Tectonics and Seismicity of the Lunar South Polar Region

No location on the Moon is free of the effects of global compressional stresses that result in tectonic deformation in the form of lobate thrust fault scarps. Lobate scarps are young, small-scale, morphologically simple landforms first detected in only limited high-resolution Lunar Orbiter and Apollo Panoramic Camera photographs (Schultz 1976; Binder 1982; Binder & Gunga 1985; Watters & Johnson 2010). From Lunar Reconnaissance Orbiter Camera (LROC) Narrow Angle Camera (NAC) images (Robinson et al. 2010), it is now known that lobate scarps are broadly distributed and found in every terrain setting on the Moon (Watters et al. 2010, 2015, 2019). In plan view, lobate scarps are typically linear to arcuate landforms. In cross section, they are generally asymmetric, with steeply sloping scarp faces (vergent sides) and more gently sloping back-scarp terrain (Watters & Johnson 2010; Watters et al. 2010; Banks et al. 2012). They average tens of meters of relief, reaching a maximum of ∼150 m, and have lengths of up to tens of kilometers (Binder & Gunga 1985; Banks et al. 2012; Clark et al. 2017). Their morphology and crosscutting relations with other landforms indicate that lobate scarps are formed by thrust faults (Watters et al. 2010, 2019; Williams et al. 2013). Mapping indicates that they are found everywhere on the Moon (Watters et al. 2010, 2015, 2019), including the south polar region in and around the proposed Artemis III candidate landing regions (Figure 1). The fault scarps are some of the youngest landforms on the Moon, indicated by their crisp morphology, crosscutting relations with small-diameter craters (<50 m in diameter; Watters et al. 2010), absolute model ages derived from buffered and traditional crater size–frequency distribution measurements (Clark et al. 2017; van der Bogert et al. 2018), and estimates of rates of infilling of small, shallow graben in the back-scarp areas of some scarps (Watters et al. 2012). Their proximity to relocated epicenters of shallow moonquakes (SMQs) recorded by the Apollo Passive Seismic Experiment (APSE) suggests that some of these small thrust faults are likely still active today (Watters et al. 2019).

Zoom In Zoom Out Reset image size

A survey of the south polar region has been made using LROC NAC polar mosaics with pixel scales of 2 m to discover any previously undetected lobate scarps. The survey has revealed 15 known and newly detected lobate scarps located within 150 km of the south pole (Figure 1) as individual scarps and in small clusters of scarps. One small cluster, the Wiechert cluster, located at ∼867 S, 1467 E, consists of at least five individual fault scarps (Figure 2). These scarps have a maximum relief of ∼10 m. The maximum slopes on the scarps in this cluster are estimated to be ∼10°. The Ibn Bajja cluster, located at ∼866 S, 1853 E, has three individual fault scarps with a maximum relief of ∼20 m (Figure 1), and the maximum slopes on the scarps are estimated to be <15°. The survey also revealed previously undetected fault scarps not included in the Watters et al. (2019) study that fall within the area of the de Gerlache Rim 2 proposed Artemis III landing region. The de Gerlache scarps (∼8802 S, 3009 E) are within 60 km of the south pole. The largest scarp in the de Gerlache cluster is ∼4 km long and occurs on a poleward regional slope (−88.02, 300.95; Figure 3). Topographic profiles generated from LOLA shot data (Smith et al. 2010) indicate that the scarp face or vergent side is on the downslope side of a gentle <10° slope, making a definitive thrust fault interpretation more difficult (Figures 3 and 4). However, the maximum relief—of about 60–80 m (in detrended profiles)—and morphology of the landform (Figures 3 and 4) are consistent with other lobate thrust fault scarps (Banks et al. 2012; Williams et al. 2013). A thrust fault origin is also supported by the presence of boulders along segments of the scarp face, an association common at lobate scarps (Watters et al. 2010, 2019), and evidence of possibly recent downslope regolith movement into two local shallow depressions along the scarp face (Figure 5), a feature also commonly associated with lobate fault scarps (see Watters et al. 2019). The two shallow depressions are interpreted to be degraded impact craters that are crosscut by the thrust fault, one ∼160 m diameter and the other ∼70 m in diameter (see Figure 5, inset). The inferred maximum displacement length ratio (D_max/L) of the de Gerlache fault scarp (0.0397) is consistent with that of the population of lunar lobate scarps (see Banks et al. 2017). A second low-relief, ∼500 m long arcuate lobate scarp is located in the back-scarp area of the larger de Gerlache scarp (Figure 5), also not uncommon for lobate scarp clusters (see Figure 2). Another prominent lobate scarp in the south polar region is the Shoemaker scarp (∼8625 S, 5455 E; see Figure 4 in Banks et al. 2012). The scarp has a maximum relief of ∼80 m, comparable to the de Gerlache scarp, and is ∼4 km in length.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

