Application of Terrestrial Laser Scanning in determining the pattern of late Pleistocene and Holocene fault displacement from the offset of pluvial lake shorelines in the Alvord extensional basin, northern Great Basin, USA

The first-order correspondence between geodetic and tectonic displacement rates across plate boundaries (DeMets et al., 1990, 1994) is in good agreement with modeled velocity and strainrate fields based on geologic, geodetic, and seismologic observations in diffuse deformation zones such as the western U.S. Cordillera (e.g., Shen-Tu et al., 1998, 1999; Holt et al., 2000; Thatcher, 2003). The models provide results consistent with plate motions averaged over time scales of 10⁶ yr, but they are not applicable for shorter time scales (10⁴ to 10⁵ yr) and do not address whether or how geodetically determined displacements relate to slip on individual faults or fault systems within the deformed belts (Holt et al., 2000). Historic patterns of earthquake rupture, such as those along the Dixie Valley, Fairview Valley, and Pleasant Valley faults of central Nevada (DePolo et al., 1991; Zhang et al., 1991; Caskey et al., 1996), point to the spatial complexity of fault interaction. Only in rare instances, such as the Wasatch fault system of Utah (Machette et al., 1991, 1992; McCalpin and Nishenko, 1996), have patterns of deformation been studied in detail over geologic time scales and even in these cases ambiguity in timing relations obscure many aspects of the spatial and temporal pattern of displacement.

Without an understanding of how displacement is partitioned within a fault system over different length and time scales, comparison of geologic and geodetic deformation rates will remain problematic. This issue is particularly acute when geodetic and geologic deformation rates are compared in extensional basins, where deformation is accommodated to varying degrees by faults bounding opposing basin margins and where some part of the deformation is taken up by displacement on antithetic and synthetic faults within the basin. In this setting, paleoseismological investigations that focus on the major basin-bounding faults (Machette et al., 1991, 1992; Wesnousky et al., 2005) will almost certainly underestimate the total vertical component of displacement, leading to a mismatch between geologically and geodetically determined deformation rates.

The magnitude of geologic rate underestimation due to complex fault interaction within extensional systems is not well understood, in large part because of the lack of a well-dated reference datum with which to compare fault displacement across and within basins. Typically, fault displacements involve geologic units or geomorphic surfaces that have limited spatial extent, which when combined with ambiguities in age and correlation, can compromise measurement of displacement patterns across individual faults and fault systems (McCalpin, 1996; Carver and McCalpin, 1996; Weldon et al., 1996; Wesnousky et al., 2005).

The Alvord extensional basin of southeastern Oregon (Fig. 1) provides an ideal setting to assess the spatial and temporal pattern of intermediate time-scale fault slip on a complex system of faults that bound and lie within the basin. Active deformation in the Alvord basin is reflected in the physiography of the region and is documented by a system of Pleistocene and Holocene faults that cut late Tertiary volcanic and sedimentary rocks, Quaternary sediments, and shorelines of the ancient pluvial Lake Alvord (Hemphill-Haley et al., 2000; Weldon et al., 2002; Singleton, 2004). The Pleistocene– Holocene shorelines rim the ancient lake and provide an areally extensive vertical datum to measure the differential and cumulative displacement on the Alvord fault system. By applying Terrestrial Laser Scanning (TLS) geo-spatially referenced by the Global Positioning System (GPS), we produced high-resolution images of the shoreline surfaces that allowed reconstruction of the paleohorizontal markers and assessment of the displacement history of faults within the basin.

In the western U.S. Cordillera, the transition from a transform to convergent plate boundary is clearly reflected in the pattern of active deformation and related physiography of the continental margin (Fig. 1). Late Cenozoic deformation along the southwestern transform margin is broadly distributed from the San Andreas fault system eastward into the Basin and Range (e.g., Hamilton and Myers, 1966; Atwater, 1970; Eddington et al., 1987; Oldow et al., 1989; Burchfiel et al., 1992; Atwater and Stock, 1998). Active deformation is marked by the pattern of earthquake epicenters and is concentrated in the California borderland and along the eastern and western flanks of the Great Basin.

Geodetic rates show that ∼20%–25% of the transform displacement between the North American plate and the Pacific plate is accommodated by northwestward motion of the Sierra Nevada block at 11–14 mm/yr (Gan et al., 2000; Dixon et al., 2000; Bennett et al., 2003). Between 70% and 80% of the differential motion between the Sierra Nevada block and North America is taken up on active structures along the western margin of the western Great Basin (Bennett et al., 2003; Oldow, 2003). Wide-aperture continuous GPS coverage in the northern Great Basin, northern Rocky Mountains, and Cascadia shows site velocities increasing from 2 to 3 mm/yr in the east to rates of nearly 20 mm/yr along the Cascadia margin (Miller et al., 2001; Bennett et al., 2003; Mazzotti et al., 2003). The displacements are attributed to displacement transfer from the Great Basin (Pezzopane and Weldon, 1993) and rotation of crustal blocks in response to the northwest migration of the Sierra Nevada block and oblique convergence on the Cascadia subduction boundary (Wells et al., 1998; McCaffrey et al., 2000; Hammond and Thatcher, 2005).

