submitted to Geophys. J. Int.

Seismotectonics and 1D velocity model of the Greater1

Geneva Basin, France-Switzerland.2

Veronica Antunes1, Thomas Planes1, Jirı Zahradnık2, Anne Obermann3,3

Celso Alvizuri4, Aurore Carrier1, Matteo Lupi1

1 Department of Earth Sciences, University of Geneva, Switzerland

2 Faculty of Mathematics and Physics, Charles University, Prague, Czech Republic

3 Swiss Seismological Service, ETH Zurich, Switzerland

4 Institute of Earth Sciences, University of Lausanne, Switzerland

4

Draft 2020 February 195

SUMMARY6

The Greater Geneva Basin, located in south-western Switzerland and neighboring France,7

is enclosed by the rotating north-western edge of the Alpine front and the Jura mountains8

chain. Recently, this basin has received increasing attention as a target for geothermal9

exploration. Historical and instrumental seismicity suggest that faults affecting the basin10

may still be active. Moderate-magnitude earthquakes have been located along the Vuache11

fault, a major strike-slip structure crossing the basin. Before geothermal exploration12

starts, it is key to evaluate the seismic rate in the region and identify possible seismogenic13

areas. In this context, we deployed a temporary seismic network of 20 broadband stations14

(from September 2016 to January 2018) to investigate the ongoing seismic activity, its15

relationship with local tectonic structures, and the large-scale kinematics of the area.16

Our network lowered the magnitude of completeness of the permanent Swiss and French17

networks from 2.0 to a theoretical value of 0.5. Using a new coherence-based detector18

2 Antunes et. al

(LASSIE - particularly effective to detect microseismicity in noisy environments), we19

recorded scarce seismicity in the basin with local magnitudes ranging from 0.7 to 2.1ML.20

No earthquakes were found in the Canton of Geneva where geothermal activities will21

take place. We constructed a local ’minimum 1D P-wave velocity model’ adapted to22

the Greater Geneva Basin using earthquakes from surrounding regions. We relocated the23

events of our catalogue obtaining deeper hypocenters compared to the locations obtained24

using the available regional velocity models. We also retrieved 8 new focal mechanisms25

using a combination of polarities and waveform inversion techniques (CSPS). The stress26

inversion shows a pure strike-slip stress regime, which is in agreement with structural and27

geological data. Combining the background seismicity with our catalogue, we identified28

seismogenic areas offsetting the basin.29

Key words: Earthquake source observations – Seismicity and tectonics – Dynamics:30

seismotectonics31

1 INTRODUCTION32

Intra-plate sedimentary basins are often characterized by low levels of seismic activity (e.g.,33

Hofstetter et al. 2012; Lacombe et al. 2019). Due to low seismic rates, strong attenuation, and34

often high levels of anthropogenic noise (in densely populated areas) the number of seismic35

stations deployed in these areas are often kept to a minimum. This affects the detection of36

microseismicity that is often above the detection threshold of 2.0M (Plenkers et al. 2015).37

Sedimentary basins are well known to hold considerable portions of mineral, energy,38

and water resources (Bethke et al. 1988; Raffensperger & Vlassopoulos 1999; Cacace et al.39

2010). In particular, those characterized by favorable hydrogeological conditions have good40

potential for geothermal energy exploitation (Cacace et al. 2010). In such cases, a significant41

amount of heat transport is provided by groundwater circulation through permeable aquifers.42

From shallow and intermediate depths (T<100◦C), valuable hydrothermal energy can be43

extracted for heating purposes. However, it is necessary to reach greater depths to generate44

electricity from sedimentary basins.45

Seismotectonics of the GGB 3

Geothermal energy extraction is a resource potentially available anywhere. However, pre-46

vious experiments have shown how important is to monitor induced seismicity, in particular47

when injection of fluids is involved (e.g. Iceland Flovenz et al. (2015) and South Korea48

Grigoli et al. (2018); Kim et al. (2018); Ellsworth et al. (2019)). In Switzerland, geothermal49

energy is a sensitive topic since the projects of Basel (2006) (Haring et al. 2008; Hillers50

et al. 2015) and St. Gallen (2013) (Kraft et al. 2009; Obermann et al. 2015). In Basel, the51

hydraulic stimulation of the crystalline basement at 5 km depth drastically increased the52

seismic activity (Haring et al. 2008), culminating in an ML3.4 earthquake (Edwards et al.53

2015) that caused damages to infrastructures (estimated in ∼ 7M US$, Kraft et al. 2009).54

Consequently, the project was aborted (Haring et al. 2008). In St. Gallen, gas from a pre-55

viously unknown Permo-Carboniferous trough underlying the Mesozoic sediments suddenly56

leaked into the borehole (Obermann et al. 2015). The well rescue operations caused a ML3.557

that was felt in the area (though with little damage reported) (Edwards et al. 2015). In58

Pohang, South Korea, hydraulic stimulations at 4 km depth caused a Mw5.5 earthquake59

(Grigoli et al. 2018; Kim et al. 2018), injuring several people and causing significant infras-60

tructure damage. Despite these examples geothermal energy in Switzerland is still perceived61

as a promising renewable energy resource. Giardini (2009) suggest that geothermal quake risk62

must be faced. Indeed, enhanced geothermal systems have already been successfully drilled63

and developed (Lu 2018). For instance, the induced seismic activity (up to 2.9 MW , Dorbath64

et al. 2009; Baisch et al. 2010) did not discourage the local authorities in Soultz-sous-Foret,65

France, where three deep geothermal wells (about 5 km deep) are being exploited. This ac-66

ceptance was facilitated by the the low population density in the region. These projects have67

in common the unexpected reservoir dynamics that culminated in undesired levels of seis-68

micity. Studying the local microseismicity and its relationship with local tectonic structures69

is thus a key step before planning geothermal exploitation projects.70

Switzerland is currently promoting the development of geothermal energy (among other71

renewable energy resources) to reduce CO2 emissions and the dependency on nuclear power72

and hydrocarbon resources. In particular, the Molasse basin in the Geneva area is in the focus73

4 Antunes et. al

of geothermal exploration, from initial shallow targets to gradually deeper ones (Clerc et al.74

2015; Allenbach et al. 2017). In 2012, the Canton of Geneva and the Industrial Services75

of Geneva (SIG) launched the GEothermie2020 ⋆ program to investigate the geothermal76

potential of the Greater Geneva Basin (GGB) (Wilhelm & PGG team 2011; Allenbach et al.77

2017; Nawratil de Bono & GEo-01 team 2018). The program includes geophysical exploration78

of the subsurface and the drilling of 4 exploratory wells by 2021. The deepest of these wells79

is expected to reach a depth of 3 km.80

The GGB shows a low level of seismicity with 168 events in 31 years of instrumental81

monitoring (Fah et al. 2011; Deichmann et al. 2010, 2011, 2012; Diehl et al. 2013, 2014,82

2015, 2018). This amounts to about 5 earthquakes/year with a magnitude range of ML1.1-83

5.3. Historical seismic catalogues (since 1500 A.D.) contain a handful of seismic events with84

intensities of IV-V (Fah et al. 2011). Until August 2016, the permanent seismic networks85

counted only a single seismometer in the GGB, resulting in an estimated magnitude of86

completeness (MC) of ∼ 2.0 (Diehl et al. 2018). Precise local velocity models are missing.87

This hindered the accurate location of the microseismicity that could better constrain the88

geometry of active faults. The currently available velocity models for event location are89

regional and not suitable for the GGB area. In France, the Sismalp velocity model (Cara90

et al. 2015) is used to locate seismicity in the French Alps. In Switzerland, the velocity91

model of Husen et al. (2003) is used by the Swiss Seismological Service (SED, Diehl et al.92

2014). However, the lack of seismic stations and seismicity detected by the Swiss and French93

networks hindered the development of a dedicated velocity model for the GGB until now.94

In this context, we deployed a temporary dense seismic network of 20 broadband stations95

around and within the GGB. The network operated for 17 months, from September 2016 to96

January 2018, reaching a detection threshold of 0.5ML. The continuous data collected has97

been used by Planes et al. (2020) to investigate the shallow shear-wave velocity structure98

of the GGB with an ambient noise tomography. We use the network to shed light on the99

natural microseismicity within and around the GGB.100

⋆ https://www.geothermie2020.ch

Seismotectonics of the GGB 5

The manuscript is built as follows. First, we introduce the geological and seismological101

setting of the GGB. Next, we describe the temporary seismic network deployed in the area102

and the methodology to obtain a preliminary seismic catalogue. We then detail the inver-103

sion procedure used to obtain a new ’minimum 1D velocity model’ adapted to the GGB.104

Afterwards, we present the relocated catalogue of events and discuss the retrieval of focal105

mechanisms. We use these focal mechanisms to perform a stress inversion in the region. We106

finally discuss our major findings and their implications in the discussion section.107

2 GEOLOGICAL AND SEISMOLOGICAL SETTING OF THE GREATER108

GENEVA BASIN109

The GGB (Gorin et al. 1993; Clerc et al. 2015, Fig. 1) is situated in the southwestern110

extremity of the western Alpine foreland basin (also referred to as Molasse Basin, yellow,111

Fig. 1), across Switzerland and France. It includes the Geneva Basin (Sambeth & Pavoni112

1988), the Rumilly Basin, and the Bornes Plateau (GB, RB and BP respectively, Fig. 1),113

and covers an area of approximately 2500 km2, reaching a maximum depth of ∼ 5 km below114

sea level (Clerc et al. 2015). The subsurface structure of the GGB is well-known thanks115

to intensive seismic reflection campaigns that occurred from 1958 to 1990, for hydrocarbon116

exploration (Gorin et al. 1993). Temperature data collected from wells in the GGB suggest117

an average geothermal gradient of about 25-30◦/km (Mosar 1999; Chelle-Michou et al. 2017).118

Geologically, the GGB is constituted of thick 3-5 km Mesozoic sediments composed of119

carbonates and marls that sit upon a crystalline basement of irregular topography, whose120

depressions are filled with Permo-Carboniferous sediments (Clerc et al. 2015). On the top121

of the Mesozoic layers stands a siliciclastic Tertiary Molasse sedimentary layer, that thins122

towards the Jura mountains. The Tertiary Molasse is overlain by Quaternary sediments of123

glacial to fluvioglacial origins (Clerc et al. 2015). Tectonically, the GGB is restricted by the124

following structures (Gorin et al. 1993; Mosar 1999; Clerc et al. 2015, Fig. 1): a) the internal125

reliefs of the Jura arc mountains in the northwest and south (blue); b) the Alpine front126

thrusts in the northeast, which includes the Penninic pre-alps (purple) and the Subalpine127

6 Antunes et. al

and Helvetic Nappes (pink); c) the Penninic nappes and the external crystalline massifs in128

the southeast (green); d) Lake Geneva in the north.129

Two main sets of faults affect the GGB accommodating the NW-SE Alpine compression.130

The first one consists of a series of thrust faults, NE-SW orientated, located in the Jura and131

in the subalpine Molasse that delineates the southeastern rim of the GGB (Fig. 1). They132

are linked with the presence of reactivated Permo-Carboniferous lineaments (Signer & Gorin133

1995). The second set corresponds to sinistral strike-slip fault systems orientated NW-SE,134

laterally accommodating the NW-SE shortening (red areas, Fig. 1). From South to North,135

geological investigations identified 4 main fault areas, namely, Vuache, Cruseilles, Le Coin136

and Arve.137

Despite the proximity to the tectonically-active western Alpine front (e.g., Gorin et al.138

1993; Clerc et al. 2015; Rabin et al. 2018; Houlie et al. 2018), little is known about the micro-139

seismic activity taking place in the GGB and surrounding regions. Before 1975, the available140

earthquake information was based purely on historical documents mentioning shaking and141

infrastructure damage and allowed magnitude and intensity estimations. The location of142

these events are biased by the location of historical villages and the available written docu-143

mentation. Such data must be considered with errors of several dozens of km (e.g., 20-50 km).144

The historical records show that the area was affected by seismic events with an intensity145

of up to about VII (Fah et al. 2011; Wiemer et al. 2015). The shores of Lake Geneva were146

flooded by at least one lake tsunami possibly caused by an earthquake in 563 AD (Kremer147

et al. 2014). The largest seismic events occurred along the Vuache fault, a well-mapped sinis-148

tral strike slip fault that accommodates the westwards rotation of the Alpine front (Baize149

et al. 2011, Fig. 1).150

Fig. 2 displays the reported background seismicity from both historical (since 1500 until151

1975, ECOS-09 catalogue, Fah et al. 2011) and instrumental seismic catalogues (from 1975152

to 31st August 2016, ECOS09 and SED public catalogues, Fah et al. 2011; Deichmann153

et al. 2010, 2011, 2012; Diehl et al. 2013, 2014, 2015, 2018). It also shows the major tectonic154

structures around and within the GGB (blue lines) previously pointed out in Fig. 1, including155

Seismotectonics of the GGB 7

the buried faults offsetting the basin and discovered in the framework of the GeoMol project156

(Clerc et al. 2015; Allenbach et al. 2017) during reflection seismic campaigns.157

From 1500 to 1850, 16 earthquakes were historically documented in the Canton of Geneva158

and nearby regions (white square on Fig. 2), with estimated intensities of IV-V that can be159

related to likely magnitudes of about 3-4, according to Fah et al. (2011) and Wiemer et al.160

