Basal seismicity cannot be attributed exclusively to glacier stick-slip motion. As shown by previous studies on temperate Alpine glaciers, there also exist basal seismic sources, which are not due to pure shear mechanisms. Their moment tensors have substantial, if not dominant, isotropic components. Based on first motion data of high-quality seismic records from Gornergletscher and Triftgletscher, Switzerland, we argue that the observed isotropic components can be explained by tensile faulting. The implied coseismic volumetric change can be both positive (fracture opening) and negative (fracture collapse). We attribute these observations to hydraulic processes near water-filled cavities, whose connectivity to the subglacial drainage system changes over time. Thus, our proposed icequake source mechanisms cannot be reconciled with pure shear sources at the glacier bed, which would be expected for basal stick-slip motion. This sliding mode has recently been proposed as a “realitively common” mechanism, which can substantially enhance subglacial erosion. The existence of several seismic source mechanisms (tensile, shear, or some combination of the two) of basal icequakes implies that a solid understanding about the nature of these events is indispensible if conclusions about glacier sliding, subglacial erosion, and other basal processes are to be drawn from observed seismicity.
 The subglacial environment is host to key ice flow processes including glacier uplift, basal sliding, and subglacial cavity dynamics [e.g., Iken, 1981]. Rigorous theoretical treatments of subglacial processes exist [e.g., Schoof, 2010]. However, the inaccessibility of the glacier bed has limited data observations to point measurements [e.g., Cohen et al., 2006] and remote-sensing techniques [e.g., King et al., 2008]. As glacier sliding [Anandakrishnan and Bentley, 1993; Wiens et al., 2008; Walter et al., 2011; Winberry et al., 2013], fracture opening [Neave and Savage, 1970; Mikesell et al., 2012], and fluid resonances [Metaxian et al., 2003; O'Neel and Pfeffer, 2007; West et al., 2010] emit elastic waves, seismic monitoring is becoming an increasingly common alternative to in situ and remote-sensing studies of subglacial processes. Resulting from such studies, there exists compelling evidence for seismogenic stick-slip motion beneath Antarctic ice streams and outlet glaciers [Anandakrishnan and Bentley, 1993; Zoet et al., 2012; Winberry et al., 2013]. These observations led Zoet et al.  to argue that subglacial erosion theories should include stick-slip motion, as this can explain rapid crack growth within bedrock. Their study furthermore refers to seismic events occurring on smaller mountain glaciers but whose source mechanisms and locations are unfortunately poorly constrained. Nevertheless, the authors suggest that these events exhibit similar mechanisms to their Antarctic counterparts and conclude that seismogenic stick-slip motion is common to various glaciers and ice streams worldwide.
 Here we present seismic data from four high-density campaign seismic networks on Gornergletscher and Triftgletscher, both located in the Swiss Alps. These seismic experiments have been documented by Walter et al. , Walter et al. , Walter , Walter et al. , and Dalban Canassy et al. , referred to from hereon as W2008, W2009a, W2009b, W2010, and D2013, respectively. Detailed description of instrumentation, data processing, and results on icequake source mechanisms is found in these references. The present paper focuses on the first motions of icequakes at the glacier base. Clusters of these basal events were recorded during all of the discussed seismic deployments. However, contrary to what was expected, our results argue against stick-slip motion as potential source processes for the vast majority of events. Instead, we suggest they are evidence for the opening and closing of tensile faults in response to subglacial cavity formation. In order to test these assertions and further clarify the role of subglacial hydraulics in the generation of basal seismicity, we provide specific suggestions for future field studies.