The transition zone between the northwestern Great Basin and the mountains of central Oregon is expressed in the physiography and is characterized by several north- to north- northeast–trending extensional basins that terminate to the north along a west-northwest–trending line that corresponds to a distributed belt of deformation known as the Brothers fault zone (Lawrence, 1976; Pezzopane and Weldon, 1993). The basin-bounding faults have trace lengths of 10–200 km and typically are characterized by between 40 m and 3000 m of throw of Pliocene sedimentary rocks and basalt that caps the late Tertiary stratigraphy of the region. Of the basins, the structure with the most pronounced physiographic expression is the Alvord basin of southeastern Oregon (Fig. 2).

The northward-trending Alvord basin is a 200-km-long fault-bound depression as wide as 15 km that stretches from northwestern Nevada into southeastern Oregon (Fig. 2). The Alvord Desert in the northern part of the basin is bound on the west by a prodigious physiographic escarpment (Fig. 3) with >1760 m of relief from the peak of Steens Mountain (2980 m) to the basin floor (1220 m). The elevation of western escarpment gradually decreases to the north (1650 m) as the basin bifurcates into a western and eastern arm separated by a central highland that locally exposes bedrock. To the south, the Alvord Desert passes into Pueblo Valley and the elevation of the western flank decreases to 2600 m in the Pueblo Mountains. The Pueblo Mountains terminate to the south in Bog Hot Valley, which is linked by a narrow pass separating the Pueblo Mountains and Pine Forest Range (Fig. 2). The eastern margin of the Alvord Desert and the Pueblo Valley is lower than the western flank of the depression and ranges from 2400 m in the south to 1400 m farther north. For most of its extent, the eastern margin is abrupt and defined by a sharp step in topography.

Bedrock exposures are dominated by Tertiary volcanic and volcaniclastic rocks that are found throughout southeastern Oregon and southwestern Idaho (Walker et al., 2003; Bond et al., 1978). Along the southwest margin of the Alvord basin, Tertiary rocks overlie Mesozoic metamorphic tectonites (Wyld and Wright, 2001), but for the most part the bottom of the Tertiary section is not exposed. The most prevalent unit of the Tertiary section is 1000 m of middle Miocene Steens Mountain volcanics (Walker, 1977; Hooper et al., 2002) that locally unconformably overlie pre-Tertiary rocks and rest upon as much as 750 m of Oligocene–Miocene interbedded tuff and tuffaceous sedimentary rocks along the eastern flank of Steens Mountain (Walker, 1977). Overlying the Steens volcanics is an upper Miocene to lower Pliocene sequence of siliceous tuff and tuffaceous lacustrine and fluvial clastic rocks, which locally are interbedded with basalt (Walker, 1965). The younger sequence varies in thickness, but locally is as thick as 500 m (Walker, 1965).

The Alvord basin is filled with poorly consolidated alluvial and lacustrine sediments of Pliocene to Pleistocene and Holocene age overlying Tertiary sedimentary and volcanic rocks. The basin is internally segmented by faults. In the north, the basin is divided into northeast and southwest subbasins by a north-northwest–trending central ridge that exposes bedrock at the northern and southern ends (Fig. 3). The sedimentary fill reaches thicknesses of at least 1.5–2.5 km in the deepest parts of the basin, but is only a few hundred meters thick over the central ridge (Whipple and Oldow, 2004).

The basin morphology is produced by a fault system that forms the eastern and western margins of the structure. The faults have a dogleg geometry and change orientation along the axis of the basin (Fig. 3). The complex array of faults forming the western margin of the Alvord basin constitutes several segments of the Steens fault system (Hemphill-Haley et al., 1999; Weldon et al., 2002; Personius et al., 2006, 2007). The eastern flank of the basin is bound by the East Pueblo Valley fault in the south and the Tule Springs Rim fault in the north.

In the central and northern parts of the basin, faults change orientation along strike and the deep axis of the basin steps from one side of the topographic depression to the other. In Pueblo Valley, the faults strike north-northwest and form a half-graben with a deep axis located along the western margin of the basin. At the transition from Pueblo Valley and the Alvord Desert, the deep axis steps east and is along the western margin of the central topographic high that segments the basin (Fig. 3). The central topographic high is formed by a system of east- and west-dipping normal faults that kinematically link the north-northeast–striking, east-facing normal faults of the Steens Mountain escarpment with west-facing faults bounding the southern Tule Springs Rim (Whipple and Oldow, 2004). In the Alvord Desert, the deep axis is along the Steens fault at the base of the prominent physiographic escarpment along the western margin of the basin.

Today the basin has no external drainage, but the low topographic barrier along the eastern margin of the northern subbasin is breached by two east-draining canyons with spill heights of 1329 and 1283 m cut into the Tule Springs Rim (Carter et al., 2006). At the north end, where the basin bifurcates into eastern and western valleys, the topographic lows terminate in passes that are at or above 1340 m. The basin is closed to the south by a height of land with a minimum elevation of 1430 m. During Pleistocene interglacial intervals, the basin was filled by ancient Lake Alvord (Morrison, 1991; Reheis, 1999; Carter et al., 2006), which produced shorelines all around the basin. The shorelines form a series of wave-cut terraces with elevations that range from 1350 m down to the basin floor at 1220 m.