(2015). The Swiss Seismological Service (SED) has been operational since 1850 and reported161

50 earthquakes in the region until 1975 (Fah et al. 2011). After 1975, the seismic network of162

Switzerland was dense enough to obtain the reliable detection and location of earthquakes163

(Fig. 2). Until 2008, the magnitude displayed in the ECOS-09 catalogue corresponds to164

Moment magnitude (MW ), for a total of 146 earthquakes. After 2009, the magnitudes of the165

SED public catalogues are displayed as local magnitudes (Ml/MLh), and account for a total166

of 22 earthquakes.167

In 1996, a ML5.3 earthquake occurred at the southern tip of the Vuache fault, followed by168

an aftershock sequence referred to as the Epagny sequence (45.9N, 6.1E; Thouvenot et al.169

1998; Courboulex et al. 1999). The source mechanism of the major event was a sinistral170

strike-slip (section 4.3). These events led to questioning whether the Vuache fault could171

generate strong magnitude events (M6, Thouvenot et al. 1998). Apart from this area, the172

seismicity in the GGB is rather diffuse (Thouvenot et al. 1998). The majority of the events173

occurs within the first 15 km depth.174

The seismicity located at the northeastern-most part of the GGB is likely associated175

with the Alpine thrusts front, and to the south with the Vuache fault and thrusts at the176

Jura mountains (Fig. 2). A few events (<10) can be found nearby the Cruseilles and Arve177

faults indicating that these faults could also still be active. Overall, very little activity (2178

earthquakes, considering the instrumental period) is actually recorded in the Canton of179

Geneva sensu-stricto (delimited by the brown line inside the white square). The earthquake180

locations might be biased by the high noise level at the stations and the poor network181

coverage prior to this study.182

8 Antunes et. al

3 EVENT DETECTION AND RELOCATION183

3.1 Temporary network184

We temporarily densified the existing seismic permanent networks (CH, FR, GU, IV, Fig. 1),185

from September 2016 to January 2018, with 20 additional broadband stations (UG tempo-186

rary network) around and within the GGB to lower the detection threshold of microseismic187

activity (<2 ML). The inter-stations distance is approximately 5-15 km, with closer station188

spacing around the city of Geneva (i.e. at the center of the network).189

We used Centaur and DataCube digitizers and three different types of Trillium Compact190

sensors: TC20s, TC120s, and TC120s-Posthole. The stations were operating offline and con-191

tinuously acquiring with a sampling rate of 100 Hz and a gain of 1. Part of the UG network192

was removed by the end of January 2018, keeping only 8 stations deployed in the GGB to193

be used for seismic monitoring purposes throughout the geothermal project. These stations,194

together with the existing networks, were used to locate two additional events that occurred195

in the Jura mountains in February 2019 (Fig. 11).196

Due to the dense population in the GGB, it was challenging to find quiet places for197

the deployment of the broadband sensors. Fig. 3 provides an overview of the noise levels198

of the UG stations, in the form of probabilistic power spectral density (PPSD, McNamara199

& Buland 2004; Beyreuther et al. 2010). Bedrock and basin stations are colour-coded. The200

noise levels from the stations represent the 75th percentile PSD values obtained for each201

station for the period of study (September 2016 to January 2018). The 75th percentile was202

chosen to take into account the periods of higher levels of noise at the stations, e.g., during203

the day, where anthropogenic sources of noise can interfere with seismic signals and affect204

the detection of earthquakes.205

We overlay the plot with synthetic Brune S-wave source spectra (Fig. 3, Brune 1970,206

1971) that were computed for local earthquake magnitudes ML between -1.5 and 2.5, using207

a stress drop of 2.0 MPa at a hypocentral distance of 7 km. 7 km corresponds to the208

largest distance that could separate an event from a seismic station inside the network.209

Seismotectonics of the GGB 9

The amplitudes were converted to octave band-passed velocity (m/s) in order to compare210

the theoretical Brune S-wave source spectra with the ambient seismic noise levels at the211

UG stations (Bormann 1998; Clinton & Heaton 2002). This synthetic computation gives an212

estimation of the minimum ML required for an earthquake to be above the noise levels and,213

therefore, to allow its detection.214

We are particularly interested in the frequency range of 1 to 30 Hz, the typical frequency215

range of local earthquakes (Bormann 2009). From Fig. 3 we note that at this frequency band,216

the stations deployed in the basin have higher levels of background seismic noise than the217

stations deployed in bedrock. Beside the much higher attenuation and site effects due to the218

sediments, the basin stations are also closer to villages and cities where the anthropogenic219

background noise is higher. Bedrock stations are deployed in more remote and hence quiet220

places in the mountains. The levels of background noise for all stations are between the221

standard low and high noise models from the U.S. Geological Survey (USGS) (Peterson222

1993). The 0.5 ML curve is above the noise level at all stations (in the frequency range of223

interest) indicating that events ML > 0.5 that occur inside the network should be detected224

by the UG network, at any time of the day.225

3.2 Preliminary seismic catalogue226

We created a manual seismic catalogue for events occurring in the area during 1.5 months227

(for September 2016 and the second half of December 2016) and used this dataset to calibrate228

an automatic detection algorithm.229

We configured an automatic detector using LASSIE† (Matos et al. 2016; Lopez-Comino230

et al. 2017a,b) an open-source tool based on the Pyrocko library (The Pyrocko Developers231

2017). It exploits the coherence of signals recorded at different stations to automatically232

detect and locate seismic events on a grid of trial points. To derive coherence measures, it233

shift-and-stacks characteristic functions (CF), i.e., time series derived from the 3-component234

recordings at each station, enhancing certain characteristics of the waveform. Time shifts235

† https://gitext.gfz-potsdam.de/heimann/lassie

10 Antunes et. al

are used to back-propagate the CF to the trial origin points, taking into account the seismic236

travel-time between trial points and station location. We use the 1D velocity profile by Husen237

et al. (2003) to compute the travel-times for P- and S-phases.238

In this application, we use two sets of characteristic functions (CF1 and CF2) in com-239

bination: CF1 is configured to be sensitive to sharp onsets in the recordings (e.g. P wave240

arrivals), CF2 is designed to rise when transient wave packets of certain duration pass (S241

waves and coda). For both characteristic functions, the signal from each channel is first242

bandpass filtered. Then, for CF1 a STA/LTA filter is applied, and the function is smoothed243

and normalized. For CF2, the signal is squared, smoothed and normalized. Finally, all sta-244

tion channels are combined and the resulting CFs are downsampled to a common sampling245

rate for efficient stacking.246

The result of the shift-and-stack of the two CFs of all stations is referred to as the image247

function (IF). It can be interpreted as a coherence value for every grid point and time sample.248

We select the maximum value over all grid points at each time step to form a time series249

of the image function maximum (IFM). The event detection is completed through peak250

detection on the IFM trace when it exceeds a given threshold value. We use a detection251

threshold of 100 with all data, a compromise value to detect micro events with ML < 1252

inside the UG network, while keeping a low number of false detections. The detection setup253

requires several parameters to be tuned (Table 1).254

Fig. 4 shows an example of a detection of a 0.8 ML event that occurred on the 18th of255

February 2017. Even in the presence of high background noise contamination, the detector256

was able to detect this earthquake occurring towards the outer limit of the UG network. The257

IFM for this event was 125, clearly above the defined threshold value of 100 (Fig. 4B). Fig. 4A258

shows the original traces for the closest stations to the source and their corresponding CFs.259

We compared the detection performance of the coherence-based detector with classical260

STA/LTA methods using the functions available in ObsPy (Beyreuther et al. 2010) for 25261

local earthquakes that occurred in June 2017. STA/LTA works only on the amplitude ratios262

and does not consider the station distance or the coherence of amplitudes. For STA/LTA263

Seismotectonics of the GGB 11

detections, we used a short time window of 0.5 s, a large time window of 35 s, a minimum264

number of stations of 3, and a detection threshold of 12. To be able to detect a 1.1 ML265

earthquake in Lake Geneva with the STA/LTA trigger, we had to decrease the detection266

threshold, increasing the number of false alarms significantly. While both methods found267

the 25 local earthquakes, with the STA/LTA trigger we obtained 2097 detections, including268

2072 false events (98.8 %). With the coherence-based detector we found 124 detections269

including 99 false alarms (79.9 %). From the 99 false detections, only 20 corresponded to270

a non-event. The remaining detections were from another type (e.g., explosions, distant or271

regional earthquakes).272

We then imported the waveforms of all detections into a SEISAN (Earthquake analysis273

software) database (Havskov & Ottemoller 1999). We visually inspected the waveforms and274

categorized them into the following types of events: local (LQ), regional (RQ), or distant275

(DQ) earthquakes, quarry or mine blasts (LE), or false alerts due to high-noise peaks at the276

stations (LU). For these microseismic events, the seismic signature from tectonic (LQ) and277

anthropogenic (LE) events is very similar. To reliably separate them and avoid misinterpre-278

tations, we labeled events as anthropogenic if they occurred within a radius equivalent to279

the location error (in km) of a known quarry (visible on satellite images).280

The phase-picking was manually executed with SEISAN. We followed the basic picking281

principles and seismogram interpretation for local events described by Kulhanek (1990),282

Bormann (2009), and Havskov & Ottemoller (2010). We located the events using the program283

HYPOCENTER (Lienert & Havskov 1995), included in the SEISAN software package, and284

using the 1D velocity model for Switzerland proposed by Husen et al. (2003). We picked285

1252 LQ events in total, with 7817 P-phases and 8864 S-phases, at an average of 6 and 7286

picks/event respectively. Only 155 of these events were located inside the study area (grey287

rectangle in Fig. 9). All picks were made by the same operator to maintain consistency. We288

used a constant VP/VS ratio of 1.78, suitable for the Molasse basin units (Vouillamoz et al.289

2016).290

The local magnitude (ML) was automatically determined using the distance attenuation291

12 Antunes et. al

function proposed by Kradolfer (1984), suitable for Switzerland (Diehl et al. 2014). The292

algorithm takes the maximum zero-to-peak amplitude on the simulated Wood-Anderson293

seismometer record of the horizontal channels. The maximum peak is taken using the max-294

imum of the modulus of the signal on a 2 s time window after the S-wave picking. The295

final magnitude value is the median of the values obtained for each station. We refer to this296

dataset as the preliminary catalogue of events.297

We used the ZMAP function (Wiemer 2001) from GISMO package (Thompson & Reyes298

2018) to determine the MC and the b-value, based on catalogue methods (Rydelek & Sacks299

1989; Woessner 2005; Amorese 2007; Mignan & Woessner 2012). The results can be seen in300

section 4.2.301

3.3 Minimum 1D P-wave velocity model302

The P- and S-wave travel times were then simultaneously inverted with the VELEST code303

package (Kissling 1988) to retrieve a ’minimum 1D velocity model’ adapted to the GGB. The304

concept of ’minimum 1D velocity model’ corresponds to an average local 1D velocity model305

that is associated with station corrections to compensate for lateral changes in velocity.306

Before the inversion, we carefully assessed the quality of the picks and assigned corresponding307

weighting values (Kissling 1988; Husen et al. 2003; Diehl et al. 2009; Husen et al. 2011).308

We developed a python algorithm that automatically computes the time uncertainty using309

an approach similar to Diehl et al. (2009) and taking the signal-to-noise ratio (SNR) into310

account. This procedure is detailed in the appendix.311

From our initial dataset, we selected events with more than 6 high quality P-phase312

readings and at least 3 readings on the UG stations; an azimuthal GAP < 180◦ and time313

misfit (RMS) error < 0.7 s. The criteria of 3 readings in the UG stations was selected314

to remove events that do not reach the basin and avoid biased results provoked by low315

magnitude events in the Alpine region.316

Our final dataset is composed of 54 events (Fig. 5) with 6-40 P-phase picks per event. The317

SNR is typically in the 8-20 dB range. The uncertainty interval of the P-wave arrival time318

Seismotectonics of the GGB 13

is generally low with an average of 0.1 s, having only a few events with uncertainties higher319

than 0.2 s. We used 761 high quality P-picks (weight 0-3) and about 550 picks were discarded320

during the weighting process (> 4). The time difference between the manual pickings (tM)321

and the automatic procedures (tA) is close to zero for the most part of the dataset. We often322

observe tM <tA. This can occur when the signal does not clearly emerge from the noise. In323

such cases the automatic routine is not able to detect it correctly. The opposite (tA <tM)324

happens less frequently. Fig. A1, in appendix, shows the data quality of the final dataset.325

Fig. 5 shows the selected events and the corresponding P-wave pseudo-ray path to each326

station. We observe a good pseudo-ray coverage for the GGB. The station OG02 was used327

as the reference for station corrections. This reference station was chosen because it is a328

permanent station deployed on bedrock, located at the center of the network, and one of the329

stations with the highest number of picks.330

We used a linear least-square regression to determine a constant VP/VS ratio of 1.70331

for the 54 selected events (Fig. 5), using PyVelest‡. The errors are close to zero and the332

data fit is high (CC=0.998), using a total of 747 station picks. The software excluded 14333

outliers (bottom panel of Fig. 5). The VP/VS ratio is computed based on the modified334

