S-wave experiments for the exploration of a deep geothermal carbonate reservoir in the German Molasse Basin

There are many successful geothermal projects that exploit the Upper Jurassic aquifer at 2–3 km depth in the German Molasse Basin. However, up to now, only P-wave seismic exploration has been carried out. In an experiment in the Greater Munich area, we recorded S-waves that were generated by the conventional P-wave seismic survey, using 3C receivers. From this, we built a 3D volume of P- to S-converted (PS) waves using the asymptotic conversion point approach. By combining the P-volume and the resulting PS-seismic volume, we were able to derive the spatial distribution of the vp/vs ratio of both the Molasse overburden and the Upper Jurassic reservoir. We found that the vp/vs ratios for the Molasse units range from 2.0 to 2.3 with a median of 2.15, which is much higher than previously assumed. This raises the depth of hypocenters of induced earthquakes in surrounding geothermal wells. The vp/vs ratios found in the Upper Jurassic vary laterally between 1.5 and 2.2. Since no boreholes are available for verification, we test our results against an independently derived facies classification of the conventional 3D seismic volume and found it correlates well. Furthermore, we see that low vp/vs ratios correlate with high vp and vs velocities. We interpret the latter as dolomitized rocks, which are connected with enhanced permeability in the reservoir. We conclude that 3C registration of conventional P-wave surveys is worthwhile.

. For instance, recording both Pand S-waves in parallel enables the derivation of Poisson's ratio.
The theoretical principles and techniques behind the use of shear waves (S-wave) in seismic exploration were developed about 40 years ago (Puzyrev et al. 1980;Dohr 1985;Tatham 1985). Pure S-wave surveys have not gained widespread popularity, but recording and processing of P-to-S converted waves are applied to a limited extent in exploration nowadays (Hardage et al. 2011DeAngelo and Hardage 2014;Donati et al. 2016;Lu et al. 2016), mostly for solving specific tasks, such as illumining beneath gas-rich sediments or inside coal seams, or to derive fracture orientations, among other things (Stewart et al. 2003;Chopra and Stewart 2010;Wei et al. 2014).
Deep geothermal energy extraction often faces the problem of induced seismicity, which may cause failure of the entire project, and many examples are documented worldwide (e.g., Deichmann 2009;Grigoli et al. 2018;Megies and Wassermann 2014). To control the seismicity (Kwiatek et al. 2019), it is helpful to delineate the reservoir beforehand (Cuenot et al. 2008). If induced seismicity occurs, precise determination of the hypocenter location is mandatory, for which knowing both P-and S-wave velocities is advantageous.
The Upper Jurassic in the German Molasse Basin (GMB) is the most important hydrogeothermal reservoir in Germany to date. The city of Munich envisages obtaining a 100% supply of sustainable heat energy in the year 2040; geothermal heat shall contribute the majority of this, using a distributed pattern of geothermal facilities in the southern part of the city (Hecht and Pletl 2015). To this end, a 3D seismic campaign that covered the southern half of Munich was carried out in the winter of 2015/16. The Leibniz Institute for Applied Geophysics (LIAG) accompanied the exploration with a comprehensive research program (Buness et al. 2016), including S-wave measurements. The S-wave acquisition consisted of active and passive experiments (Fig. 1). The active experiment comprised generating S-waves using the 'Shover' technique, i.e., vertical vibrators sweeping with opposing phases (Edelmann 1981), as described by Wawerzinek et al. (2017). In this paper, we refer to the passive experiment that recorded the shots of the conventional 3D survey using 3-component MEMS (micro-electro-mechanical systems). The 3D multicomponent data were used to derive the spatial distribution of v p /v s within both overburden and reservoir. We show here the results of this experiment and demonstrate their significance for lithological characterisation of the reservoir, as well as their implications to determine the hypocenters of induced seismicity.

