1 Introduction

Past earthquake activity in Switzerland and surrounding regions has been documented in a series of annual reports since 1879. A detailed overview on the history of past reports and studies covering different aspects of the recent seismicity of Switzerland is given e.g. by Diehl et al. (2014). The present report first summarizes the state of the seismic network and documents changes in its configuration during 2014. Then we give a short overview on the methods used for earthquake analysis. This is followed by a description of the seismic activity and significant earthquakes in 2014. Finally, we discuss the focal-depth distribution, the occurrence of earthquake sequences, and the unusually large number of earthquakes located in 2014.

2 Data acquisition and analysis

2.1 Seismic stations in operation during 2014

As described in former reports (e.g. Diehl et al. 2014), the Swiss Seismological Service (Schweizerischer Erdbebendienst, SED) operates two separate nationwide seismic networks, a high-gain seismometer network predominantly consisting of broad-band instruments (Table 4 in Appendix) and a low-gain accelerometer network (Table 5 in Appendix). In addition, the SED operates a number of temporary stations for various projects (Table 6 in Appendix). SED stations with on-line data acquisition that were operational at the end of 2014 are shown in Fig. 1.

Fig. 1
figure 1

Seismograph stations in Switzerland with on-line data acquisition operational at the end of 2014. The stations defined as high-gain (HG) are mostly equipped with broad-band or 5-s sensors, whereas the strong-motion stations (SM) are accelerometers (see also Tables 4, 5, 6 in Appendix)

Full size image

Within the Swiss Strong Motion Network (SSMNet) renewal project, 100 free-field, real-time, very broad-band accelerometer stations are being installed over a 10-year timeframe between 2010 and 2019 (Clinton et al. 2011; Michel et al. 2014). In 2014, 9 new accelerometer stations were installed (SARE, SBAK, SBAM2, SBAS, SBIF2, SBIK, SLUK, SNES2, SSCN). Two new stations (SMFL, STRW), funded by the state of Liechtenstein, were also installed and are operated by the SSMNet.

The station at ACB was upgraded from a 5 s-sensor to a broadband sensor. No other significant changes were made to the high-gain network, though the SED began a project to upgrade all stations. Each existing station will be upgraded to include (1) a very-broad-band sensor (if not yet existing); (2) a modern datalogger with significant local storage to improve final data completeness and lessen the impact of communication outages; and (3) a co-located strong motion accelerometer (if not yet existing). Where necessary, sites will also be re-located or improved to reduce site noise and improve earthquake detectability and signal quality. Additionally, 8 new stations (including 2 new borehole sites in the Alpine foreland) will be added to the network in order to achieve homogenous network spacing across the country.

The NAGRA network, which was installed under a contract with the National Cooperative for the Disposal of Radioactive Waste in northeastern Switzerland (Diehl et al. 2014), is expected to be a long-term network operated by the SED. The 10 new stations (7 surface, 3 borehole) are now included in the broad-band network (Table 5 in Appendix). The SED has begun monitoring microseismicity near the two undergound rock physics laboratories in Switzerland on a long-term basis, funded by various consortia that include the Federal Office of Topography (Swisstopo) and the Swiss Competence Center for Energy Research—Supply of Electricity (SCCER-SoE). At the Mont Terri in NW Switzerland, a strong motion and a broad-band station (MTI01 and MTI02) were installed inside the underground laboratory, and a surface strong motion station (MTI03) directly above it. A surface short period sensor, GSF01, was installed near Grimsel Rock Laboratory, which is the site of the existing SDSNet broad-band station GRIMS, near the Grimsel Pass. The temporary networks built to monitor geothermal projects in Basel (2006) and Sankt Gallen (2013) continue to operate, albeit with a reduced number of sensors.

Temporary stations are deployed to improve understanding of on-going earthquake sequences in Switzerland. In 2014, a broad-band station continued to operate near Sargans following the ML 4.1 Balzers (LI) event on December 12th 2013 (Diehl et al. 2014), and a number of stations were installed near Diemtigen (BE), following the vigorous sequence described in this report. In addition to improving the location of local events, the installation of these aftershock stations that have real-time communications also significantly reduces the magnitudes of earthquakes that can be detected.

To improve the reliability of locations for events at the periphery of or outside of Switzerland, the SED continues to be engaged in an on-going cross-frontier cooperative effort to exchange seismic data in real-time with foreign seismic networks as documented in detail by Diehl et al. (2014). With two new stations in Austria and one new station in Italy a total of 44 foreign stations were monitored at the SED in 2014.

2.2 Routine earthquake analysis

Methods and software currently used for routine earthquake analysis are described in detail in Diehl et al. (2013, 2014). All events are assigned a local magnitude (ML) that is calculated from the maximum amplitude of the horizontal components of the seismograms filtered to simulate the response of a Wood-Anderson seismograph (Diehl et al. 2014). In addition, a procedure has been implemented, which routinely and automatically provides estimates of the moment magnitude (MW) for a subset of events, using a spectral fitting technique (Edwards et al. 2010). Focal mechanisms in 2014 were determined from the azimuthal distribution of first-motion polarities (fault-plane solutions) using the HASH algorithm of Hardebeck and Shearer (2002).