2 Glacier-Seismological Experiments
 The present study focuses on data collected during three separate seismic campaigns on Gornergletscher (Figures 1 and 2) in the summers 2004 (W2008), 2006 (W2008), and 2007 (W2009b), and one seismic campaign on Triftgletscher (Figures 1 and 2) during summer 2010 (D2013). With apertures of a few hundred meters, the seismometer networks contained between 8 and 24 short period seismometers, which were installed in direct contact with the glacier ice (Figure 2). The Gornergletscher networks also contained at least one deep borehole seismometer installed at depths between 40 and 250 m. Due to the high-frequency character of englacial seismic sources, the recording systems were operating at sampling frequencies of 1000 Hz or higher. All four networks were located in the glacial ablation zone, where surface melt of several meters each summer is typical [Huss et al., 2008]. Such high melt rates demanded daily station visits to assure sensor alignment or the deployment of borehole instruments, even for near-surface seismometers.
 The Gornergletscher and Triftgletscher catalogs contain up to several thousand “icequakes” per day. Applying linearized (W2008) or probability-based (D2013) inversion algorithms to arrival time observations, icequake sources have been located with uncertainties of a few tens of meters. This has revealed that most observed icequake sources locate near the glacier surface. The predominant icequake source process is therefore the formation or extension of surface crevasses, which are typically confined to the top 20 m of the ice column [Paterson, 1994]. Even in the absence of high-precision locations, surface-crevassing icequakes are readily identified by their dominant Rayleigh phase [Deichmann et al., 2000].
3 Basal Icequake Seismicity
 Although far less numerous than near-surface events, icequakes occurring near the glacier base were confirmed during all deployments. In contrast to near-surface icequakes, their waveforms are characterized by impulsive P- and S-arrivals (Figures 3 and 4). At the same time, they lack a strong Rayleigh phase, which can be recognized by its retrograde elliptical particle motion [Deichmann et al., 2000]. Basal icequakes typically form clusters, active from a few days to at least several weeks (Table 1; see also W2008 and D2013). In the context of the present work, a cluster was defined in a spatial sense, only, while seismograms from a single basal cluster may show a high degree of similarity, too (W2008; D2013). Occasionally, however, there exist important differences in first motion polarity (W2009b; D2013), frequency content, and seismogram length (W2010).
Table 1. Summary of the Activity, Approximate Depth, and Polarity of Icequakes Belonging to Basal Clusters Beneath Gornergletscher (GG) and Triftgletscher (TG)a
aThe 2006 Gornergletscher cluster consists of five subclusters discussed in Walter et al. . Polarity of the Triftgletscher events is also given in D2013. Start and end dates refer to the first and last icequake detection of the corresponding cluster. As these dates are typically close to the beginning or end of the seismometer deployments, the clusters were likely active for a longer time. “Up” and “down” refer to the number of compressional and dilatational single-polarity events. The number of mixed polarity events is given in the last column.
15 June 2004
1 July 2004
16 June 2004
1 July 2004
21 June 2006
23 July 2006
16 June 2007
22 July 2007
20 July 2010
31 July 2010
24 July 2010
5 August 2010
 An example of a basal icequake cluster beneath Gornergletscher is shown in Figures 2 and 5. Figure 5 shows Gornergletscher's surface and radar-derived bed topography [Huss, 2005] along an east–west cross section. The cross section intersects the epicenter of the basal icequake cluster detected in 2007 (Table 1 and Figure 2) whose hypocenter locations are also shown. The figure illustrates an important weakness of basal icequake studies. For most events, the location uncertainties are too large to determine whether the events occurred directly at the ice-bedrock interface, just above, or just below it. This problem is furthermore amplified by uncertainties and limited spatial resolution of radar measurements used to determine the glacier bed topography [Huss, 2005].
 Despite the uncertainties in hypocentral locations, there exists evidence that icequakes recorded beneath Gornergletscher, Triftgletscher, and Unteraargletscher [Deichmann et al., 2000] occur within the glacier. The waveforms of basal icequakes qualitatively resemble seismograms of other events, whose locations unequivocally place the source above the glacier bed and below the surface-crevassing zone [Deichmann et al., 2000; W2009a]. Furthermore, waveform modeling using sources placed a few meters above the bedrock successfully reproduced the direct P- and S-phases of basal icequakes beneath Gornergletscher (W2010).