Geological setting
The basement in this area is Palaeozoic in age; locally, it contains small Permo-Carboniferous grabens (Bachmann et al. 1987). Triassic, Early and Middle Jurassic sediments transgressed over the basement, although around Munich, the first Mesozoic sedimentation was Middle Jurassic (Dogger) fine-grained clastics with carbonate beds (Bachmann et al. 1987). The Upper Jurassic (Malm) constitutes a 450-600 m thick series of limestones and marls. The upper 350 − 400 m comprises sponge, algal and coral reef systems with intervening lagoonal deposits. These beds were deposited in a shallow and warm marine environment on the European Tethyian continental shelf (Lemcke 1988;Meyer and Schmidt-Kaler 1990). The Malm carbonates form the major reservoir for geothermal energy, as investigated in this work.
Information about the lithology of the Malm reservoir is available either from outcrops about 100 km to the north, or from a limited number of deep boreholes in the surrounding area (cf. Fig. 1). The main constituents of the carbonate platform are limestones and dolomites of varying facies. Traditionally, the limestone facies is divided into a massive (or reef ) facies and a number of bedded facies. The former has a low content of clastics; whereas, the later may contain a considerable portion of clay (Schmid et al. 2005). Two processes have altered locally these primary facies: karstification and dolomitization. Dolines, which are also visible in the seismic volume (Sell et al. 2019), evidence the paleo-karstification during Cretaceous-early Paleogene times. Dolomitization is known from surrounding boreholes, and on the basis of 17 wells that penetrate the reservoir, Böhm et al. (2013) concluded that the main reservoir consists of dolomitized massive facies. Mraz (2018) discovered that different types of dolomites found in the reservoir (planar-anhedral and planar-euhedral) show strongly varying permeability, depending on the porosity of the primary limestone facies and the existence of fault and/or fractures. Since secondary dolomitization Fig. 1 Overview map of seismic measurements in the Greater Munich area. The 3D P-wave conventional survey was carried out in the southern and western part of Munich and accompanied by S-wave experiments. The eastern half of the shots (green dots) was also recorded by 3C receivers (red dots). The black dashed line marks the part of the 3D volume covered by S-wave information. The thick dashed grey line gives the position of the geological section shown in Fig. 2a, the yellow line gives the position of the seismic section shown in Fig. 2b. The large blue rectangle outlines the 3D cube used for PS processing (inlines 1-230, from WSW to ENE; crosslines 1-172, from SSE to NNW). The small blue rectangle marks the area of the v P /v S analysis in the Upper Jurassic unit  needs pathways for Mg-rich fluids, the dolomitization process is more probable in a facies with high primary porosity and in tectonically deformed regions.
From the late Jurassic to Early Cretaceous times, the water depth receded, ultimately causing exposure of the carbonates and, thus, strong karstification. At the end of the Cretaceous, and with the renewed onset of sedimentation, the area turned into a foredeep in front of the Alpine Orogen (Ziegler et al. 1995;Shipilin et al. 2020). In the study area, the Jurassic carbonate platform is overlain in places by thin (< 50 m) Cretaceous Purbeck units and Eocene Priabonian limestones, which are considered in the seismic image as the top of the carbonate platform sensu lato, because of their strong and continuous reflection character.
The formation of the Molasse foreland basin (i.e., the GMB) from the Eocene onwards was due to either simple collision of the African and Adriatic microplates (Frisch 1979;Allen et al. 1991;Ziegler et al. 1995;Ziegler 1987), or the African, Iberian and European plates, previous to the Alpine Orogeny (Kley and Voigt 2015). As a consequence of the down-flexure of the European crust, synthetic and antithetic, Alpine orogen-parallel striking normal faults developed (e.g. Meyer and Schmidt-Kaler 1990;Roeder and Bachmann 1996;Dussel et al. 2016;von Hartmann et al. 2016;Shipilin et al. 2020).