A recent study by Mishra & Kumar (2023) mapped many more landforms as lobate scarps in the south polar region. Many are not interpreted to be lobate thrust fault scarps in this study because (1) the landform morphology is not consistent with typical lobate scarps, (2) the inferred D_max/L ratio is not consistent with typical scarp faults, or (3) the morphology cannot be characterized with the available data (see Watters et al. 2010; Banks et al. 2012; Williams et al. 2013).

Based on the young ages of the fault scarps, it has been suggested that slip events on recently active thrust faults may have been recorded by the APSE (Watters et al. 2019). Four seismometers placed at the Apollo 12, 14, 15, and 16 landing sites operated from 1969 to 1977 and recorded 28 SMQs (Nakamura et al. 1979). These SMQs registered Richter equivalent magnitudes ranging from 1.5 to ∼5, with maximum and body wave magnitudes reaching >5.5. Stress drops are estimated to be ≤10 MPa for 16 of the events (Oberst 1987). The stress drop is an indicator of the magnitude of the regional ambient stress field (Dawson et al. 2008). The epicenter of one of the SMQs (designated the N9 event) occurred on 1973 March 13 near the south pole. The epicenter was located at ∼84°S, 134°E (Nakamura et al. 1979).

Current models for the near-surface stress of the lunar crust that best fit the spatial distribution and orientations of the young thrust faults involve a combination of compressional stresses from global contraction due to ongoing interior cooling and tidal stresses generated predominately by orbital recession with a small component from solid-body diurnal tides (Watters et al. 2015, 2019; Matsuyama et al. 2021). The magnitude of the superimposed compressional stresses is estimated to be >2 MPa, with the maximum stresses around the tidal axes when the Moon is near apogee (Watters et al. 2019; Figure 6). Tidal stresses shown in Figure 6 were determined using the equations of Matsuyama & Nimmo (2008) and a value of h₂, the Love number defining the vertical displacement in response to a gravitational potential, of 1.8 (see Watters et al. 2019). The N9 SMQ is contrary to the stress model predictions, as it occurred near the south pole, not near a tidal axis, and near perigee (Watters et al. 2019). The model suggests that stresses at the poles are predominantly from global contraction with little tidally contributed stress (Figure 6).

Zoom In Zoom Out Reset image size

If the N9 SMQ was the result of a slip event on or the formation of the de Gerlache thrust fault, an analysis of the N9 event may provide insight into the magnitude of the ambient stress and the expected magnitude of seismic shaking the event generated in the south polar region. The quality of the Apollo seismic data imposes limits on the accuracy of epicenter locations, but connections between the recorded SMQs and young lobate thrust fault scarps may indicate recently active lunar faults. Standard methods to locate an epicenter use a known velocity model and the observed arrival times of the direct P and S waves, and so the reported large uncertainties in arrival times directly translate into large uncertainties in the quake locations. To determine if coseismic slip events on the young faults were the source of some of the SMQs, Watters et al. (2019) applied a relocation algorithm (LOCSMITH) specifically adapted to use inaccurate data from very sparse seismic networks that requires only arrival time uncertainties (Hempel et al. 2012; Weber et al. 2015). Rather than solving for a best-fit location, this approach divides the solution set into candidate locations using an adaptive grid search and accounts for the arrival time uncertainty using windows around the true arrival time. The result for each event is not a single location but a cloud of candidate locations with surface solutions (Watters et al. 2019). Of the 28 SMQs identified by Nakamura et al. (1979), 13 have confirmed locations; the location cloud contains the original epicentral location. In that study, was found that the epicenters of 13 of the SMQs fall within 90 km of a mapped thrust fault scarp (Watters et al. 2019).

