with the investigation of small magnitude events (ML<0.0)(e.g., Abercrombie, 1995;von Seggern et al., 2003;Ruiz et al., 2006), whereby most researchers in-vestigate aftershocks and not the background seismicity or use borehole sensors for event detection.Butler (2003) suggests the term «nanoearthquakes» for events with ML<0.0.
The concept of Nanoseismic Monitoring (Joswig, 2008), a technique developed to detect and characterize small magnitude sources, is tested as a short-term alternative for semi-permanent seismic networks to reduce network recording time.It is successfully tested for On-Site-Inspections of the Preparatory Commission for the Comprehensive Nuclear-Test-Ban Treaty Organization and was applied to detect and characterize small magnitude events triggered by material impacts in sinkholes along the western Dead Sea shores (Wust-Bloch and Joswig, 2006).Nanoseismic Monitoring as a kind of seismological microscope shall finally help to shed light on small earthquake trigger mechanisms.Fault weakness can be caused by increased fluid pressure that reduces the effective normal stress (Hainzl et al., 2006).In this model, stress and pore pressure redistribution after large earthquakes come along with higher permeability for fluids causing possible new nucleation points and a characteristic migration scheme (Cox, 1995;Miller et al., 2004).The detection and location of small magnitude seismicity may support this model of fluid transport and a shear stress behavior driven by porosity reduction (Johnson and McEvilly, 1995;Miller et al., 1996).
For the study, a seismically active section of the Betic Cordillera (near Murcia in Spain) with favorable signal-to-noise conditions was selected.Since we started to build up the system at that time one recording unit was available for field use.The small magnitude seismicity detected and partly localized by the single, smallaperture tripartite array within two successive monitoring nights, at two different sites, is compared with the regional 1984-2003 bulletin of the Instituto Geográfico National, Madrid (IGN).Brune and Allen (1967) have shown along the San Andreas Fault system that usually a two-day measurement of seismic activity is sufficient to make an approximate estimation about the local rate of microseismicity.

Geological and tectonic setting
The Betic Cordillera, which is situated in the southern part of Spain, is a collision zone generated by the nearby African-Eurasian plate boundary.This boundary is defined by a high seismicity which is distributed over several hundreds of kilometers (Calvert et al., 2000).The area selected for the feasibility study lies within the Subbetic Zone which, together with the Prebetic Zone, represents the External Zone of the Betic Cordillera.The thickness of the crust beneath the Betic is 25-39 km (Banda et al., 1993).Focal depths of regional events is restricted to the top 40 km (fig. 1) with moderate magnitudes generally less than 5.5 (Buforn et  1984-2003(Source: IGN, 2004).The two measuring sites (Capres and Burete) are indicated by triangles pointing up (not in scale).Our located events are marked by black dots and the two co-detected events by triangles pointing down, connected with lines.The dashed black line sketches the trace of the Crevillente Fault Zone (CFZ).al., 2004).The shallow seismicity is associated with the dense spread of fractures in this region (Sanz de Galdeano et al., 1995).
Nanoseismic Monitoring was carried out at two different locations in the vicinity of the Crevillente Fault Zone (CFZ): one near Capres, at the fringe of the Fortuna basin; the other near Burete, in the east of the Sierra de Espuna (fig.1).The CFZ strikes NE-SW parallel to the axis of the Subbetic Zone and extends laterally over 600 km.The main activity of the CFZ was during Late Miocene (Alfaro et al., 2002), and it is still active.Focal mechanisms indicate that the CFZ is a right-lateral strike slip fault which can also be observed on geological features (Buforn et al., 1988).Estimates for the total displacement along the CFZ range between 75-100 km (Nieto and Rey, 2004) and 400 km (de Smet, 1984).
The Subbetic Zone consists of deposits situated far from the South Iberian Margin (Ruano et al., 2004) which comprise parautochthonous to allochthonous, non-metamorphic sediments.
The main geological units in these areas beside Quaternary unconsolidated sediments and Miocene alluvial fan deposits include Triassic marls, claystones, gypsum, dolomites as well as Jurassic and Cretaceous limestones and marls.To consider potential site effects, the thickness of sediments to the top of the basement at each of the two monitoring sites has been estimated on the basis of available data and field evidence: near Capres the thickness of sediments reaches 100 m (IGN, 1972a), near Burete it is about 50 m (IGN, 1972b;Poisson and Lukowski, 1990).

