Paleoseismology of the 2016 MW 6.1 Petermann earthquake source: Implications for intraplate earthquake behaviour and the geomorphic longevity of bedrock fault scarps in a low strain‐rate cratonic region

The 20 May 2016 MW 6.1 Petermann earthquake in central Australia generated a 21 km surface rupture with 0.1 to 1 m vertical displacements across a low‐relief landscape. No paleo‐scarps or potentially analogous topographic features are evident in pre‐earthquake Worldview‐1 and Worldview‐2 satellite data. Two excavations across the surface rupture expose near‐surface fault geometry and mixed aeolian‐sheetwash sediment faulted only in the 2016 earthquake. A 10.6 ± 0.4 ka optically stimulated luminescence (OSL) age of sheetwash sediment provides a minimum estimate for the period of quiescence prior to 2016 rupture. Seven cosmogenic beryllium‐10 (10Be) bedrock erosion rates are derived for samples < 5 km distance from the surface rupture on the hanging‐wall and foot‐wall, and three from samples 19 to 50 km from the surface rupture. No distinction is found between fault proximal rates (1.3 ± 0.1 to 2.6 ± 0.2 m Myr−1) and distal samples (1.4 ± 0.1 to 2.3 ± 0.2 m Myr−1). The thickness of rock fragments (2–5 cm) coseismically displaced in the Petermann earthquake perturbs the steady‐state bedrock erosion rate by only 1 to 3%, less than the erosion rate uncertainty estimated for each sample (7–12%). Using 10Be erosion rates and scarp height measurements we estimate approximately 0.5 to 1 Myr of differential erosion is required to return to pre‐earthquake topography. By inference any pre‐2016 fault‐related topography likely required a similar time for removal. We conclude that the Petermann earthquake was the first on this fault in the last ca. 0.5–1 Myr. Extrapolating single nuclide erosion rates across this timescale introduces large uncertainties, and we cannot resolve whether 2016 represents the first ever surface rupture on this fault, or a > 1 Myr interseismic period. Either option reinforces the importance of including distributed earthquake sources in fault displacement and seismic hazard analyses.

Commonly, surface rupture traces and/or related surface deformation features (e.g., fault scarps, folds) produced in contemporary and/or pre-historic ('paleo') earthquakes are excavated by hand or machine to expose subsurface structural-stratigraphic relationships, from which earthquake displacements and chronologies may be characterized (e.g., Khajavi et al., 2016;Sieh, 1978;Stahl et al., 2016). Nearsurface fault exposures in natural settings (e.g., stream cut-banks; Quigley et al., 2006;Sandiford, 2003) and progressive offsets of surface features (e.g., stream channels and terraces; Amos et al., 2011;Gold et al., 2017;Little et al., 2010) may also be utilized for this purpose. Determination of an earthquake recurrence interval requires estimates of the timing of multiple discrete seismic displacements, although other approaches that combine geodetic and seismologic data, slip rates, and earthquake scaling parameters are also used (e.g., Atwater et al., 2003;Dixon et al., 2003;Leonard & Clark, 2011;Nicol et al., 2016;Zielke, 2018). The accuracy and precision with which recurrence intervals may be estimated depends on many factors, including (i) the suitability of fault-related sediments and surface features for age dating, (ii) the analytical uncertainties associated with the age data, (iii) the epistemic uncertainties associated with interpretation and attribution of structural-stratigraphic data to distinct earthquakes at distinct times, and (iv) the assumption that the geologic record is a complete archive of seismic activity on the studied fault(s).
Many paleoseismic studies are limited by several of these factors.
