Abstract

Normal faults in basalt have distinctive surface-trace morphologies and earthquake evidence that provide information about the slip behavior and earthquake potential. The 47-km-long Hat Creek fault in northern California (USA), a useful case example of this fault style, is a segmented fault system located along the western margin of the Modoc Plateau that is a regional earthquake hazard. In response to interaction with sporadically active volcanic systems, surface ruptures have progressively migrated westward since the late Pleistocene, with older scarps being successively abandoned. The most recent earthquake activity broke the surface through predominantly ca. 24 ka basaltic lavas, forming a scarp with a maximum throw of 56 m. Past work by others identified 7–8 left-stepping scarp segments with a combined length of 23.5 km, but did not explicitly address the throw characteristics, fault evolution, slip history, or earthquake potential. We address these deficiencies in our understanding of the fault system with new field observations and mapping that suggest the active scarp contains 2 additional segments and is at least 6.5 km longer than previously mapped, thus increasing the knowledge of the regional seismic hazard. Our work details scarp geomorphic styles and slip-analysis techniques that can be applied to any normal-faulted basalt environment. Applied to the Hat Creek fault, we estimate that a surface-breaking rupture could produce an earthquake of ∼Mw (moment magnitude) 6.7 and a recurrence interval of 667 ± 167 yr in response to a rapid slip rate in the range 2.2–3.6 mm/yr, creating a moderate risk given a lack of historical earthquake events.

INTRODUCTION

Normal faults are prevalent in basalt environments in response to the common association between basaltic volcanism and rifting. Such faults have distinctive surface morphologies where they cut through near-surface lavas (Peacock and Parfitt, 2002; Grant and Kattenhorn, 2004; White and Crider, 2006; Rowland et al., 2007; Ferrill et al., 2011) and remain active during volcanic periods such that variably aged lava flows cut by the fault can be used as temporal markers of slip rates and slip history. We use the case example of the Hat Creek fault in northeastern California (USA) to illustrate the efficacy of using offset lava flows to constrain slip histories and earthquake potential for this fault style, as well as to provide an improved regional seismic hazard assessment related to this fault. These techniques provide a viable alternative to traditional paleoseismologic analyses, such as trenching, which are ill-suited for the analysis of faulted lavas.

The Hat Creek fault is located within a volcanic corridor between Mount Shasta and Lassen Peak, near the southern end of the Cascade Range and its associated underlying subduction system (Fig. 1) (Wills, 1991; Muffler et al., 1994; Blakely et al. 1997; Walker, 2008). Normal faulting and recurring volcanic activity from more than 500 vents over the past 7 m.y. created a pervasively faulted volcanic region in the vicinity of Lassen Peak (Muffler et al., 1994). The fault is located in the extending arc and/or backarc transition of the Cascadia subduction zone, marking the approximate western margin of a Miocene and younger volcanic highland called the Modoc Plateau (White and Crider, 2006). The Modoc Plateau hosts numerous Neogene and Quaternary normal faults, similar in style to the Basin and Range province to the east (LaForge and Hawkins, 1986). The north-northwest-trending, west-dipping Hat Creek fault is the most prominent normal fault in the region. It is highly segmented and is composed of three subparallel systems of scarps of different ages (Fig. 2A) that accrued a cumulative throw in excess of 600 m (Muffler et al., 1994; Walker, 2008). The oldest and largest system of scarps, referred to as the Rim, has as much as ∼350 m of throw and defines the easternmost extent of the fault system. The 47-km-long Pleistocene Rim consists of seven right-stepping, northwest-oriented segments ranging in length from ∼1–16 km (Walker, 2008). These scarps are heavily vegetated and have prominent talus piles that reflect gradual geomorphic modification of the scarps to slope angles of ∼30°–45°. There is no evidence of disruption of the talus slopes by recent surface rupture, consistent with these scarps reflecting an older, abandoned portion of the fault system. Lava flows at the top of the footwall east of Murken Bench have been K-Ar dated as 924 ± 24 ka (Clynne and Muffler, 2010), constraining the maximum age of the fault system as Calabrian (late Pleistocene).

The intermediate-aged fault scarps west of the Rim are colloquially referred to here as the Pali (a Hawaiian term for eroded basaltic cliffs) and have accrued as much as ∼175 m of throw (Walker, 2008). The Pali, also Pleistocene in age, extends for ∼24 km and is made up of five left-stepping segments with generally north orientations in the southern part of the fault system where the Pali intersects the Rim, but changing to north-northwest orientations in the north where the Pali segments approach the volcanic edifice at Cinder Butte (Fig. 2A). Many of the segments are overlapping and exhibit physically connected (i.e., breached) relay ramps, creating a mechanically continuous system of interacting segments.

The youngest system of scarp segments, referred to here as the Active Scarp, has a maximum displacement of 56 m just north of Murken Bench (Fig. 2B) and exhibits evidence of repeated earthquake activity since the late Pleistocene. Surface-breaking ruptures of the Active Scarp follow the older Pali scarps within a few tens of meters of the base of the Pali, except the southernmost Active Scarp segment, which occurs a few tens of meters from the base of a southern Rim segment (Fig. 2A). The majority of the Active Scarp offsets the Hat Creek Basalt, a low-potassium olivine tholeiite that covers much of the hanging wall valley floor west of the fault scarp. The basalt originated from a cluster of vents 30 km to the south of the throw maximum and has been 40Ar/39Ar dated as 24 ± 6 ka (Turrin et al., 2007). These lavas flowed north down the Hat Creek valley, which is bounded on its western side by a ca. 500–800 ka volcanic escarpment (there is no western bounding fault antithetic to the Hat Creek fault).