The Tertiary Molasse sediments consist of a succession of marine and fluviatile layers (Sissingh 1997). The thickness of the Molasse increases from north to south in general; in the study area, the Molasse reaches a thickness of 2.0-2.8 km (Fig. 2). The Molasse can be subdivided into five groups, in order of stratigraphic age, from oldest to youngest

Survey
The 3D seismic survey covered 170 km 2 in the southern and western parts of Munich and was carried out from November 2015 to March 2016 (Fig. 1). The acquisition was targeted on the Upper Jurassic (Malm) reservoir. The top of the Malm carbonates varies at depths between 2000 m in the north and 2800 m in the south. To account for this, the survey layout used increasing line distances from 400 m in the northwest to 500 m in the southeast (Schuck and Hecht 2017), with shot and geophone distances of 50 m. Three 12 t vibrators excited a 12 s long sweep with a frequency of 12-96 Hz, nine times at each location. The shots were recorded by groups of 12 1C geophones spread over 50 m, with a sampling rate of 2 ms and a record length of 5 s to include S-wave signals.
About half of the ~ 7000 shots were additionally recorded at 467 3C receivers that were deployed along one main profile, running WNW-ESE and on two perpendicular crosslines, in the eastern part of the survey (Fig. 1). This layout was chosen because of the active S-wave experiment (Wawerzinek et al. 2017). The 3C recordings used single micro-electro-mechanical systems (MEMS, Sercel DSU3 sensors). The horizontal H1 components were aligned along the direction of each line, i.e. the H1-components on the main line point towards 100°N, geophones on the crosslines point towards 190°N. To discriminate both datasets, the normal 3D seismic survey is referred to hereafter as 'conventional' survey and the multicomponent survey is referred to as '3D3C' survey.

Methods
The processing of the P-wave data of the whole conventional survey recorded by geophones followed the usual steps including refraction and residual static, spherical divergence correction, deconvolution, time-variant band-pass filtering, velocity analysis and CRS stacking (e.g., Yilmaz 2001). The final sections (Fig. 2) were derived by pre-stack depth migration, the velocity model was produced by a tomographic inversion of residual moveouts. In our study, the conventional data set is used for the facies classification and the P-wave reflectivity analysis.
To ensure comparability of the individual components of the 3D3C data set to each other, we applied a similar processing sequence to the individual components, but a few differences did occur (e.g., rotation, binning). The 3C data recorded by MEMS were first integrated to transfer the acceleration recordings of the MEMS into velocity values comparable to the geophone recordings. The next step was a rotation of the horizontal components of the recordings into a radial and transversal coordinate system, according to the receiver-shot positions. We also tested rotation based on the maximization of energy around the P-wave onset on one of the horizontal components. Only 7% of the receivers showed deviations larger than 30°. Since the first-break signal quality breaks down at offsets > 1500 m, we chose to keep rotation based on geometry.
Furtheron, we applied elevation datum statics, true amplitude recovery, a bandpass filter of 12-60 Hz, automatic gain control of 500 ms, and performed an interactive velocity analysis on the 3C data (for P-and PS-waves) before stacking. Despite rotation, no clear reflection events were evident in the PS-waves, neither in the shot-or CDP domains, nor even in CDP supergathers. Since we did not know at the beginning whether we had registered (1) reflected S-waves, generated at or in the vicinity of the seismic source, or (2) signals converted at the reflecting point, we tested a binning schema developed for P-to-S conversion (Tessmer and Behle 1988). This takes into account the different wavetypes for the downgoing-and upgoing legs by calculating an asymptotic conversion point (ACP, Fig. 3a). The distance between the conversion point and the receiver (x p ) is calculated according to: where v p and v s are velocities for P-and S-waves, respectively. Strictly, this point gives the exact position only for an infinite depth of the reflector and leads to reflector point dispersal and hence smearing effects, if the offset exceeds the reflector depth. Whereas this condition is met for most of the traces regarding the horizons below the top of the (1) Fig. 3 Acquisition principles and parameters. a Principle of ACP-binning (modified after Behle & Dohr 1985). If a downgoing P-wave from shotpoint SP is converted at a reflector, the conversion point (CP) moves towards the receiver (G), compared to a reflection point of the same wavetype (CMP). Black and blue arrows mark polarizations, i.e., particle motions, x is the offset between SP and G, x p and x s are the distances from SP to CP and from CP to G, respectively. b Distribution of absolute offset of our 3D3C dataset. The offset varies from 0 to 6852 m, with a median value (black line) of 2756 m. c Map of CMP and ACP fold of our 3D3C dataset. In the case of ACP-binning, the fold is higher but covers a smaller area. The cyan line marks the position of the cross section in Fig. 7. The blue star marks the position of the ACP bin shown in Fig. 3d and the green star marks the center of the supergather shown in Fig. 6. d Offset and azimuthal coverage of an example ACP bin. The azimuthal coverage is limited due to the linear receiver deployment carbonate platform (Fig. 3b), lateral blurring of horizons above has to be taken into account. A further problem was that the v p /v s ratio was not known beforehand. In this study, we use a nearby VSP survey to derive initial values (see below). Since the reflection points move towards the receiver, we observe a higher fold in a smaller area (Fig. 3c). To improve the signal-to-noise ratio, we increased the bin size from 25 m used in the conventional P-wave processing data to 50 m for the 3C data processing. After stacking, we applied a mild f-xy deconvolution to attenuate random noise and a FD time migration on the 3C data (for P-and PS-waves).