Data acquisition and processing
Data was acquired by one Seismic Navigation System (SNS).This six-channel SNS is a portable, sparse array consisting of four shortperiod sensors: a central, three-component instrument and three one-component seismometers arranged as a tripartite array.High resolution and coherency of microseismic events are attained by utilizing a small aperture of 200 m and 400 Hz sampling rate.Nanoseismic Monitoring was performed at night to reduce the ef-fect of anthropogenic noise sources.Event detection and location were carried out by Son-oDet and HypoLine modules of the SparseNet software (Joswig, 1999;2008).
Measurements were performed during four nights.Due to the first operation of the system some adaptations to the equipment had to be made in the field.Two of the four nights were successfully completed and were taken for analysis.During this period, a total of 19 seismic events in the magnitude range -2.6 Յ M L Յ 1.5 could be detected and discriminated from noise bursts by sonogram analysis (Joswig, 1990) (see table I). Figure 2 shows two examples of table I, events nos. 5 and 11, demonstrating the usefulness of sonograms for event detection.Hypocentral locations could be estimated for 15 events.Four other weak events did not present clear P-and S-phase onsets.However, M L magnitudes could be estimated for all 19 events as maximum amplitude and distance from ts-tp time differences could be estimated with confidence by sonogram analysis (Catalog A of table II).
The apparent velocities of most of the events, derived from array analysis, were not in accordance with much faster velocities of the standard velocity model for Spain (Dãnobeitia et al., 1998).Location residuals reduced significantly using a data-adapted half-space model with velocities ranging from vP = 1 to 5 km/s for 0.3 to 14 km depth.
Figure 1 shows the location of the 15 events (solid black dots) on the background of the regional seismicity from 1984-2003 (open gray circles).Both positions of the SNS deployments are marked with triangles pointing up (not in scale).Most of the events south of Burete are aftershocks and were generated by the 2002 Bullas (ML = 5.0) earthquake (IGN, 2004) that occurred 607 days before our measurement.Event locations were calculated with one single array which results in a large location error of a few kilometers.Depths were estimated with the intersection of hyperboloids by tP-tP information.Two events were co-detected by the local network (nos.3 and 12 in table I) which are displayed as triangles pointing down in fig. 1.For these two events, the mean horizontal location difference is about 20 km and the mean magnitude variation 0.4.Figure 3 plots magnitudes versus distances for all 19 events and the distance-correction curve used for ML calculation fitted empirically to the data (note: slantdistance instead of epicentral distance for <10 km).Additionally, the two co-detected events are shown with triangles pointing down.
The observed detection threshold indicates that signal-to-noise conditions were better at Burete than at Capres.The overall sensitivity limit was about ML = -1.0 at 10 km, and ML = -2.0 at 2.5 km.

Characterization of regional seismicity
The 1984-2003 regional seismic catalog (IGN, 2004) was used to assess the performance of our measurement.Magnitudes of both datasets, are equivalent.By fig.3, and as will be shown later in fig.5, our magnitude of completeness is ML = -1.0 in about 10 km distance.However, the analysis of regional seismicity within a 10 km radius presents several challenges.
Artifacts like modifications in network geometry and density, station hardware and processing software result in a rather heterogeneous distribution of seismicity, both in space (fig.1), and in time.
Figure 4 shows Catalog C of table II with 1040 events as open gray circles (right vertical axis) and the annual event frequency by the black curve (left vertical axis).
The annual event frequency and the detection level increase with time.Numerous investigations deal with catalog completeness and its statistical fluctuations (e.g., Rydelek and Sacks, 1989;Zúñiga and Wiemer, 1999;Woessner and Wiemer, 2005).
The first change took place in 1997 when a digital recording system was installed; the second jump occurred after 2000 when another two stations (ETOB and EMUR) were added.An additional increase in number of events is caused by the aftershock activity of the 2002 Bullas M L = 5.0 earthquake.Between 1984 and 1998, there is a constant magnitude of completeness (MC) of 2.6, calculated with the entire-magnitude-range method (Woessner and Wiemer, 2005).It has decreased since 1998.Further investigations have shown that there is no catalog contamination by quarry blasts.