2.3 Supplementary earthquake analysis

To lower the detection threshold and to provide robust magnitudes of earthquakes that are too weak to be reliably and consistently detected by the standard automatic detection algorithms of the SED, supplementary analysis methods were applied to selected earthquake sequences in 2014. Template signals (e.g. seismograms of previously detected earthquakes) were compared to continuous data streams by cross-correlation (CC). Such CC-based algorithms can lower the earthquake detection threshold by one magnitude unit compared to common trigger algorithms like the one implemented in the SeisComP3 monitoring software of the SED (e.g. Schaff and Waldhauser 2010). It should be noted that independent hypocenters cannot be determined for the majority of the events detected by CC, due to the insufficient number of available arrival-time picks. In some cases, however, the number of manually identified P- and S-wave arrivals allows an independent hypocenter location of an event detected by CC. This was the case for 160 events in 2014. The ML magnitudes of these events are almost exclusively less than 1.0 and, because of the logarithmic nature of the magnitude scale, even attain negative values. A relative magnitude M (corr)L is assigned to events without independent hypocenter locations using a procedure described in Edwards et al. (2015).

3 Seismic activity during 2014

3.1 Overview

During 2014, the Swiss Seismological Service detected and located a total of 817 earthquakes in the region shown in Fig. 2. Based on criteria such as the time of occurrence, the location, and signal character or on direct communication, 180 additional seismic events were identified as quarry blasts. In addition to these events, 160 earthquakes with ML ≤ 1.0 have been located by supplementary analysis as described in Sect. 2.3. Magnitude values of the events recorded in 2014 range from ML −0.4 to 3.3 (Fig. 3). The events with ML ≥ 2.5 and the criteria used to assign the quality rating for the given locations as well as the corresponding estimated location uncertainties are listed in Tables 1 and 2. Table 1 also includes the available Mw values derived using the spectral fitting method of Edwards et al. (2010). None of the events in 2014, however, generated sufficient long-period energy to produce a high-quality full-waveform moment-tensor inversion. Fault-plane solutions based on first-motion polarities are shown in Figs. 4 and 5 (see also Fig. 2) and their parameters are listed in Table 3.

Fig. 2
figure 2

Epicentres and focal mechanisms of earthquakes recorded by the Swiss Seismological Service during 2014. Epicentres of events mentioned in the text are Biel (Bie), Bürglen (Bür), Cheyres (Che), Cortébert (Cor), Diemtigen (Die), Effretikon (Eff), Frauenfeld (Fra), Stalden (Sta), Vallorcine (Val), Walensee (Wal), Werdenberg (Wer). FL marks the territory of Liechtenstein

Full size image
Fig. 3
figure 3

Earthquake activity during 2014: a timeline showing magnitude of each event and cumulative number of events (arrows mark the onsets of the Biel, Frauenfeld, and Diemtigen sequences); b magnitude histogram. Removing events detected by supplementary algorithms and events associated with the two sequences results in 607 events in 2014 (blue curve in a), which is close to the average number of 541 events/year (horizontal dashed grey) over the last 12 years

Full size image
Table 1 Earthquakes with ML ≥ 2.5
Full size table
Table 2 Criteria and location uncertainty corresponding to the quality rating (Q) of the hypocentral parameters in the event list in Table 1
Full size table
Fig. 4
figure 4

Fault-plane solutions based on first-motion polarities for eight events in 2014 (Biel ML 2.8 and 3.2, Stalden, Frauenfeld, Effretikon, Werdenberg, Bürglen, Cheyres; see Table 3). All stereograms are lower hemisphere, equal-area projections. Solid circles correspond to compressive first motion (up); empty circles correspond to dilatational first motion (down). The take-off angles were computed with the NonLinLoc software (Lomax et al. 2000), using the 3D velocity model of Husen et al. (2003). Grey lines show sets of acceptable solutions derived by the HASH algorithm (Hardebeck and Shearer 2002); black bold lines indicate the (preferred) average focal mechanisms of all accepted solutions; red bold lines mark the active plane as determined from high-precision relative earthquake relocations

Full size image
Fig. 5
figure 5

Fault-plane solutions based on first-motion polarities for two events in 2014 (Diemtigen ML 2.7 May 2014; Diemtigen ML 3.2 October 2014; see Table 3). Symbols and explanation as in Fig. 5

Full size image
Table 3 Focal mechanism parameters based on first-motion polarities of ten earthquakes in 2014 (see Figs. 2, 4, 5)
Full size table

Figure 6 shows the epicentres of the 933 earthquakes with ML ≥ 2.5, which have been recorded in Switzerland and surrounding regions over the period 1975–2014. These events represent about 7 % of the total number of events detected during that time in the same area. The chosen magnitude threshold of ML 2.5 ensures that the data set is complete for the given period (Nanjo et al. 2010) and that the number of unidentified quarry blasts and of mislocated earthquakes is negligible.