The N9 SMQ was one of the largest-magnitude events recorded by the Apollo Passive Seismic Network (PSN), with an estimated body wave magnitude M_b of 5–5.6 (Goins et al. 1981; Oberst 1987). The location cloud of the relocated epicenter for the N9 event is taken from those determined in Watters et al. (2019). The relocated epicenters for this moonquake are shown in Figure 1. These points represent possible epicenter locations in a cloud of probable locations. One of the candidate epicenters in the cloud is within ∼24 km of the de Gerlache scarps (Figure 1). A much smaller scarp in the Newton E cluster (−8002, 32735) is within 8 km of an epicenter in the cloud. Although other fault scarps within the probabilistic space for the N9 event cannot be ruled out as the source, the length and relief of the largest de Gerlache scarps makes it a plausible candidate.

It is therefore plausible that the de Gerlache lobate thrust fault scarp was the source of the N9 event. The magnitude of an SMQ from a slip event on the de Gerlache thrust fault can be estimated using the relation M_o = GDA (Aki & Richards 2002; Shearer 2019) for the seismic moment, where G is the shear modulus; A is area of the fault given by A = LW, where L is the length of the fault and W is the down dip fault width given by T/sin(θ), where T is the depth of the fault estimated from elastic dislocation modeling of the scarp faults (Williams et al. 2013; Watters et al. 2019); and the maximum average fault displacement D is estimated by D = h/sin (θ), where h is the maximum relief of the scarp and θ is the fault plane dip. From this, the M_o from a single slip event that formed the scarp is estimated to be ∼1.29 × 10¹⁷ nm. This slip-based moment magnitude is based on the observed fault geometry, estimated material rigidity, and slip and thus is an independent estimate of moment and not dependent on the parent body. Using the maximum relief of the scarp to estimate D assumes that the total displacement on the thrust fault occurred in a single slip event. Thus, M_o can be considered an upper limit of the magnitude of a quake generated by the formation of the fault scarp. It is not unprecedented for meter-scale structural relief to be generated on thrust fault scarps from ruptures generated by single seismic events, particularly with intraplate earthquakes like those that have occurred in Western Australia (Clark et al. 2012). A value of G of 100 MPa is assumed for the near-surface lunar regolith, which is based on characterization of the mechanical properties of a lunar regolith simulant (He et al. 2013). With a maximum relief of 80 m, the estimated moment magnitude M_w is ∼5.3. The static stress drop from such an event can be estimated assuming a rectangular fault geometry by

$$\begin{eqnarray}&&{\rm{\Delta }}\sigma =\tfrac{2{M}_{o}}{\pi {W}^{2}L}\end{eqnarray} \tag{ 1 }$$

(Shearer 2019). The Δσ from such an event is estimated to be ∼5 MPa (Table 1). A model stress drop of ∼5 MPa from a slip event on the de Gerlache fault is consistent with model estimates of the maximum global compressional stress (Watters et al. 2019; Figure 6). Larger estimates of the stress drop from the N9 quake have been suggested ranging from 40 MPa (Goins et al. 1981) to >150 MPa (Oberst 1987) based on estimates of the corner frequency of the shear wave spectrum, significantly greater than predicted here for a single slip event on the de Gerlache thrust fault. Oberst (1987) suggested that lunar scattering and internal attenuation are factors that bias the spectral shape of corner frequency estimates, an effect that could potentially explain the discrepancy between fault scarp-based estimates here and waveform propagation estimates of stress drop. Future recordings by new seismometers deployed on the Moon by the Farside Seismic Suite (Panning et al. 2022), Artemis astronauts, or a Lunar Geophysical Network (Haviland et al. 2021) will be essential for understanding the faulting mechanism in more detail.