Australia has experienced 11 surface-rupturing earthquakes since 1968  and references cited therein) that collectively comprise the richest domestic intraplate record of historical reversefault surface ruptures globally. Five of these earthquakes have been investigated using machine-dug (Clark & Edwards, 2018;Crone et al., 1997) or hand-dug Lewis et al., 1981;this study) excavations combined with the structural-stratigraphic approaches outlined earlier. For four of these five events, structural and stratigraphic evidence for paleo-earthquakes is not recorded in the exposed Quaternary/Tertiary sediments and/or Precambrian bedrock. In one case, the Lake Surprise East (Tennant Creek) rupture (Crone et al., 1992), the excavated trench exposed a pre-existing bedrock scarp of ambiguous origin, and a non-seismic interpretation is equally permissible . It is therefore unresolved whether these earthquakes occurred on 'active' faults (e.g., faults with evidence of preceding Holocene or Pleistocene surface ruptures; Clark et al., 2008Clark et al., , 2012Crone et al., 2003;Machette, 2000;Quigley, Clark & Sandiford, 2010), whether these faults were newly activated (e.g., along pre-existing planar structures) or newly created in these earthquakes (e.g., had no brittle failure prior to historical rupture; Clark & Edwards, 2018;Machette et al., 1993), or whether these faults exhibit incipient, emergent rupture behaviours (e.g., fault emergence at the ground surface through upward fault tip migration and fault propagation folding; Finch et al., 2003;Hardy & Finch, 2007;Livio et al., 2014Livio et al., , 2019Tindall & Davis, 1999). Such distinctions bear directly on questions of how best to characterize seismic hazard in slowly deforming continental interiors Clark, 2018;Clark et al., 2012;Leonard et al., 2007) and evaluate potential driving forces responsible for intraplate seismicity (Calais et al., 2005Craig et al., 2016;. In slowly deforming continental interiors, fault recurrence data are often challenging to obtain because likely recurrence intervals for surface rupturing earthquakes (greater than several 10 4 to greater than 10 5 years) may exceed the age of available near-surface sediments (Clark et al., 2008Crone et al., 1997), and/or may exceed the temporal range over which earthquake-associated sediments may be dated by radiocarbon (ca. 0-40 kyr; Ramsey, 2008) and optically stimulated luminescence (OSL) techniques (commonly less than 70 ka for U-Th-K (uranium, thorium, potassium) rich sediments in the Australian context; Clark, Griffin, et al., 2017;Quigley et al., 2006). Furthermore, landscape modification rates associated with erosion and/or sedimentation processes may exceed fault-slip rates associated with the maintenance of fault-associated topography (Hornblow et al., 2014). We attempt to tackle these challenges here by combining environmental effects observed in the 2016 M W 6.1 Petermann earthquake  with a multi-methods approach aimed to quantify landscape evolution rates in the near-fault region. We integrate landscape analyses using satellite imagery (e.g., Worldview data) and beryllium-10 ( 10 Be) cosmogenic nuclide-derived erosion rate estimates with more typical structural-stratigraphic approaches (trenching) to expand the confidence of paleoseismic interpretations and the timerange over which such studies may be undertaken.

| Seismotectonic setting
Australia is one of the most seismically active Stable Continental Regions (SCRs) globally (Braun et al., 2009;Johnston et al., 1994;Sandiford & Quigley, 2009), with a rich historical record of surfacerupturing earthquakes King et al., 2019). The identification of more than 360 faults and folds with topographic and/or geologic evidence for Quaternary earthquake displacement(s) extends contemporary seismicity over geological timescales (Figure 1; Clark, 2012;Clark et al., 2012;Quigley, Clark & Sandiford, 2010). Investigating earthquake surface rupture hazard is important in the Australian context for multiple reasons. Firstly, highpopulation urban centres (e.g., Adelaide, Melbourne and Sydney) and critical infrastructure (e.g., large dams, energy schemes, transport and utility corridors) are located near known and/or suspected Quaternary-active faults. The seismic risk is unknown for many of these faults; particularly the faults which are suspected of being active during the Quaternary. These uncharacterized faults present the largest potential seismic hazard in Australia. Secondly, the current national probabilistic seismic hazard assessment (NSHA18)  and seismic guidelines for large dams (ANCOLD (Australian National Committee on Large Dams), 2019) rely on sparsely resolved source-based inputs for faults. Finally, geologic data provides constraints on, and aids in, the development of new models to explain intraplate seismicity Clark, Griffin, et al., 2017;Stirling et al., 2011). Within Australia these geological data require further investigation to prove useful, as they are complicated by preliminary evidence indicating highly episodic rupture recurrence behaviours, spatial and temporal earthquake clustering Crone et al., 1997), and surface-rupturing earthquakes on faults without clear evidence for analogous predecessors (Clark & Edwards, 2018;Clark et al., 2020;King et al., 2019 Clark et al., 2014;Machette et al., 1993), the 1989 M L 4.99 Uluru earthquake (Michael-Leiba et al., 1994), and 2013 M W 5.38 Mulga Park earthquake (Clark & McPherson, 2013; Figure 1; magnitudes from ). No earthquakes greater than M W 3.0 have been instrumentally recorded within a 50 km radius of the epicentre of the Petermann earthquake over the period of the instrumental record (see Leonard, 2008).