3.1 First Motion Polarity
 The polarity of direct P-waves can vary between icequakes belonging to the same basal cluster. As shown in Figures 3 and 4, this “first motion polarity” can be either compressional (“up”) or dilatational (“down”). The sources of the shown waveform pairs locate within the same cluster, and therefore, the reversal of first motion polarity can be best explained by differences in seismic source mechanisms.
 Following previous investigations of icequake source mechanisms [e.g., Anandakrishnan and Bentley, 1993; Carmichael et al., 2012; DALINPRESS], we investigate first motion polarity for basal icequakes recorded on Gornergletscher. We focused on two basal clusters from the year 2004 (W2008), one cluster from 2006 (consisting of 5 subclusters documented in W2008), and one cluster from 2007, which is also shown in Figures 2a and 5 (W2009b). We only considered events, which were recorded by the majority of seismometers, having high signal-to-noise ratios and no precursory distortion. The results of the first motion analysis are presented in Table 1. In total, 340 out of the 500 analyzed basal icequakes had clear first motions, with 337 (99.1%) exhibiting the same polarity at all recording stations. Out of these 337 events, 239 (70.9%) had compressional first motions at all recording stations, whereas 98 (29.1%) had exclusively dilatational arrivals. Only 3 (0.9%) of the 340 high-quality events showed both compressional and dilatational first arrivals.
 We refer to the predominant events as “single-polarity events”, in contrast to “mixed-polarity events”, for which at least one seismometer recorded a first motion polarity different from the rest of the stations. Our results are in line with the findings of D2013, who found 28 single-polarity events (7 compressional and 21 dilatational) and only 1 mixed-polarity event at the base of Triftgletscher (clusters TG 2010-1 and TG-2010-2 in Table 1).
3.2 Seismic Source Mechanisms
 In the following, we discuss the first motion polarity of our basal icequakes under the hypothesis that their sources are some superposition of shear and tensile faulting, as suggested by D2013. This allows for both pure shear and pure tensile events as end members of the possible mechanisms. Both of these end members have been shown to exist in Alpine glacier ice: Using full-waveform inversions, W2010 determined pure tensile cracks as source mechanisms of 14 events of basal cluster GG-2004-1 (Table 1). On the other hand, near the glacier surface, W2009a found both pure shear and pure tensile-opening dislocations. We furthermore adopt the view that stick-slip motion manifests itself as pure shear faulting icequakes near the glacier base. Basal tensile faulting, on the other hand, is likely related to different subglacial processes.
 Whereas pure shear icequakes are described by deviatoric “double-couple” moment tensors, pure tensile sources contain an isotropic moment tensor component in addition to a deviatoric component (W2009a). The superposition of the two is called the crack-plus-double-couple (“CDC”) model [Minson and Dreger, 2007]. In this sense, pure tensile sources radiate only one first motion polarity, whereas pure shear (double-couple) sources radiate a quadrantal pattern [Aki and Richards, 2012]. Projecting the first motion polarity onto the focal sphere (small hypothetical sphere centered at the icequake hypocenter) produces the well-known “beachball diagrams” (Figures 6b and 6c). Beachball diagrams exhibit equal areas of compressional and dilatational first motion reflecting shear fault geometry. A set of first motion observations that cannot fit such a constraint indicates a source that includes nonshear components [Ross et al., 1996]. As the basal icequakes recorded on Gornergletscher and Triftgletscher emit predominantly single-polarity first motions, we argue against pure shear dislocations caused by basal stick-slip motion. This is a crucial result because shear dislocations emitting the expected mixed first motion polarities had previously been observed beneath Antarctic ice streams [e.g., Anandakrishnan and Bentley, 1993].