Once a PS-wave section is constructed, the derivation of the v p /v s ratio is possible, if the reflectors can be associated with those in the P-wave section. In this case, the interval traveltimes allow the derivation of the v p /v s ratio, according to Garotta (1987) and Tessmer and Behle (1988) where ∆T PS and ∆T P are two-way-traveltime (TWT) differences of PS-and P-waves, respectively. The velocities of P-, S-and PS-waves are related as follows (Garotta 1987): In our study, we used the 3D3C P-and PS-wave volume to determine the two-waytraveltime differences.
In principle, multicomponent surveys are suited to infer azimuthal anisotropy or horizontal transverse isotropy (HTI) caused by (near) vertical fractures or horizontal stress, amongst other things (Liu and Martinez 2012). Azimuthal anisotropy splits the medium into a fast and slow direction and leads to azimuthal-dependent arrival times and amplitudes (e.g. Bale et al. 2009, Far and Hardage 2014, Donati et al. 2016). In our survey, the azimuthal coverage was not sufficient to observe traveltime or amplitude variations (Fig. 3d), since only a limited number of receivers was deployed along the 2D lines (cf Fig. 1). Existing anisotropy would cause deterioration of the P-SV stack due to azimuthal-dependent traveltime deviations.

Initial v p /v s ratio derived from VSP measurements
The Stadtwerke München conducted a near-vertical VSP at the borehole Freiham TH-1, just outside the 3D-seismic area (Fig. 1). We rotated the horizontal components into radial and transverse directions, based on maximization of energy around the P-wave first arrivals. After rotation, the data show remarkably clear S-waves, despite the use of a vertical vibrator source. The vertical component (Fig. 4a) shows a strong reflection at the top of the carbonate platform, i.e., Top Priabonian (~ 1600 m below sea level). This reflection is also prominent on the radial component (Fig. 4b), where the reflection of the downgoing P-wave splits into an upgoing P-wave and an upgoing S-wave, i.e., a P-to-S converted wave. Several downgoing S-waves are visible as well, some of them were generated by the transition of the downgoing P-wavefield at interfaces at depth, as confirmed by P-wave reflectivity (Fig. 4a). However, one of the downgoing S-phases can be followed to the position of the vibrator at the surface, indicating that S-waves were also excited directly at the (P-wave) vibrator. Downgoing S-waves also appear on the transverse component, indicating heterogeneity of the uppermost subsurface layer. Similar observations were described by . Theoretically, the differently polarised S-waves of the radial and transverse components allow to estimate anisotropy. In practice, however, this was not feasible, due to the relatively large errors in S-wave picking, due to interference with other wavetypes. P-and S-wave interval velocities (Fig. 5a) were derived using smoothed first-break onsets, as well as a smoothing operator of 60 m to account for pick uncertainties. The resulting v p /v s values (Fig. 5b) for the Molasse sediments are very high (about 3) at the surface and decrease below 2 at 900 m depth. Below 900 m, we observe stronger variations in the velocity ratios. v p /v s were averaged for the known stratigraphic sequences: The uppermost 500 m has a median value of 2.66; it varies between 1.9 and 2.1 for the deeper Molasse units (Fig. 5b). Unfortunately, only v p /v s ratios for the Cenozoic sediments were determined, since no clear S-wave onsets are visible within the carbonate platform. Another VSP recorded at Geretsried GEN-1 borehole, about ~ 25 km south of Munich (Dussel et al. 2019), revealed a similar high median value of 1.9 for Chattian and Rupelian sediments at a depth interval of 2000-4000 m. Perhaps the greater depth of the formation and consequent stronger consolidation of the Chattian and Rupelian caused the v p /v s ratio at Geretsried to be lower than that at Freiham. Thus, we used the averaged v p /v s ratio of 2.0 (depth interval 515-2015 m) for the initial converted-wave processing (Fig. 6). This resulted in a clear improvement of the stacked signals and enabled a further areal adjustment of the values using normal moveout velocity analysis.