Although this region of California is seismically active (Fig. 3), including small events (M<3.5) in the vicinity of the Hat Creek fault, the Active Scarp has not experienced a surface-breaking earthquake event in recorded history (∼200 yr for northern California). Nonetheless, in the southern portion of the fault system, the Active Scarp offsets glacial deposits (younger than 15 ka) by 20 m (Muffler et al., 1994; U.S. Geological Survey, 1996), indicating that 35% of the total maximum throw in the northern part of the Active Scarp and perhaps as much as 80% of the average throw along the entire fault length has accrued since these deposits formed, and suggesting that motion along the fault likely continued from the late Pleistocene into the Holocene.

The U.S. Geological Survey (USGS) Quaternary fault database (http://earthquake.usgs.gov/hazards/qfaults) currently gives an incomplete picture of the fault and its recent activity. The Hat Creek fault is listed as having been active in the past 15 k.y.; however, the database does not discriminate between different parts of the fault system that have been active at different times (i.e., Rim, Pali, and Active Scarp). The USGS also provides various earthquake hazard estimates for the Hat Creek fault system in the 2008 National Seismic Hazards Maps database (http://geohazards.usgs.gov/cfusion/hazfaults_search/disp_hf_info.cfm?cfault_id=8,%209); however, the earthquake magnitude estimates (which range from M6.5–7.2) combine three different faults (Hat Creek, McArthur, and Mayfield faults; 97 km cumulative length) that are unlikely to rupture in tandem. Details about the slip characteristics of the Hat Creek fault were documented in the USGS Quaternary fault database by Sawyer (1995) (http://geohazards.usgs.gov/cfusion/qfault/); however, that analysis also combines multiple faults into one system (59 km cumulative length), and the details of the fault slip history are poorly constrained, with a suggested recurrence interval in the range 1000–3000 yr and a slip rate of 1–5 mm/yr.

Our investigation of the Hat Creek fault tightly constrains offset and timing history that can be used to refine and advance seismic hazard assessment. In conjunction with a revised cumulative length of the Active Scarp, the history provides a more accurate estimate of the earthquake potential of that portion of the fault system that is likely to rupture in a single event. At risk are numerous local towns (Burney, Fall River Mills, Susanville, Red Bluff, and Redding are all within 90 km of the Hat Creek fault), a hydroelectric infrastructure within 20 km of the fault system, the Allen Telescope Array at the Hat Creek Radio Observatory, only 1.5 km west of the Active Scarp, and Lassen Volcanic National Park, 25 km to the south.

The long-term geometric and kinematic evolution of the fault system in the context of regional tectonics was described in Walker (2008) and Walker and Kattenhorn (2008), outlining temporally variable tectonic and magmatic influences that affected fault trace geometries. For example, whereas the Rim is made up of seven right-stepping segments, the later scarps of the Pali and the Active Scarp all have a consistent left-stepping en echelon pattern (Fig. 2). This change in geometry likely reflects the influence of regional tectonic patterns, particularly the Walker Lane belt to the southeast, which transferred dextral shear into the Lassen region (Blakely et al., 1997; Muffler et al., 2008), creating the left-stepping segment patterns and potentially driving the ongoing activity of the later scarps (Walker, 2008).

ACTIVE SCARP OBSERVATIONS

Field and Analytical Methods

Our observations of the morphology of the surface trace of the Hat Creek fault are based on seven separate field visits to all segments of the fault (Rim, Pali, and Active Scarp; Fig. 2A) between 2003 and 2012. The scarp can be reached via numerous access roads and hiking trails that lead off of State Route 89, which comes within 1 km of the fault at segment 6 (Fig. 2B). A gravel road that is along much of the top of the Rim provides numerous vantage points along the Rim footwall. These visits allowed us to collect extensive field descriptions of the scarp morphology along the entire length of the fault, with emphasis on the most recent surface ruptures along the Active Scarp.

In all locations visited along the Active Scarp, we made detailed observations of the manner in which the fault had interacted with the youngest lava flows. The two main elements of the fault trace, hanging wall monoclines and vertical scarp faces (described in the following), were examined to unravel the progressive sequence of disruption of the lavas by repeated fault motions. We noted the dimensions of columnar blocks of lava that had been disrupted or moved out of place along both the vertical scarp and the upper surface of the hanging wall monocline. Relative exposure ages of different portions of the vertical scarp were qualitatively linked to the amount of lichen coverage of exposed columns in the scarp face.

We did not date samples of lava flows cut by the Active Scarp, relying instead on robust ages provided by previously published works. General flow directions of lavas were determined based on flow extents relative to known source vents for flows of different ages, also documented previously (Muffler et al., 1994; Turrin et al., 2007). These flow directions were used to identify where lavas had flowed toward, and draped across, existing portions of fault scarps located at lower elevations than the source vents. Where lavas of different ages abut or overlap, we attempted to discern the locations of the lava contacts in the field; however, the vegetation density commonly makes such distinctions difficult, in which case aerial photos and Google Earth were used to approximate the contacts.