F I G U R E 1 (a) Seismotectonic map of Australia showing cratonic regions (Leonard et al., 2014), historic onshore earthquakes from the NSHA18 catalogue , neotectonic features (Clark, 2012), historic surface rupturing earthquakes , geographic locations as discussed in text, and significant earthquakes in the vicinity of the 2016 Petermann earthquake (i), 1986 M W 5.7 Marryat Creek (Machette et al., 1993) (ii), 2012 M W 5.18 Pukatja surfacerupturing earthquakes ) (iii), 1989M L 4.99 Uluru (Michael-Leiba et al., 1994 (iv), 2013 M W 5.38 Mulga Park (Clark & McPherson, 2013). (b) Shaded relief map of region around the 21 May M W 6.1 Petermann earthquake with published epicentres and surface rupture as in King et al. (2019), Woodroffe Thrust location (Scrimgeour et al., 1999), location of samples (also show in Figure 2), trench locations and bedrock outcrops in the vicinity of the 2016 surface rupture as mapped from available satellite imagery [Color figure can be viewed at wileyonlinelibrary. com] The Petermann earthquake generated a discontinuous surface rupture extending for a total length of 21 km, as identified from analysis of interferometric synthetic aperture radar (InSAR) data and field observations (Gold et al., 2019;King et al., 2018;Polcari et al., 2018;Wang et al., 2019; Figure 1). The surface rupture trace has an average strike of 294 ± 29 (Attanayake et al., 2020). The rupture plane dips between 25 and 35 to the northeast as indicated by focal mechanisms, modelled by InSAR, and mapped from surface data (Attanayake et al., 2020;Polcari et al., 2018;Wang et al., 2019;this study). The total length of discrete ground surface ruptures identified through field studies is 20 km (Gold et al., 2019;King et al., 2019). with an average uplift of the hanging-wall of 23 cm relative to the footwall as measured at the surface rupture by a real-time kinematic global positioning system (RTK GPS) (Attanayake et al., 2020).
Hanging-wall bedrock outcrops within 5 km of the Petermann surface rupture experienced strong ground motion induced rock falls, in situ fracturing, and displacement of thin (2-5 cm thick) rock fragments sourced primarily from exfoliation sheets (as visible in most outcrop images of Figure 2 and described in King et al., 2018). Rock fall damage on foot-wall outcrops was only observed at less than 2 km distance from the surface rupture, and the intensity of bedrock fracturing and rock fragment displacements on the foot-wall was greatly reduced relative to hanging-wall outcrops (King et al., 2018).
The surface rupture trace obliquely intersects a 055 orientated fault scarp (Hanks, 2000). Erosion has been locally enhanced by bedrock shattering, rock fragment displacement (Figure 2), and rockfalls, particularly on the hanging-wall where coseismic shaking damage was more intense (King et al., 2018).   isolation of the quartz fraction using heavy liquid separation, and etching with 40% hydrogen fluoride (HF) for 45 min followed by HCl treatment. We apply an infrared-depletion test to check for the presence of feldspar (Duller, 2003) followed by a single-aliquot regeneration (SAR) protocol from Murray and Wintle (2000) to determine D e values for 24 sample aliquots (1 mm) ( Quoted errors represent both analytical uncertainty and model uncertainty associated with the chosen production rate model (Balco et al., 2008).