Fig. 6
figure 6

Epicentres of earthquakes with magnitudes ML ≥ 2.5, during the period 1975—2014. Grey circles denote earthquakes in the period 1975–2013, bold red circles indicate earthquakes in 2014

Full size image

3.2 Significant earthquakes of 2014

3.2.1 Biel/Bienne (BE)

Two earthquakes of ML 2.8 and ML 3.2 occurred within 15 min on January 8th about 7 km W of Biel/Bienne (BE) (Fig. 2). Both events were felt by the local population. The stronger of the two reached degree IV on the European Macroseismic Scale EMS-98 (Grünthal 1998). With the closest station BOURR at a distance of 28 km, the focal depth is not well resolved. Including Pg, Sg, and Pn phases in the location procedure results in focal depths of 8–10 km for both events. Although uncertainties in focal depths are in the order of ±2–3 km, it is certain that the earthquakes occurred in the crystalline basement beneath the southern edge of the Jura. Focal mechanisms were determined from the analysis of first-motion polarities of both events (Figs. 2, 4; Table 3). The mechanisms are dominated by normal-fault and strike-slip components with T-axes roughly oriented in NE–SW direction, typical for mid and lower crustal earthquakes in the northern Alpine foreland of Switzerland (e.g. Kastrup et al. 2004; Singer et al. 2014). Uncertainties of the mechanisms are significant (Fig. 4), mainly due to sparse station coverage in the Swiss and French Jura. Nevertheless, the two mechanisms cannot be identical (see for example the change in polarity at station BOURR in Fig. 4), which suggests that either the slip direction on the fault changed between the ML 2.8 and the ML 3.2 event or the events occurred on different faults. The two earthquakes triggered a small sequence of 21 locatable earthquakes over the following 12 days with magnitudes ranging between ML 0.6 and ML 2.3. Waveforms of these 21 aftershocks show at least two different groups with P-wave signals similar to either the ML 2.8 or the ML 3.2 event at station BOURR and CHMF. This observation may support the hypothesis that the ML 2.8 and the ML 3.2 events have occurred on different faults.

3.2.2 Cortébert (BE): the first known event in the upper mantle below northern Switzerland

On January 29th, 2014, 22:37 UTC, a small (ML 1.4) but unusually deep event occurred 4 km SSE of Cortébert (about 9 km W of Biel below the ridge of the Chasseral; see Fig. 2). Despite the small magnitude of the event and the low signal-to-noise ratio at several stations, reliable first onsets could be picked out to a distance of 120 km (station SLE). The location using all available data (18 P- and 7 S-onsets, including two additional foreign stations) and the 3D velocity model of Husen et al. (2003) gives a focal depth of 38.5 km, with an azimuthal gap of 65° and an RMS of 0.22 s. Other combinations of onsets with distance cut-offs at 50 and 80 km as well as with and without the use of S-arrivals give focal depths between 35.6 and 36.3 km. The stability of the results with varying data is the first indication for the reliability of the calculated focal depth. Based on the argument that the onsets at larger epicentral distances are generally weaker, and that model errors have a larger effect over longer distances, the location obtained with the available 11 P- and 5 S-onsets out to epicentral distances of 80 km is the preferred location (47.156/7.129, Z = 36 km, GAP = 63° and RMS = 0.07 s). According to Waldhauser et al. (1998) and Wagner et al. (2012), the crust-mantle boundary (Moho) below the epicenter lies at about 28–29 km. Consequently, our location places the hypocenter of this small earthquake clearly in the upper mantle and thus confirms the hypothesis mentioned by Stange et al. (2014). To illustrate this fact, Fig. 7 shows a seismogram-section along a WSW–ENE trending profile for the Cortébert event, together with a seimogram-section along a W–E trending profile for the ML 3.2 Courrendlin (JU) event of July 3rd, 2013, 14:13 UTC. According to Diehl et al. (2014), the epicenter of the latter event is about 20 km N and 20 km E of the Cortébert event, with a hypocenter at 25 km in the lower crust. Superimposed on these two record-sections are travel-time curves calculated with a 2D ray-tracing program, that takes into account the Moho topography as determined by Waldhauser et al. (1998), assuming a focal depth of 24 km for the Courrendlin event and of 36 km for the Cortébert event. As one would expect for a lower-crustal event, beyond about 70 km, the first arrival of the Courrendlin event changes from the slower direct wave (Pg) to the faster wave refracted in the upper mantle (Pn). The prominent second arrival visible in the three traces beyond 80 km is the superposition of the Pg and the reflection at the Moho discontinuity (PmP). In the seismograms of the Cortébert event, on the other hand, there is no visible kink in the travel-time curve of the first arrivals in this distance range, and a distinct secondary phase is entirely lacking. As the calculated travel-time curve shows, this is what is expected for a hypocenter below the Moho. Given the remaining uncertainties of the focal depths, the Moho topography and the seismic velocities, the calculated travel-time curves in Fig. 7 are schematic and merely serve to illustrate the categorical differences between the two events. Nevertheless, the low apparent velocity of the Pg/PmP phase seen in the record-section of the Courrendlin event (6.2 km/s) is evidence for the lack of high-velocity material present over any substantial depth interval in the lower crust of northwestern Switzerland. Moreover, the large amplitude of the reflected phase (PmP) requires a sharp velocity discontinuity at the crust-mantle transition.