We recorded all observation locations along the fault using a handheld Trimble GeoExplorer 6000 GPS device, which was also used to record throw along the Active Scarp with a vertical precision of as little as ∼10 cm. Throw profiles for the majority of the Active Scarp are based on data derived from stereo imagery (Walker, 2008); however, GPS elevations were also obtained to derive throw profiles along portions of the Active Scarp by physically walking out both the hanging wall and footwall cutoffs and collecting GPS data points at 5 s intervals. All data were imported into an ArcGIS environment and superposed on a 30 m digital elevation model basemap (Fig. 2).

GPS field data record locations and elevations and thus only account for the vertical component of fault motion, or throw (T). Our analysis of the fault motion history also considers the component of motion in the plane of the fault itself, referred to as the slip or displacement (D). These parameters are related using the fault dip in the subsurface (θ) such that T = D sin θ (Fig. 4).

Scarp Geometry and Surface Morphology

The Active Scarp was previously mapped as consisting of 7 (Walker, 2008) or 8 (Muffler et al., 1994) left-stepping segments spaced from 0.79 to 1.83 km apart and with a cumulative length of 23.5 km. The discrepancy in number of segments documented in past work simply reflects the interpretative nature of separating out connected segments. We adopt the interpretation of 7 segments along the previously identified extent of the Active Scarp (Fig. 2B). The strike of the Active Scarp is subparallel to the Rim and Pali segments in the southern third of the fault system (segments 5–7 in Fig. 2B) where the Rim and Pali converge; however, in the northern part of the fault system, both the Pali and the Active Scarp diverge from the dominant trend of the older scarp system, following a northwest trend toward the ∼335-m-tall, 8.5-km-wide shield volcano Cinder Butte (Fig. 2). The westward migration of successive scarp system segments of the Hat Creek fault suggests that Cinder Butte and its underlying magmatic system may have focused the development of the active portion of the fault in the proximity of Cinder Butte. Such a model is consistent with documented examples of magmatic systems affecting fault growth and orientation in response to local stress perturbations related to magma pressure (Clifton and Schlische, 2003; Clifton and Kattenhorn, 2006; Rowland et al., 2007; Gudmundsson et al., 2009) or the topographic and mechanical attributes of volcanic constructs (Friese, 2008; Jenness and Clifton, 2009). Lava flows from Cinder Butte covered the northern end of the Pali scarp; however, the Active Scarp subsequently dissected these lavas and curved toward the center of Cinder Butte (Fig. 5), indicating more recent motion on the fault relative to Cinder Butte volcanism. The final stage of Cinder Butte volcanism ended at 38 ± 7 ka with the eruption of the basaltic andesite of Cinder Butte (Turrin et al., 2007).

The majority of the Active Scarp ruptures through the 24 ± 6 ka Hat Creek Basalt, which covers much of the Hat Creek valley and accumulated at the base of preexisting scarps of the Rim (in the south) and the Pali (further north). Rupture of the Active Scarp through the Hat Creek Basalt resulted in surface features that are characteristic of active normal faults cutting basaltic lava flows (Gudmundsson, 1987a, 1987b, 1992; Grant and Kattenhorn, 2004; Martel and Langley, 2006), including vertical scarps where the fault breached the surface along columnar joints and subsequently accrued throw, dilational cracks and fissures along the base of the scarp, and as much as ∼40 m wide zones of basalt rubble at the base of the scarp created by repeated episodes of fault rupture (Fig. 6A).

Although geomorphic processes have affected scarps of all ages along the Hat Creek fault, evident in the rubble piles that collected along Rim and Pali scarps in response to mass wasting, earthquake-related ground shaking effects are the dominant geomorphic modifier along the relatively youngest Active Scarp. For example, significant disaggregation of the network of small columns (generally <20 cm across and <30 cm tall) within the rapidly cooled upper surface of the Hat Creek lavas is common even where lava surfaces are subhorizontal, implying strong ground shaking (Fig. 6C). In some instances, individual columns were lifted up out of place and onto the surface of the flow, leaving cavities within the flow top that can be matched to individual column blocks like puzzle pieces, implying ground accelerations sufficient to overcome the weight of the blocks (i.e., >1 gn). Large basalt columns (1–2 m wide and several meters tall) within the hanging wall have tilted, fallen over, or broken apart adjacent to the scarp, suggesting these areas have undergone significant ground shaking. This shaking also caused similar-sized columns in the footwall to fall from the vertical scarp and infill the fissure below, exposing less-weathered and relatively lichen-free surfaces along the scarp wall. Older surfaces are coated by lichen of species Rhizocarpon with maximum diameters of ∼24 mm that imply an exposure time of at least 250 yr (e.g., Bull, 2000), although lichenometric dating techniques are not robust.

The Active Scarp is also characterized by a hanging wall fault-trace monocline (Fig. 6B), representing the near-surface flexing of the Hat Creek Basalt above an upward-propagating fault tip prior to initial surface breaching by the fault. Such monoclines are common along normal faults that rupture to the surface through young basalt flows (Grant and Kattenhorn, 2004; Martel and Langley, 2006; White and Crider, 2006; Rowland et al., 2007). Along the Active Scarp, the monocline accounts for as much as 33 m of throw (Walker, 2008), implying numerous fault slip events prior to breaching of the surface along the upper hinge line of the fold. Once the surface was breached by the fault, monocline growth ceased, and all subsequent surface throw accumulation was directed along the vertical scarp.