| Optically stimulated luminescence
content and the field hydrological context were used to estimate average water content (Table 1) during sediment burial.

| Beryllium-10 cosmogenic nuclide erosion rates
Seven bedrock samples were collected for 10 Be cosmogenic nuclide erosion rate analysis along an 8.5 km long traverse approximately perpendicular to the 2016 surface rupture (Figure 2). Three additional samples were collected distal to the surface rupture (19-50 km) to provide proxies for background erosion rate ( Figure 2). Samples were obtained by hammer and chisel from subhorizontal weathered surfaces at the top of bedrock outcrops ( Figure 2). Given the low relief, topographic shielding was minimal (Supporting Information Table S3). Sample thicknesses were recorded in the field (Table 2).
Samples A and J were collected to provide estimates of regional Quartz was extracted from bedrock samples at the University of Melbourne using standard rock crushing, magnetic and heavy mineral separation techniques. Separation of Be from quartz grains was conducted at the University of Melbourne, following the method described in Quigley et al. (2007). The 10 Be/ 9 Be isotopic abundance ratio of the beryllium oxide samples was measured with XCAMS, the Accelerator Mass Spectrometry facility at GNS Science in New Zealand (Zondervan et al., 2015) (details in Table S4).
Beryllium-10 erosion rates (  Table S2). Reported data (Table 2) follow the recommended protocol of Frankel et al. (2010) and Dunai and Stuart (2009). Erosion rates reported in Table 2  Other stationary and time-variant model estimates Stone, 2000) are reported in Table S1. The latter deviate no more than 5% from the reported rates and are not discussed further.
Our erosion rate data combined with offset measurements of the Petermann surface rupture (Attanayake et al., 2020;Gold et al., 2019) provide an opportunity to quantify the potential time required to erode a scarp and uplifted topography in this bedrock-dominated landscape. To do this we make a number of assumptions. Firstly, we propose that the long-term post-earthquake hanging-wall (scarp) erosion rate may be approximated by the maximum 10 Be erosion rate observed in our sample cohort (Table 2). We acknowledge that past and future scarp degradation could involve a diversity of erosional mechanisms (e.g., alluvial and aeolian mechanical weathering, in situ chemical weathering) in rock and sediment that vary in time and space and thus acknowledge large epistemic uncertainties are inherent in this assumption. Secondly, we assume that our average and minimum 10 Be erosion rates (Table 2) represent long-term erosion rate estimates for the foot-wall. And finally, we assume that the differential erosion rates (ΔE; Figure 5b) obtained from subtracting the average and minimum rates from the maximum rate provide crude proxies for the rate at which the hanging-wall re-achieves topographic F I G U R E 5 (a) Graph of erosion rate versus rock sheet/fragment thickness, where lines represent 1σ standard deviation in steady-state erosion rate for varying sheet thickness and mean erosion rate. Description of Boxes (i), (ii), (iii) and point (iv) as described in the text. equilibrium with the foot-wall (i.e., the fault scarp is denuded to a low slope feature indistinct from regional topography, and thus is no longer recognizable as a distinct, linear, stepped topographic anomaly that could be attributed to ground surface rupture). We use Equation (1): 1 mx Myr * (Er max -Er min [or avg] ) to estimate the residence of a 1 m high scarp in the landscape (note that offsets of ≈ 1 m are only observed in the central 10 km of the scarp; Figure 5b).