Wadati diagram analysis that assumes that the S and the P station travel times are directly335

proportional. We obtained a very similar value to the one of Diehl et al. (2014): 1.71. We336

found that the value used in the preliminary catalogue, 1.78, is suitable for sedimentary337

environments (Vouillamoz et al. 2016) but not for our dataset. This might be due to the338

fact that the majority of our events occur below a mountain range (either the Alps or Jura339

mountains).340

We ran VELEST with two ’a priori velocity models’ (initial models) retrieved from the341

literature: Husen et al. (2003), used in Switzerland, and Sismalp model (Cara et al. 2015),342

used in the French Alps (Fig. 7).343

The ’minimum 1D P wave model’ is then used together with the locally determined344

‡ https://github.com/saeedsltm/PyVelest

14 Antunes et. al

VP/VS ratio to relocate the preliminary catalogue, referred herein as ’relocated catalogue’.345

The results can be seen in section 4.1.346

3.4 Focal mechanism determination347

We estimated double-couple focal mechanisms for all possible earthquakes in the area using348

a combination of methods as described below, together with the 1D model computed in349

section 3.3. We selected events with the largest magnitudes (to ensure reasonable SNR) and350

sufficient azimuthal coverage of the focal sphere. We additionally included two earthquakes351

(ML of 3.2 and 3.4) that occurred in February 2019, outside the period of our experiment.352

The data for these events were recorded by the temporary stations that remained in place353

for seismic monitoring purposes, together with the permanent networks in the area. In total,354

we selected 8 candidate events, with ML ranging between 1.3 and 3.4 and hypocenter values355

ranging within 7-13 km.356

Standard waveform-matching techniques, used without any polarity pre-constraint, were357

found unstable for these small events. Therefore, we applied different methods in parallel358

to obtain reliable results and to cross-check the robustness of the solutions. Firstly, we used359

FOCMEC (Snoke 2003), a software based on polarity readings to determine double-couple360

focal mechanisms. Due to the small magnitude of the events, only about 10 traces per event361

show clear polarities, leading to highly non-unique solutions (e.g., more than 200 mechanism362

solutions for a 6-10 degree increment of the P-T axes, when allowing for one polarity error,363

Fig. 6). To obtain better constrained polarities, we complemented the polarity data with364

results from waveform matching techniques. We used the CSPS method (cyclic scanning365

of the polarity solutions, Fojtikova & Zahradnik 2014; Zahradnık et al. 2015) integrated366

in ISOLA (Sokos & Zahradnik 2008, 2013; Zahradnık & Sokos 2018b) using broad-band367

stations at epicentral distances shorter than 30 km. The CSPS method uses the multiple368

FOCMEC solutions and, for each of them, performs waveform inversion for source depth,369

time, and scalar seismic moment. The waveform fit is quantified by the variance reduction370

(VR). We selected solutions with VR >0.95 VRopt (where VRopt is the best obtained fit).371

Seismotectonics of the GGB 15

The output of the CSPS technique is a suite of focal mechanisms and magnitudes that372

equally fit the waveforms, giving an estimation about the uncertainties (Fig. 6).373

We sought to estimate the moment tensors of the largest events with the algorithm of374

Alvizuri et al. (2018), which performs a grid search over the full moment tensor space to find375

the best-fitting mechanism. In our focal mechanism estimates, most of the events are small,376

shallow, with relatively high frequencies (0.3-2 Hz), and their waveforms are dominated by377

S-waves and surface waves. Our velocity model for the region was derived from P-waves378

and may not account for un-modeled 3-D structure such as topography and site effects. As379

a consequence, the fit between observed and synthetic waveforms suffered from anomalous380

time shifts, amplitude discrepancies, and some later arrivals in the observed waveforms.381

Therefore, we decided to apply the waveform envelope technique (Zahradnık & Sokos 2018a;382

Carvalho et al. 2019), which is less sensitive to phase delays and site amplifications. The383

envelope fits show high correlations and similar focal mechanism solutions among events,384

except for 2 of them (F6 and F7). The waveform fits and the results from the complementary385

methods can be seen in the supplementary material.386

4 RESULTS387

4.1 Velocity model and station corrections388

The ’minimum 1D velocity model’ retrieved from the velocity inversions can be seen in389

Fig. 7 and corresponds to the best model derived from the initial model of Husen et al.390

(2003). To verify the stability of the solution, we also show the best solution derived from391

the initial Sismalp model (Cara et al. 2015), and the 30 ’best-fit’ solutions obtained for each392

of the two initial models (60 total, Fig. 7A). Even though the two initial velocity models393

are different, all best-fit models converge into the ’minimum 1D velocity model’, confirming394

that the final results are independent from the initial models. The 0 km depth corresponds395

to the mean sea level. The new ’minimum 1D velocity model’ shows a lower P-wave velocity396

for the upper layers (until 8km depth) when compared with the initial models. The velocity397

is nearly 6 km/s down to the depth of ∼ 35 km, where the velocity abruptly increases to ∼398

16 Antunes et. al

8 km/s. This abrupt increase in velocity corresponds to the depth of the Moho discontinuity,399

estimated to be at 32-34 km below the GGB (Kastrup et al. 2004).400

The station corrections associated with the ’minimum 1D velocity model’ are shown in401

Fig. 8. Table 2 includes station coordinates, elevation and geological setting according to402

Fig. 1. This provides a better understanding of the station terms and highlights the major403

velocity anomalies caused by topography and/or geology. Strong positive delays can be seen404

towards the center of the UG network (Molasse basin) and strong negative delays are found405

at the eastern limit of the study area and beyond (in the Alps). Small positive station406

corrections are found in the stations located in the Jura mountains. The time delays range407

from -0.23 s to 0.27 s for the GGB stations and between -0.27 s and -0.72 s for the remaining408

stations. The S-wave delays are stronger than the P-wave delays and range from -2.44 s to409

-0.73 s in the Alpine stations and from -0.46 s to 0.72 s in the GGB stations. In general,410

station corrections seem to correspond to topography and geology anomalies.411

4.2 Earthquake relocation412

We relocated the full catalogue of earthquakes (LQ reloc) and a selection of the best quarry413

blast events (LE reloc) with the HYPOCENTER code (part of the SEISAN software pack-414

age). We used our final 1D P-wave velocity model and respective station corrections with415

a VP/VS ratio of 1.70. The preliminary locations (LQ prelim and LE prelim, Fig. 9) were416

obtained using the Husen et al. (2003) velocity model with station corrections (computed417

using OG02 as reference station for consistency). Fig. 7B and C show the location errors418

in depth (in km) and time residuals (RMS, in s), respectively, for the preliminary locations419

and for the relocations of both types of events (LQ and LE). We significantly decreased the420

errors in the locations, especially for the LE type of events. The median of the errors in421

depth decreased from 4.6 km to 3.2 km for the LQ. The average of the RMS errors of the422

LQ did not change significantly (∼ 0.4 s for both preliminary locations and relocations) but423

changed from 0.36 s to 0.29 s for the LE type of events.424

Fig. 9 shows the preliminary locations and relocations of LQ and LE events where it is425

Seismotectonics of the GGB 17

possible to observe small differences in latitude and longitude on both types of events. The426

azimuthal gap between stations is <180◦ and the distribution of the network is homogeneous427

and dense enough to cover the events distribution. However, the differences in the depth428

location are significant. After relocation, the LQ events are located at greater depths (average429

of 2 km), reaching a maximum of 20 km, and occur slightly earlier than the preliminary430

locations (1 or 2 s). From the N-S and the E-W profiles in Fig. 9, we can observe an431

improvement on the relocations of the LQ events. In general, the epicentres of the relocated432

events cluster more than in the original locations. The depth locations are only reliable inside433

the grey rectangle in Fig. 9, where station corrections have lower values. In the Alpine area,434

the depth locations are biased by the selection of reference station.435

The 97 LE events located in the study area are weak and highly attenuated in the basin,436

leading to poor SNR ratios. The signal is only clear in a few stations, usually the closest to437

the event, which is not enough to constrain their depth properly. For this reason, we decided438

to fix their depth at zero, as we know that the quarry blasts occur at the surface. Thus,439

no depth errors can be estimated (Fig. 7B). Yet, we estimate the RMS errors associated to440

this assumption (Fig. 7C). We analyzed the time occurrence of the LE events (Fig. 10D)441

and we observed that they occur exclusively during working hours, confirming the initial442

identification as anthropogenic activities. They are more likely to happen in the morning,443

from 7-11 am (UTC time) and no events were identified between 6 pm and 7 am.444

We located 1252 LQ events in the whole region shown on the map of Fig. 9, but only445

155 in the study area (inside the grey rectangle). The seismicity is moderate with several446

swarms occurring in the Alpine region (east of the study area). Swarms are defined as seismic447

sequences with more than 10 events and if a maximum number of events per day is greater448

than twice the square root of the swarm duration in days (Mogi 1963; Matos et al. 2018). The449

swarms occurred in specific time periods and are visible as ”steps” in the cumulative number450

of events plot (Fig. 10A). The largest of the swarms in the Alpine region was located in the451

northeast of the lake Geneva and includes the largest seismic event recorded in the whole452

18 Antunes et. al

region (4.2 ML) during our experiment. We detected more than 500 earthquakes associated453

with this swarm.454

Considering the study area alone (grey rectangle, Fig. 9), only 155 LQ events were455

located within the 1.5 years period of the experiment. The seismic swarm on the southeast456

of the lake Geneva occurred in December 2016. We detected 43 events within 2 days (22nd457

and 23rd of December 2016). This swarm started with a ML 3.2 event. This was the largest458

event in the study area within the time range of this experiment. The magnitude of the459

events in the area is low and ranging from 0.7 to 2.5 ML, with a few larger events going460

up to 3.2 ML (Fig. 10B). The seismic rate in the study area is low, with an average of 100461

earthquakes/year.462

We obtained a MC value of 1.3 and a b-value of 1.14 for the study area (Fig. 10C) using463

the Goodness of Fit Test with a probability of 90 % (GFT - Mc90) (Wiemer & Wyss 2000;464

Wiemer 2001). Please note that the estimated MC is still relatively high compared to a465

possible detection threshold of ML of 0.5 (Fig. 3). This is a consequence of the low number466

of events in the GGB area — i.e., close to the center of our network — during the relatively467

short time period of our experiment.468

4.3 Regional tectonics and stress inversion469

The seismicity in the GGB is weak and scarce with only 17 earthquakes located during the470

time range of our experiment. Fig. 11 shows the final catalogue of the relocated events and471

the estimated local magnitudes together with the focal mechanisms retrieved in this study.472

We also included all focal mechanisms from the literature (L events in Table 3 and Fig. 11,473

Kastrup et al. 2004, and references therein). The earthquakes from the relocated catalogue474

are associated with known local tectonic structures. We recorded 3 earthquakes associated475

with the Vuache fault (F1, F4, and F5) and 2 micro-events at the Cruseilles and Arve (F8)476

faults. The majority of the events (F1-F5 and L1-L4) show an almost pure sinistral strike-477

slip mechanism. One of the two possible plane solutions is aligned or at least sub-parallel478

to the strike of the Vuache fault. Mechanisms L6, L7, and F8 are strike slip as well but the479

Seismotectonics of the GGB 19

possible plane solutions orientation cannot be directly associated with the orientation of a480

fault. Mechanism L8 can be associated with the thrusts in the Jura units and L5 shows a481

transpressional kinematic. We additionally obtained a normal-fault mechanism in the Pre-482

alps region (F6), where the December 2016 swarm occurred, and a reverse-fault event that483

can be associated with the Alpine thrusts front and located in Lake Geneva (F7). Here, we484

detected a small cluster of 6 earthquakes.485

Table 3 shows relevant information about the newly retrieved focal mechanisms (F1-F8),486

together with the ones retrieved from the literature, taken from Kastrup et al. (2004) and487

references therein (L1-L8). This information was used to calculate the stress field in the area488

with STRESSINVERSE (Vavrycuk 2014) for a total of 14 events. F6 and F7 mechanisms489

(Figs. 6 and 11) were excluded from the inversion due to major problems with waveform fits490

and result consistency. We decided to show the mechanisms nevertheless, due to the amount491

of polarities available. For these 2 events, waveform fits were difficult to obtain due to the492

large inter-station distance.493

The final result of the stress inversion can be seen at the top of Fig. 11 showing quasi pure494

strike slip kinematic acting in the GGB. We observe that the horizontal maximum principal495

stress, σ1, is orientated N291◦E with a dip of 4◦. The intermediate principal stress, σ2, is496

oriented N150◦E with a dip of 85◦, and the least principal stress, σ3, is orientated N22◦E497

and with a dip of 3◦. The σ1 orientation is sub-parallel to the direction of the major fault498

systems crossing the GGB: Vuache, Le Coin, Cruseilles and Arve which are characterized499

by a strike ranging from N120◦E to N150◦E (Thouvenot et al. 1998; Baize et al. 2011; de La500

Taille et al. 2015).501

The two principal fault mechanisms derived from stress inversion (Vavrycuk 2011, 2014)502

are shown in Fig. 11. The arrows point out the principal fault nodal lines that correspond503

to the fault planes satisfying the failure criterion with high probabilities of being activated504

(Vavrycuk 2011). The nodal planes are those with an angle less than 45◦ from the axis of σ1.505

The upper focal mechanism fault plane is aligned with the major strike-slip fault structures506