Horizons
The 3D3C P-and PS-wave volume were recorded only along three 2D lines, which resulted in a far lower coverage than the conventional 3D P-wave volume. Nevertheless, the 3D3C P-wave volume imaged similar reflectors to the conventional 3D seismics (cf. Wawerzinek et al. 2017). The signal-to-noise ratio in the PS-section is lower than in the P-wave volume, since the PS conversion is generally weaker than a pure P-reflection (McCamy et al. 1962;Stewart et al. 2002). Nevertheless, we were also able to pick several horizons in the 3D3C P-and PS-volume, to a varying extent for PS-waves (Fig. 7), whereby Top Chattian and Top Priabonian were the most traceable. Therefore, we focus on three horizons to construct traveltime maps: (1) Top Chattian, (2) Top Priabonian and (3) base reservoir. These horizons were selected to investigate the spatial v p /v s distribution in both the Molasse and the reservoir. Using Eq. (2), we derived the v p /v s ratio down to these horizons and between Top Chattian and Top Priabonian as well as between Top Priabonian and base reservoir.

V P /V S ratio in the Molasse
The reflection at the top of the carbonate platform is the most prominent throughout the dataset; it could be picked well on both the P-and the PS-sections of the 3D3C volume. The structures of the TWT maps of both sections show good correlation (Fig. 8a, b), indicating reliable picking of the PS section. We could derive the v p /v s ratio (Fig. 8c) for the entire Cenozoic pile (Molasse and Quaternary). The spatial display shows that the velocity ratio increases from the western part to the southeastern part of the study area. The derived v p /v s ratios vary between 2.0 and 2.3, with the median at 2.15, which is in very good accordance to the median value of 2.08 (0-2 km depth) found in the VSP at Freiham (Fig. 5). However, this value is influenced strongly by the very high (> 2.5) ratios in the first few hundred meters.
To receive a more unbiased spatial distribution, an interval v p /v s ratio between Top Chattian and Top Priabonian was computed. The calculation of interval v p /v s ratio has  Fig. 3c. Note the different scales of the sections. Light orange: Top Chattian, orange: Top Rupelian, dark orange: Top Priabonian, blue: base reservoir (Base Tithonian). The stronger noise on the PS-wave volume is due to the lower reflection coefficients of the PS-waves the advantage that potential time errors (e.g., incompletely corrected static effects) are eliminated. In the lower part of the Molasse about 90% of the values vary smoothly between 1.8 and 2.1, with a median of 1.92 (Fig. 9). The residual 10% of v p /v s ratio spread over a range of 1.5-2.4, probably caused by pick uncertainties due to noisy traces. The median v p /v s ratio is again in good agreement with the ratio of 1.99 found at VSP Freiham in the same stratigraphic interval, but at a slightly shallower position (1350-2015 m). Similar to VSP Freiham, the traveltime-derived v p /v s ratio of the entire strata down to Top Priabonian is generally higher than in the lower part of the Molasse (cf. Figures 8 and 9), reflecting the influence of the higher near-surface v p /v s ratio.