Applying our erosion rate results in this way requires assumptions about how long the surface has been exposed, and that erosion rate has been constant through time (i.e., steady-state erosion). The steady-state erosion assumption is often supported through estimating a timescale for which a calculated erosion rate average is representative, in combination with geomorphological evidence. The sampled surface may be assumed to have been eroding at a constant rate over a timescale represented by Equation (2): Λ ∕ ε, where ε is the rate of erosion and Λ is attenuation length. This represents the time required to erode a layer of thickness equal to the attenuation length and assumes no contradictions from field observations and geomorphological arguments (further explored in the Discussion section). The ratio Λ ∕ ε is, to first order, approximated by the ratio between concentration N and production rate P of the cosmogenic isotope (Granger et al., 1996).
Here we aim to constrain the time-range over which those assumptions, and therefore our erosion rates, are applicable. This timerange is in practice limited by a combination of external processes (removal of overlying rock through erosion, attenuation of cosmic-ray particles by that material, radiometric decay). For the specific case of this study, the dominant external process is best described by gradual removal of overlying material. It is unlikely that these samples have The land surface in the vicinity of the Petermann rupture slopes gently south to southwest and is characterized by sparse evidence for channelized drainage. Broad, shallow paleo-channels draining towards the southwest are truncated by the northwest-trending dune set, forming strings of playas in interdune regions (Figures 1 and 3).
Paleo-channel margins are locally framed by groundwater carbonate horizons, which form low escarpments, and in some cases have

| Trench logs
Trench 1 was excavated across the Petermann Ranges earthquake surface rupture within a playa occupying the axis of the shallow paleovalley (co-located with profile Y-Y 0 in Figure 3). The trench exposed coarsely layered to un-layered silty fine sand and fine sandy Trench 2 was dug within an inter-dune location with bedrock outcrops on both hanging-wall and foot-wall (location is co-located with profile X-X' in Figure 3). Trench 2 also exposed a faulted calcrete horizon developed within aeolian and alluvial sediments. Medium sand is more abundant and silt less abundant in the clastic sediment profile in comparison to the profile in Trench 1. This is interpreted to reflect a primarily sheetwash origin with aeolian input. At 0.25 m depth a nodular pedogenic carbonate is developed with nodules up to 5 cm in diameter. This calcrete is dissimilar in texture to the massive calcrete   (Fitzsimmons et al., 2007), and alluvial sediment transport around the Flinders Ranges (Gliganic et al., 2014) under moist climatic conditions (Quigley, Horton, et al., 2010) dominated by summer rainfall (Singh & Luly, 1991). This age provides a minimum time interval for any prior ruptures on this fault, as trenching found no prior offset recorded in the dated sediments (Figure 4).

| Cosmogenic nuclide erosion rates
Bedrock erosion rates derived from time-dependent 10 Be production rate model LSDn (Lifton et al., 2014) range from 1.27 ± 0.10 to 2.57 ± 0.20 m Myr −1 with 7-12% relative uncertainties (Table 2, column LSDn in Table S1). Minimal (less than 1.5%) rate variation is observed between these rates and those derived using the time-independent production model based on Stone (2000) and Lal (1991) (column St in Table S1). Erosion rates calculated with the production model based on Lifton et al., 2008 (column Lm in Table S1)  We cannot identify any correlation between erosion rate and (i) proximity of the sample to the surface rupture zone, (ii) samples located on the hanging-wall versus foot-wall, or (iii) samples in the epicentral area versus far-field regions. We also find no correlation between erosion rate and height of the sample above the sandplain.
This lack of spatially variable erosion rates is explored in more detail in the Discussion section (Section 4.3).
Equation (1) Equation (2) Λ ∕ ε provides an approximate value for the timescale over which our assumptions of steady-state erosion and infinite exposure are applicable (and hence over which our erosion rates might be reasonably extrapolated). This equation is solved by equating Λ with the attenuation length of the mechanism that dominates production near the surface. For 10 Be, that mechanism is spallation by fast neutrons, for which Λ ≈ 150 g cm −2 /2.7 g cm −3 ≈ 55 cm (Gosse & Phillips, 2001). Applying Λ ∕ ε for our highest (2.57 m Myr −1 ) and lowest (1.27 m Myr −1 ) erosion rates, suggests our calculated erosion rate average is representative over approximately the last 0.2-0.4 Myr.