Gutenberg-Richter relationship: regional seismicity and Nanoseismic Monitoring
We make two assumptions in order to compare the Gutenberg-Richter relationship estimated from the regional seismicity (Catalog C of table II) with the same relationship based on nanoseismic data (Catalog A of table II).First, Catalog C characterizes a representative b-value for the whole region.This assumption is rooted in fig. 4 which shows an average constant seismicity without any high seismicity cycles above MC = 2.6.Second, Catalog A can be obtained by concatenating the data recorded over two nights near Capres and near Burete without loss of statistical significance.Figure 5 compares the frequency-magnitude relationships for the different catalogs of table II.Note that Catalogs B and C of table II are downscaled to the measurement period of this feasibility study of two nights.Due to the small number of events of Catalogs A and B the bvalue for the larger area, Catalog C, was determined and applied to our study area, Catalog A. Analyzing Catalog C (black dots) a b-value (solid black line) of 1.16±0.07was derived according to the formula of Aki (1965) with MC =2.6.This result corroborates the b-value of 1.1±0.1 estimated by López Casado et al. (1995) for magnitudes larger than 3.5 in the Murcia region between 1930-1992.In hazard assessment, the Gutenberg-Richter relation, log N = a -bM, (5.1) with N the cumulative number of earthquakes of magnitude M or greater and a and b constants (Ishimoto and Iida, 1939;Gutenberg and Richter, 1944), is used to predict the frequency of occurrence of large earthquakes on the basis of smaller events.Inversely, few studies investigate the extrapolations made from the stronger to the weaker events.Studies concerned with the detection of very small earthquakes (e.g., Iio, 1991;Piccinini et al., 2003) or with the constancy of the b-value to lower magnitudes and self-similarity of seismic events (von Seggern et al., 2003) failed to show a b-value decrease towards small magnitudes.Abercrombie and Brune (1994) verified that there was no significant decrease down to magnitude 0.0 on three major fault zones in California.Furthermore there is good agreement between b-values extrapolated from regional bulletins and those of microseismic activity (Abercrombie, 1995;1996).Although an investigation of induced seismicity (Trifu et al., 1993) showed a non-similar frequency-magnitude distribution between magnitudes -0.5 and 0.0, there is no reason why a decrease in b-value should be expected for natural seismicity at local distance.
In conclusion, the b-value of 1.16 from Catalog B was extrapolated to small magnitudes.Although it is not possible to calculate a b-value for our events and the existence of high uncertainties in the statistical estimation, i.e. the normalization of the frequency of events to the measurement period of two nights, the extrapolated b-value fits remarkably well with Catalog A. There might also be a slight shift in magnitudes due to the different applied velocity models which are used for location and the resulting variation in epicentral distances.However, the agreement of Catalog B with A is obvious, even with an assumed maximum magnitude deviation of 0.4.This result suggests that it might be possible to infer on the basis of bulletin data on the expected amount of events in a certain area prior to a field campaign.A similar observation was made by Brune and Allen (1967) who found out that the amount of microseismicity could be approximately predicted by extrapolation of frequency-magnitude curves from 29year records of larger earthquakes.

Conclusions
In this feasibility study, the new concept of Nanoseismic Monitoring was applied to characterize small magnitude natural seismicity in the Betic Cordillera (Spain).A total of 19 events (-2.6 Յ M L Յ 1.5) were recorded and detected within an observation period of two nights, indicating the high sensitivity of Nanoseismic Monitoring.The analysis of the frequency-magnitude distributions shows a good approximation between the amounts of recorded events with those extracted from local catalogs.However, it must be further proven by investigations in other geological and tectonic settings if the amount of small magnitude seismicity can be anticipated from existing catalogs by linear extrapolation.
The performance of Nanoseismic Monitoring of about ML = -1.0 at 10 km and ML = -2.0 at 2.5 km demonstrates its potential for a costeffective technique for active fault mapping.Fault mapping could be realized at high resolution within weeks instead of years.Due to the use of only one small array, there is a high location error of a few kilometers.A further challenge was the discrepancy between the standard velocity model for Spain and our observations.Therefore it was not possible to identify specific fault segments.At least two small arrays must be deployed to reach this aim.Both small arrays can be combined as a kind of network to reduce the azimuthal gap and hence to increase the location accuracy.Cross bearing and the intersection of two t s-tp circles provide additional location constraints and support the determination of an appropriate velocity model.It can then be tested if relative location methods might be applicable for further improvement of the location results.

Fig. 1 .
Fig. 1.Spatial distribution of 1040 earthquakes from 1984-2003 (Source: IGN, 2004).The two measuring sites (Capres and Burete) are indicated by triangles pointing up (not in scale).Our located events are marked by black dots and the two co-detected events by triangles pointing down, connected with lines.The dashed black line sketches the trace of the Crevillente Fault Zone (CFZ).

Fig. 2 .
Fig.2.Seismograms and the corresponding sonograms of the four vertical components for two events.Seismograms are filtered between 3 and 30 Hz (optimized filter setting).P and S onsets are indicated with arrows.The event in a) corresponds to no. 5, the event in b) to no. 11 in table I.

Fig. 3 .
Fig. 3. Magnitude-distance relationship for the 19 events detected at Burete and at Capres as well as the distance correction for ML.The two co-detected events are shown with triangles pointing down, linked to the corresponding events by gray lines.

Fig. 5 .
Fig. 5. Frequency-magnitude distributions of Catalogs A, B and C of table II.The cumulative distributions are marked in black, the normal distributions in grey.Note that Catalogs B and C are normalized to the measurement period of two nights.The solid black line shows the b-value for Catalog C that is applied to Catalog B (dashed black line).

Fig. 4 .
Fig. 4. Time-event (left vertical axis) and time-magnitude distribution (right vertical axis) of Catalog C of table II.

Table I .
Parameters of the recorded events.