Fig. 7
figure 7

Record sections and travel-time curves for the Cortébert event of 2014 with a focal depth at 36 km in the upper mantle (lower panel) and the Courrendlin event of 2013 with a focal depth of 24 km in the lower crust (upper panel)

Full size image

The Cortébert event is thus the first documented earthquake with a hypocenter in the upper mantle below the northern Alpine foreland of Switzerland. Supplementary analysis of signals at station BOURR (epicentral distance ~27 km) detected 3 additional events in 2014, which all occurred on February 19th (see Sect. 2.3 for details). These events were slightly smaller than the ML 1.4 event and could not be located. The similarity of waveforms with the ML 1.4 event, however, indicates that the three detected events were located very close to this event. Though the magnitudes of these four events are small (ML ≤ 1.4), it shows that, contrary to previous observations, the upper mantle in this region is not entirely aseismic.

3.2.3 Stalden (VS)

Many people in the Rhône and Visper Valley felt the ML 3.0 earthquake, which occurred on February 25th (00:52 UTC) close to the village of Stalden (VS). About 120 felt reports were collected and the maximum observed intensity was IV. The focal depth of 6 ± 1 km is well constrained by several nearby weak- and strong-motion stations and is consistent with a source in the upper Penninic Nappes. The extensional focal mechanism shown in Fig. 4 (see also Fig. 2) is well constrained by first motion polarities of P-phases and is consistent with previous seismicity and the general extensional regime expected south of the Rhône Valley (e.g. Kastrup et al. 2004).

3.2.4 Frauenfeld (TG)

Between April 16th and April 18th a sequence of 38 earthquakes occurred about 5 km S of Frauenfeld (TG). Magnitudes in this sequence range from ML 0.4 to ML 2.0 and routinely calculated focal depths of 10 ± 1 km show that the source of these earthquakes is within the crystalline basement beneath the Plateau Molasse. The dense seismic network in NE Switzerland permits the calculation of a focal mechanism for the strongest event of the sequence. The distribution of first-motion polarities suggests an almost N–S striking normal-fault mechanism (Figs. 4, 2). The high similarity of waveforms across the sequence allowed the measurement of differential times by cross-correlation, which were used for relative relocation of the hypocenters applying the double-difference technique of Waldhauser and Ellsworth (2000). The relocated hypocenters, shown in Fig. 8, image a planar structure dipping to the west. The dip of about 45° agrees well with the westward dipping plane of the focal mechanism of the strongest event in the sequence (Figs. 4 and 8) and we conclude that this westward dipping plane was also the active rupture plane of the ML 2.0 event. The sequence initiated at the up-dip edge of the structure and propagated in down-dip direction (Fig. 8). The ML 2.0 event, which occurred about 32 h after the first located event of the sequence, locates at the northern edge of the seismicity, about halfway between the up-dip and down-dip end of the structure. The plane outlined by the seismicity corresponds to a fault in the crystalline basement, which also hosted the ML 2.0 event.

Fig. 8
figure 8

Relative relocations of the 3-day Frauenfeld sequence in April 2014 using double-difference techniques in combination with waveform cross-correlation: a map view; b vertical cross-section along an EW profile. Colors indicate origin time of the events. The focal mechanism of the ML 2.0 event of April 17th is shown as lower hemisphere projection in map view and projected to the depth profile in cross-section

Full size image

Supplementary analysis of signals at station WEIN (epicentral distance ~6 km) detected 361 additional earthquakes with magnitudes between M (corr)L −0.8 and M (corr)L 1.0 (see Sect. 2.3 for details). They all occurred between April 15, 15:39, and April 18, 05:17, except for one M (corr)L 0.6 event that occurred about 26 h before that period. What caused the identified fault to rupture in a sequence of about 400 small earthquakes in less than 62 h remains to be investigated.