Extending our erosion rate results past this time-range (such as in our 0.5-1 Myr estimate for time required to erode the Petermann scarp) introduces large epistemic uncertainties into our interpretations. This is explored in more detail in the Discussion section (Section 4.5).

| Erosion rate variations
We observe a bimodality in our erosion rates with ranges of 1.27 to 1.71 m Myr −1 and 2.21 to 2.57 m Myr −1 . This may be attributed to outcrop scale differences in erodibility (e.g., variable lithology and fracture density) and/or erosion (e.g., variable exposure to erosive forces such as wind and water). Alternatively (or additionally), the cohort of higher erosion rates could relate to different spatial and temporal scales of episodic erosion (Figure 7) via periodic or stochastic removal of previously overlying rock fragments of unknown thickness, but with a minimum probable thickness less than 30-50 cm in order to generate the observed variability (Small et al., 1997).
Differences in erosional processes (e.g., wind vs. water) are unlikely at the scale considered since no relationship between site elevation and erosion rate, nor field evidence for variations in fluvial or aeolian erosional processes, were found. However, at many sample sites we identified proximal evidence for in situ exfoliating rock sheets and eroding boulders greater than 30 to 50 cm thick, as well as eroded rock debris near sampled outcrops with similar properties (Figures 2 and 7).
Our favoured hypothesis is that bedrock in the study area erodes via a combination of (i) exfoliation of thin (thickness less than 5 to 10 cm) sheets that can be approximated by the steady-state erosion rate assumption (Lal, 1991; Figure 7) and (ii) episodic, and probably stochastic, erosion of thicker (greater than 30 to 50 cm) overlying rock fragments (i.e., boulders, Figure 7) that locally increase the apparent erosion rate (Small et al., 1997). We argue that an unresolved combination of these processes is responsible for most of the variability in the 10 erosion rates determined from in situ produced 10 Be.

| Coseismic erosion during the Petermann earthquake and implications for cosmogenic 10 Be concentrations
Previous work on Australian faults with Quaternary offsets in the similarly arid Flinders Ranges found that bedrock and sediment proximal to the faults recorded increases in the mean and standard deviation of 10 Be erosion rates (Quigley et al., 2007). This work suggested that earthquake-induced erosion rate perturbations could persist above background (pre-seismic) rates for greater than 30 kyr within the arid Australian landscape. This hypothesis suggested multiple processes could lead to enhanced delivery of material low in 10 Be to the sampled catchments, increasing the apparent erosion rate. These include: seismically-triggered landslides (e.g., Niemi et al., 2005); seismic shaking and in situ rock fracturing (e.g., Quigley et al., 2016); and steepening of stream profiles and catchment flanks via upstream propagation of earthquake-induced knickpoints (Wobus et al., 2005).
These mechanisms could influence 10 Be erosion rates over sustained (> 10 3 to 10 6 ) timescales, dependent upon conditions including the nature of the perturbation relative to background erosion rates, lithologies, topography, vegetation, rates of natural and anthropogenic surface processes, landscape interactions, and climate-weather variability. Quigley, Clark and Sandiford (2010) proposed a positive correlation between regional seismic strain rate and average 10 Be erosion rates across the Australian continent. Within this context, it is valuable to consider whether coseismic erosion induced by the 2016 Petermann earthquake, in this significantly more subdued landscape relative to those studied elsewhere in Australia, could be expected to perturb bedrock 10 Be concentrations in a manner consistent with that proposed by Quigley, Horton, et al. (2010) and Quigley, Clark and Sandiford (2010).