The Alvord basin is in the seismically quiescent northwestern Great Basin, and only four earthquakes (M <4) have been recorded in the region over the past 100 yr (Fig. 1). Nevertheless, active extension is expressed by the basin physiography and by normal faults with alluvial scarps mapped along the margins and within the basin (Weldon et al., 2002). The extension direction for the region is poorly documented, but offset on active faults in the Alvord basin is dominantly dip slip. Where well-preserved drainages cross fault scarps with several meters of vertical separation, our reconnaissance studies show no significant horizontal offset.

Contemporary displacement is recorded across the basin by two continuous GPS sites forming the southern extent of the Pacific Northwest Geodetic Array (Miller et al., 2001). The eastern GPS site is located 42 km to the northeast of the Alvord basin and the western site is 60 km to the southwest. The two sites provide a displacement baseline over a distance of 50 km measured perpendicular to the trend of the basin-bounding fault systems. Based on ten years of observation, the continuous sites indicate that the western site is moving west-northwest at 1.75 ± 0.3 mm/yr (Fig. 2) with respect to the eastern site (Mazzotti et al., 2003).

The fault displacement history within the Alvord basin was studied by Hemphill-Haley (1987) and Hemphill-Haley et al. (2000), who mapped several faults cutting Quaternary sediments in the central part of the basin, and by Personius et al. (2006, 2007), who mapped and trenched the southern part of the Steens fault system exposed in Bog Hot Valley (Fig. 2). Two trenches across the 27-km-long Alvord fault, a strand of the Steens fault system in the central part of the Alvord basin (Fig. 3), record at least two surface ruptures of 1.1–2.0 m. Displacement occurred since 8190 ± 2240 ka and before 470 ± 350 ka, based on ¹⁴C ages of charcoal (Hemphill-Haley et al., 2000). The upper age is not considered a good minimum constraint because the sediments were reworked, but scarp morphology analysis suggests that displacement occurred ca. 2 ka (Hemphill-Haley, 1987). A similar age was determined by scarp morphology analysis on a fault near the northeastern flank of the basin (Lindberg, 1989). In Bog Hot Valley (Fig. 2) at the southern extremity of the Alvord extensional basin system, Personius et al. (2006, 2007) trenched the southern end of the Steens fault system and recognized multiple earthquake ruptures with vertical displacements of 1.1–2.2 m that occurred before 4.6 ka. Taken together, trench studies and scarp morphology analysis point to displacements on Holocene faults no younger than 2 ka, and probably no younger than 4.6 ka.

Our work in the region expands on earlier results (Hemphill-Haley, 1987; Hemphill-Haley et al., 2000) and focuses on active faults exposed in the central Alvord basin. In this area, active structures with well-developed surface expression consist of three major fault systems (Fig. 3). The Serrano fault system is along the western margin of the central basin ridge and the Alvord fault system is along the east flank of this physiographic high. The two fault systems bound bedrock exposed as two south-southeast–trending ridges forming the southwestern margin of the Alvord Desert playa. The ridges form the northern part of the central basement high that divides the Alvord basin into two subbasins. The easternmost range-front fault is along the Tule Springs Rim, forming the eastern flank of the basin. The fault along the eastern margin of the basin is largely covered by sand dunes and is best located by a steep gravity gradient along the range front (Cleary et al., 1981a, 1981b; Whipple and Oldow, 2004).

The Alvord basin contains at least two series of well-developed shorelines exposed in a band rising ∼130 m above the modern playa and valley floor. The shorelines formed during ancient lake-level stillstands and exhibit both erosional and depositional geomorphic features (Lindberg, 1999; Carter et al., 2006). Of these features, wave-cut terraces are the most prevalent and are exposed discontinuously around the basin (Reheis, 1999; Lindberg, 1999; Singleton, 2004; Carter et al., 2006; Personius et al., 2006).

The morphology of the ancient shorelines led previous workers to conclude that the pluvial lake records at least two cycles of fill and desiccation (Reheis, 1999). The shorelines are exposed around the basin and assigned to late Pleistocene and an older lake cycles based on soil profiles, weathering analysis, and cosmogenic age determinations (Carter et al., 2006; Personius et al., 2006, 2007). In several parts of the basin, flights composed of five (Lindberg, 1999; Carter et al., 2006) and as many as eight (Singleton, 2004) individual shorelines are recognized.

Lindberg (1999) studied flights of five shorelines in three locations along the northern and northeastern flanks of the Alvord basin. Similarly, Carter et al. (2006) recognized five shorelines that partially coincide with those of Lindberg (1999), and proposed that shoreline elevations around the northern Alvord basin are relatively constant, with an older cycle of lake shorelines at altitudes of 1310–1305 m and late Pleistocene shorelines at characteristic altitudes of 1292, 1287, and 1280 m ⁰¹.

In Bog Hot Valley, located at the southern extremity of ancient Lake Alvord (Fig. 2), mapping by Personius et al. (2006) showed shorelines with altitudes ranging from ∼1308 to 1279 m ⁰¹. In this part of the Alvord basin, the mapped surfaces constitute two flights of four shorelines each, exposed east and west of the Steens Mountain fault. Along the western flank of the valley in the footwall of the fault, shoreline elevations range from 1308 to 1295 m, whereas in the hanging wall to the east, the shorelines range from 1291 to 1279 m. The formation of the lower shorelines exposed in the footwall just above the Steens fault was dated as 17.83 ± 1.1 ka from luminescence ages on samples from a soil pit and paleoseismology trench (Personius et al., 2007).