3.2.5 Diemtigen (BE)

A remarkable sequence of earthquakes commenced on April 13 close to the village of Diemtigen (BE), W of the Lake Thun (Fig. 2). By the end of 2014, 274 events with magnitudes between ML −0.1 and ML 3.2 were located, and seismic activity continues into 2015. In 2014, the sequence can be divided into three major periods of increased activity: the main pulses in April to May; mid-June to mid-July; and another phase in October. The three largest events occurred on May 10th (ML 2.7), June 25th (ML 2.7), and October 15th (ML 3.2). Regional residents felt all three events, with the strongest event reaching degree IV on the European Macroseismic Scale EMS-98 (Fig. 9).

Fig. 9
figure 9

ShakeMap with intensities converted from instrumentally recorded ground motions of the ML 3.2 Diemtigen event of October 15th. Colored dots mark macroseismic intensities (EMS-98) assessed by the SED based on about 430 felt reports submitted via the SED website. The star indicates the event epicentre. The felt intensities tend to indicate higher intensity levels than the instrumental ShakeMap close to the epicenter and in the urban areas of Bern and Interlaken

Full size image

We used the ML 3.2 earthquake to compare felt intensities derived from the analysis of macroseismic web questionnaires with instrumental intensities converted from instrumental records of ground motions (ShakeMap; Wald et al. 1999; Worden et al. 2010). The SED ShakeMap shown in Fig. 9 is produced using the state-of-the-art in engineering seismology in Switzerland (ground-motion predictive models, site effect estimates, etc.) as described in Cauzzi et al. (2014). In the case of the ML 3.2 Diemtigen event, the felt reports suggest higher intensity levels than the instrumental ShakeMap close to the epicenter and in the urban areas of Bern and Interlaken. A good agreement between the ShakeMap and the felt reports is apparent in the town of Thun, on the western shores of Lake Thun, and in the Aare valley, with macroseimic intensities approaching or reaching degree IV. Elevated intensities in the lower Kander valley and the Interlaken region are consistent with areas of higher site amplification, only partly captured by the resolution of the ShakeMap, which is targeted to the national scale. The peak-ground-velocity (PGV) of about 0.04 cm s−1 recorded by the accelerometric station SBERN in Bern corresponds to average intensity levels approaching degree II at most. Note however that the calibration dataset of Faenza and Michelini (2010), used in the SED ShakeMap, shows a large variability of PGV values for intensity IV, spanning two orders of magnitude from 0.01 to 1 cm s−1. With only one recording SDSNet station (WIMIS) within 20 km of the epicenter, the Diemtigen event demonstrates that the automatic inclusion of high-quality felt reports into SED ShakeMaps would produce a more complete and possibly better constrained shaking scenario.

To improve the quality of hypocenter locations and to lower the detection threshold of earthquakes associated with the sequence, three additional temporary seismic stations were installed at changing locations in the region between June 2014 and April 2015 (for locations see Fig. 10a). All stations were equipped with short-period seismometers and real-time communication. Focal mechanisms could be determined for the May 10th (ML 2.7) and October 15th (ML 3.2) events (Figs. 5, 10b). Compared to the event in May, the additional stations in the source region helped to reduce the uncertainties in the mechanisms of the ML 3.2 event. Both mechanisms correspond to a combination of thrust and strike-slip components (Figs. 5, 10b) with P-axis oriented NE–SW, which is the typical stress-regime observed along the northern front of the Alps and Helvetic domains NE of Diemtigen (Kastrup et al. 2004).

Fig. 10
figure 10

The Diemtigen sequence of 2014: a overview map with SED routine locations (red dots indicate earthquakes in 2014) and local seismic stations (brown triangle: SDSNet station WIMIS; Yellow triangles: temporary stations deployed by the SED). Orange box marks area shown in Fig. 10b; b relative relocations of a subset of 187 events of the sequence solving the double-difference equations by singular-value decomposition (SVD). Colors indicate origin time of the events. Crosses indicate relative location errors of epicentres

Full size image

To image the active rupture plane and the spatio-temporal evolution of this unusual sequence, we performed a relative relocation of hypocenters similar to the procedure described for the Frauenfeld sequence. We applied waveform cross-correlation and selected 187 events for the double-difference relocation shown in Fig. 10b. After relocation, the epicenters cluster much more tightly and cover an area of about 0.7 × 0.4 km (Fig. 10b). Figure 11a, b show two vertical cross-sections that cut two nodal planes of the focal mechanisms, shown in Figs. 5 and 10b. Cross-section AA’ (Fig. 11a) suggests that the earthquakes occur on a structure dipping towards ENE, consistent with plane number two (see Figs. 5, 10b) of the focal mechanisms. The seismicity in the vicinity of the October 15th ML 3.2 event forms a particularly sharp, distinctively dipping structure, which is offset with respect to the May 10th and June 25th events (Fig. 11a). The fine-scale structure of the relocated seismicity in Fig. 11a indicates that the seismicity associated with the sequence is not located on one simple planar structure. Instead, a kink or a bend along the structure seems to separate the structures hosting the events in May and July from the ML 3.2 event in October (Fig. 11a). Seismicity below the May 10th event appears more diffuse and seems to form an almost horizontal structure at its down-dip end (Fig. 11a).