 For single-polarity icequake sources (the majority of the discussed events), the tensile component of the moment tensor has to exceed a certain strength compared to the shear component, in order to mask the mixed-polarity P-wave radiation of the latter. Furthermore, a single-polarity compressional event implies that the source undergoes a positive volumetric change (crack opening), whereas a single-polarity dilatational event implies a crack collapse. This concept can be fit into a quantitative statement by describing a CDC source as a dislocation with parallel fault planes and a dislocation vector, which lies between the fault normal and fault parallel direction (supporting information). Thus, the angle θ between dislocation vector and fault normal determines the relative strength of the tensile and double-couple components. With θ = 0° (180°), the CDC source is a pure tensile opening (closing); with θ = 90°, it is a pure shear dislocation. Using general properties of CDC moment tensors [Tape and Tape, 2012], one can derive a condition for θ when the CDC source produces single-polarity P-waves, only (supporting information). For θ < 70° and θ > 150°, an icequake source will only radiate compressional and dilatational P-waves, respectively. For CDC events with opening crack component, the condition that θ < 70° is likely a conservative lower bound because full waveform inversions of selected events (W2010) revealed tensile-opening mechanisms with negligible shear components (θ ≈ 0°).
 There are well-known problems with polarity measurements if the array recording an event only samples a portion of the focal sphere. This typically occurs when the azimuthal coverage of the recording seismometer network is small. In this case, mixed first motion polarity does not necessarily confirm a shear fault, because there exist nondouble-couple seismic sources, which emit mixed polarity P-waves [e.g., Julian et al., 1998]. The ambiguity is even more severe if the recording array covers only those quadrants of the focal sphere in which all P-waves happen to have the same polarity. In this case, the array records waves with only a single first motion polarity, although the source radiates both polarities.
 The effect of limited station coverage is illustrated in Figure 6, which shows the radiation pattern of a tensile crack opening source (Figure 6a) and of a double-couple source (Figures 6b and 6c). The locations of two hypothetical seismometers are projected onto the focal spheres as indicated by red dots in Figures 6a–6c. The tensile crack opening source (Figure 6a) radiates compressional first motion, only, which is recorded by both stations. However, the same station configuration will also record exclusively compressional first motion for the pure shear source shown in Figure 6b. Therefore, for this source-station geometry, first motion polarity cannot distinguish between a tensile crack opening and a shear source with fault plane orientations as in Figure 6b. Source discrimination is possible only if the station locations project onto two neighboring quadrants of the focal sphere (Figure 6c).
 The station coverage of the Gornergletscher and Triftgletscher basal clusters is shown in Figure 6d. The network monitoring the 85 events from Gornergletscher's 2006 basal cluster (W2008) covers an extended region of the focal sphere. The remaining clusters are recorded at somewhat poorer azimuthal coverage. However, it is unlikely that for all clusters, the fault planes of a hypothetical shear source were oriented in such a way as to emit single-polarity P-waves in exactly the focal sphere quadrant where seismometers were installed. For the Triftgletscher clusters (TG-2010-1 and TG2010-2), D2013 furthermore present an argument against those shear sources that are associated with basal stick-slip motion: The authors show that for most of their events, the observed first motion is incompatible with pure shear sources whose fault planes are parallel to the glacier bed and whose slip direction is consistent with glacier surface displacement. We therefore reject the possibility that the discussed single-polarity icequakes are due to double-couple sources (in particular stick-slip events) recorded at limited network coverage.
4 Hydraulic Model
 The opening and closing of basal fractures can be viewed as a change to the damage state of the basal ice layer. In the basal environment, high shear stresses [Dahl-Jensen, 1989] and high hydrostatic pressures are likely the dominant influence of the damage state. In the following, we present a conceptual hydraulic model that focuses on a link between basal ice damage and changing hydrostatic pressures. The occurrence of basal icequakes during low or falling subglacial water pressures supports such a link, as does the pronounced diurnal fluctuation of subglacial water pressures observed beneath Gornergletscher (W2008).
 In addition to the dependence of basal icequakes on subglacial water pressures, our hydraulic model is constrained by three central observations: (1) the majority of basal icequakes recorded on Alpine glaciers are tensile-opening and closing events, potentially superimposed on shear dislocation (as suggested in the present work), (2) W2008 show that the events of cluster GG-2006 occurred during times when the glacier surface was lowering, and (3) full moment tensor inversions of cluster GG-2004-1 reveal pure tensile-opening events with bed-parallel fault planes (W2010).