20 Antunes et. al

in the area. The lower focal mechanism fault plane is aligned with fault structures mapped507

to the north of the Arve fault area.508

5 DISCUSSION509

The 17 months of seismic data show a low seismic rate for the GGB, with 17 detected events510

counting a total of 12 earthquakes per year (Figs. 9 and 11). No events were located in511

the canton of Geneva, where the geothermal activities will take place. The quasi-absence512

of instrumental seismicity in the canton of Geneva (2 events between 1975 to August 2016,513

Fig. 2) suggests that the historical major events (MW > 3 from 1500 to 1975) felt in the514

canton of Geneva probably occurred along the seismically active faults crossing the GGB515

(Figs. 2 and 11). We speculate that the higher population density of the city of Geneva516

(compared to the surrounding regions) may have biased the location of these moderate517

earthquakes.518

The pre-selection of events used to retrieve our new ’minimum 1D velocity model’519

(Fig. 7A) yield to a complete pseudo-ray coverage for the GGB (Fig. 5). The station correc-520

tions are in agreement with local geology (Fig. 8), generally suggesting lower velocities for521

the Basin infill and higher velocities in the alpine area. We note that the pattern of station522

correction in the Jura mountains is more complex, showing a mix of positive and negative523

anomalies. Interestingly, the pattern of S-wave corrections correlates with the shear-wave524

velocity model of the GGB, obtained from a local ambient-noise tomography study (Planes525

et al. 2020).526

After relocation, we significantly decreased the errors in the locations (Fig. 7B and C) for527

both earthquakes and quarry blast events. In the Alpine region, the depth of the relocated528

events is biased towards greater depths due to strong station delays (Figs. 8 and 9 and529

Table 2). As a test, we a ran a VELEST inversion using an alpine reference station (e.g.,530

AIGLE), which resulted in shallower depths for the Alpine events. However, the Alpine area531

is outside the main focus of this study. Thus, we keep OG02 as a central reference station532

for the GGB.533

Seismotectonics of the GGB 21

Keeping this in mind, we want to exclude a possible bias of the velocity model induced534

by the distant Alpine events. For this purpose, we tested VELEST on a subset of 18 events535

(occurring within or in closer proximity to the GGB) and using our new ’minimum 1D536

velocity model’ as input (supplementary material). The ’subset model’ obtained with the537

subset of event is in fact equivalent to the ’minimum 1D velocity model’ with the exception538

of the very first layer (> 0 km), where the velocity is 4.41 km/s. The ’subset model’ leads539

to a significant decrease of the RMS errors of the quarry blast events (to an average of540

0.25 s), but not for the earthquakes. This implies that the ’subset model’ could be more541

accurate to locate events at the center of the GGB (Molasse basin). For what concerns the542

measured earthquakes, located mainly towards the limits of the GGB (mountain areas), our543

new ’minimum 1D velocity model’ seems to be more suitable (supplementary material).544

Of particular interest to us is whether the depth of the Vuache events is affected by the545

choice of the retrieved velocity model and respective reference station. We compared the546

relocations in the following case scenarios: original dataset with reference stations AIGLE,547

UG10, and UG03; and the subset of events using stations OG02 and UG04 as reference. We548

verified that the depth of the Vuache events remains similar for all these models, located549

about 4 km deeper than the preliminary locations (using the initial models) and with lower550

errors associated (supplementary material).551

The MC value of 1.3 computed for the study area is already a considerable improve-552

ment compared to the MC of 2.0 before our experiment. However, this value seems to be553

influenced by the seismic swarm of December 2016 located outside the UG network. The554

nearest stations are at epicentral distances of 14 km and 15 km (UG10 and AIGLE stations)555

and the remaining stations above 24 km. This affects the detection and location of such556

weak events. Inside the dense UG network, where we estimate an average of 7 km epicentral557

distance to the nearest station, more earthquakes with lower magnitudes could have been558

detected. For this reason the MC would have been lower if such events had occurred. As559

pointed out previously, when comparing the background noise level at the UG stations with560

the theoretical S-wave source spectra (Fig. 3), we verify that a theoretical value of 0.5 ML561

22 Antunes et. al

earthquake could be detected inside the UG network. For example, the coherence-based tool562

detected a 0.8 ML earthquake (Fig. 4) above the limit threshold.563

The determination of the focal mechanisms for such low-magnitude events in a complex564

tectonic region is challenging. The lateral variations of velocity, amplification effects, and565

shallow depths strongly affect synthetic computation, thus affecting the waveform correla-566

tions. Such variations at a local scale cannot be taken into account in a simple 1D velocity567

model. Besides, we constantly observed differences between the hypocenters from the re-568

located catalogue and the centroid depths from waveform inversion, with the latter being569

shallower. This can be a bias due to the selection of stations at close epicentral distances570

(<30km). This usual selection allows lower frequency ranges, with better SNR, and to be571

less dependent on the local structure and site effects. Using stations at greater epicentral572

distances, the centroid depths match better with the hypocentral values (from our new cat-573

alogue). This was verified for the largest events (F2 and F3 in Figs. 6 and 11) but not for574

the remaining events due to the poor SNR at large epicentral distances.575

The stress inversion results point out a quasi-pure strike-slip regime. This is in agreement576

with structural geology (e.g., Gorin et al. 1993; Baize et al. 2011; Charollais et al. 2013; Rabin577

et al. 2018) and geophysical (Allenbach et al. 2017) investigations suggesting that strike-slip578

faults are accommodating the converging and rotation of the Alpine front and of the Jura579

mountains.580

STRESSINVERSE yields two principal focal mechanisms (concept from Vavrycuk 2011,581

indicated by the red focal mechanisms on Fig. 11). In each focal mechanism, the principal582

plane is the one striking less than 45◦ from the maximum principal stress, σ1 (indicated by583

the arrows in the red focal mechanisms). The Vuache, Cruseilles, Le Coin, and the Arve584

faults are oriented along one of these principal planes (upper red mechanism, Fig. 11). The585

fault systems occurring to the north of the Arve fault are oriented along the other one586

(bottom red mechanism, Fig. 11). This implies that the major faults crossing the GGB are587

active or may be prone to being reactivated under the ongoing tectonic regime (Vavrycuk588

2011).589

Seismotectonics of the GGB 23

Fig. 11 points out that the seismic activity occurs mainly beneath the Alps and the Jura590

mountains. This is in agreement with GPS data pointing out the uplift of the chains (c.f.,591

Rabin et al. 2018; Houlie et al. 2018) and therefore an active deformation. We observed 1592

earthquake at the Arve fault (at 11 km depth), and 1 at the Cruseilles fault. The Vuache593

fault showed a seismic activity marked by 3 events with a ML between 2.0 and 2.1, and594

hypocenters ranging from 6.4±1 km and 7.1±2.1 km. An ongoing debate is whether the595

major fault structures offsetting the GGB reach the basement, at ∼ 4.5 km depth below596

the Vuache fault (Sambeth & Pavoni 1988; Baize et al. 2011; Thouvenot et al. 1998). The597

hypocenter values obtained in this study (shown in Fig. 9 and Table 3) point out that598

the Vuache fault (and by analogy possibly also the Arve, Le Coin, and Cruseilles faults)599

are crustal structures reaching at least 6-7 km depth. This was previously postulated by600

Sambeth & Pavoni (1988) and Baize et al. (2011). The northernmost distribution of micro-601

seismicity detected along the Vuache fault (Fig. 11) also supports Donzeau et al. (1997)602

who suggested that this fault splits into two segments when offsetting the Jura Mountains603

(Fig. 11).604

No earthquakes were associated with the Le Coin fault. Similarly, the fault systems605

identified from active seismic data in the framework of previous studies (Allenbach et al.606

2017, blue lines in Figs. 2 and 11) did not show seismic activity (Figs. 9 and 11). The short607

inter-station distance and the low detection threshold of 0.5 ML estimated for the centre of608

the network would have been sufficient to detect microseismic activity along those faults.609

This implies that such faults are either inactive or that their period of recurrence of events is610

larger than the period of our experiment. Looking at the instrumental seismicity, we observe611

only one event of ML2 occurring in the center of the GGB (in green, Fig. 2), 3 events that612

seem be associated with the Arve fault, at least 3 earthquakes falling within the limits of the613

Cruseilles fault area, and another 6 earthquakes with a disperse distribution at the NE of614

the canton of Geneva (white rectangle, Fig. 2). Regarding the 2 earthquakes located near the615

Le Coin fault, we cannot discriminate which fault systems they are associated with (either616

the Le Coin strike-slip fault or the SW-NE thrust faults).617

24 Antunes et. al

The instrumental seismicity combined with our recent data suggest that the Vuache,618

Cruseilles, and Arve faults need to be considered as active seismogenic areas reaching at619

least 5 km depth (i.e. relocated depth of 7.1±2.1 km). As for the Le Coin fault, we did620

not find clear evidences of seismic activity. The potential crustal-reaching of the local faults621

could have important implications for the seismic hazard of western Switzerland (Wiemer622

et al. 2016). The ML5.3 Epagny earthquake (L3 in Fig. 11 and Table 3) that occurred in 1996623

along the Vuache fault (Thouvenot et al. 1998) shows that this strike slip structure is able624

to generate moderate-magnitude seismic events. In particular, if considering the empirical625

formulas of Wells & Coppersmith (1994) to estimate the magnitude of an earthquake based626

on their rupture surface (in km2), we suggest a possible M6 along the Vuache fault (assuming627

the Vuache as a strike-slip fault with 30 km long and reaches 7 km depth, the deepest628

hypocenter value in this fault). Thouvenot et al. (1998) declared a possible 12 km long629

potential seismic gap in the Vuache fault and also pointed out a possible M6 in this fault.630

The depth obtained for the Vuache events can be another indicator that a possible M6 might631

occur in this strike-slip fault.632

6 CONCLUSIONS633

The study of microseismicity in sedimentary basins is challenging due to the scarce oc-634

currence of seismic events combined with low signal-to-noise ratios and the often strong635

attenuation. However, the investigation of the sporadic (yet present) natural seismicity with636

dedicated dense networks can provide useful information, even with short-term experiments.637

We deployed a dense temporary network composed of 20 broadband stations, from638

September 2016 to January 2018, around and within the Greater Geneva Basin reaching639

a detection threshold of ML of 0.5. During this time we recorded 155 events with magni-640

tudes of 0.7-3.2ML in our study area, but only 17 events were located in the Greater Geneva641

Basin, with ML ranges of 0.7-2.1. No earthquake was located in the Canton of Geneva. In-642

strumental and historical seismic catalogues show the same results (with only 2 earthquakes643

at north, in the lake).644

Seismotectonics of the GGB 25

We used the recorded seismicity to retrieve a minimum 1D velocity model for the Greater645

Geneva Basin. This new velocity model allowed us to relocate micro-seismic activity up to646

11 km depth along the main faulted areas (i.e. Vuache, Cruseilles, Le Coin, Arve) offset-647

ting the basin. Among them, the Vuache fault is the major tectonic structure crossing the648

basin. Empirical relations suggest that this fault may possibly generate M6 earthquakes.649

The inversion of the focal mechanisms has shown a tectonic deformation dominated by a650

quasi-pure strike slip regime. This is in agreement with geological data.651

From a geothermal exploration perspective, we found no events in the Canton of Geneva652

where the geothermal project is currently ongoing. However, monitoring the evolution of653

the seismicity throughout the geothermal project is highly recommended. We propose using654

the newly-computed 1D velocity model for the Greater Geneva Basin in future seismological655

investigations.656

Quantifying the seismic rates occurring before geothermal operations also helps quanti-657

fying the impact that geothermal energy extraction may have on the Greater Geneva Basin.658

ACKNOWLEDGMENTS659

This work was supported by the Services Industriels de Geneve (SIG) and partially by the660

Swiss National Fund GENERATE project (166900). Matteo Lupi is a SCCER-SoE Profes-661

sor and thanks the KTI for financial support. Jirı Zahradnık was supported by the Czech662

Science Foundation grant GACR-18-06716J. We would like to especially thank the people663

and municipalities that hosted our temporary seismic stations during the time frame of this664

experiment. We are thankful to Nicolas Clerc for providing the seismic faults interpolated665

from seismic profiles within the GeoMol project (Allenbach et al. 2017). We thank Sebastian666

Heimann and Simone Cesca for all support with LASSIE tool, to Toni Kraft for providing667

the script to compute the S-wave Brune source spectra, and to Catarina Matos for the in-668

sightful discussions. Finally, a especial thanks to Sebastiano D’Amico and an anonnymous669

reviwer for constructive comments that highly helped to improve the manuscript.670

To improve the analysis, we complemented our temporary network with the following671

26 Antunes et. al

permanent stations: AIGLE, BRANT, GIMEL, SENIN, TORNY, GRYON, SALAN, DIX,672

SCOD (CH, SED 1983); OG02, OG35, OGMY, OGSI, OGRV, RIVEL, RSL (FR, RESIF673

1995); CERN1, CERN5, CERNS (C4, CERN 2016); REMY, LSD (GU, University of Genova674

1967); MRGE (IV, INGV 2006). For the location of the 2 extra events we additionally675

used new seismic stations installed in the GGB: SAVIG (CH, SED 1983); and CI08, CI10676

and CI11, from the Grenoble temporary network (XT, ISTerre 2018-2020). The data was677

downloaded via ArcLink/WebDC request client using ObsPy (Beyreuther et al. 2010). The678

seismic data from our temporary network will be made available by the corresponding author679

on request. All maps presented were produced using the software package GMT (Wessel et al.680