Fig. 11
figure 11

Depths distribution of relative relocations shown in Fig. 10. a vertical cross-section along a WSW–ENE striking profile; b vertical cross-section along a NW–SE striking profile; c vertical migration of seismicity between April and mid November 2014. Vertical bars indicate relative location errors of focal depths and colours indicate time evolution following colourbar of Fig. 10

Full size image

Supplementary analysis of signals at station WIMIS (epicentral distance ~4 km) detected 1720 additional earthquakes with magnitudes between M (corr)L −1.5 and M (corr)L 0.8 in 2014 (see Sect. 2.3 for details). Results of this analysis indicate, that the Diemtigen sequence already started on March 19, 13:01 (UTC), with an earthquake of magnitude M (corr)L −0.1, which was too small to be located. Seven template events had to be used to find all previously located events of the sequence. All detected events were associated with one of these templates by means of waveform similarity and therefore belong to one of seven waveforms clusters. Comparing the result of the template-matching analysis to the double-difference relocation shown in Fig. 11, we found that the template events are located in different parts of the imaged structure. The spatial volumes covered by the individual clusters hardly overlap, which allows to estimate the location of all detected events—down to magnitude M (corr)L −1.5—to a precision of a few hundred meters by comparing waveforms at a single station.

The relocated seismicity shown in Fig. 11 reveals clear along-dip migration patterns during the evolution of the sequence. As shown in Fig. 11c seismicity initiated at about 8.1 km depth on April 13th and migrated down-dip in the following weeks. The ML 2.7 event on May 10th occurred around the initial depth of 8.1 km and seismicity following May 10th migrated in up-dip direction. Seismicity in November occurred then on the up-dip periphery of the sequence and the entire sequence covered a depth interval of 0.6 km. Vertical errors in Fig. 11 are relative errors and it should be noted that the accuracy of the absolute depth of the sequence is less well constrained. The accuracy of the absolute source depth is mainly controlled by the quality of the velocity model used for location. Even focal depths determined with 3D regional models can have errors of few kilometers due to the imperfect knowledge of the local velocity structure (e.g. Husen et al. 2012).

The top of the autochthonous crystalline basement in the region of Diemtigen is expected to lie at a depth of about 6–7 km (e.g. Pfiffner et al. 1997a). SW of Diemtigen, geological profiles suggest sediments reach down to 8 km depth below the Niesen nappe (e.g. Pfiffner et al. 1997b). Because of the uncertainties in the absolute focal depth and in the geological interpretation of the seismic data (including along-strike deviation from the 2D profiles) it is not possible to determine whether the sequence occurred in the crystalline basement or in the lowermost part of the overlying sediments. The apparent kinks or bends in the seismogenic structure suggest that the structure might be intersected by sub-horizontal lithological boundaries, which would be indicative for a source in sedimentary units (e.g. Mesozoic sediments).

The Diemtigen sequence shows the typical characteristics of an earthquake swarm, e.g. lack of a distinct main shock, unusually large spatial extent compared to the moment release of the largest individual event, and magnitudes that fail to decay with time (e.g. Roland and McGuire 2009). However, the underlying mechanisms, which have initiated and driven this swarm, are not yet understood. Processes including temporary fluid overpressure, perhaps associated with creep might be drivers of the swarm, but the complex structure imaged by the relocated seismicity suggests that stress-transfer and sequential rupture of several small fault segments might play a role in the evolution of the sequence.

The ENE dipping structure highlighted by the sequence can neither be associated with known faults mapped at the surface nor with other instrumentally located seismicity prior to April 2014. However, the estimated epicentre of the historic Mw 5.2 Frutigen earthquake of 1729 is located only about 4 km to the SE of the sequence (Fäh et al. 2003). Considering the significant uncertainties in the location of historic events, it is possible that the fault that ruptured during the 1729 Mw 5.2 event is related to the fault system, which was active during the Diemtigen sequence.

3.2.6 Effretikon (ZH)

On November 8th, an ML 2.8 earthquake occurred close to the village of Effretikon in the canton of Zürich (Fig. 2). The earthquake was felt in the region and triggered about 75 felt reports submitted to the SED website. The focal depth of the event is about 12 km, and therefore—as the events of the Frauenfeld sequence, about 20 km to the East—located in the crystalline basement below the Plateau Molasse. In contrast to the strongest event of the Frauenfeld sequence, the focal mechanism of the Effretikon earthquake is predominately strike-slip with an almost E–W oriented T-axis (Fig. 4; Table 3), typical for mid to lower crustal earthquakes in this region (e.g. Kastrup et al. 2004). The mechanisms of the Frauenfeld and Effretikon events reinforce the tendency of the azimuth of extensional T-axes to be oriented parallel to the strike of the Alpine orogen. Well-aligned T-axes, with a moderate plunge, are a predominant feature of the crust in the northern Alpine foreland (Deichmann 1990; Deichmann et al. 2000b; Kastrup et al. 2004; Singer et al. 2014).