The altitude of shorelines constituting the prominent flights around the northern Alvord basin varies spatially, underscoring unresolved issues about the lateral correlation of individual surfaces ⁰¹. Although recognized by Lindberg (1999) and Hemphill-Haley (1987) as being involved in faulting, offsets of the shorelines were thought to be only a few meters (Carter et al., 2006). Similarly, the degree to which shorelines are tilted since formation (Lindberg, 1999) remains equivocal.

The greatest obstacle to assessing the lake-level history in the Alvord basin stems from an ambiguity in the number and lateral continuity of shorelines over distances of several kilometers and around the ancient lake as a whole. Carter et al. (2006) interpreted a flight of five shorelines to be laterally continuous around the northern Alvord basin, based on the analysis of 1:72,000-scale aerial photographs supplemented by local mapping. The correlation of shorelines over distances of tens of kilometers was tied to elevation control, often for single shorelines, determined either as part of their work or taken from published reports (Hemphill-Haley, 1987; Lindberg, 1999). The interpretation carries the implicit assumption that faulting or tilting had little impact on shoreline altitudes in and around the basin. The results of Carter et al. (2006) are at odds with the variation in shoreline elevations reported by Lindberg (1999) and are inconsistent with our results ⁰¹.

To fully appreciate the spatial and temporal relations between shoreline development, the relationship between shoreline altitude and late Cenozoic faulting, and the possibility of local or regional tilt will require an exhaustive study of the structures and terraces along the margins and within the 200-km-long extensional basin system. In our study, we focused on assessing the degree to which terrace elevation is affected by displacement on late Pleistocene–Holocene normal faults. To this end, we studied shorelines and faults within an east-west transect across the central Alvord basin (Fig. 4), where we document the lateral continuity of a flight of eight shorelines over distances of several kilometers and measured tens of meters of vertical displacement of the surfaces on individual faults and across the basin. In this work, we expand upon our earlier results (Singleton, 2004; Oldow and Singleton, 2005; Oldow et al., 2005) and produce a comprehensive reconstruction of shoreline development through the application of high-resolution imaging. As part of this project, we were able to document that eight shorelines are preserved in the Little Sand Gap area (Fig. 4), where five surfaces were originally mapped (Lindberg, 1999). We discovered that shoreline elevations record a pattern indicating complex interplay between fault displacement and water-level variation within the pluvial lake. The spatial and temporal relationship between shorelines and faults is discussed at length in following sections, but the degree of altitude variability for each of the shorelines is apparent in the summary of our results presented in ⁰¹.

Our transect was chosen to build on earlier work and to cross well-developed shorelines, many of which were known to be cut by faults (Hemphill-Haley, 1987; Hemphill-Haley et al., 1999; Lindberg, 1999; Carter et al., 2006). We used ground-based LiDAR (Light Detection and Ranging) also known as Terrestrial Laser Scanning (TLS) to image shoreline terraces at the centimeter scale along the transect from Little Sand Gap on the eastern side of the Alvord Desert to Serrano and Alvord Points along the western margin of the basin (Fig. 4).

Through detailed mapping in the Serrano-Alvord area and Little Sand Gap, using digital orthophoto quadrangle coverage (pixel resolution of 1.0 m) and 10 m digital elevation models (DEMs) supplemented by the high-resolution LiDAR surface images, we recognized two series of wave-cut terraces (Singleton, 2004). The terraces are assigned to an upper series of three shorelines, the Serrano terraces, and a topographically lower succession of five shorelines assigned to the Alvord terraces. At the western end of the transect (Fig. 5), the eight shorelines were mapped around the headlands at Serrano and Alvord Points and across the intervening low-relief topography. These areas provide expressions of the surfaces developed on west-, south-, and east-facing exposures and provide insight into the natural variation of the shorelines as a function of original slope and facing direction. In this area, shorelines are offset by active faults and record differential vertical displacements ranging from about 1 to 29 m. Two complete sequences of Serrano and Alvord shorelines are preserved in the hanging-wall and footwall blocks of the Serrano fault (Fig. 5), and a partial sequence is preserved east of the Alvord fault. A complete sequence of shorelines is also preserved at Little Sand Gap along the Tule Springs Rim (Fig. 6), where they were studied in some detail by Lindberg (1999).

Interpretation of shoreline continuity around the basin is predicated on the expectation that, within the limits imposed by lateral variability in development, stillstand terraces have a regular expression around the lake perimeter. Within a tectonically quiescent environment or one where deformation occurs on a time scale much greater than that of lake-level variation, the shoreline pattern around the basin should be relatively symmetric. In contrast, where fault displacement occurs at a time scale comparable to that of lake-level variation, shoreline spacing will vary across individual faults and from one side of the basin to the other. In a setting where unequal magnitudes of fault slip and water-level variation occur with time, a complex array of permutations is possible for pattern of coeval shorelines around the lake.