Progressive disaggregation of the monocline occurred along the Active Scarp in response to the local effects of repeated fault rupture, with the final stage of monocline history being its complete collapse, rendering the monocline to a pile of rubble (Fig. 6A). The rubble pile is commonly composed of intact columns of basalt with individual dimensions of several meters and with large open cavities between the blocks. Monocline disaggregation is most advanced along fault segments with the largest cumulative throws (segments 2–6 in Fig. 2), and is least advanced where throw decreases toward the segment tips or at relay zones, implying a relationship between cumulative fault activity and monocline breakdown.

Taken together, these characteristics of monoclines in various stages of breakdown suggest repeated earthquake-induced ground shaking events. Intact or only partially disaggregated breached monoclines can be seen along portions of the fault where the scarp height is at least 6 m, implying that multiple surface-breaching earthquake events are needed to induce monocline disaggregation or destruction after initial breaching of the monocline upper hinge line.

Maximum Fault Throw

The northern portion of the Active Scarp (segments 1 and 2; Fig. 2B), north of Murken Bench, crosses the northern extent of the Hat Creek Basalt along the fault (Fig. 5), beyond which the fault ruptures through basaltic andesite (38 ± 7 ka) originating from several vents concentrated at Cinder Butte (Turrin et al., 2007). The maximum observed throw of 56 m (determined from GPS data with 10 cm vertical precision) offsets the northern extent of the Hat Creek lavas at the transition to older Cinder Butte lavas. Larger offsets are apparent for the upper surface of the Cinder Butte lavas (at least 70 m but potentially as much as 83 m), indicating 14–27 m of throw accrual along the fault during the ∼14 k.y. period between the eruption of the Cinder Butte lavas and the Hat Creek lavas near the end of the Pleistocene.

Southward-flowing lavas from Cinder Butte were deformed and offset by the northern portions of the Pali (and later the Active Scarp), which propagated northwest toward Cinder Butte as the fault system seemingly responded to the Cinder Butte magmatic system (Figs. 7A, 7B) (Walker, 2008). The uncertainty in the amount of Active Scarp offset of the Cinder Butte lavas derives from the fact that later lavas of this eruptive period flowed south along the axis of the Pali scarps, resulting in bifurcation of lava flows into two lobes that flowed onto the fault footwall and down onto the hanging wall (Fig. 7C). The resultant vertical offset of the two lobes of the lava flow was thus not tectonic, but geomorphic. Subsequent activity along the fault after the eruption of the Cinder Butte lavas requires some portion of the total offset to be fault-related throw. In contrast, Hat Creek lavas flowed from the south and so accumulated only in the hanging wall of the Pali scarp (Fig. 7D). These lavas were subsequently deformed as the Active Scarp broke to the surface through them, such that the entire 56 m of throw within these lavas is decidedly fault-related (Fig. 7E).

Ordinarily, it may be impossible to distinguish between the geomorphic and tectonic components of offset where lavas have flowed along both the footwall and hanging wall sides of an active fault because the height of the scarp would need to be known at the time of lava emplacement across the fault (which is unlikely to be determinable). In such cases, the scarp height during lava emplacement would be subtracted from the cumulative offset at the time of measurement to ascertain the total tectonic offset. In the case of the Hat Creek fault, however, the Active Scarp postdates both the Cinder Butte and Hat Creek lava flows and formed on the hanging wall side of the older Pali scarps. Hence, any offset of the Cinder Butte lavas across the Active Scarp must be purely tectonic. The southern margin of the south-flowing Cinder Butte lavas on both the footwall and hanging wall sides of the Active Scarp, where it is onlapped by the north-flowing Hat Creek lavas, exhibits at least 70 m of cumulative fault throw. A second measurement of 83 m of throw was taken ∼150 m north of the first measurement. In this location, the Active Scarp and Pali scarp appear to merge (Fig. 7E); therefore, a portion of this 83 m offset may be geomorphic, indicating that 70 m is the more reliable measurement of minimum total tectonic throw for the Cinder Butte lavas. The combination of geomorphic and tectonic offset of young lavas highlights the importance of incorporating the relative timing and propagation directions of fault-growth events and lava-flow episodes and advancement directions into fault-offset analyses in tectonically active volcanic settings in order to obtain appropriate slip rates.

Revised Active Scarp Length

The previously identified northern extent of the Active Scarp terminates within Cinder Butte (Muffler et al., 1994). Our investigation of the fault system and field mapping northeast of Cinder Butte suggests the continuation of young rupture activity, implying the length of the Active Scarp has been underestimated. In this region, the base of the Rim scarp created a buttress for lavas flowing across the hanging wall of the Rim (Fig. 5). Some of these lavas flowed eastward away from eruptive center at Cinder Butte, covering an older flow surface of the basalt west of Six Mile Hill, dated as 53.5 ± 2 ka (Muffler et al., 2012). The lavas of the basalt west of Six Mile Hill flowed southward along the base of the Rim, following a local slope toward the throw maximum along the Rim 8.5 km south of the source vents immediately south of the Cassel–Fall River Mills road (Fig. 5). The surface morphologies of these lavas are distinct from older Pleistocene flows exposed in the Rim, defining a youthful but rugged lava surface with numerous tumuli and deflation pits. The roughness of this lava surface complicates the identification of the eastern extent of the Cinder Butte lavas. The Cinder Butte lavas appear to abut the base of the Rim scarp due east of the high point of Cinder Butte but likely did not reach the Rim to the northeast of the high point (Muffler et al., 2012) as a result of emplacement across the southward sloping surface of the older lavas of the basalt west of Six Mile Hill (Fig. 5).