3.2.7 Walensee (SG)

An interesting earthquake of ML 3.1 occurred at the northern shore of the Walensee in eastern Switzerland (Fig. 2) on November 14th at 04:14 UTC. With a depth of 0 km, the event appears to have a very shallow source with a location a few hundred meters south of a closed quarry, which is situated about 2 km W of the village of Quinten. A sequence of ten earthquakes with ML between 1.4 and 3.8 occurred 1–2 km below and within about 0.5 km horizontally from the same quarry in 1997. Deichmann et al. (1998) speculated that these events may have been triggered by activity at the nearby mine. Figure 12 shows a comparison of the 2014 event with the three strongest events in 1997 recorded at stations WIL and WILA. The distance between the two stations is less than 100 m and can be neglected for this comparison. The similarity between the 2014 and the 1997 waveforms suggest that the 2014 event is part of the same cluster, with a similar location and focal mechanisms as the events in 1997 (Fig. 12). The distribution of available first motion polarities does not allow a well-constrained determination of the focal mechanism. Strike-slip as well as thrust mechanisms with N–S oriented P-axes fit the observed polarities equally well.

Fig. 12
figure 12

Comparison of vertical velocity seismograms of the ML 3.1 Walenstadt event of November 14th 2014 recorded at station WILA (bottom trace, bold) and the Walenstadt sequence of November 1997 recorded at station WIL (top three traces)

Full size image

3.2.8 Vallorcine (France)

With 270 felt reports, the ML 3.2 earthquake that occurred close to the village of Vallorcine (France) (Fig. 2) on November 23rd, was the second-most strongly felt earthquake in 2014. This earthquake is part of the on-going earthquake sequence that followed the ML 4.9 earthquake of September 2005 (Deichmann et al. 2006; Fréchet et al. 2010; Diehl et al. 2013). Together with an ML 3.2 event in December 2010, the ML 3.2 of November 2014 is the second largest earthquake in the sequence since the mainshock of 2005. As in previous years, the region of Vallorcine was one of the seismically most active regions in 2014.

3.2.9 Werdenberg (SG)

An ML 2.7 earthquake was located about 5 km E of Buchs (SG) in eastern Switzerland on December 7th (Fig. 2). The event is situated about 3 km SE of the Wildhaus sequence of 2009 (Deichmann et al. 2010; Marschall et al. 2013). The routinely determined focal depth of 4 km is well constrained by P and S-wave arrivals recorded at three strong-motion stations with epicentral distances from 3 to 6 km. The shallow focal depth suggests that the source is most likely located in the sediments, similar to the Wildhaus earthquakes in 2009 (Deichmann et al. 2010; Marschall et al. 2013). The derived focal mechanism is strike-slip with a normal-fault component (Fig. 4), with orientations of P- and T-axes similar to the mechanisms in 2009.

3.2.10 Lower crustal earthquakes of Bürglen (TG) and Cheyres (FR)

The strongest earthquake in 2014 was the ML 3.3 earthquake near Bürglen (TG) on December 15th (Fig. 2). With a focal depth of about 25 km it is located in the lower European crust. It was felt only sporadically (about 20 felt reports). The focal mechanism of this event is strike-slip with orientations of T- and P- axes typical for lower crustal earthquakes in northern Switzerland (Fig. 4). The ML 3.1 Cheyres (FR) earthquake of December 17th, which is located close to the SW corner of Lake Neuchâtel (Fig. 2), is another example of a lower crustal earthquake. The Moho in this area is at about 31 km depth and the focal depth of 29 km indicates that the earthquake occurred just a few kilometers above the Moho. The predominately strike-slip mechanism shown in Fig. 4 is consistent with other lower crustal earthquakes in the region (Kastrup et al. 2004; Singer et al. 2014).

3.2.11 Seismicity associated with the former Deep Heat Mining project in Basel

Over the time period 2008–2011, the seismic activity induced by the geothermal project in Basel in 2006 and 2007 (e.g. Deichmann and Giardini 2009) remained below the current automatic detection threshold of the SED (approx. ML 0.9). In 2012 and 2013, the activity picked up again with seven detected events with magnitudes between ML 0.9 and 1.8 (Diehl et al. 2014). For almost a whole year thereafter, activity, if at all, remained below the detection threshold until the occurrence of an ML 1.6 event on December 23rd, 2014. With the exception of the ML 1.1 event of October 3rd 2012, which was located about 500 m ESE and 400 m below the casing shoe, and thus clearly detached from all other events, the hypocenters of all the events recorded in 2012 and 2013 were located above the casing shoe at the SSE periphery of the previously stimulated rock volume (Deichmann et al. 2014). The hypocenter of the event of 2014, on the other hand, is located in a part of the stimulated rock volume that had already been activated in December 2006 and January 2007. A subsequent cross-correlation based analysis (for details on method see Sect. 2.3) at borehole station MATTE (epicentral distance ~4 km) detected 9 additional events with magnitudes between M (corr)L −0.4 and M (corr)L 0.6 in December 2014. Four of these events had very similar waveforms to those of the event of December 23rd, 2014, and were, therefore, located close to this event.