Episodic removal of rock fragments of different thicknesses affects steady-state erosion rate estimates differently depending on the rate of erosion. Lines of equal relative uncertainty (i.e., the ratio of uncertainty to erosion rate) are shown on Figure 5(a) (adapted from figure 6, Small et al., 1997). These lines vary depending on rock fragment thickness and mean erosion rate (Small et al., 1997). For example, for very low erosion rates (e.g., < 1.0 m Myr −1 ) 10 cm thick rock removal results in high variability around the sample mean (relative uncertainty ≈ 7-20%). For high erosion rates (e.g., > 10 m Myr −1 ) the erosion rate recovers quickly from removal of 10 cm and their relative uncertainty is lower (≈ 5%).
The thickness range of exfoliated rock sheets/fragments (2-5 cm) displaced in the Petermann earthquake (King et al., 2018) is plotted against our 10 Be erosion rates (1.27-2.57 m Myr −1 ; Table 2) in Box (i) of Figure 5(a). Box (i) crosses the lines of equally relative uncertainty from ≈ 1.5 to 3%. This indicates that the potential range of new erosion rates following removal of 2 to 5 cm sheet is low, even if sampling prior to or after an exfoliation event (assuming steady-state erosion). Box (ii) of Figure 5(a) shows the range of our reported relative uncertainties (7.5-11.8%; Table 2. Box (iii) of Figure 5(a) represents the thicknesses of rock fragments (greater than 13 to 15 cm, labelled *(iv) on x axis) that would be required to induce 1σ variations from steady-state erosion.
Displaced rocks with thicknesses exceeding this 13 to 15 cm threshold value were only very rarely observed (1 × 10 −4 % of area of observed outcrops) on steep bedrock faces (i.e., rockfalls) and never from flat-lying bedrock surfaces (King et al., 2018). Coseismic rock fragments displaced from sub-horizontal surfaces in the Petermann earthquake are thus of insufficient thickness (< 13 cm) to perturb 10 Be concentrations in the underlying bedrock, and no 10 Be evidence for coseismic processes on these surfaces will be evident. While isolated occurrences of seismically triggered rockfalls are of sufficient depth to perturb 10 Be concentrations, these outcrops collectively account for << 0.1% of the total outcrop area in the epicentral region.
The utility of these surfaces to constrain past rupture behaviours on the Petermann fault via 10 Be analysis is therefore limited without independent evidence (e.g., Quigley et al., 2007). to the surface rupture relative to regional rates, and in hanging-wall versus foot-wall rates (see next section). We hypothesize that spatially variable erosion rates resulting from prior ruptures of this fault may persist over a longer time range than fault-related offsets analogous to the 2016 event (i.e., < 1 m).

| Expected longevity of the Petermann earthquake scarp and implications for active fault recognition
The Petermann earthquake fault scarp traverses both bedrock and sedimentary landscape elements (King et al., 2018). Sheetwash and channel-related alluvial erosional processes were observed (King et al., 2018) to erode the scarp more rapidly than adjacent surfaces due to the enhanced topographic slope at the scarp (Colman & Watson, 1983;McCalpin, 2009). The erodibility and rate of migration of knickpoints into the uplifted topography may be further enhanced by coseismic fracturing and fissuring of hanging-wall outcrops (Hsu & Pelletier, 2004;McCalpin, 2009  The Petermann earthquake therefore appears to be the first surfacerupturing earthquake on this fault, or any other potential proximal faults in the epicentral region, in the last 10 ka to 1 Myr or longer.
We note that as the time over which we extrapolate our results increases, the epistemic uncertainty associated with data limitations and assumptions increases. It is therefore difficult to speculate on the pre-0.5 to 1 Ma earthquake history of the Petermann fault. We con-

| Implications for seismic hazard modelling
Conceptually, for the purposes of probabilistic seismic hazard assessment, a fault source is a seismogenic fault that has produced earthquakes in the past and can be expected to continue doing so (Musson, 2012). Long-term slip rates are a key component of fault source input data for seismic hazard modelling (e.g., Allen et al., 2020;Stirling et al., 2012). While the Petermann fault demonstrably represents a historical seismic source, the lack of a determinable slip-rate challenges its inclusion in such models. If moderate to large earthquakes are assumed to recur on the faults that comprise the 2016 Petermann rupture, the long-term average uplift rate (i.e., the vertical component of the slip rate) is limited by our very low regional bedrock erosion rate determinations of 1 to 3 m Myr -1 (Table 2).