Young ruptures through the lavas along the base of the northern Rim scarp show striking similarities to the previously documented surface ruptures along the fault segments south of Cinder Butte. This portion of the fault was not previously identified as part of the Active Scarp, with presumed late Pleistocene most recent activity in the USGS Quaternary fault database. The young ruptures along the base of the northern Rim scarp motivate us to test if they are part of the Active Scarp system (i.e., rupture in tandem with previously identified active segments), resulting in a longer fault length that should be considered in a seismic hazard analysis, or if they represent independently rupturing parts of the overall Hat Creek fault system.

To map the young scarps along the northern Rim northeast of Cinder Butte, we walked the entire length of the rupture (both footwall and hanging wall), mapping it with differential GPS to capture the segment geometries and throw distribution. We identified two new rupture segments (Fig. 2B) extending a total of 4 km, breaking the surface ∼50 m west of the base of the Rim scarp with a maximum throw of 30 m in the southern segment. The surface rupture location relative to the base of the Rim is identical to the previously documented Active Scarp segments south of Cinder Butte, which break vertically to the surface through the Hat Creek lavas several tens of meters from the base of the Pali scarps. The southern termination of the newly identified young rupture is located 4.5 km northeast of the northern termination of the previously identified Active Scarp (segment 1) within Cinder Butte (Fig. 2B). The northern termination is ∼0.8 km north of where the fault is crossed by the Cassel–Fall River Mills road. Additional Rim segments of the Hat Creek fault continue for at least 6 km north-northwest beyond this point to just north of the Pit River; however, no evidence of recent activity was confirmed by this study.

Similar to the geometry of cracks along the Active Scarp segments south of Cinder Butte, the newly discovered rupture zone contains fracturing that has a left-stepping geometry, indicating similar rupture kinematics (normal with a small right-lateral component). Displacement of the late Pleistocene lavas has created features identical to those observed along the Active Scarp within Hat Creek Basalt, such as vertical scarps, dilational cracks and fissures, the presence of a large fault trace monocline in the hanging wall, and a large rubble zone created by ground shaking and the breakdown of the monocline (Fig. 8). The vertical fault scarp exposes older Pleistocene basalt near its southern end where late Pleistocene lavas did not reach the base of the Rim; however, where the youngest lavas abut the Rim, all of the throw is taken up within those lava flows, partially by the monocline and partially by throw along the vertical scarp. Therefore, the accumulation of the total throw must have occurred at least within the past 53.5 ± 2 k.y. (i.e., since the eruption of the basalt west of Six Mile Hill) and possibly within the past 38.5 ± 7 k.y., if the basaltic andesite lavas of Cinder Butte reached the Rim in the vicinity of the young ruptures.

The northernmost segment of the newly identified scarp definitively cuts through the basalt west of Six Mile Hill. Where exposed along the fault scarp, this lava appears noticeably older than basaltic andesite lavas from Cinder Butte, with a significantly greater amount of surface lichen. Nonetheless, the fault scarp features in the basalt west of Six Mile Hill are similar to other portions of the Active Scarp, including a variably disaggregated hanging wall monocline and a vertical fault scarp. The throws within this unit are as much as 23 m within the northernmost recently active segment of the fault. The combination of evidence based on surface morphologies suggests that the Hat Creek fault ruptured through at least three late Pleistocene lava flows of different ages, with young surface rupture both south of Cinder Butte (the previously identified Active Scarp) and northeast of Cinder Butte, along the base of old Rim segments.

DISCUSSION

If the newly mapped rupture trace is the continuation of the Active Scarp to the northeast of Cinder Butte, it indicates that the most recent surface ruptures along the Hat Creek fault extend further north than previously considered. Therefore, the Hat Creek fault has remained active along the Rim in the northern part of the fault system, despite having abandoned the Rim portion of the fault system further south. The additional segments of the Active Scarp northeast of Cinder Butte define a 4.5 km right step and a 2.5 km along-strike gap within the fault system, increasing the total length of young ruptures by 6.5 km compared to prior estimates. Although predominantly a left-stepping fault system in response to dextral-oblique extension, this right step in the fault geometry is interpreted to result simply from the effect of the Cinder Butte magmatic system on the temporal evolution of the southern portion of the fault. Large steps are not uncommon in segmented normal fault systems and are not necessarily hindrances to earthquake ruptures. For example, steps from 3 to 8 km wide were associated with both the 1954 Dixie Valley (moment magnitude, Mw 6.8) and the 1915 Fairview Peak (Mw 7.2) earthquakes in Nevada (Zhang et al., 1991).

There are numerous lines of evidence to suggest that the newly identified segments are part of the Active Scarp system and rupture in tandem. For example, the various morphologic features consistent with the Active Scarp, such as vertical scarps flanked by a fault-trace monocline, the disaggregated appearance due to the collapse of the monocline during earthquakes, and offsets of relatively young lava flows, suggest these fault segments have both undergone recent rupture with significant ground shaking and should thus be considered together when evaluating the credible earthquake magnitude and seismic hazard potential of the region. To test this assertion, we explore fault evolution and earthquake scenarios in which the newly mapped segment is first treated as an independent fault and then considered to be incorporated into the entire Active Scarp system.