2013).681

REFERENCES682

Allenbach, R., Baumberger, R., Kurmann, E., Michael, C., & Reynolds, L., 2017. Geomol: Modle683

gologique 3d du bassin molassique suisse rapport final, Accessed: 2019-02-19.684

Alvizuri, C., Silwal, V., Krischer, L., & Tape, C., 2018. Estimation of Full Moment Tensors,685

Including Uncertainties, for Nuclear Explosions, Volcanic Events, and Earthquakes, Journal of686

Geophysical Research: Solid Earth, 123(6), 5099–5119.687

Amorese, D., 2007. Applying a Change-Point Detection Method on Frequency-Magnitude Distri-688

butions, Bulletin of the Seismological Society of America, 97(5), 1742–1749.689

Baisch, S., Voros, R., Rothert, E., Stang, H., Jung, R., & Schellschmidt, R., 2010. A numerical690

model for fluid injection induced seismicity at Soultz-sous-Forets, International Journal of Rock691

Mechanics and Mining Sciences, 47(3), 405–413.692

Baize, S., Cushing, M., Lemeille, F., Gelis, C., Texier, D., Nicoud, G., & Schwenninger, J.-L.,693

2011. Contribution to the seismic hazard assessment of a slow active fault, the Vuache fault694

in the southern Molasse basin (France), Bulletin de la Societe Geologique de France, 182(4),695

347–365.696

Bethke, C. M., Altaner, S. P., Harrison, W. J., & Upson, C., 1988. Supercomputer Analysis of697

Sedimentary Basins, Science, 239(4837), 261–267.698

Beyreuther, M., Barsch, R., Krischer, L., Megies, T., Behr, Y., & Wassermann, J., 2010. ObsPy:699

A Python Toolbox for Seismology, Seismological Research Letters, 81(3), 530–533.700

Bormann, P., 1998. Conversion and comparability of data presentations on seismic background701

noise, Journal of Seismology , 2(1), 37–45.702

Seismotectonics of the GGB 27

Bormann, P., 2009. New Manual of IASPEI Seismological Observatory Practice, New Manual of703

IASPEI Seismological Observatory Practice (NMSOPM), 1(2002), 1162.704

Brune, J. N., 1970. Tectonic stress and the spectra of seismic shear waves from earthquakes,705

Journal of Geophysical Research, 75(26), 4997–5009.706

Brune, N., 1971. Correction, Journal of Geophysical Research, 76(20), 5002.707

Cacace, M., Kaiser, B. O., Lewerenz, B., & Scheck-Wenderoth, M., 2010. Geothermal energy708

in sedimentary basins: What we can learn from regional numerical models, Chemie der Erde,709

70(SUPPL. 3), 33–46.710

Cara, M., Cansi, Y., Schlupp, A., Arroucau, P., Bethoux, N., Beucler, E., Bruno, S., Calvet, M.,711

Chevrot, S., Deboissy, A., Delouis, B., Denieul, M., Deschamps, A., Doubre, C., Frechet, J.,712

Godey, S., Golle, O., Grunberg, M., Guilbert, J., Haugmard, M., Jenatton, L., Lambotte, S.,713

Leobal, D., Maron, C., Mendel, V., Merrer, S., Macquet, M., Mignan, A., Mocquet, A., Nicolas,714

M., Perrot, J., Potin, B., Sanchez, O., Santoire, J.-P., Sebe, O., Sylvander, M., Thouvenot, F.,715

Van Der Woerd, J., & Van Der Woerd, K., 2015. SI-Hex: a new catalogue of instrumental716

seismicity for metropolitan France, Bulletin de la Societe Geologique de France, 186(1), 3–19.717

Carvalho, J., Barros, L. V., & Zahradnık, J., 2019. Inversion for Focal Mechanisms Using Wave-718

form Envelopes and Inaccurate Velocity Models: Examples from Brazil, Bulletin of the Seismo-719

logical Society of America, 109(1), 138–151.720

CERN, 2016. CERN Seismic Network. ETH Zurich. Other/Seismic Network.721

Charollais, J., Wernli, R., Mastrangelo, B., Metzger, J., Busnardo, R., Clavel, B., Conrad, M.,722

Davaud, E., Granier, B., Martin, M. S., & Weipmann, M., 2013. Presentation d’une nouvelle723

carte geologique du Vuache et du Mont de Musieges (Haute-Savoie, France) stratigraphie et724

tectonique, Archives des Sciences, 66(1), 1–63.725

Chelle-Michou, C., Do Couto, D., Moscariello, A., Renard, P., & Rusillon, E., 2017. Geothermal726

state of the deep Western Alpine Molasse Basin, France-Switzerland, Geothermics, 67(2), 48–65.727

Clerc, N., Rusillon, E., Moscariello, A., Renard, P., Paolacci, R., & Meyer, M., 2015. Detailed728

Structural and Reservoir Rock Typing Characterisation of the Greater Geneva Basin, Switzer-729

land for Geothermal Resource Assessment, in Proceedings World Geothermal Congress 2015 ,730

p. 10, Melbourne, Australia.731

Clinton, J. F. & Heaton, T. H., 2002. Potential Advantages of a Strong-motion Velocity Meter732

over a Strong-motion Accelerometer, Seismological Research Letters, 73(3), 332–342.733

Courboulex, F., Deichmann, N., & Gariel, J.-C., 1999. Rupture complexity of a moderate in-734

traplate earthquake in the Alps: the 1996M5 Epagny-Annecy earthquake, Geophysical Journal735

International , 139(1), 152–160.736

de La Taille, C., Jouanne, F., Crouzet, C., Beck, C., Jomard, H., de Rycker, K., & Van Daele,737

28 Antunes et. al

M., 2015. Impact of active faulting on the post LGM infill of Le Bourget Lake (western Alps,738

France), Tectonophysics , 664, 31–49.739

Deichmann, N., Clinton, J., Husen, S., Edwards, B., Haslinger, F., Fah, D., Giardini, D., Kastli,740

P., Kradolfer, U., Marschall, I., & Wiemer, S., 2010. Earthquakes in switzerland and surrounding741

regions during 2009, Swiss Journal of Geosciences, 103(3), 535–549.742

Deichmann, N., Clinton, J., Husen, S., Edwards, B., Haslinger, F., Fah, D., Giardini, D., Kastli,743

P., Kradolfer, U., & Wiemer, S., 2011. Earthquakes in Switzerland and surrounding regions744

during 2010, Swiss Journal of Geosciences, 104(3), 537–547.745

Deichmann, N., Clinton, J., Husen, S., Edwards, B., Haslinger, F., Fah, D., Giardini, D., Kastli,746

P., Kradolfer, U., & Wiemer, S., 2012. Earthquakes in Switzerland and surrounding regions747

during 2011, Swiss Journal of Geosciences, 105(3), 463–476.748

Diehl, T., Kissling, E., Husen, S., & Aldersons, F., 2009. Consistent phase picking for regional749

tomography models: application to the greater Alpine region, Geophysical Journal International ,750

176(2), 542–554.751

Diehl, T., Deichmann, N., Clinton, J., Husen, S., Kraft, T., Plenkers, K., Edwards, B., Cauzzi, C.,752

Michel, C., Kastli, P., Wiemer, S., Haslinger, F., Fah, D., Kradolfer, U., & Woessner, J., 2013.753

Earthquakes in Switzerland and surrounding regions during 2012, Swiss Journal of Geosciences,754

106(3), 543–558.755

Diehl, T., Clinton, J., Kraft, T., Husen, S., Plenkers, K., Guilhelm, A., Behr, Y., Cauzzi, C.,756

Kastli, P., Haslinger, F., Fah, D., Michel, C., & Wiemer, S., 2014. Earthquakes in Switzerland757

and surrounding regions during 2013, Swiss Journal of Geosciences, 107(2-3), 359–375.758

Diehl, T., Deichmann, N., Clinton, J., Kastli, P., Cauzzi, C., Kraft, T., Behr, Y., Edwards, B.,759

Guilhem, A., Korger, E., Hobiger, M., Haslinger, F., Fah, D., & Wiemer, S., 2015. Earthquakes760

in Switzerland and surrounding regions during 2014, Swiss Journal of Geosciences, 108(2-3),761

425–443.762

Diehl, T., Clinton, J., Deichmann, N., Cauzzi, C., Kastli, P., Kraft, T., Molinari, I., Bose, M.,763

Michel, C., Hobiger, M., Haslinger, F., Fah, D., & Wiemer, S., 2018. Earthquakes in Switzerland764

and surrounding regions during 2015 and 2016, Swiss Journal of Geosciences , 111(1-2), 221–244.765

Donzeau, M., Wernli, R., Charollais, J., & Montjuvent, G., 1997. Carte geologique de la France766

a 1/50 000, feuille 653: Saint-Julien-en-Genevois - Orleans, BRGM - Notice explicative 62p.767

Dorbath, L., Cuenot, N., Genter, A., & Frogneux, M., 2009. Seismic response of the fractured and768

faulted granite of Soultz-sous-Forets (France) to 5 km deep massive water injections, Geophysical769

Journal International , 177(2), 653–675.770

Edwards, B., Kraft, T., Cauzzi, C., Kastli, P., & Wiemer, S., 2015. Seismic monitoring and771

analysis of deep geothermal projects in St Gallen and Basel, Switzerland, Geophysical Journal772

Seismotectonics of the GGB 29

International , 201(2), 1022–1039.773

Ellsworth, W. L., Giardini, D., Townend, J., Ge, S., & Shimamoto, T., 2019. Triggering of the Po-774

hang, Korea, Earthquake (Mw5.5) by Enhanced Geothermal System Stimulation, Seismological775

Research Letters , 90(5), 1844–1858.776

Fah, D., Giardini, D., Kastli, P., Deichmann, N., Gisler, M., Schwarz-Zanetti, G., Alvarez-Rubio,777

S., Sellami, S., Edwards, B., Allmann, B., Bethmann, F., Wossner, J., Gassner-Stamm, G.,778

Fritsche, S., & Eberhard, D., 2011. ECOS-09 Earthquake Catalogue of Switzerland Release779

2011, Tech. Rep. SED/ECOS/R/001/20110417, Swiss Seismological Service ETH Zurich.780

Flovenz, O. G., Agustsson, K., Arni Gunason, E., & Kristjansdottir, S., 2015. Reinjection and781

Induced Seismicity in Geothermal Fields in Iceland, in Proceedings World Geothermal Congress782

2015 , p. 15, Melbourne, Australia.783

Fojtikova, L. & Zahradnik, J., 2014. A New Strategy for Weak Events in Sparse Networks: The784

First-Motion Polarity Solutions Constrained by Single-StationWaveform Inversion, Seismological785

Research Letters , 85(6), 1265–1274.786

Giardini, D., 2009. Geothermal quake risks must be faced, Nature, 462, 848–9.787

Gorin, G., Signer, C., & Amberger, G., 1993. Structural configuration of the western Swiss Molasse788

Basin as defined by reflection seismic data, Eclogae Geologicae Helvetiae, 86(3), 693–716.789

Grigoli, F., Cesca, S., Rinaldi, A., Manconi, A., Lopez-Comino, J., Clinton, J., Westaway, R.,790

Cauzzi, C., Dahm, T., & Wiemer, S., 2018. The november 2017 mw 5.5 pohang earthquake: A791

possible case of induced seismicity in south korea, Science, 360(6392), 1003–1006.792

Haring, M. O., Schanz, U., Ladner, F., & Dyer, B. C., 2008. Characterisation of the Basel 1793

enhanced geothermal system, Geothermics, 37(5), 469–495.794

Havskov, J. & Ottemoller, L., 1999. SeisAn Earthquake Analysis Software, Seismological Research795

Letters, 70(5), 532–534.796

Havskov, J. & Ottemoller, L., 2010. Routine Data Processing in Earthquake Seismology , Springer797

Netherlands, Dordrecht.798

Hillers, G., Husen, S., Obermann, A., Planes, T., Larose, E., & Campillo, M., 2015. Noise-based799

monitoring and imaging of aseismic transient deformation induced by the 2006 Basel reservoir800

stimulation, Geophysics, 80(4), KS51–KS68.801

Hofstetter, R., Dorbath, C., & Calo, M., 2012. Crustal structure of the Dead Sea Basin from local802

earthquake tomography, Geophysical Journal International , 189(1), 554–568.803

Houlie, N., Woessner, J., Giardini, D., & Rothacher, M., 2018. Lithosphere strain rate and stress804

field orientations near the Alpine arc in Switzerland, Scientific Reports, 8(1), 2018.805

Husen, S., Kissling, E., Deichmann, N., Wiemer, S., Giardini, D., & Baer, M., 2003. Probabilistic806

earthquake location in complex three-dimensional velocity models: Application to Switzerland,807

30 Antunes et. al

Journal of Geophysical Research: Solid Earth, 108(B2).808

Husen, S., Kissling, E., & Clinton, J. F., 2011. Local and regional minimum 1D models for809

earthquake location and data quality assessment in complex tectonic regions: Application to810

Switzerland, Swiss Journal of Geosciences, 104(3), 455–469.811

Ibele, T., 2011. Tectonics of the western Swiss Molasse Basin during Cenozoic times, Ph.D.812

thesis, University of Fribourg, Switzerland.813

INGV, S. D. C., 2006. Rete Sismica Nazionale (RSN). Istituto Nazionale di Geofisica e Vulcanolo-814

gia (INGV).815

ISTerre, 2018-2020. Cifalps (China-Italy-France Alps seismic transect) - 2 (Cifalps-2).816