In the simplest situations, where shorelines are formed either in a transgressive or regressive system of lake-level variation and where changes in water level are greater than cumulative fault slip over contemporaneous time intervals, older terraces will show larger cumulative displacement than younger surfaces. In transgression, topographically higher terraces will have smaller aggregate displacement than older shorelines at lower elevations. In the more commonly expected setting where shorelines are formed during water-level regression, higher older surfaces will record greater cumulative displacements than lower younger surfaces.

Complexity in the pattern of terrace formation increases if water level alternates between periods of regression and transgression or if water-level change and fault slip conspire to produce reversals in the pattern of successive shorelines formation across a fault. Where reversals in the vertical succession of coeval terraces occur across a fault, a downward monotonic correlation will engender miscorrelation. To explore the implications of this scenario and its impact on our ability to model offset of shorelines across faults, we modeled end-member permutations where reversals in shoreline order across a fault are unrecognized. In all but a few cases, the cumulative displacement measured for successively younger terraces (with miscorrelation) will not decrease systematically and can be recognized by a reversal in apparent displacement (greater cumulative displacement than preceding and succeeding correlated terraces). In a few situations involving specific patterns of relative transgression and regression, a succession of decreasing cumulative displacement of miscorrelated shorelines is possible, but produces an underestimate of total fault displacement. Thus, in most situations miscorrelation of shorelines across faults is recognizable through reversal of the successive pattern of cumulative slip. Even in the worst-case scenario, where events conspire to produce a monotonically decreasing pattern of cumulative slip, miscorrelation will result in an underestimation of total fault displacement.

Fortunately, the complexity and potential for error in correlation only are fully developed if shorelines have no distinguishing characteristics. Where extrinsic recognition criteria exist for individual surfaces or for groups of surfaces, the degrees of freedom are reduced and successful correlation is greatly aided. The series of terraces around the northern Alvord basin consist of individual shorelines that are characterized by differences in geomorphic expression related to development and degradation. When viewed as a group, the shorelines have expressions that exhibit a systematic pattern of prominent and subtle morphology that varies in a predictable order. These attributes enhance the ability to correlate the shorelines across the lake with some confidence.

In the central Alvord basin, the older Serrano terraces are composed of three erosional shorelines found at elevations ranging from ∼1320 m to 1280 m. The terraces are designated in descending order of elevation as S1, S2, and S3, and occur at higher elevations than the younger Alvord terrace series. Around Serrano Point and Alvord Point (Fig. 5), S1 ranges from a high of 1318 m to a minimum elevation of 1298 m and the lower shoreline, S3, ranges from a high of 1304 m to a low of 1280 m ²⁰. At Little Sand Gap on the eastern flank of the basin, the S1, S2, and S3 surfaces are found at 1308, 1292, and 1284 m, respectively ²⁰.

Serrano terraces are discontinuously exposed throughout the study area; some of their best exposures occur at headlands and bayhead beaches. For the most part, these older surfaces can be distinguished from the younger Alvord shorelines by the clast size of sediments deposited during beach formation. Serrano terraces typically expose boulder to pebble clasts in a sand matrix and the surfaces often are rough, exposing boulder to cobble lag deposits. The surfaces show variable weathering, the highest shoreline (S1) containing coarse deposits with stage II pedogenic carbonate development (Carter et al., 2006). Elsewhere, lower units (S3) have clasts with discontinuous varnish and stage I pedogenic carbonate accumulations (Carter et al., 2006).

The uppermost terrace, S1, typically shows greater erosional degradation than topographically lower surfaces, S2 and S3, which are prominent geomorphic features in the northern Alvord basin. All shorelines of the Serrano series show erosional degradation to varying degrees and all members of the Serrano series terraces are channeled (Fig. 7). In some cases channeling is extensive, particularly in regions of high relief, and occasionally individual shorelines are buried by younger alluvial fans and landslides.

The relative timing of the Serrano and Alvord terraces is well established in the Serrano and Alvord Point area. The lower part of Serrano terrace S3 is reworked by the highest Alvord terrace (A1), and channels cut in the older surface are buried by the Alvord shoreline A1 (Fig. 7). Similarly, at locations at the southern end of Alvord Point, the highest Alvord series terrace (A1) overprints a fault scarp that cut terraces S1 through S3 (Fig. 5). The Alvord terraces truncate the southern extent of the fault and clearly were developed during a younger regression cycle when the fault was no longer active.

The Alvord terraces record at least five still-stands of paleo–Lake Alvord and are designated A1, A2, A3, A4, and A5, in descending order of elevation. All five terraces are recognized throughout the study area. The terraces are offset by faults ²⁰ and have altitudes ranging from 1294 to 1256 m at their maximum elevations near Serrano Point to their minimum elevations of 1270 to 1244 m at Alvord Point. At Little Sand Gap, the Alvord terrace elevations range from 1278 to 1259 m.

Alvord terraces are well preserved and form some of the most obvious geomorphic surfaces in the region. The terraces preserve a surface expression indicating little erosional degradation and, for the most part, are composed of poorly consolidated sand and gravel. Terrace tops typically are regular surfaces exposing occasional cobbles and in some areas are composed of gravel deflation surfaces. The Alvord terraces are characterized by little or no varnish or carbonate accumulation and underwent only minor degradation by stream channels, downslope sediment transport, and subaerial weathering processes. At lower elevations, near the present-day playa, some terraces within the Alvord series are partially obscured by sand dunes.