V P /V S ratio in the Upper Jurassic
Reflections inside the reservoir are not as continuous as in the Molasse units, in both 3D3C P-and PS-wave seismic volumes. We could only reliably pick a horizon below the Top Priabonian in the southernmost part of the PS-wave seismic volume, were the fold is higher (cf. Figs. 4 and 7). This horizon is situated ~ 360 ms (PS time) below Top Priabonian; it corresponds presumably to the base of the reservoir. The v p /v s ratio of the reservoir shows a considerably stronger variation than in the lower Molasse units, it varies between 1.5 and 2.2. The variation is organized in a specific pattern, i.e., a circle of low (1.50-1.80) values is surounded by high (1.85-2.20) values (Fig. 10).
The imaging of the picked horizon required velocity analysis of the PS-seismic volume; we did this using a grid of 10 × 10 lines. Due to the relatively low signal-to-noise ratio, supergathers of 25 ACPs were built and yet it was still not possible to determine the velocity function at all grid points. Thus, a refinement of analysis positions did not seem meaningful. Since the shear wave velocity can not directly be determined from velocity analysis, v s was derived using Eq. (3). In terms of absolute values, the v p /v s ratio derived from the velocity analysis at sparse locations in both seismic volumes do not agree exactly with the velocity ratios derived from traveltime differences. But in terms of relative trends, the both ratios correlate with each other (Fig. 10). Since we focus on the spatial distribution of the velocity ratio and its relation with the facies and the lithology, the v p /v s ratio derived from traveltime differences (Eq. 2) will be used in the following.

Comparison with classification and reflectivity
To gain insight into the very heterogeneous facies distribution in the Malm units, we The PS-wave section shows the same structural trends as the P-wave section, indicating consistent horizon interpretation. The resulting v p /v s ratio shows a median value of 2. 15 Wawerzinek et al. Geotherm Energy (2021) 9:6 and (3) similarity. Each of these quantities were divided into two (similarity) or three (amplitude and frequency) categories. The differentiation of the seismic signals into classes gives a clearer overview of the reflectivity distribution of the subsurface (Fig. 11). Especially circular structures with high coherent amplitudes (reddish colors) enclose a core of lower reflectivity (bluish colors), which are interpreted as carbonate build-ups or bioherms surrounded by sedimentary troughs (Fig. 11; von Hartmann et al. 2019). During the diagenetic processes, the higher primary porosity of these later structures enabled their pronounced dolomitization, in contrast to the adjoining sediments (Mraz et al. 2019).  Figure 11 shows the resulting seismic facies classification map of the eastern part of the survey, from a layer in the middle level of the carbonate platform (von Hartmann et al. 2019). Its pattern shows good correlation with the v p /v s ratio distribution that was derived in this work for the whole Upper Jurassic reservoir. High amplitudes in the seismic facies classification along an arcuate-shaped structure coincide remarkably well with high v p /v s ratio values. Since the seismic facies classification is dominated by amplitude values, we infer that high v p /v s ratios correlate with high reflectivity. The correlation of the v p /v s ratio and the facies classification extends to the adjacent layers, since the main feature, the arcuate-shaped structure encircling a low amplitude area, is also visible there.
The dependency of reflectivity, the seismic velocities v p and v s , and their ratio can be further clarified by comparing the conventional P-wave seismic sections and the attributes of the 3D3C P-and S-wave volumes (Fig. 12). Two sections are chosen, one dominated in its middle part and to the right by very low v p /v s ratios and another one dominated by high ratios. The left one shows that the variation from 1.5 to 2.0 correlates closely with reflectivity: low ratios corresponds to a diffuse reflectivity and higher values with more coherent reflectivity. Moreover, low values of the ratio correspond to high values of both v p and v s . However, the correlation with reflectivity is much better for the v p /v s ratio than for v p or v s on their own. Similar behavior is shown by the right section in Fig. 12: Strong and continuous reflectivity in the NNW part is characterized by a high v p /v s ratio, although neither the v p nor the v s velocity reach high values. This confirms the superior role of the v p /v s ratio as a discriminator with respect to the values of both seismic velocities.