To evaluate the magnitude and effect of seismic shaking in the south polar region and estimate the potential hazard posed by such a moonquake, we generated a model for a notional slip event on the de Gerlache thrust fault. The simulation is for the M_w = 5.3 event located at the de Gerlache fault at a source depth of 350 m. We modeled the elastic wave propagation at the de Gerlache fault using the WPP computer code (Petersson 2010). WPP uses an anelastic finite difference second-order accuracy scheme that includes mesh refinement and a grid that allows for a topography-defined free surface. The code is stable for stochastic heterogeneous materials and therefore appropriate for lunar-like crustal conditions where there is strong scattering and heterogeneity in the megaregolith. We generated a 60 km × 60 km grid with 100 m grid point spacing centered on the de Gerlache fault and incorporated the topography of the region from the LDEM_64 LOLA topography (Smith et al. 2010). The background P-wave velocity (and density) was 500 m s⁻¹ (1500 kg m⁻³) in the first 1 km, 1000 m s⁻¹ to 15 km depth (2500 kg m⁻³), and 3000 m s⁻¹ (2800 kg m⁻³) down to the Moho at 38 km depth, with a mantle velocity of 5000 m s⁻¹ (3300 kg m⁻³). S-wave velocity was scaled from the P-waves by assuming a Poissonian solid (Garcia et al. 2011). We then introduced 25% random heterogeneity in the uppermost 1 km to simulate the strong scattering effects of the lunar megaregolith using a van Karman random distribution with a characteristic scale length of 100 m (Shearer 2015). Grid refinements were issued at depths of 15 and 1 km depth for a final grid length of 25 m in the uppermost 1 km. The simulation is accurate to where there are at least 15 points per wavelength, resulting in a maximum resolved frequency of 1 Hz in the uppermost layers. The simulation was run for 300 s. We position stations every kilometer in the model at the surface and evaluate the peak ground motion by measuring the envelope of the waveform and finding the maximum amplitude in acceleration. By using the United States Geological Survey Instrumental Intensity Scale estimates for ground motion (e.g., Kramer & Upsall 2006), we show that the de Gerlache fault event is predicted to have generated strong to moderate ground shaking out to at least ∼40 km and moderate to light shaking extending beyond ∼50 km (Figure 7).

Zoom In Zoom Out Reset image size

A waveform comparison between our computed ground motions for models with scattering and lunar topography, topographic effects only, and recordings of an SMQ (N9) from Apollo is shown in Figure 8. The 3D wave propagation simulation is limited to distances of <40 km due to computational restrictions. To simulate waveforms at >1000 km distances for the observed SMQs is not feasible. Furthermore, the simulation is ended at 300 s owing to interactions of the expanding wave front with the absorbing boundaries of the box affecting amplitudes and producing nonphysical behaviors in later arriving energy. This large difference in the distance of the stations is manifested in Figure 8 by the time separation of the S- and P-wave arrivals, demarcated at 4 minutes for the N9 moonquake (∼2700 km) and ∼1 minute for the synthetics at a 38 km distance. The simulations are low-pass filtered at 1 Hz to isolate the numerically stable part of the simulation and so do not contain the higher-frequency energy present in the Apollo data. Despite these differences, the scattering medium simulation produces a coda behavior that replicates some key characteristics of the SMQs. For example, the simulation with scattering and topography exhibits a weaker contrast between the average P- and S-wave coda amplitudes. For the N9 event, this factor is ∼4.7 (Figure 8(a)); for the scattered synthetics, it is 6.5 (Figure 8(b)); and it is ∼36 in the unscattered wavefield (Figure 8(c)). Diffusion in the scattering layer should reduce the contrast between the P- and S-wave arrivals and dissipate the surface waves (Dainty et al. 1974). The scattered synthetics also exhibit a longer-lived coda of both S and P energy that extends to the end of the simulation and show depolarized S-wave energy across the horizontal components, a feature seen in the N9 data. It should be noted that near-field strong shaking from a large event has yet to be recorded by a seismometer on the Moon. Although the simulations capture some of the properties exhibited in lunar seismograms, it is likely that the intensity of scattering in the lunar crust and regolith will further reduce the amplitudes and extend the duration of codas beyond what is observed in the models. Thus, we consider these simulations to be a feasible upper bound on the intensity of shaking from a nearby SMQ.

Zoom In Zoom Out Reset image size

Seismic shaking is expected to have the greatest impact on steep slopes in the south polar region by triggering regolith landslides and boulder falls, particularly on the walls of impact craters. In order to predict areas at the south pole most susceptible to landslides from seismic shaking, we used an approach introduced by the civil engineering community to estimate the vulnerability of terrain to landslides from earthquakes. Slope stability is estimated from an infinite-slope stability model, and results are given as the factor of safety (Fs), which describes the ratio of shear strength to shear stress for downslope movement. For a dry, cohesionless material, F_S is given by

$$\begin{eqnarray}&&{Fs}=\displaystyle \frac{\tan \phi }{\tan \theta },\end{eqnarray} \tag{ 2 }$$

where ϕ is the angle of internal friction and θ is the surface slope. Unstable slope conditions are when F_s < 1, and stable slope conditions are when F_s > 1. For a dry material with finite cohesion like the lunar regolith, the factor of safety or slope stability F_S as a function of θ is given by

$$\begin{eqnarray}&&{Fs}(\theta )=\left(\displaystyle \frac{c}{{\rho }_{r}{gt}\,\sin \theta }+\displaystyle \frac{\tan \phi }{\tan \theta }\right),\end{eqnarray} \tag{ 3 }$$