Estimation Method for Earthquake Slip and Recurrence

Empirical relationships between rupture length and maximum surface displacement during discrete earthquake events (e.g., Wells and Coppersmith, 1994; Wesnousky, 2008) permit the estimation of slip rates and recurrence intervals between earthquakes if the ages of offset layers are known. For example, Wells and Coppersmith (1994) presented empirical data for normal fault lengths in the range 3.8 km to 75 km to generate a regression line for maximum displacement (MD, in meters) versus surface rupture length (SRL, in kilometers). The relationship is given by log (MD) = –1.98 + 1.51 × log (SRL) for normal faulting. Maximum displacement per event can be used to determine the number of events required to accrue the cumulative displacement along the fault. If the age of the oldest offset unit is known, the number of events in this time interval informs us about the recurrence interval between events. This method uses the simplifying assumption of characteristic earthquake events (equal displacement per event and constant rupture length). Nonetheless, it provides a reasonable insight into earthquake activity in the absence of additional information such as paleoseismologic evidence from scarp trenching, which is typically limited to only the last several events and is not well suited to paleoseismological analysis in faulted basalt. Moreover, the Active Scarp represents the reactivation of an existing fault system at depth (the Pali and the Rim), which controlled the rupture length of the Active Scarp as it developed through young lavas. The surface trace length thus likely remained approximately constant through time, as did the maximum displacement per event.

Regression relationships are based on maximum surface displacement, not throw (which is the vertical component of the fault displacement). Faults scarps at Hat Creek are vertical within ∼50 m of the surface, where they break through the Hat Creek lavas (Muffler et al., 1994; Walker, 2008); however, slip along the fault below this depth occurs along a dipping fault plane, resulting in a component of dilation at the surface along the vertical scarp. This phenomenon is typical of dilational faults in basalt lavas (e.g., Grant and Kattenhorn, 2004; Ferrill et al., 2011). Throw is converted to displacement assuming a typical subsurface normal fault dip of 60° (Anderson, 1951). This assumption is reasonable given that geomorphically modified scarps of the Pali and Rim have typical dips of ∼45°, indicating originally higher dips.

Earthquake Potential of Newly Mapped Segments

The newly mapped, recent rupture segments along the northern portion of the Rim have a total rupture length of 4 km (i.e., within the range used in the Wells and Coppersmith regression). If these segments rupture independently of any other segments of the Hat Creek fault, the slip rate and recurrence interval must accommodate the offset of the 53.5 ± 2 ka basalt west of Six Mile Hill. Given the maximum throw of 30 m (corresponding to 34.6 m of displacement along the fault plane; Fig. 4), the newly identified young fault scarps would have an associated slip rate of ∼0.65 mm/yr (or in the range 0.6–0.7 mm/yr when accounting for lava age uncertainty). Using the Wells and Coppersmith (1994) regression, a normal fault with a length of 4 km (which is somewhat low for a surface-breaking fault rupture) should have an average maximum displacement of only ∼8.5 cm per rupture event (equivalent to ∼7.5 cm of throw), implying a low recurrence interval of 134 ± 5 yr to accrue 30 m of throw since the faulted lava was erupted.

Rupture of these fault segments could produce a Mw 5.8 earthquake, using the regression relationship between moment magnitude and maximum displacement per event (MD, in meters), given as M = 6.61 + 0.71 × log (MD) (Wells and Coppersmith, 1994). Such an event would probably be felt regionally in northeastern California. For example, a M5.7 earthquake along a normal fault located ∼65 km south-southeast of the southern end of the Hat Creek fault on 23 May 2013 was felt as far south as the San Francisco Bay area, as well as north into southern Oregon and east into central Nevada (USGS earthquake database: http://earthquake.usgs.gov). However, no seismic events have been attributed to the fault in recorded human history in the region (∼200 yr), and there is no field evidence of a very recent rupture. It is therefore more likely that the newly mapped segments are part of a larger system that ruptures less frequently—specifically, the entire Active Scarp portion of the Hat Creek fault.

Slip Rate Analysis

As additional evidence in support of this assertion, we consider the throw distribution along the entire length of the Active Scarp. It is well documented that interacting normal fault segments distribute throw throughout the length of a fault system (i.e., kinematic coherence; Figs. 9A, 9B), typically producing the maximum throw at the center of the fault trace (e.g., Walsh and Watterson, 1991; Dawers et al., 1993; Cartwright et al., 1995; Willemse, 1997). The throw commonly exhibits an approximately elliptical distribution, attenuating from the maximum at the center of the surface trace to zero at the fault tips, with local variability at segment boundaries within the fault system related to either fault growth history (Childs et al., 1995; Kattenhorn and Pollard, 2001) or the partial accommodation of fault throw by relay ramp deformation (Huggins et al., 1995; Blakeslee, 2012).