Kastrup, U., Zoback, M. L., Deichmann, N., Evans, K. F., Giardini, D., & Michael, A. J., 2004.817

Stress field variations in the Swiss Alps and the northern Alpine foreland derived from inversion818

of fault plane solutions, Journal of Geophysical Research: Solid Earth, 109(B1), 1–22.819

Kim, K.-H., Ree, J.-H., Kim, Y., Kim, S., Kang, S. Y., & Seo, W., 2018. Assessing whether820

the 2017 mw 5.4 pohang earthquake in south korea was an induced event, Science, 360(6392),821

1007–1009.822

Kissling, E., 1988. Geotomography with local earthquake data, Reviews of Geophysics , 26(4),823

659–698.824

Kradolfer, U., 1984. Magnitudenkalibrierung , Ph.D. thesis, ETH, Zurich.825

Kraft, T., Mai, P. M., Wiemer, S., Deichmann, N., Ripperger, J., Kastli, P., Bachmann, C., Fah,826

D., Wossner, J., & Giardini, D., 2009. Enhanced Geothermal Systems: Mitigating Risk in Urban827

Areas, Eos, Transactions American Geophysical Union, 90(32), 273.828

Kremer, K., Marillier, F., Hilbe, M., Simpson, G., Dupuy, D., Yrro, B. J., Rachoud-Schneider, A.-829

M., Corboud, P., Bellwald, B., Wildi, W., & Girardclos, S., 2014. Lake dwellers occupation gap in830

Lake Geneva (FranceSwitzerland) possibly explained by an earthquakemass movementtsunami831

event during Early Bronze Age, Earth and Planetary Science Letters, 385, 28–39.832

Kulhanek, O., 1990. Anatomy of seismograms, Elsevier B.V., Netherlands.833

Lacombe, O., Mazzoli, S., von Hagke, C., Rosenau, M., Fillon, C., & Granado, P., 2019. Style834

of deformation and tectono-sedimentary evolution of fold-and-thrust belts and foreland basins:835

From nature to models, Tectonophysics , 767(July), 228163.836

Lienert, B. R. & Havskov, J., 1995. A Computer Program for Locating Earthquakes Both Locally837

and Globally, Seismological Research Letters, 66(5), 26–36.838

Lopez-Comino, J., Cesca, S., Kriegerowski, M., Heimann, S., Dahm, T., Mirek, J., & Lasocki, S.,839

2017a. Monitoring performance using synthetic data for induced microseismicity by hydrofrack-840

ing at the Wysin site (Poland), Geophysical Journal International , 210(1), 42–55.841

Lopez-Comino, J. A., Cesca, S., Heimann, S., Grigoli, F., Milkereit, C., Dahm, T., & Zang, A.,842

Seismotectonics of the GGB 31

2017b. Characterization of Hydraulic Fractures Growth During the Aspo Hard Rock Laboratory843

Experiment (Sweden), Rock Mechanics and Rock Engineering , 50(11), 2985–3001.844

Lu, S.-M., 2018. A global review of enhanced geothermal system (egs), Renewable and Sustainable845

Energy Reviews, 81, 2902–2921.846

Matos, C., Heimann, S.and Grigoli, F., Cesca, S., & S., C., 2016. Seismicity of a slow deforming847

environment: Alentejo, south portugal.848

Matos, C., Custodio, S., Batlo, J., Zahradnık, J., Arroucau, P., Silveira, G., & Heimann, S., 2018.849

An Active Seismic Zone in Intraplate West Iberia Inferred From HighResolution Geophysical850

Data, Journal of Geophysical Research: Solid Earth, 123(4), 2885–2907.851

McNamara, D. E. & Buland, R. P., 2004. Ambiente noise levels in the continental United States,852

Bulletin of the Seismological Society of America, 94(4), 1517–1527.853

Mignan, A. & Woessner, J., 2012. Estimating the magnitude of completeness for earthquake854

catalogs, Community Online Resource for Statistical Seismicity Analysis.855

Mogi, K., 1963. Some discussions on aftershocks, foreshocks, and earthquake swarms, Bulletin of856

the Earthquake Research Institute, University of Tokyo, 41(3), 615–658.857

Mosar, J., 1999. Present-day and future tectonic underplating in the western Swiss Alps: recon-858

ciliation of basement/wrench-faulting and decollement folding of the Jura and Molasse basin in859

the Alpine foreland, Earth and Planetary Science Letters, 173(3), 143–155.860

Nawratil de Bono, C. & GEo-01 team, a., 2018. GEo-01: The first GEothermie 2020 P&D well861

in the Canton of Geneva - Preliminary results.862

Obermann, A., Kraft, T., Larose, E., & Wiemer, S., 2015. Potential of ambient seismic noise tech-863

niques to monitor the St. Gallen geothermal site (Switzerland), Journal of Geophysical Research:864

Solid Earth, 120(6), 4301–4316.865

Peterson, J., 1993. Observations and modelling of seismic background noise. Open-File Report866

93-322, Tech. rep., US Geological Survey.867

Planes, T., Obermann, A., Antunes, V., & Lupi, M., 2020. Ambient-noise tomography of the868

Greater Geneva Basin in a geothermal exploration context, Geophysical Journal International ,869

220(1), 370–383.870

Plenkers, K., Husen, S., & Kraft, T., 2015. A Multi-Step Assessment Scheme for Seismic Network871

Site Selection in Densely Populated Areas, Journal of Seismology , 19(4), 861–879.872

Rabin, M., Sue, C., Walpersdorf, A., Sakic, P., Albaric, J., & Fores, B., 2018. Present-Day873

Deformations of the Jura Arc Inferred by GPS Surveying and Earthquake Focal Mechanisms,874

Tectonics , 37(10), 3782–3804.875

Raffensperger, J. P. & Vlassopoulos, D., 1999. The potential for free and mixed convection in876

sedimentary basins, Hydrogeology Journal , 7(6), 505–520.877

32 Antunes et. al

RESIF, 1995. RESIF-RLBP French Broad-band network, RESIF-RAP strong motion network878

and other seismic stations in metropolitan France. RESIF - Reseau Sismologique et geodesique879

Francais.880

Rydelek, P. A. & Sacks, I. S., 1989. Testing the completeness of earthquake catalogues and the881

hypothesis of self-similarity, Nature, 337(6204), 251–253.882

Sambeth, U. & Pavoni, N., 1988. A seismotectonic investigation in the Geneva Basin, southern883

Jura Mountains, Eclogae Geologicae Helvetiae, 81(2), 433–440.884

SED, S. S. S. A. E. Z., 1983. National Seismic Networks of Switzerland.885

Signer, C. & Gorin, G., 1995. New geological observations between the jura and the alps in886

the geneva area, as derived from reflection seismic data, Eclogae Geologicae Helvetiae, 88(2),887

235–265.888

Snoke, J. A., 2003. 85.12 FOCMEC: FOCal MEChanism determinations, in International Geo-889

physics, Volume 81, Part B , pp. 1629–1630, Elsevier.890

Sokos, E. & Zahradnik, J., 2013. Evaluating Centroid-Moment-Tensor Uncertainty in the New891

Version of ISOLA Software, Seismological Research Letters, 84(4), 656–665.892

Sokos, E. N. & Zahradnik, J., 2008. ISOLA a Fortran code and a Matlab GUI to perform893

multiple-point source inversion of seismic data, Computers & Geosciences, 34(8), 967–977.894

Spicher, A., 1980. Tektonische karte der schweiz, 1 : 500’000.895

The Pyrocko Developers, 2017. Pyrocko: A versatile seismology toolkit for Python.896

Thompson, G. & Reyes, C., 2018. GISMO - a seismic data analysis toolbox for MATLAB [software897

package].898

Thouvenot, F., Frechet, J., Tapponnier, P., Thomas, J.-C., Le Brun, B., Menard, G., Lacassin, R.,899

Jenatton, L., Grasso, J.-R., Coutant, O., Paul, A., & Hatzfeld, D., 1998. The M L 5.3 Epagny900

(French Alps) earthquake of 1996 July 15: a long-awaited event on the Vuache Fault, Geophysical901

Journal International , 135(3), 876–892.902

University of Genova, 1967. Regional Seismic Network of North Western Italy. International903

Federation of Digital Seismograph Networks. Dataset/Seismic Network.904

Vavrycuk, V., 2011. Principal earthquakes: Theory and observations from the 2008 West Bohemia905

swarm, Earth and Planetary Science Letters, 305(3-4), 290–296.906

Vavrycuk, V., 2014. Iterative joint inversion for stress and fault orientations from focal mecha-907

nisms, Geophysical Journal International , 199(1), 69–77.908

Vouillamoz, N., Wust-Bloch, G. H., Abednego, M., & Mosar, J., 2016. Optimizing Event Detection909

and Location in LowSeismicity Zones: Case Study from Western Switzerland, Bulletin of the910

Seismological Society of America, 106(5), 2023–2036.911

Wells, D. L. & Coppersmith, K. J., 1994. New empirical relationships among magnitude, rupture912

Seismotectonics of the GGB 33

length, rupture width, rupture area, and surface displacements, Bulletin of the Seismological913

Society of America, 84(4), 974–1002.914

Wessel, P., Smith, W. H. F., Scharroo, R., Luis, J., & Wobbe, F., 2013. Generic Mapping Tools:915

Improved Version Released, Eos, Transactions American Geophysical Union, 94(45), 409–410.916

Wiemer, S., 2001. A Software Package to Analyze Seismicity: ZMAP, Seismological Research917

Letters, 72(3), 373–382.918

Wiemer, S. & Wyss, M., 2000. Minimum magnitude of completeness in earthquake catalogs:919

Examples from Alaska, the Western United States, and Japan, Bulletin of the Seismological920

Society of America, 90(4), 859–869.921

Wiemer, S., Kraft, T., Diehl, T., Obermann, A., Dublanchet, P., Mignan, A., & Hetenyi, G., 2015.922

Projet Geothermie Profonde a Geneve, Tech. rep., SED, ETH, Zurich.923

Wiemer, S., Danciu, L., Edwards, B., Marti, M., Fah, D., Hiemer, S., Wossner, J., Cauzzi, C.,924

Kastli, P., & Kremer, K., 2016. Seismic Hazard Model 2015 for Switzerland (SUIhaz2015), Swiss925

Seismological Service (SED) at ETH Zurich, p. 164.926

Wilhelm, J. & PGG team, 2011. Evaluation du potentiel geothermique du canton de Geneve, Etat927

de Geneve.928

Woessner, J., 2005. Assessing the Quality of Earthquake Catalogues: Estimating the Magnitude929

of Completeness and Its Uncertainty, Bulletin of the Seismological Society of America, 95(2),930

684–698.931

Zahradnık, J. & Sokos, E., 2018a. Fitting Waveform Envelopes to Derive Focal Mechanisms of932

Moderate Earthquakes, Seismological Research Letters, 89(3), 1137–1145.933

Zahradnık, J. & Sokos, E., 2018b. ISOLA Code for Multiple-Point Source ModelingReview, in934

Moment Tensor Solutions: A Useful Tool for Seismotectonics , pp. 1–28, ed. D’Amico, S., Springer935

International Publishing, Cham.936

Zahradnık, J., Fojtıkova, L., Carvalho, J., Barros, L., Sokos, E., & Jansky, J., 2015. Compromising937

polarity and waveform constraints in focal-mechanism solutions; the Mara Rosa 2010 Mw 4938

central Brazil earthquake revisited, Journal of South American Earth Sciences, 63, 323–333.939

34 Antunes et. al

LIST OF FIGURES940

1 Simplified geological map of the region surrounding the Greater Geneva Basin941

(GGB) together with the dense seismic network deployed from September 2016942

to January 2018 (UG network). The study area is outlined by the dashed grey943

rectangle. The red zones represent the 4 major fault areas in the GGB: 1-Vuache,944

2-Cruseilles, 3-Le Coin, and 4-Arve faults. The red lines represent the active fault945

segments in the fault areas (GGB A.F.) The basins composing the GGB are: GB-946

Geneva Basin, BP-Bornes Plateau, and RB-Rumilly Basin. The city of Geneva is947

represented by a star. The brown line represents the national borders: FR-France,948

CH-Switzerland, and IT-Italy. The major tectonic structures presented here were949

modified and combined considering the Tectonic map of Switzerland (Spicher 1980),950

Ibele (2011), and Kremer et al. (2014). The tectonic maps from Spicher (1980) and951

Ibele (2011) were based on geological interpretation and outcrops evaluation. The952

faults in Lake Geneva were taken from Kremer et al. (2014) and mapped using953

seismic reflection data.954

2 Seismicity map for the GGB area with information about the background955

seismicity according to the ECOS-09 and Swiss Seismological Service public cata-956

logues. Historical seismicity is represented as black pentagons, the size correspond-957

ing to the estimated MW . Instrumental seismicity is represented as circles, size958

representing MW or ML (depending on the catalogue) and color representing the959

depth. The evolution of events through time (historical and instrumental period)960

that occurred in the canton of Geneva (delimited by the brown line) and nearby961

(white square), is shown in the bottom panel. The estimated MC for the area is962