where the shear stress is ${\rho }_{r}{gt}\ {\sin }\theta$, ρ_r is the material density, g is the acceleration due to gravity, t is the landslide thickness, and the shear strength is the cohesion of the material C (Newmark 1965; Gallen et al. 2015). Thus, the key parameters for slope stability of the lunar regolith are the surface slopes, angle of internal friction, regolith cohesion, and bulk density. Estimates of ϕ vary from ∼30° to 50° (Mitchell et al. 1972). The majority of steep slopes >32°–35° on the Moon are associated with young impact craters (Kreslavsky & Head 2016). The cohesion and angle of friction of the lunar regolith are largely estimated from observations of the failure of regolith associated with boulder falls and the failure of trench walls (Mitchell et al. 1972). The likely range of C is estimated to be from 0.1 to 1.0 kPa (Mitchell et al. 1972). The bulk density of the near-surface regolith will vary with depth. The density has been measured directly from Apollo core tube samples (Carrier et al. 1991) and most recently from infrared data from the Diviner Lunar Radiometer (Hayne et al. 2017), as well as in situ measurements using ground-penetrating radar (Fa et al. 2015, Fa 2020). The bulk density used in this study is 1660 ± 50 kg m⁻³ for the top 60 cm (Mitchell et al. 1973; Carrier et al. 1991). An analysis of the sensitivity of the key parameters shows that C has the greatest influence on F_S (Figure 9). At the upper limit of the range of C (1.0 kPa), slopes as high as nearly 50° may be stable (Figure 9). An increase or decrease of the density to 1610–1710 kg m⁻³ has a minor effect on the slope stability models. Even an increase to 1800 kg m⁻³ at a depth of 1 m (Hayne et al. 2017) does not significantly change the predicted slope stability.

Zoom In Zoom Out Reset image size

Using Equation (3) and global surface slopes from the SLDEM2015 64PPD computed from the 512ppd SLDEM2015+LOLA (Barker et al. 2016), we calculated the susceptibility of slopes in the south polar region for ϕs of 30°, 32°, and 35° and C = 0.1 kPa. The lower limit of the range of C was selected for a landslide thickness of t = 1 m (Figure 10). We chose a landslide thickness of t = 1 m because the recent regolith landslides observed in LROC NAC temporal images (Watters et al. 2022) do not form detectable detachment scarps. The lower end of the range of cohesion (C = 0.1 kPa) is chosen to identify areas where the model predicts that slopes may be the most suspectable to regolith landslides.

Zoom In Zoom Out Reset image size

The landslide susceptibility model suggests that large portions of the interior walls of Shackleton crater could experience landslides in response to seismic shaking. Areas of the northwestern rim of Shackleton are also susceptible, along with some areas on the edge of the Connecting Ridge site near the rim of Shackleton (Figure 10). As suggested by Figure 10, the greatest area of instability is predicted for a ϕ of 30°, and C is 0.1 kPa.

Recent boulder falls have been identified on the interior wall of Shackleton crater in newly obtained images from ShadowCam (www.shadowcam.asu.edu; Figure 11); however, evidence of recent regolith landslides is not observed. The absence of recent landslides where predicted on the walls of Shackleton and other permanently shadowed regions may indicate regolith with greater cohesion, possibly due to the presence of a cementing agent like water ice. Gertsch et al. (2006) found that as little as 0.3% water ice added to a lunar regolith simulant significantly increases the regolith cohesion. The similarity in relief and length of the Shoemaker scarp to the de Gerlache scarp and its proximity to the Nobile Rim 1 candidate Artemis III landing region (∼30 km) suggests the potential for a comparable magnitude SMQ to the de Gerlache scarp. A comparable slip event on the Shoemaker scarp could result in moderate to light shaking in the area of the Nobile Rim 1 landing region, where the landslide susceptibility model indicates areas with unstable slopes (Figure 9). Areas of slope instability are predicted in two permanently shadowed regions, one north of the de Gerlache Rim 2 landing region and another on the rim of Nobile crater <20 km from the Shoemaker scarp (Figure 9).