 Figure 7 shows a schematic of diurnal variations in subglacial water pressure (Figure 7a) and a cartoon of the glacier base with the lee side of a bed undulation (Figures 7b–7e). Above the undulation, there exists some degree of damage in the form of microfractures, similar to what is generally encountered on glacier surfaces. Even in the absence of local hydraulic forces, such damage may be advected from further up-glacier or it may be induced by simple shear near the bed. When the basal water pressure is low, the basal ice is well coupled to the glacier bed (Figure 7b). During increasing subglacial water pressures, a water-filled cavity can open on the lee side of the bed undulation (Figure 7c). This assumes that water from the subglacial drainage system can efficiently reach the lee side of the bed undulation. Conversely, we assume here that no such connectivity exists between the preexisting fractures and the subglacial drainage system or the cavity. The pressure may increase to a point when the glacier reaches floatation level and locally decouples from the bed (W2008). When subglacial water pressure falls again, the basal ice next to the cavity recouples to the bed (Figure 7d). At this point, the basal ice is still at overburden. The cavity pressure, on the other hand, may fall below this value to the point that it creates suction on the ice above it. Additionally, gravity pulls on the ice constituting the lower crack faces of any preexisting near-horizontal fractures. The two forces (suction and gravity) can induce tensile-opening icequakes as the fractures above the cavity open further. Pressurized water occupying the preexisting fractures will further expedite fracture opening.
 A different situation exists if the cavity is suddenly sealed from its surroundings (Figure 7e). Such sealing can occur due to shifting subglacial till or collapsing subglacial channels in response to basal sliding. In this case the glacier will also recouple to its bed as the subglacial water pressure falls. However, the water cavity will remain pressurized and act as a “pinpoint” (an isolated zone of high water pressure), which supports more than overburden pressure as long as the surrounding ice has not completely coupled to the bed. The resulting compressive deviatoric stress will close the previously formed fractures producing icequakes with tensile closing mechanism.
 In summary, the conditions for tensile closing and opening icequakes are reversed. In the former case, the counterforce to the ice weight is ‘focused’ above the cavity. In the latter case, this counterforce is partially shifted away from the cavity to the surrounding bed. In both cases, high basal shear stresses [Dahl-Jensen, 1989] can impose some shear faulting component in addition to tensile dislocation.
 A central conclusion of this work is that basal stick-slip motion cannot account for the basal seismicity we have observed beneath Gornergletscher and Triftgletscher. Instead, we find that the presented basal icequake sources must contain a tensile component in addition to any minor shear faulting. The relative magnitude of the two is poorly constrained. For single-polarity events, the source mechanism lies between a pure tensile crack and dislocation along a direction crossing the fault normal at 70°.
 The presented hydraulic model is highly conceptual and likely oversimplified. It is based on the idea that subglacial water pressures dominate the damage evolution of ice near water-filled cavities. Numerical studies of basal sliding and cavity formation [e.g., Gagliardini et al., 2007] could be used to clarify to what degree basal shear forces and cavity water pressures influence the ice damage state.
 It is interesting to note the temporal changes in event types observed for the 2007 Gornergletscher cluster (Figure 8). On 27–28 July 2007, tensile closing events dominate the seismicity of this cluster. Over the following week, tensile-opening events dominate. In terms of the proposed hydraulic model, this suggests that the responsible subglacial cavity was first disconnected from the subglacial drainage system and became connected after 28 July 2007. These two states of subglacial cavities have been proposed as limiting cases of the hydraulic regime in which subglacial cavities generally reside [Lliboutry, 1976].