Using the same cross-correlation based detection as for 2014, we searched for additional undetected events in previous years. In the period from May 2012 to end of 2014, we found a total of 32 events in addition to the nine events detected by the SED monitoring system. The magnitudes of these additional events range between M (corr)L −0.6 and M (corr)L 0.8. Whereas the large majority of these events can be associated with three clusters located at the periphery of the previously stimulated rock volume, two small events (ML around 0), which occurred in February 2013, must be located close to the cluster that was active in December 2014. Thus it seems that the apparent migration of the seismicity back into the previously activated rock volume closer to the borehole observed at the end of 2014 started with some weak precursors already in the previous year. It has to be emphasized that the adopted set of template signals only covers four small, separate parts of the previously seismically active rock volume, and that the cross-correlation detection is therefore insensitive to events that might have occurred in other parts of this volume. Whether such events occurred is currently under investigation.

4 Discussion and conclusion

In 2014, as in previous years, a large portion of the seismic activity was concentrated in the Valais, Graubünden, and their immediately adjacent regions (Fig. 2). Routinely calculated focal depths for all but 32 events recorded in 2014 are less than 16 km. Except for the Cortébert earthquake, all the deeper hypocenters occurred in the crust below the Molasse Basin and the Jura of northern Switzerland and southern Germany, with the main activity clustering between 7°E and 10°E (e.g. Deichmann et al. 2000a; Singer et al. 2014). The Cortébert event of January 2014, however, is the first documented earthquake with a hypocenter in the upper mantle in the northern Alpine foreland of Switzerland. Though the magnitude is only ML 1.4, it shows that, contrary to common belief, the upper mantle in this region is not entirely aseismic. The Cortébert mantle event occurred 21 days after the ML 2.8 and ML 3.2 Biel earthquakes; its epicenter is located only about 2 km W of the Biel sequence. Given the relatively small magnitudes of the Biel events in combination with a vertical distance of about 30 km, a connection between the Biel and the Cortébert earthquakes appears very unlikely despite the noticeable proximity of epicenter and origin time.

Seismicity in 2014 was largely dominated by the occurrence of several earthquake sequences. On-going activity in the regions of Vallorcine and Balzers is related to the ML 4.9 and ML 4.1 earthquakes of the year 2005 and 2013, respectively. The sequences of Diemtigen and Frauenfeld, on the other hand, show the typical characteristics of earthquake swarms. High-precision hypocenter relocation in combination with available focal mechanisms identified a westward dipping normal fault in the crystalline basement in the Frauenfeld case. Seismogenic structures imaged in the Diemtigen sequence show ENE dipping structures associated with mechanisms composed of thrust and sinistral strike-slip components. The underlying mechanisms initiating and driving such earthquake swarms are not yet understood. The SED has therefore formed a working group to investigate these issues in more detail.

Although the maximum magnitude in 2014 reached only ML 3.3, the unusually large number of 817 earthquakes was detected and located by routine analysis of the national seismic network over the year (Fig. 3a). An additional 160 earthquakes with ML ≤ 1.0 have been located by supplementary analysis described in Sect. 2.3. The total of 977 earthquakes is the largest number of events located within a year since the first installation of a seismic network in Switzerland in 1975. Between the year 2002 (completion of the SDSNet) and 2013, the average number of events recorded per year was about 540 (grey horizontal line in Fig. 3a). Compared to the average of previous years, the number of earthquakes (977) appears to have increased by more than 50 % in 2014. As illustrated in Fig. 3a, this increase was mainly caused by the combination of two phenomena: (1) The exceptionally vigorous earthquake sequences of Frauenfeld and Diemtigen contributed more than 200 events (Fig. 3a). (2) On-going efforts to lower the thresholds for earthquake detection in Switzerland starting to have a significant impact on the total number of recorded events. Supplementary detection algorithms, on-going densification of the weak and strong-motion networks, as well as the deployment of temporary local networks (such as in the Diemtigen region) has improved the detection thresholds in certain parts of Switzerland. Excluding events of the two earthquake sequences and events located by supplementary earthquake analysis results in 607 earthquakes in 2014, a number close to the average of previous years. In summary it should be noted that the occurrence of two exceptional earthquake sequences and the on-going lowering of detection thresholds mainly explain the unusually large number of earthquakes in 2014. The total number of 22 earthquakes with ML ≥ 2.5, on the other hand, closely matches the yearly average of 23.0 events over the previous 39 years in this magnitude range. The total number of located earthquakes does therefore not allow conclusions on the overall seismic hazard in Switzerland.