This low rate is unlikely to contribute significantly to ground-motion hazard for return periods that may affect ordinary-use structures (e.g., 475 or 2475 years) .
More broadly there is no compelling evidence for prior Quaternary events on any of the faults that have hosted Australia's 11 historical surface rupturing earthquakes King et al., 2019). These faults have either hosted 'one-off' ruptures or have interseismic periods extending beyond 10 5 to 10 6 years (this study; Clark et al., 2020;King et al., 2019).
In contrast, paleoseismic investigations of pre-historic fault scarps in the Precambrian SCR of Australia have identified recurrence of large Quaternary earthquakes that enable the estimation of slip-rates.
In both of these settings there is evidence that large earthquake occurrence on some faults might be temporally-clustered (e.g., Crone et al., 1997;Clark, McPherson, et al., 2017).the current lack of defensible mechanism to explain these observations (e.g., Calais et al., 2016) introduces significant uncertainty when such faults are included as sources into seismic hazard models (Clark & Edwards, 2018;Clark et al., 2020).
Seismic hazard models commonly account for the potential for unmapped active faults, and for 'one-off' ruptures, by permitting moderate-to-large magnitude (i.e., M w 5.0-7.0) 'distributed earthquakes' to occur anywhere (e.g., Stirling et al., 2008). These ensemble models of seismic hazard weight both fault source-based versus distributed seismicity logic tree branches Clark et al., 2016;Griffin et al., 2018). Multi-methods studies such as ours are necessary to parameterize fault source models, to provide an evidencial base for deciding the balance of weighting between logic tree branches for specific tectonic environments within low-strain rate regions (e.g., Clark et al., 2020).
Thus while many of the faults discussed in this article are very remote and represent a low risk to Australia's spatially concentrated population, the knowledge gained in studying the behaviour of such faults contributes directly to the development of more robust and defensible seismic hazard models, and thereby benefits populated centres. Moreover, insight gained through studies of Australian surface rupturing faults and paleoseismic scarps may be applied to analogous low-strain rate regions globally (e.g., Crone et al., 1997Crone et al., , 2003.

| CONCLUSIONS
This article presents a multi-methods approach to quantifying landscape evolution rates in the near-fault region of the 2016 M W 6.1 Petermann earthquake to understand the prior rupture history of this fault and to expand the time-range and confidence over which paleoseismic studies may be undertaken. Interpretation of results from landscape analysis (using pre-and post-earthquake satellite imagery), OSL dating, and cosmogenic 10 Be nuclide erosion rate sampling across the near-and far-field of the modern rupture area show an absence of evidence for penultimate rupture. We estimate that the Petermann scarp and associated hanging-wall uplift could be erosionally removed in 0.5-1 Myr. This suggests the Petermann earthquake may be the first surface-rupturing event on this fault in the last 10 ka to 1 Myr or longer. Extrapolating our assumptions and single nuclide erosion rates across timeframes > 0.4 Ma introduces large epistemic uncertainties.
We therefore cannot definitively say whether the 2016 event was the first surface rupturing event on this fault, or that this fault has a > 1 Myr interseismic period. Either option negates the derivation of a meaningful long-term slip-rate or recurrence model for this fault, which are key inputs for fault sources in seismic hazard models. We suggest that the occurrence of future analogous surface-rupturing earthquakes on unidentified faults may be addressed by including distributed seismicity components in seismic hazard.

DATA AND RESOURCES
The DEMs used in this article are subject to third party restrictions and were originally produced by Gold et al. (2019) DOI: 10.
1029/2019GL084926. Restrictions apply to the availability of these data. All cosmogenic nuclide and OSL sample data are provided in Table 1 and/or the Supporting Information. Additional trench images are provided in Supporting Information, with more available upon request from the author.