Significance for lithology
Numerous studies have shown that the v p /v s ratio is a good indicator of carbonate lithology, most of them using laboratory experiments (Pickett 1963;Domenico 1984;Marion and Jizba 1997;Johnston and Christensen 1992;Wang and Szata 1999), but also log data (Miller 1992) and reflection seismic data (Miller et al. 1994;Robertson 1987;Pardus et al. 1990). On the basis of laboratory measurements on core samples that cover effective stresses from 0-415 bar and porosities of the carbonates that range from 2-25%, Pickett (1963) concluded that the v p /v s ratio can be used well as a discriminator, since limestones always show a higher ratio than dolomites.
Another major factor that determines the v p /v s ratio, besides lithology, is porosity (e.g., Rafavich et al. 1984). However, the relation between porosity and v p /v s is complex and depends, among other things, on the shape of the pores and/or fractures that form the porosity. According to the commonly used petrophysical model of Toksöz et al. (1976), a high aspect ratio (> 0.1) decreases the v p /v s ratio; whereas, a low aspect ratio increases the v p /v s ratio (Robertson 1987). With regards to experimental data, Robertson (1987) claims positive correlation of porosity and v p /v s ratio, based on a few reflection seismic data, whereas Anselmetti and Eberli (1997) deduced a positive correlation between v p (> 4000 m/s), taken as a proxy for porosity, and the v p /v s ratio, on the basis of laboratory data. Likewise, Ding et al. (2019) found in laboratory experiments v p /v s ratio increases with increasing density of fluid-filled, randomly oriented fractures (and thus decreasing v p ). The last two findings (Anselmetti and Eberli 1997;Ding et al. 2019) contradict the hypothesis that higher porosity caused by, e.g., stronger subseismic fracturing might Fig. 12 Correlation of reflectivity, velocity and v p /v s . Two P-wave seismic sections (left and right) from the conventional 3D seismic, overlain with v p /v s ratio, v p and v s velocities. Center; seismic facies classification and v p /v s ratio from Fig. 11, showing position of the sections. The sections were chosen because of their different v p /v s ratios and seismic facies classifications. v p /v s shows a good correlation with P-wave reflectivity. The v p / v s ratio was derived using the interval traveltimes (Eq. 2), whereas the smoothed velocities (v p , v ps ) were determined using interactive velocity analysis, from which v s was derived using Eq. 3. Top Upper Jurassic (Top Priabonian) horizon is indicated by the green line. The sections are flattened to Top Priabonian cause the observed decrease of the v p /v s ratio in our data (Fig. 12), since in our case, it is accompanied by an increase in both v p and v s velocity.
So far, we have considered the v p /v s ratio. The velocities of both wave modes in carbonate rocks were investigated by Marion and Jibza (1997). They found that both Pand S-wave velocities of dolomite are higher than those for limestone, independent of porosity.
An example of the application of the technique used here was presented by Pardus et al. (1990) for oil exploration in the Scipio field in the US. They used data of separately measured P-and S-wave profiles to derive v p /v s ratios above and within the reservoir. Although both of their profiles do not show any structural or stratigraphic features, a significant decrease of the velocity ratio coincides with a dolomitized part of the target layer, and even a relative estimate of the amount of dolomitization was possible, verified by a number of boreholes.
In our case, the reason(s) for the strong variation of the v p /v s ratio in the reservoir will definitely be clarified only after the reservoir is drilled and/or cored. However, if our hypothesis is proved, the data will then be valuable for further reservoir analysis.