Zoom In Zoom Out Reset image size

The amount of ground acceleration necessary to trigger a regolith landslide of a given thickness can also be evaluated using the F_S relationship. The critical ground acceleration A_C necessary to trigger a regolith landslide of thickness t as a function of surface slope (Newmark 1965) is given by 

$$\begin{eqnarray}&&{A}_{c}(\theta )=\left(\displaystyle \frac{C}{{\rho }_{r}{gt}\ \sin \theta }+\displaystyle \frac{\tan \phi }{\tan \theta }-1\right)\ g\ \sin \theta .\end{eqnarray} \tag{ 4 }$$

Analysis of A_C as a function of θ shows that for the range in C and a fixed ϕ of 30°, the ground acceleration needed to trigger a regolith landslide decreases linearly with increasing θ, with greater regolith cohesion providing greater slope stability (Figure 12, Table 2). For example, to trigger a 1 m thick landslide in regolith with low cohesion on slopes greater than 30°, very little ground acceleration is needed (∼2.3% g; Figure 12), or light shaking on the modified Mercalli intensity scale. The de Gerlache thrust fault is less than 60 km from the rim of Shackleton crater, which is within a distance predicted to experience weak to light shaking from an SMQ event with the fault as its source (Figure 7(A)). The model suggests that with greater regolith cohesion, greater surface slopes approaching 50° are stable, and greater seismic shaking is required to trigger a regolith landslide at surface slopes of <∼30°. We use as an example the ∼21 km diameter Focas crater, south of Orientale Basin (Figure 13(A)). Evidence of numerous landslides is found on the interior walls of Focas crater. The slope stability model predicts that much of the interior walls of the crater are unstable using C = 0.1 kPa (Figure 13(B)), while with C = 1.0 kPa, only a very small area of the interior walls is predicted to be unstable. These results support that lower values of the regolith cohesion are consistent with the ubiquitous occurrence of regolith landslides and granular flows (Bickel et al. 2022; Watters et al. 2022).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Analogs to lunar SMQs are found on Earth. Shallow earthquakes, within the upper 1 km, have been recorded; the intraplate earthquakes in Western Australia are an example. One of the most well known is the Meckering quake that occurred in 1968 with an M_w of ∼6.3. The quake resulted in an arcuate series of surface-breaking thrust faults ∼37 km long and fault scarps with at least 1.5 m of relief. The faults are arranged in a series of en echelon stepping segments from hundreds of meters to >10 km in length (Gordon & Lewis 1980; Johnston & White 2018), analogous to lunar lobate scarp clusters. The static stress drop of this and other terrestrial shallow earthquakes in Western Australia is estimated to be ∼10 MPa (Denham et al. 1980). Shallow intraplate earthquakes like those in Western Australia are potential analogs to lunar SMQs. The relatively low relief (tens of centimeters to meters) of some of the thrust fault scarps of Western Australia (Clark et al. 2012) suggest that there may be another scale of lunar thrust fault scarps (see Figure 5), a population that may not be easily detected even in LROC NAC images with optimum lighting conditions and that may only be clearly recognized in ground surveys by robotic and human exploration.

Other efforts have been made to characterize the tectonics and potential seismicity of the south polar region (Mishra & Kumar 2022). Based on a deterministic seismic hazard analysis, Mishra & Kumar (2022) conclude that only magnitudes M_w from 1 to 4 moonquakes are expected from lobate scarp faults and that seismic ground motions from these moonquakes would not affect many areas in the south polar region, including the majority of the Artemis III candidate landing areas. We suggest that the lobate thrust fault scarps in the south polar region in and around the areas of the proposed Artemis III landing regions, particularly the de Gerlache Rim sites and Nobile Rim 1 regions, are potential sources for future seismic activity that could produce strong regional seismic shaking. If slip events on these young faults occur in the south polar region and elsewhere on the Moon, regolith landslides and potential boulder falls (Kumar et al. 2016; Mohanty et al. 2020) can be expected at distances of tens of kilometers from the source faults. Small amounts of water ice in the lunar regolith are expected to significantly increase the cohesion, stabilizing steep slopes against shallow landslides from seismic shaking. Based on our analysis of an N9-level event in the south polar region, we conclude that such an event poses a potential hazard to future robotic and human exploration in the region.

We thank the editor and two anonymous reviewers for their thoughtful comments and suggestions that greatly improved the manuscript. We gratefully acknowledge the LRO engineers and technical support personnel. T.R.W. was supported by the LRO Project and an ASU LROC contract. N.C.S. was supported by NASA SSERVI GEODES grant 80NSSC19M0216. C.L.J. acknowledges support from the Natural Sciences and Engineering Resource Council of Canada, Discovery Grants Program.