Terraces in the northern Alvord basin are not dated directly, but we are able to establish first-order age estimates on the basis of regional correlations. Dated shorelines are exposed in Bog Hot Valley (Personius et al., 2006, 2007) at the south end of the Alvord basin (Fig. 2) and provide incomplete but critical age constraints. At the regional scale, comparison to dated shorelines in pluvial Lake Lahontan in northwestern Nevada (Lindberg and Hemphill-Haley, 1988; Reheis, 1999; Carter et al., 2006) and Lake Chewaucan in south-central Oregon (Allison, 1982; Freidel, 1993; Negrini, 2002) is possible because the ancient lakes have similar patterns of Pleistocene fill and desiccation.

The shorelines dated in Bog Hot Valley are the topographically lowest terraces exposed in the footwall of the Steens fault, and Personius et al. (2007) established the date of their formation as 17.83 ± 1.1 ka on the basis of luminescence ages from lacustrine sediments exposed in a paleoseismological trench. The dated shorelines have well-developed morphologies and are topographically below a set terraces preserved discontinuously around the western margin of the valley that, according to Personius et al. (2006), may have formed during an older lake cycle. We cannot make a direct correlation between the Bog Hot Valley shorelines and those in the northern part of the basin, but the morphology and relative position of the dated surface suggest that it is part, and possibly the highest, of the Alvord series of shorelines.

Although pluvial Lakes Alvord and Lahontan were separated by a height of land and were not in direct communication during their fill and desiccation cycles, they share a pattern of at least two lake-level cycles (Reheis, 1999; Carter et al., 2006; Morrison, 1964). The regional pattern of lake-level rise and fall suggests formation in response to climatic forcing (Benson, 1978, 1981, 1999), and as such supports some degree of synchroneity. In pluvial Lake Lahontan, terraces are divided into an older suite, Eetza, and a younger succession of late Pleistocene shorelines, Sehoo, that appear to be analogous to the Serrano and Alvord series in Lake Alvord. For most of the Lahontan basin, the younger Sehoo terraces overprint the Eetza shorelines (Adams and Wesnousky, 1998; Adams et al., 1999), but in the southernmost extent of the basin (Walker Lake subbasin) the Eetza and topographically lower Sehoo are physically separated (Reheis, 1999; Reheis et al., 2003). The age of the late Pleistocene Sehoo highstand is well dated as 13.28 ± 0.11 ¹⁴C Ky B.P. to 13.11 ± 0.11 ¹⁴C Ky B.P. (Adams and Wesnousky, 1998), which equates to calendar ages of 15.5–15.3 ka (Fairbanks, 2007). In contrast, the age of the Eetza highstand is poorly established, but appears to coincide with Oxygen Isotope Stage 6, providing a range of between 200 and 130 ka (Reheis et al., 2003).

A first-order correlation between the Alvord and Lahontan terraces is generally accepted (Reheis, 1999; Carter et al., 2006), but must be tempered by differences in timing relations established in pluvial Lake Chewaucan. Lake Chewaucan is located 150 km west of the Alvord basin and records a history of late Pleistocene highstands that is older than that found in Lake Lahontan. Lake Chewaucan achieved a maximum lake highstand between 17.5 ± 0.3 ¹⁴C Ky B.P. to 16.4 ± 0.5 ¹⁴C Ky B.P. (Allison, 1982; Freidel, 1993; Negrini, 2002) and a younger highstand dated as 11.75 ± 0.8 ¹⁴C Ky B.P. (Licciardi, 2001). Conversion to calendar years (Fairbanks, 2007) indicates two highstands for Lake Chewaucan at 20.7–19.5 ka and 13.6 ka.

Taken together, the regional timing constraints for pluvial lake highstands permit a range of upper age limits for the Serrano and Alvord shorelines in the northern Alvord basin. As a first approximation, we adopt the correlation of the Lahontan Eetza and Sehoo shorelines with the Serrano and Alvord terraces, respectively (Reheis, 1999; Carter et al., 2006). However, in light of evidence that similar patterns of water-level cyclicity in the pluvial lakes are not necessarily synchronized, we do not subscribe to a direct age correlation. Based largely on the luminescence ages for shorelines in the Bog Hot Valley (Personius et al., 2006) and the difference in highstand ages of Lakes Chewaucan and Lahontan, we bracket the upper limit of the Alvord terraces as between 20 and 15 ka. Applying age constraints to the Serrano shorelines is even more tenuous, but correlation with the Eetza terraces of Lahontan suggests an age of between 200 and 130 ka for the Serrano shorelines, a range that subsumes any phase lag in highstand age.

An estimate of the minimum age for the Alvord shorelines is provided by the discovery of prehistoric arrowheads around the Alvord basin at elevations of 15–23 m above the present-day playa (Pettigrew, 1984). The arrowhead ages were estimated as between 10 and 12 ka and are thought to record land use patterns of prehistoric humans around the last vestige of Lake Alvord. The archaeological age constraints in the Alvord basin are consistent with the findings of a regional synthesis by Negrini (2002) that points to a pluvial lake-level minimum ca. 11.5 ka for southern Oregon and northwestern Nevada. Succeeding water-level increases within lake basins across the region did not reach late Pleistocene depths, and the lowest prominent Alvord shorelines are probably not younger than 11 ka.