Implications for seismicity
In the vicinity of our study area, seismicity induced by two geothermal projects has gained public attention (Seithel et al. 2019). They are the Poing project (Ziegler and Heidbach 2020), approximately 15 km to the east, and the Unterhaching project that adjoins the Munich area to the south (see Fig. 1). In the latter area, Unterhaching, seismicity, with M l up to 2.4, occurred shortly after the production started in 2007, which led to a setup of a local passive seismic network. Megies and Wassermann (2014) found that the majority of the epicenters fall within 500 m of the deepest part of the open-hole section of the injection well Unterhaching GT-2, suggesting it originates from injection. Hypocenter depth was determined to be 1.4-1.9 km below the open-hole section (Fig. 13a). This calculation was based on a v p /v s ratio of 1.65, which Lueschen (2012) derived from a 3D seismic volume in Unterhaching (Lueschen et al. 2011). However, an analysis of earthquake localisation residuals by Megies and Wassermann (2014) resulted in a higher v p /v s ratio: For the case of a half-space model, the minimum of residuals was reached at value of 1.94. This value is very close to the median value of 1.92 that we calculated in this study for the lower part of the Molasse above the carbonate platform. It is lower than the mean v p /v s ratio of 2.1, estimated for the entire Cenozoic and Malm strata combined (depth interval 0-3.1 km). However, this mean value is highly affected by the high near-surface v p /v s . The high v p /v s ratio values are further supported by a S-wave survey carried out in 1985, about 70 km east of Munich: for the majority of the Cenozoic sediments, a value of 1.9-2.0 was derived (Edelmann 1985). The consequences for earthquake hypocenter depth are drastic (Fig. 13b): Hypocenters are translated from deep into the basement (~ 1.5 km) to a position very close to the top of the basement, only some hundreds of meters below the open-hole section.

Conclusions
We recorded about half of the shots of a conventional 3D seismic survey on 467 single 3C receivers spread over three 2D lines. Our results show that most of the recorded energy on the horizontal components is from P-waves that converted to SV-waves. Despite the limited number of 3C receivers, we were able to build a seismic 3D PS-wave volume around the 2D lines that allowed the picking of horizons in the Molasse units as well as in the carbonate platform. We were able to identify the v p /v s ratio for different depth intervals and stratigraphic units. We use it to discriminate between dolomitized  Fig. 13 Effect of v p /v s on the hypocentral depth (Megies and Wassermann 2014, rearranged). a Original hypocenter depths of seven events at Unterhaching based on a v p /v s ratio of 1.65 (black) and hypocenter depths based on a half-space model (red) with a ratio of 1.94 that was derived from rms residual minimization. b Individual variation of hypocenter depths of the seven events shown for a half-space model. As the v p /v s ratio approaches 2, hypocenter depths move closer to the top of the crystalline basement. Reprinted from Geothermics, 52, Megies T, Wassermann J, Microseismicity observed at a non-pressure-stimulated geothermal power plant, 36-49, (2014), with permission from Elsevier and non-dolomitized reservoir rocks. Strong dolomitization in the Malm reservoir is connected with enhanced permeability (Böhm et al. 2013;Mraz et al. 2019). Additionally, the knowledge of the v p /v s ratio of the Molasse units above the reservoir enables more reliable hypocenter determination of induced seismicity.
A starting point for our processing originated from the evaluation of VSP measurements in the surrounding of the study area. In spite of only a P-wave source being used in all cases, some of the VSP show S-waves on the horizontal components. The reason why they only appear on some of the VSPs is unknown and should be further researched. Since almost all VSPs are realized using 3C geophones (except the emerging technique of DAS recording), we recommend they are inspected for S-waves and therefore the recording time during the surveys should be approximately twice as long as necessary for P-waves. This is low cost, and it may give valuable S-wave information. Wawerzinek et al. (2017) showed that there is no significant loss in data quality using single 3C sensors instead of geophone groups for P-waves. Due to the additional benefit of S-wave information, we argue for the use of 3C single sensors in future surveys in this area, the GMB (and perhaps elsewhere too). By recording 3C on the entire 3D receiver grid, one can achieve a much better-defined S-wave volume than for the sparsely sampled data in this study. Regular distribution of this kind of data would also open the door for more sophisticated interpretation techniques that involve angle and azimuth effects.