2.0 (Diehl et al. 2018). Only since 1975 have earthquakes been recorded by instru-963

mental networks, and only since then has the network density allowed a data-based964

magnitude determination (Fah et al. 2011). The buried faults (B.F.) retrieved from965

seismic campaigns are represented as blue lines.966

Seismotectonics of the GGB 35

3 PPSD plots for the UG stations using the statistical moment of 75 percentile.967

Stations deployed in the basin (blue lines) show noisier records than bedrock sites968

(black dashed lines). The theoretical Brune S-wave spectra for events was estimated969

at 7 km epicentral distance and for a magnitude range of ML-1.5 to ML2.5 (red970

dashed lines). The units were converted to the octave wide band-passed velocity in971

m/s for comparison. The USGS noise models (low and high) from Peterson (1993)972

are represented in grey.973

4 Detection of a 0.8 ML magnitude event that occurred on the 18th of February974

2017 at 23:16 and located on the Cruseilles fault: A) closest recorded waveforms975

to the source location and respective characteristic functions. The characteristic976

functions (CF) used in the shift-and-stack approach are shown in red (CF1) and977

blue (CF2) lines. The traces are ordered considering the distance between station978

and detection. B) Map of all stations used in the detection process (upper figure,979

colour scale representing the coherence values) and Image Function (IF, lower fig-980

ure) used in the detection process. The point of image function maximum (IFM),981

125 for this event, is still higher than the detection threshold used (100), indicating982

that lower magnitude events could be detected as well.983

5 Selected events used for the velocity inversion with the corresponding pseudo-984

ray paths between events and stations. Only paths corresponding to a good quality985

picking (0-3) were considered. The VP/VS ratio for the selected events is 1.70986

(bottom panel) using 747 station picks and discarding 14 picks (black points) from987

the original 761 P-picks.988

6 Focal mechanism solutions for the selected events in the area of study: com-989

parison between the output from FOCMEC allowing one polarity error, and CSPS990

results, including all solutions with VR > 0.95 VRopt. Combining polarities with991

waveform inversion methods allowed us to reduce uncertainties and improve the992

focal mechanism solutions, even for weak and shallow events. For the detailed po-993

larity pick for each event, please consult the supplementary material.994

36 Antunes et. al

7 Velocity models and error analysis for the events located in the area of study.995

A) ’Minimum 1D velocity model’ for the GGB (red bold line) and best-fit solutions996

(grey and blue solid lines) derived from the initial models (dashed lines) (Husen997

et al. 2003; Cara et al. 2015). The result is independent from the input. B) Errors998

in depth for the preliminary locations using the Husen et al. (2003) initial model999

(dashed line) and for the relocations using the final ’minimum 1D velocity model’1000

(solid line) for the earthquake events (LQ, red bold line). C) Residual time errors1001

(RMS) for the preliminary locations and the relocations of the LQ and the quarry1002

blast events (LE, thin orange line).1003

8 P-wave (upper image) and S-wave (lower image) station corrections for the1004

’minimum 1D velocity model’. We observe positive anomalies for the stations lo-1005

cated in the basin, composed of sediments, and negative anomalies on the stations1006

deployed in the mountains on bedrock. Larger anomalies are found on remote sta-1007

tions. The S-wave corrections are stronger than the P-wave. REF - reference station1008

used with VELEST; BB - broadband station; SP - short period station; SM - strong1009

motion station.1010

9 Map of the earthquakes (LQ) and quarry blasts (LE) epicenters. The pre-1011

liminary locations were performed using the Husen et al. (2003) velocity model1012

(black dots for LQ and grey circles for LE) and the relocations were performed1013

with the new 1D velocity model (red dots for LQ and orange circles for LE). The1014

latitude and longitude of the relocated events do not vary significantly. In depth,1015

the relocated events appear deeper (∼ 3km). The area where the relocation with1016

the minimum 1D model is expected to be more reliable is shown within the grey1017

rectangle. The depth of the LE events is fixed as they are expected to be at the1018

surface.1019

Seismotectonics of the GGB 37

10 Quantitative analysis of the seismic catalogue. Upper panels show the mag-1020

nitude of events vs time (blue dots) and cumulative number of events (red line) for1021

LQ events, during the acquisition period: A) whole region; B) area of study (note1022

different vertical scale). C) The Gutenberg - Richter relation and MC for the area1023

of study using the Goodness-of-fit Test with a probability of 90 % plotted with Z-1024

MAP (GFT - Mc90) (Wiemer & Wyss 2000; Wiemer 2001). The scarce seismicity1025

in the area is affecting the MC value. D) Origin time (UTC, in hours) of the events1026

identified as LE (quarry blasts). LE events occur mainly during working hours.1027

11 Geological and tectonic map with the relocated events, their magnitudes ML,1028

and retrieved focal mechanisms (F, in black). The earthquakes retrieved from the1029

literature (L) are taken from Kastrup et al. (2004) and are shown in grey. Their1030

sizes scale with magnitude and their values can be consulted in Table 3. The two1031

new earthquakes (F2 and F3), that occurred in February 2019, are represented1032

as open circles. The stress inversion result is on the top of the map, with σ1, σ2,1033

and σ3 in red, green, and blue respectively. The orientation of the principal fault1034

mechanisms obtained from stress analysis are represented as red mechanisms. The1035

associated black arrows indicate the principal fault nodal lines. The white dashed1036

line represent the limits of the GGB. Legend of the map features can be found on1037

Figs. 1 and 2. ∗ - excluded from the stress-inversion.1038

A1 Evaluation of the picking quality process before the new velocity model cal-1039

culation: tM - manual picking; tL - the latest time the pick can be done (1.5x1040

maximum of noise); tE - the earliest time automatically computed (point where1041

tangent is zero); tA - average between tE and tL; Unc - uncertainty of the pick1042

(biggest interval between tM, tL and tE); 2 - weight value assigned to this example.1043

Dataset quality to be used with VELEST and P-wave quality information based1044

on the automatic procedure developed are shown in the left bottom panel.1045

38 Antunes et. al

A2 Examples of pick assignment. Red line is tM, green line is tA. tE and tL are1046

represented by the two green dots. The uncertainty is the greatest interval between1047

tM, tA, tE and tL. The numbers above the plots represent the respective weight1048

value assigned.1049

LIST OF TABLES1050

1 Parameters used in the setup of the automatic detector. A very different1051

weight was needed to balance the influence of the two CFs which are normalized1052

in different ways.1053

2 Station coordinates (latitude, in °N and longitude, in °E), elevation (meters1054

above the sea level), station type (BB-broadband; SP-short period; ST-strong mo-1055

tion), station corrections for P- and S-waves (both in seconds) and geological setting1056

where the station is deployed (GS).1057

3 List of the selected events for which we computed focal mechanisms (F),1058

together with the ones retrieved from the literature (L), according to Kastrup et al.1059

(2004) and references therein. The Table headers correspond to: OT - origin time;1060

Lat Lon HZ - catalogue location (latitude, in °N; longitude, in °E; and hypocentral1061

values, in km); ML - local magnitude; Filter - frequency range used for waveform1062

inversion, in Hz; MW - moment magnitude; CZ - centroid depth, in km; s/d/r1063

- strike, dip and rake of possible fault planes, in degrees, obtained using CSPS1064

method; Comments - F1-F8: alternative name, L1-L8: reference number in Kastrup1065

et al. (2004).1066

A1 Weighting values proposed by Diehl et al. (2009) for the P-waves and S-waves1067

based on the uncertainty interval (in seconds). No uncertainty value is obtained1068

(NaN) if the routine is not able to detect tL or the if the time residual of the1069

picking is higher than 0.8 s (considered as a phase misindentification). The SNR is1070

used to increase the weight. Weight higher than 5 is assigned as 5.1071

Seismotectonics of the GGB 39

Figure 1. Simplified geological map of the region surrounding the Greater Geneva Basin (GGB)

together with the dense seismic network deployed from September 2016 to January 2018 (UG

network). The study area is outlined by the dashed grey rectangle. The red zones represent the

4 major fault areas in the GGB: 1-Vuache, 2-Cruseilles, 3-Le Coin, and 4-Arve faults. The red

lines represent the active fault segments in the fault areas (GGB A.F.) The basins composing the

GGB are: GB-Geneva Basin, BP-Bornes Plateau, and RB-Rumilly Basin. The city of Geneva is

represented by a star. The brown line represents the national borders: FR-France, CH-Switzerland,

and IT-Italy. The major tectonic structures presented here were modified and combined considering

the Tectonic map of Switzerland (Spicher 1980), Ibele (2011), and Kremer et al. (2014). The tectonic

maps from Spicher (1980) and Ibele (2011) were based on geological interpretation and outcrops

evaluation. The faults in Lake Geneva were taken from Kremer et al. (2014) and mapped using

seismic reflection data.

40 Antunes et. al

Figure 2. Seismicity map for the GGB area with information about the background seismicity

according to the ECOS-09 and Swiss Seismological Service public catalogues. Historical seismicity

is represented as black pentagons, the size corresponding to the estimated MW . Instrumental

seismicity is represented as circles, size representing MW or ML (depending on the catalogue) and

color representing the depth. The evolution of events through time (historical and instrumental

period) that occurred in the canton of Geneva (delimited by the brown line) and nearby (white

square), is shown in the bottom panel. The estimated MC for the area is 2.0 (Diehl et al. 2018).

Only since 1975 have earthquakes been recorded by instrumental networks, and only since then has

the network density allowed a data-based magnitude determination (Fah et al. 2011). The buried

faults (B.F.) retrieved from seismic campaigns are represented as blue lines.

Seismotectonics of the GGB 41

Figure 3. PPSD plots for the UG stations using the statistical moment of 75 percentile. Stations

deployed in the basin (blue lines) show noisier records than bedrock sites (black dashed lines). The

theoretical Brune S-wave spectra for events was estimated at 7 km epicentral distance and for a

magnitude range of ML-1.5 to ML2.5 (red dashed lines). The units were converted to the octave

wide band-passed velocity in m/s for comparison. The USGS noise models (low and high) from

Peterson (1993) are represented in grey.

42 Antunes et. al

Figure 4. Detection of a 0.8 ML magnitude event that occurred on the 18th of February 2017 at

23:16 and located on the Cruseilles fault: A) closest recorded waveforms to the source location and

respective characteristic functions. The characteristic functions (CF) used in the shift-and-stack

approach are shown in red (CF1) and blue (CF2) lines. The traces are ordered considering the

distance between station and detection. B) Map of all stations used in the detection process (upper

figure, colour scale representing the coherence values) and Image Function (IF, lower figure) used

in the detection process. The point of image function maximum (IFM), 125 for this event, is still

higher than the detection threshold used (100), indicating that lower magnitude events could be

detected as well.

Seismotectonics of the GGB 43

Figure 5. Selected events used for the velocity inversion with the corresponding pseudo-ray paths

between events and stations. Only paths corresponding to a good quality picking (0-3) were con-

sidered. The VP /VS ratio for the selected events is 1.70 (bottom panel) using 747 station picks

and discarding 14 picks (black points) from the original 761 P-picks.

44 Antunes et. al

Figure 6. Focal mechanism solutions for the selected events in the area of study: comparison

between the output from FOCMEC allowing one polarity error, and CSPS results, including all

solutions with VR > 0.95 VRopt. Combining polarities with waveform inversion methods allowed

us to reduce uncertainties and improve the focal mechanism solutions, even for weak and shallow

events. For the detailed polarity pick for each event, please consult the supplementary material.

Seismotectonics of the GGB 45

Figure 7. Velocity models and error analysis for the events located in the area of study. A)

’Minimum 1D velocity model’ for the GGB (red bold line) and best-fit solutions (grey and blue

solid lines) derived from the initial models (dashed lines) (Husen et al. 2003; Cara et al. 2015). The

result is independent from the input. B) Errors in depth for the preliminary locations using the

Husen et al. (2003) initial model (dashed line) and for the relocations using the final ’minimum 1D

velocity model’ (solid line) for the earthquake events (LQ, red bold line). C) Residual time errors

(RMS) for the preliminary locations and the relocations of the LQ and the quarry blast events

(LE, thin orange line).

46 Antunes et. al

Figure 8. P-wave (upper image) and S-wave (lower image) station corrections for the ’minimum

1D velocity model’. We observe positive anomalies for the stations located in the basin, composed

of sediments, and negative anomalies on the stations deployed in the mountains on bedrock. Larger

anomalies are found on remote stations. The S-wave corrections are stronger than the P-wave. REF

- reference station used with VELEST; BB - broadband station; SP - short period station; SM -

strong motion station.

Seismotectonics of the GGB 47

Figure 9. Map of the earthquakes (LQ) and quarry blasts (LE) epicenters. The preliminary lo-

cations were performed using the Husen et al. (2003) velocity model (black dots for LQ and grey

circles for LE) and the relocations were performed with the new 1D velocity model (red dots for

LQ and orange circles for LE). The latitude and longitude of the relocated events do not vary

significantly. In depth, the relocated events appear deeper (∼ 3km). The area where the relocation

with the minimum 1D model is expected to be more reliable is shown within the grey rectangle.

The depth of the LE events is fixed as they are expected to be at the surface.

48 Antunes et. al

Figure 10. Quantitative analysis of the seismic catalogue. Upper panels show the magnitude

of events vs time (blue dots) and cumulative number of events (red line) for LQ events, during

the acquisition period: A) whole region; B) area of study (note different vertical scale). C) The

Gutenberg - Richter relation and MC for the area of study using the Goodness-of-fit Test with a

probability of 90 % plotted with Z-MAP (GFT - Mc90) (Wiemer & Wyss 2000; Wiemer 2001).