To be useful in estimating the pattern of deformation and deformation rates, a geologic surface must have a known age and orientation prior to distortion by tectonic processes. As a vertical datum, ancient shorelines provide an unparalleled opportunity to measure the vertical component of geologic deformation rates over time scales of 10⁴ to 10⁵ yr. Unlike many geomorphic surfaces, the initial geometry of the lake shorelines is related to water level (paleogeoid), and for large lakes the surfaces are aerially extensive. Where lateral correlation of surfaces is well known, vertical displacements can be determined over wide regions by measurement of the differential elevations of the same surface. This reduces age uncertainty as well as the propagated displacement-magnitude uncertainties associated with orientation and age ambiguity encountered where multiple discontinuous surfaces are used as a reference datum.

The caveat to this scenario, however, stems from several considerations. Uncertainty in the geomorphic stability of a shoreline surface and its relation to the ancient water level must not exceed the resolution required to measure tectonic displacement. Degradation of the surface by erosional and constructional surface processes must be recognizable, and for the surface to be used as a vertical datum, quantifiable. Both of these issues are compounded by the inherent roughness of most geomorphic surfaces and the uncertainty this imparts to quantitative measurement of the landform.

Historically, shoreline features in the Alvord basin were mapped using standard U.S. Geological Survey topographic coverage supplemented in some cases with hand-held GPS receivers and the use of engineering levels and/or total station control (Lindberg, 1999; Carter et al., 2006). Combined hand-held GPS and topographic coverage provides a horizontal positioning resolution of no better than ±3 m and typically greater than ±5 m. Altitudes taken from 1:24,000-scale topographic sheets are limited to half of the contour interval and range from ±2.0 to ±3.5 m. Use of total station or engineering level surveys referenced to geodetic benchmarks can greatly improve positioning uncertainty, but does not address sampling aliasing when imaging a rough topographic surface.

A more vexing problem for accurate measurement of the geomorphic surfaces is the difficulty in producing the high-resolution topographic data needed for quantitative analysis (Burke and Dixon, 1988; Dietrich et al., 1993). Spot elevations used to determine the altitude of inherently rough topographic surfaces, no matter how determined, are susceptible to unquantifiable errors arising from the selection of measurement site. Taken together with quantifiable positioning errors, these intangible uncertainties conspire against accurate assessment of the differences in altitude of correlated surfaces over distances of hundreds of meters and certainly over distances of kilometers.

Ancient lake shorelines, although dramatic subhorizontal features when viewed from a distance (Fig. 8), typically are subtle and topographically noisy when viewed at close hand. For most applications, using existing topographic coverage is inadequate to produce accurate assessments of the morphology and elevation of specific features. Hare et al. (2001) used static and real time kinematic (RTK) GPS measurements, locally combined with total station surveying, to develop the protocols needed to adequately characterize ancient shorelines in Lake Lahontan of west-central Nevada. We expanded upon these techniques by using differential GPS positioning and TLS to provide the resolution needed to compare differences in the altitudes of correlated surfaces.

Although a major advance in the quest to quantitatively image geomorphic surfaces, the capacity to produce high-resolution measurements carries uncertainties related to reference datum, instrument precision, and computational analysis; fortunately these sources of error (discussed below) are relatively small. When dealing with pluvial lake shorelines formed during water-level fluctuations, the physical response of the land surface to lake-level fall will contribute to changes in elevation over time. Where displacements are measured across individual faults, this has no impact, but for comparisons of differential elevations over several kilometers, isostatic rebound must be considered.

In this study we focused on wave-cut features as a means of determining the paleogeoid, which serves as a vertical datum in this deformation study. Wave-cut terraces are best preserved in regressive shore settings and typically are destroyed or strongly modified during lake transgression (Morrison, 1964). If terrace construction is followed by a cycle of lake desiccation, the terrace is left in a subaerial environment and no longer subject to erosion by wave action. This cycle may repeat several times, leaving behind a record of multiple lake stillstands in the form of terraces (Morrison, 1964, 1991).

The wave-cut terrace at the highest elevation in a group is representative of a paleolake highstand, whereas lower elevation terraces typically record subsequent stillstands during the regression cycle. If lake levels subsequently increase, the record of previous stillstands may be modified or obliterated by the redistribution of nearshore sediments by wave action (Atwood, 1994). If lake level increases, younger terraces can develop at higher elevations than terraces formed earlier in the lake history. In the Alvord basin, terraces do not show evidence that younger highstands overprinted the higher elevations of the Alvord series, and it is clear that the Alvord series did not rise above the highest terraces of the Serrano series.

The relation between the water surface (paleogeoid) and the formation of the wave-cut terrace is of critical importance to this study. Models of wave-cut platform formation developed by Trenhaile (1987, 2002) and Sunamura (1975, 1976, 1977, 1978a, 1978b, 1982, 1992) are based on theory, scale models, and actual measurements. Terraceforming erosion is primarily driven by breaking wave action and is most effective near the still-water line. Thus, erosion occurs above a certain level, leaving a platform or terrace beneath. The resultant abrupt change in slope is known as the shoreline knickpoint or shoreline angle (Fig. 9), and the depth and slope of a terrace below the still-water level are proportional to the wave energy, with most shoreline erosion occurring during periods of high wave energy (i.e., storms). The development of wave-cut features is dependent upon the shoreface slope, orientation of the shoreface with respect to the dominant wave-propagation direction, degree of protection of the shoreline, and substratum composition (Sunamura, 1978a). As these parameters vary, the morphology of an individual terrace can differ spatially in breadth and slope.