 Our hydraulic model has further implications for subglacial erosion. Pressure reduction in water cavities at the lee side of bed undulations induces deviatoric stresses in the underlying bedrock. This promotes crack growth and hence quarrying [Iverson, 1991; Cohen et al., 2006]. Zoet et al.  proposed that cavity pressure drops in response to seismogenic stick-slip motion significantly enhance this mechanism. For a 10 cm high and 30 cm long idealized cavity, the authors show that stick-slip motion results in cavity pressure drops as high as 10 MPa. Even for a brief moment, this is enough to substantially accelerate crack growth in bedrock.
 Following the discussion in Zoet et al. , we investigate if tensile cracks near the glacier bed can enhance subglacial erosion just like stick-slip events. We consider an end-member of the CDC events, namely a pure tensile icequake with an isotropic moment magnitude of −1.0 [Walter et al., 2010]. This corresponds to a volume change of 0.004 m3 (using equation (1) in Müller ). If this icequake occurs close enough to a water-filled cavity, the cavity volume will change by the same amount but with opposite sign. In case of a pure tensile crack collapse, the cavity volume will grow by 0.004 m3, a 4% increase assuming an initial 0.1 m3 cavity volume. At this point we further assume that no water can be exchanged between the crack and the cavity, at least not on the small time scales of icequake rupture. In other words, as the tensile crack collapses, no water enters the cavity and hence the associated cavity volume increase results in a cavity pressure decrease. Multiplying the 4% fractional volume increase with the reciprocal of water compressibility [Goodman, 1989] yields an 80 MPa pressure drop inside the cavity (assuming a 5*10−10 Pa−1 water compressibility at 0° as given in Fine and Millero ). This pressure drop is likely a lower bound, because the chosen initial cavity volume is large compared to the idealized two-dimensional geometry employed by Zoet et al. . An 80 MPa drop exceeds the magnitude of melt-induced subglacial water pressure fluctuations, which can be as high as 1 MPa, corresponding to a 100 m water column (W2008). Furthermore, 80 MPa corresponds to the ice-overburden pressure for ice thicknesses of about 8.9 km. Consequently, for ice thicknesses below this value, an 80 MPa pressure drop temporarily results in vanishing cavity pressures assuming that the subglacial water pressure is normally near ice-overburden pressure. Collapsing tensile cracks near the glacier bed can thus enhance subglacial erosion just like basal stick-slip events [Zoet et al., 2013]. However, in contrast to the tensile crack collapse, a tensile crack opening event will decrease the volume of a nearby water-filled cavity. The resulting pressure increase will hamper rather than favor crack growth in bedrock and resultant quarrying.
5.1 Directions of Future Research
 In order to explain the existence of tensile faulting with positive and negative volumetric change during falling subglacial water pressures (W2008; DALINPRESS), we proposed a hydraulic model of subglacial cavity evolution. This model is furthermore supported by previous results of full moment tensor inversions, which showed that tensile fault planes of events belonging to cluster 2004-1 (Table 1) were approximately bed parallel (W2010). The limited number of basal events, for which information about fault geometry exists, is a weakness of our current study. In this sense, it is not clear if there also exist basal events with vertical fault planes, for example. If this were the case, then there may also exist other tensile source processes, such as vertical propagation of basal crevasses [van der Veen, 1998; Harper et al., 2010]. Unfortunately, moment tensor inversions of basal icequakes are computationally expensive, as they require the generation of 3-D Green's Functions to account for bed topography (W2010). Furthermore, such full waveform inversions are contingent upon high signal-to-noise ratios and are only feasible for relatively simple waveforms. At the same time, deriving fault plane orientations with alternative techniques, such as P-to-S ratio measurements is complicated by phase conversions at the glacier base (W2010). For these reasons, more information about fault mechanisms and geometries of basal icequakes are hard to come by, although they could certainly improve our understanding of the underlying seismic source processes.