Along the previously identified seven segments of the 23.5 km Active Scarp, throw is distributed among segments by mechanical interaction, whereby the fault segments are affected by the presence of one another despite spacings or underlaps between segments of hundreds of meters to several kilometers (Fig. 2B). However, the throw versus distance profile along these seven segments (Walker, 2008) does not show a maximum throw at the center of the total length (Fig. 9C). Instead, the throw profile is greatly skewed toward its northern end, where the 56 m maximum throw displaces the Hat Creek Basalt within segment 2, just south of Cinder Butte (Fig. 5). The throw profile shape suggests that the previously identified seven segments of the Active Scarp do not represent the full length of the active fault system, requiring mechanical interaction with more segments north of segment 1. The addition of the newly mapped, young segments northeast of Cinder Butte extends the active fault system northward, producing a maximum throw more centered along the rupture length and an overall throw distribution with a more symmetric, somewhat elliptical shape (Fig. 9C). The throw pattern implies that the newly mapped segments are kinematically coherent with the previously identified Active Scarp, rather than being part of an independent fault. Nonetheless, the throw profile is somewhat skewed toward the north end, raising the possibility of additional active segments farther north (Figs. 2 and 3), where the Hat Creek fault approaches the Pit River; however, no clear evidence for young rupture (e.g., through alluvial deposits that cross portions of the fault) was observed in that area during a reconnaissance survey.

Mechanical interaction and resultant slip partitioning in segmented normal fault systems results in higher throws (both cumulative and per earthquake event; e.g., Dawers and Anders, 1995; Willemse, 1997) and higher slip rates for segments near the centers of fault systems. For example, the 387 km Wasatch fault in Utah has a slip rate of 1–2 mm/yr in the central segments, decreasing to 0.5 mm/yr in the distal segments (Machette et al., 1991). Analogously, the active portion of the Hat Creek fault exhibits variable slip rates in different segments along its length. To characterize the earthquake potential, including slip rates, slip-per-event, recurrence intervals, and earthquake magnitude along a fault system consisting of multiple segments that rupture in tandem, we use the assumption that the segment containing the maximum cumulative throw should be used. In so doing, we account for the maximum amount of throw that necessarily accumulated during a determined time interval: in this case, the age of the Hat Creek lavas. Segment 2 has the maximum throw (56 m in 24 ± 6 ka lavas) and hence the highest slip rate averaged since the late Pleistocene: 2.7 mm/yr, or in the range 2.2–3.6 mm/yr given the uncertainty of the Hat Creek lava age. This estimate assumes a subsurface fault dip of 60° and so a total displacement of 64.7 m in the plane of the fault (Fig. 4). We conservatively assume pure dip-slip motion along the fault, although a slight dextral component of motion may be present based on the presence of left-stepping fractures along the surface rupture trace. Hence, actual slip rates may be slightly higher than we calculate.

The maximum throw of the Cinder Butte lavas (at least 70 m) is less definitive; however, this estimate would require a minimum slip rate of 2.1 mm/yr (or in the range 1.8–2.6 mm/yr taking into account lava age uncertainty) since these slightly older lavas erupted, approximately consistent with the slip rate deduced from the offset of the Hat Creek lavas. Compared to other normal fault slip rates, such as the Wasatch fault, the Hat Creek fault slip rate is relatively high. The 0.65 mm/yr slip rate computed for the newly mapped segments northeast of Cinder Butte is consistent with those segments being at the distal end of the coseismic rupture segments of the Active Scarp. Analogously, at the southern end of the Active Scarp, segment 6 has an effective slip rate 1.0 mm/yr (or in the range 0.8–1.4 mm/yr taking into account lava age uncertainty). The observed attenuation of the slip rate to the distal segments is thus consistent with our assertion that the newly mapped segment is part of the Active Scarp system.

Slip rate estimates provide some insights into the timing of the long-term evolution of the fault system. For example, a hypothesized stress perturbation induced by the Cinder Butte magmatic system caused a new branch of the fault to propagate away from the Rim and redirect fault activity toward Cinder Butte, forming the Pali scarp system west of the original Rim scarps. Given the 175 m maximum throw along the Pali system and the simplifying assumption of a constant a slip rate of 2.7 mm/yr, the evolution of the Pali may have commenced ca. 65 ka, indicating ∼30 k.y. of activity in the Cinder Butte magmatic system prior to the eruption of the youngest lavas at 38 ± 7 ka. In this time interval, the Pali scarps could have acquired maximum throws of ∼70 m by the time the basaltic andesite lavas erupted. Although the highest Pali scarps are ∼5.4 km south of the southern extent of the Cinder Butte lavas (Fig. 5), the geomorphic offset of bifurcating lava lobes flowing onto the footwall and hanging wall blocks of the Pali scarp closer to Cinder Butte (Fig. 7C) may nonetheless have been quite significant. This possibility strengthens the argument that only the throw offset of Cinder Butte lavas across the relatively younger Active Scarp can be reliably used to estimate fault slip rates.

One final consideration regarding slip rate analysis relates to the seemingly anomalous nature of the throw peak along segment 2 (Fig. 9C) where the maximum throw of 56 m was measured by GPS. Although approximately elliptical throw profiles have been noted along other mechanically interacting segmented normal fault systems (Dawers and Anders, 1995; Willemse, 1997), the throw peak along segment 2 appears to fall above any choice of ellipse to approximate the overall throw profile. There are many reasons why this throw peak may have occurred. One possibility is that throw has been underestimated along the Active Scarp segments south of the throw peak as a result of the throw being partitioned onto other fault segments during earthquakes. For example, in the vicinity of Active Scarp segments 3 and 4, there are two north-oriented fault scarps that appear to link the Pali with the Rim (Fig. 2). Although we found no clear evidence of recent rupture along those segments, the Hat Creek lavas did not reach these segments, which may make recent rupture evidence difficult to identify. Another possibility is that the overall pattern of throw is highly skewed toward the northern end of the Active Scarp, with a maximum at segment 2. Such skewed cumulative throw profiles have been noted (e.g., Willemse et al., 1996) where fault segments mechanically interact with a nearby perturbing influence, such as another fault (or in this case, perhaps Cinder Butte).