The scarce seismicity in the area is affecting the MC value. D) Origin time (UTC, in hours) of the

events identified as LE (quarry blasts). LE events occur mainly during working hours.

Seismotectonics of the GGB 49

Figure 11. Geological and tectonic map with the relocated events, their magnitudes ML, and

retrieved focal mechanisms (F, in black). The earthquakes retrieved from the literature (L) are

taken from Kastrup et al. (2004) and are shown in grey. Their sizes scale with magnitude and

their values can be consulted in Table 3. The two new earthquakes (F2 and F3), that occurred

in February 2019, are represented as open circles. The stress inversion result is on the top of the

map, with σ1, σ2, and σ3 in red, green, and blue respectively. The orientation of the principal

fault mechanisms obtained from stress analysis are represented as red mechanisms. The associated

black arrows indicate the principal fault nodal lines. The white dashed line represent the limits

of the GGB. Legend of the map features can be found on Figs. 1 and 2. ∗ - excluded from the

stress-inversion.

50 Antunes et. al

Table 1. Parameters used in the setup of the automatic detector. A very different weight was

needed to balance the influence of the two CFs which are normalized in different ways.

Parameter P-wave S-wave

weight 30.0 1.0

f-min [Hz] 1.0 1.0

f-max [Hz] 15.0 8.0

short window [s] 1.0 –

window ratio 8.0 –

f-smooth [Hz] 0.2 0.05

f-normalize [Hz] 0.02 –

Seismotectonics of the GGB 51

Table 2. Station coordinates (latitude, in °N and longitude, in °E), elevation (meters above the

sea level), station type (BB-broadband; SP-short period; ST-strong motion), station corrections

for P- and S-waves (both in seconds) and geological setting where the station is deployed (GS).

Netw. Stat. Lat. Lon. Elev. Type P-corr S-corr GS

UG UG01 46.1755N 6.0393E 415 BB 0.20 0.64 Molasse

UG UG02 46.1183N 6.1717E 1162 BB 0.05 -0.02 Jura

UG UG03 46.0853N 6.0255E 702 BB 0.13 -0.01 Molasse

UG UG04 46.1767N 5.8812E 1056 BB 0.07 0.08 Jura

UG UG05 46.2420N 5.9657E 652 BB 0.09 -0.04 Jura

UG UG06 46.1527N 6.1270E 477 BB 0.27 0.59 Molasse

UG UG07 46.1968N 5.9838E 465 BB 0.14 0.25 Molasse

UG UG08 46.1050N 5.9172E 522 BB 0.07 0.30 Jura

UG UG09 46.1850N 5.6843E 783 BB 0.12 0.31 Jura

UG UG10 46.2690N 6.6148E 809 BB -0.15 0.25 Pre-alps

UG UG11 46.0030N 5.7852E 484 BB -0.03 0.03 Jura

UG UG12 46.2483N 6.2673E 463 BB 0.24 0.60 Molasse

UG UG13 46.3087N 5.8420E 1130 BB 0.24 -0.06 Jura

UG UG14 46.3773N 5.9123E 1046 BB 0.09 -0.37 Jura

UG UG15 45.9580N 5.9382E 562 BB -0.01 -0.17 Molasse

UG UG16 45.9688N 6.0742E 839 BB -0.16 -0.31 Jura

UG UG17 46.0898N 5.8013E 469 BB 0.16 0.18 Molasse

UG UG18 46.0647N 6.2657E 866 BB 0.09 0.12 Molasse

UG UG19 46.3042N 6.4473E 578 BB 0.18 0.49 Pre-alps

UG UG20 46.1927N 6.3327E 950 BB 0.19 0.52 Pre-alps

CH DIX 46.0811N 7.4084E 2370 BB -0.38 -1.13 Alps

CH SCOD 46.4705N 7.1287E 923 ST -0.48 -0.94 Pre-alps

CH AIGLE 46.3418N 6.9530E 785 BB -0.36 -0.45 Pre-alps

CH SENIN 46.3634N 7.2993E 2020 BB -0.50 -0.82 Alps

CH GIMEL 46.5336N 6.2655E 1094 BB 0.14 0.72 Jura

CH TORNY 46.7736N 6.9587E 710 BB 0.08 0.28 Molasse

CH BRANT 46.9380N 6.4730E 1145 BB 0.23 -0.14 Jura

CH GRYON 46.2505N 7.1111E 1282 SP -0.46 -0.73 Alps

CH SALAN 46.1442N 6.9730E 1881 SP -0.44 -1.06 Alps

FR OG02 46.1542N 6.2202E 606 BB 0.00 0.09 Jura

FR OGSI 46.0567N 6.7561E 750 BB -0.72 -1.21 Alps

FR RSL 45.6882N 6.6245E 1590 BB -0.61 -1.78 Alps

FR OG35 46.0449N 5.5718E 900 BB -0.07 0.19 Jura

FR OGMY 45.8810N 5.8893E 656 BB -0.16 -0.39 Jura

FR RIVEL 46.7590N 5.9225E 871 BB -0.21 0.17 Jura

FR OGSM 45.6093N 5.6972E 546 BB -0.17 -0.46 Jura

FR OGRV 46.3459N 4.7522E 293 BB -0.27 -0.44 Massif-central

C4 CERN1 46.2359N 6.0548E 356 BB 0.17 -0.09 Molasse

C4 CERN5 46.3099N 6.0761E 418 BB -0.23 -0.45 Molasse

C4 CERNS 46.2669N 6.0662E 463 BB 0.09 -0.17 Molasse

GU REMY 45.8378N 7.1565E 2448 BB -0.39 -1.22 Alps

GU LSD 45.4595N 7.1343E 2285 BB -0.49 -2.44 Alps

IV MRGE 45.7698N 7.0610E 1660 BB -0.22 -1.15 Alps

52 Antunes et. al

Table 3. List of the selected events for which we computed focal mechanisms (F), together with

the ones retrieved from the literature (L), according to Kastrup et al. (2004) and references therein.

The Table headers correspond to: OT - origin time; Lat Lon HZ - catalogue location (latitude, in

°N; longitude, in °E; and hypocentral values, in km); ML - local magnitude; Filter - frequency range

used for waveform inversion, in Hz; MW - moment magnitude; CZ - centroid depth, in km; s/d/r -

strike, dip and rake of possible fault planes, in degrees, obtained using CSPS method; Comments

- F1-F8: alternative name, L1-L8: reference number in Kastrup et al. (2004).

ID OT Lat Lon HZ ML Filter MW CZ s/d/r s/d/r Comments

F1 2017-08-24 21:47 46.15 5.83 6 2.0 0.3-0.6 2.3 8 161/83/-25 254/65/-172 VuacheT

F2 2019-02-05 21:32 46.03 5.68 13 3.4 0.3-0.6 2.7 6 241/87/170 332/80/3 New1

F3 2019-02-05 21:52 46.03 5.68 13 3.2 0.3-0.6 2.4 5 246/80/-174 155/84/-10 New2

F4 2017-04-10 13:30 46.04 5.95 7 2.0 1-1.5 2.0 5 332/61/7 239/84/151 VuacheM

F5 2017-12-04 18:49 45.95 6.07 7 2.1 0.3-0.6 2.5 1 35/84/-163 303/73/-6 VuacheB

F6 2016-12-22 19:50 46.35 6.76 13 3.2 0.3-0.6 3.2 3 330/42/-90 150/48/-90 Dec-alps

F7 2017-02-15 20:48 46.42 6.43 11 1.7 0.6-1 1.8 3 228/66/90 48/24/90 Lake

F8 2017-07-18 17:36 46.34 6.09 11 1.3 1-2 1.4 5 351/85/-30 84/60/-174 Arve

L1 1983-11-16 00:27 46.03 5.96 4 2.6 – – – 349/90/0 79/90/-180 145

L2 1975-05-29 00:32 46.04 6.02 0 4.2 – – – 242/70/174 334/84/20 141

L3 1996-07-15 00:13 45.94 6.09 2 5.3 – – – 316/70/-11 50/80/-160 153

L4 1980-12-02 05:58 45.83 6.28 1 4.3 – – – 302/76/-4 33/86/-166 143

L5 1994-12-14 08:56 45.96 6.43 10 4.5 – – – 332/44/29 220/70/130 152

L6 1982-11-08 13:02 46.15 6.27 4 3.8 – – – 97/62/-167 1/79/-29 144

L7 1971-06-21 07:25 46.40 5.80, 3 4.4 – – – 99/57/-166 1/78/-34 140

L8 1968-02-05 02:28 46.60 5.80 6 3.5 – – – 224/38/90 44/52/90 139

Seismotectonics of the GGB 53

APPENDIX A: PICKING WEIGHT ASSIGNMENT1072

To verify the quality of the picking, ensure consistency, compute the uncertainty, and reduce1073

human error, we used a combination of manual and automatic processes. We re-picked the1074

preliminary catalogue of events with close attention around the P phase, especially when1075

the signal and the noise may be mixed and where automatic routines can fail. Then, we1076

developed a python algorithm that automatically assigns weight values and evaluates the1077

dataset quality. The algorithm assigns the weight value considering the picking uncertainty1078

and the signal to noise ration (SNR) around the manual pick.1079

For each event, our algorithm reads the waveform corresponding to the vertical channels,1080

removes their mean, tapers and bandpass filters by 1-15 Hz Butterworth 4th order filter. Then1081

it selects a time window of 3 s before and 1 s after the manual picking (tM) for analysis,1082

Fig. A1. To keep consistency between the manual pick (using SEISAN) and the automatic1083

routine we did not implement a zero-phase filter.1084

The SNR is taken using the first 2.3 s as noise and the 1 s after tM as signal (red and green1085

areas respectively, Fig. A1). We exclude a 0.7 s interval before tM to remove the interval1086

where the signal and the noise may be mixed, even if a phase misidentification happens to1087

occur (e.g., Pg picked as Pn). We found that using only the SNR to define the quality of1088

the picking was not enough to assign the quality. Thus, we adapted the approach described1089

by Diehl et al. (2009) to compute the interval of uncertainty (unc.) around tM.1090

The uncertainty is the time interval where the pick can always be correctly done.1091

We define the ”latest picking time” tL as the earliest time where the signal amplitude1092

crosses/overpasses a threshold of 1.5 times the noise level. We recursively determined the1093

tangent slope for each point, from tL to earlier times until we obtain a tangent of zero (0.011094

times the maximum of the function) as the ”earliest picking time”, tE. tA is the average of1095

tE and tL. The uncertainty is the maximum interval between tM, tE and tL.1096

The weighting is assigned from 0 to 5 as it is proposed by Kissling (1988); Husen et al.1097

(2003); Diehl et al. (2009); Husen et al. (2011), Table A1. The algorithm additionally uses1098

the SNR to decrease the quality, by adding weight as it is shown in Table A1. Weights higher1099

54 Antunes et. al

than 5 are assigned as 5. The routine automatically writes the final weight associated with1100

the picking in the respective s-file (Nordic format; used by SEISAN). Some examples of the1101

picking assignment can be found in Fig. A2.1102

We assigned the quality of the S-picks using the same approach as for the P-waves but1103

adapting the parameters to the characteristics of the S-waves, as shown in Table A1, and1104

using a bandpass filter of 1-8 Hz. Due to lower signal-to-noise ratios, the majority of the1105

S-waves picked were discarded, remaining a total of 514 S-picks.1106

Combining manual and automatic procedures, we include the advantages of the manual1107

picks (control on the picking when the signal and the noise are mixed) with the advantages1108

of the automatic routines (consistency and stability). Thus, we were able to use picks that1109

otherwise would be discarded, increasing our final dataset.1110

Seismotectonics of the GGB 55

Figure A1. Evaluation of the picking quality process before the new velocity model calculation:

tM - manual picking; tL - the latest time the pick can be done (1.5x maximum of noise); tE - the

earliest time automatically computed (point where tangent is zero); tA - average between tE and

tL; Unc - uncertainty of the pick (biggest interval between tM, tL and tE); 2 - weight value assigned

to this example. Dataset quality to be used with VELEST and P-wave quality information based

on the automatic procedure developed are shown in the left bottom panel.

56 Antunes et. al

Figure A2. Examples of pick assignment. Red line is tM, green line is tA. tE and tL are represented

by the two green dots. The uncertainty is the greatest interval between tM, tA, tE and tL. The

numbers above the plots represent the respective weight value assigned.

Seismotectonics of the GGB 57

Table A1. Weighting values proposed by Diehl et al. (2009) for the P-waves and S-waves based

on the uncertainty interval (in seconds). No uncertainty value is obtained (NaN) if the routine is

not able to detect tL or the if the time residual of the picking is higher than 0.8 s (considered

as a phase misindentification). The SNR is used to increase the weight. Weight higher than 5 is

assigned as 5.

Weight Quality P-Unc.[s] S-Unc.[s] P-SNR [db] S-SNR [db] Weight increment

0 best < 0.05 < 0.1 < 10 < 7 2

1 good 0.05 - 0.1 0.1-0.175 10-20 7-15 1

2 fair 0.1 - 0.2 0.175-0.25 > 20 > 15 0

3 worst 0.2 - 0.4 0.25-0.3

4 discarded > 0.4 > 0.3

5 discarded NaN NaN