 Since our hydraulic model invokes water-filled cavity dynamics and glacier uplift, it can be tested with seismic monitoring of Alpine glaciers outside the melt season. Lack of basal seismicity in the absence of melt water will support the model. Extended time series of basal icequake detections and subglacial water pressure may also allow for more reliable statistics of the diurnal activity of basal icequakes. Previous results on Gornergletscher have so far only shown that basal icequakes occur during cold day times when subglacial water pressures are low or falling (W2008). However, the data did not reveal a phase shift in diurnal activity of tensile crack opening and closing events. For simplicity, our hydraulic model therefore explains both tensile crack opening and closing events by low-pressure periods. So far, the occurrence of opening or closing events depends only on the connection of the underlying cavity to the subglacial drainage system. However, if the two types of events were systematically associated with different subglacial water pressures and/or rates of change thereof, our model would have to be modified accordingly.
 Longer detection time series of basal icequakes will also clarify the role of uplift. Complying with the observation of basal icequakes during glacier surface lowering (W2008), our model requires substantial uplift in order to produce basal icequakes. However, even in the absence of substantial uplift, the diurnal evolution of subglacial cavities leads to diurnal redistribution of basal stresses (see, e.g., Figure 9 in Iken ). Long-term seismic monitoring in combination with measurements of uplift and subglacial water pressures will therefore clarify if uplift is a necessary condition or if evolving subglacial cavities suffice in producing basal icequakes.
 Most importantly, like other studies of basal seismicity, our investigation suffers from our inability to determine whether icequakes occur directly at the ice-bed interface, just below, or above it. We argue for the latter, based on moment tensor inversions (W2010) and the resemblance of basal and intermediate-depth icequakes [Deichmann et al., 2000; W2009a; W2010]. The occurrence of icequakes near the bed of till-based Unteraargletscher also argues against tensile dislocation within bedrock [Deichmann et al., 2000]. A more solid argument for in-ice faulting could be made if seismic head waves [Aki and Richards, 2012] traveling along the ice-bedrock interface were observed. Although head waves have been observed in response to explosions (DALINPRESS), the networks on Gornergletscher and Triftgletscher were too small to study the development of head waves emitted by the basal icequakes presented here. Therefore, larger seismic networks would likely provide additional constraints on basal icequake hypocenters.
 Finally, we acknowledge that the deployment of high-frequency seismometers (natural frequency at or above 1 Hz) as used on Gornergletscher and Triftgletscher (W2008; DALINPRESS) constitutes another limitation of the present data set. On Antarctic ice streams, basal stick-slip motion generates a low-frequency signal (below 0.04 Hz) [Walter et al., 2011] in addition to high-frequency seismicity (>1 Hz) [Winberry et al., 2013]. However, it is questionable if such low-frequency signals should be expected on our Alpine glaciers, for which stick-slip motion would likely take place on considerably smaller fault planes.
 Returning to the question posed in the title of this manuscript, we stress that stick-slip motion is not the only possible source of basal icequakes. The type of basal seismicity presented here likely contributes little to glacier motion and depending on the coseismic volumetric change, it will accelerate or hamper subglacial erosion. If this type of seismicity is typical for Alpine glaciers, our results may indicate fundamental differences of the subglacial environments of Alpine glaciers studied here and previously investigated ice streams in Antarctica, where seismogenic stick-slip events are common. Accordingly, basal seismicity has yet to be fully understood before using it as a proxy for glacier sliding, subglacial erosion, or other processes near the glacier base.
 This work was financed by the European Union through the IceDISC project (grant PIEF-GA-2011-299195). The Gornergletscher and Triftgletscher field campaigns were funded by the Swiss National Science Foundation (grants 200021-103882/1, 200020-111892/1) and by the EU-FP7 “ACQWA” Project 650 (www.acqwa.ch, contract 212250), respectively. Vaclav Vavrycuk provided his MATLAB® code for plotting focal mechanisms. We thank N. Deichmann for help with acquisition and processing of the Gornergletscher data. The discussions with W. Tape and C. Tape were extremely helpful in the derivation of a polarity condition for CDC moment tensors. M. Funk and K. Hutter provided crucial input for basal ice fracturing and the conceptual cavity model. The comments and suggestions of Erin Pettit, Josh Carmichael, and two anonymous reviewers helped improve the manuscript.