Ultimately, throw profiles simply reflect the cumulative effects of interactions between different components of a constantly evolving segmented fault system; therefore, no faults have truly elliptical throw profiles, and local peaks are not uncommon (Fig. 9B) (cf. Dawers and Anders, 1995; Cartwright and Mansfield, 1998). Our careful matching of Hat Creek lavas across the Active Scarp at segment 2 using field-based GPS measurements leave us confident in the accuracy of the 56 m throw accumulation (and hence our calculated slip rate) since the Hat Creek lavas erupted, pooled against the existing Pali scarp, and were subsequently offset by the development of the Active Scarp (Fig. 7).

Revised Credible Earthquake Magnitude

The additional active segments increase the originally mapped rupture length by 6.5 km to a total of 30 km. Applying the surface rupture length to maximum displacement scaling relationship for normal faults described earlier (Wells and Coppersmith, 1994), a 30 km fault should manifest a maximum displacement of 1.78 m per rupture event (or 1.54 m of maximum throw assuming a fault dip of 60° in the subsurface), implying a recurrence interval of 667 ± 167 yr for the 56 m throw maximum in 24 ± 6 ka lavas. This recurrence interval necessitates ∼15 Holocene rupture events along the Active Scarp in addition to the late Pleistocene activity that postdated the eruptions of the basalt west of Six Mile Hill, Cinder Butte basaltic andesite, and the Hat Creek Basalt.

The concomitant potential earthquake magnitude is Mw 6.7 based on maximum displacement versus magnitude relationships for normal faults, as previously described (Wells and Coppersmith, 1994). Magnitude versus rupture area relationships (M = 3.93 + 1.02 × log (RA), with RA in square kilometers) will provide the same result for a seismogenic thickness of 15 km (down-dip rupture width of 17.3 km); however, greater seismogenic thicknesses will increase slightly the maximum credible earthquake. For example, an 18–20 km seismogenic thickness would only increase the potential earthquake magnitude to Mw 6.8. Given the relatively low recurrence interval and the lack of historic earthquake events or knowledge of the timing of the last event, the Hat Creek fault may thus provide a greater probabilistic seismic hazard than has been previously realized.

CONCLUSIONS

Field observations and mapping along the Hat Creek fault lead us to propose the existence of two previously overlooked northern segments of the Active Scarp, providing new insights about the geometry, evolution, and seismic potential of the fault. The newly mapped segments are kinematically coherent with the previously identified Active Scarp segments, with a surface morphology in 53.5 ± 2 ka lava flows identical to surface ruptures along the previously mapped Active Scarp within the 24 ± 6 ka Hat Creek Basalt to the south. Near the center of the reinterpreted Active Scarp system, a maximum of 56 m of offset of Hat Creek Basalt implies a late Pleistocene–Holocene slip rate of 2.2–3.6 mm/yr, implying a very active extensional fault system with high strain rates, possibly reflecting the contribution of a local magmatic extension component to the overall strain budget.

Based on earthquake magnitude and slip-per-event scaling relationships for the reinterpreted 30 km rupture length, we estimate a credible earthquake magnitude of Mw 6.7 and a recurrence interval of 667 ± 167 yr to account for the cumulative throw. As one of the most prominent faults in the area, but one that lacks historical earthquake events (the timing of the last event is unknown but is likely in excess of 200 yr ago in order to predate historical records of earthquakes in the region), the active portion of the fault is thus a potential significant earthquake hazard for northeastern California.

The Hat Creek fault study provides a context for the evaluation of earthquake hazards, slip rate analysis, and recurrence interval determination in tectonically active volcanic environments. Young basalt lavas can be reasonably well dated using Ar-Ar analysis, and thus provide useful temporal offset markers. True tectonic offsets can be distinguished from any geomorphic offset caused by lavas flowing across existing scarps through field analysis of flow directions and characteristics relative to developing fault scarps. Using the simplifying assumption of characteristic earthquake events during the period of tectonic offset of young lavas (i.e., constant slip-per-event and earthquake recurrence), we provide a reasonable methodology for evaluating seismic hazard in faulted lavas. This technique is particularly useful given the difficulty of trenching analysis in lavas and the limitations of other techniques such as lichenometry and cosmogenic nuclide analysis (which may be incapable of distinguishing between events that are relatively closely spaced in time).

We thank Erin Walker for developing the fault evolution history in our related study, Marie Jackson for contributing the photogrammetry database of the Active Scarp heights of the Hat Creek fault used in our throw profiles, Nicole Bellino for lichen sampling and analysis, Leslie Fernandes and Tom Sawyer for field assistance, and Patrick Muffler and Robert Krantz for helpful discussions. Portions of this work were funded under National Science Foundation grant EAR-1113677 and a Seed Grant from the University of Idaho. Digital elevation models in Figure 2 were derived from 30 m resolution data obtained from the U.S. Geological Survey National Elevation Dataset (http://nationalmap.gov/viewer.html). Earthquake data used in the production of Figure 3 were obtained from the Northern California Earthquake Data Center (http://quake.geo.berkeley.edu/). We thank Juliet Crider and Patrick Muffler for their thoughtful reviews of the original manuscript.