A terrace is composed of components that contribute to the landform as a whole. In general, a wave-cut terrace is a bench with a large relatively flat upper surface that breaks into relatively steep slope faces. The slope face is referred to as the riser and the flat upper bench surface is the bench top. The riser crest is the point of maximum upward convexity, and the knickpoint is the point of maximum upward concavity (Fig. 9).

The utility of the shoreline knickpoint as an indicator of paleo–water level is well established (Keller and Pinter, 2002) and serves as a good paleohorizontal marker (Locke and Meyer, 1994). The study by Meyer and Locke (1994) at Yellowstone Lake showed that knickpoints measured for more than 100 modern wave-cut shorelines terraces had a mean height of ∼1.9 ± 0.3 m above the average lake level. Although the knickpoint serves as the best analog for water depth, other elements of the shoreline terrace serve as a good paleovertical datum. The analysis of Lahontan terraces by Hare et al. (2001) using GPS and total station measurements indicates that noise levels for the riser crest, knickpoint, and riser inflection are statistically indistinguishable and have a standard deviation of ∼0.5 m. In our study using TLS and GPS, an increase of sampling density by 5 orders of magnitude reduced the noise levels a factor of 10 to ∼5 cm (Singleton, 2004).

Possible variability in the characteristic heights of specific geomorphic features developed along shorelines is an important question when using the surface as a vertical datum. The degree to which shoreline features vary spatially is a source of debate (Singleton, 2004; Hare et al., 2001; Adams et al., 1999; Adams and Wesnousky, 1998; Atwood, 1994). Adams and Wesnousky (1998) and Adams et al. (1999) proposed that for a given water level of Lake Lahontan, the height of constructional shoreline features varied by as much as 2.6 m over distances of 4–5 km. Similarly, Atwood (1994) reported variability from 0.2 m to 1.9 m of shoreline highstands above the still-water level for the 1980 lake level of the Great Salt Lake of Utah. In contrast, our application of the quantitative techniques outlined by Hare et al. (2001) coupled with the ability to sample geomorphic surfaces at a density of several thousand measurements per meter squared allowed us to document that specific geomorphic features on wave-cut terraces display characteristic altitude continuity with a standard deviation of 0.19 m and standard error of 0.04 m over distances of 2.4 km.

For the most part, we suspect that the elevation noise reported by Atwood (1994), Adams and Wesnousky (1998), and Adams et al. (1999) stems from the types of surfaces measured and methods used in measurement. Where characteristic elevations were taken from crests of constructional barriers (Adams and Wesnousky, 1998) and water-level indicators such as debris lines, erosional steps, gravel ridges and beaches, and vegetation lines (Atwood, 1994), significant variability is reported. In contrast, where wave-cut terrace surfaces are measured, vertical differences of only tens of centimeters or centimeters are recognized over kilometer distances (Hare et al., 2001; Singleton, 2004). In addition to the noisy character of constructional features, some of the elevation variation may stem from lowresolution methods of measurement. In the Atwood (1994) study, positional uncertainty, or how different types of features varied in elevation, was not addressed. Although insightful for development of shoreline features, the elevation differences cannot be rigorously analyzed. Adams and Wesnousky (1998) used a total station, locally controlled by GPS, to determine elevations for specific shoreline features. A comprehensive analysis of repeatability and error was not reported, but the published uncertainty values appear to be optimistic given the equipment and techniques used.

The study by Hare et al. (2001) and our work in the Alvord basin focused on imaging all characteristic features of wave-cut terraces and on the production of high-resolution DEMs, which allow quantitative assessment of vertical uncertainty and document centimeter-scale stability in wave-cut surfaces. Our experience also shows that dense elevation data aid in the isolation and exclusion of areas of high surface noise, such as stream channels and low-relief downslope transport processes. As an example, high- resolution DEMs of several sites in the Alvord basin allowed detailed surface analysis and revealed low-amplitude (1.5 m), long-wavelength (50 m) erosional channels that cut down through several shorelines. The channels were not obvious in the field, but could be recognized readily during quantitative surface analysis.

Because lake terraces are related to water surfaces, terrace elevation is related to an equipotential surface at the time of lake occupation. From precise imaging of the terraces, the elevation of the equipotential surface can be determined, and deviations of the terraces from the equipotential surface can be attributed to tectonic and isostatic processes. The equipotential surfaces will be nearly parallel to the geoid existing at the time of terrace formation. Differences between the geoid during terrace formation and the present geoid may result from changes in topography and subsurface density distributions. Given the relatively short time duration since shoreline formation and the much greater characteristic time scale for tectonic mass-redistribution processes, the effects of these changes are expected to be quite small and of no significance to this study.

GPS provides a three-dimensional (3-D) position relative to an ellipsoid surface referenced to the center of the Earth and is unaffected by local variations in density and associated measured gravity. In order to compare GPS positions to conventional survey results (orthometric height), the ellipsoidalgeoidal separation must be determined (Balde, 1995). There are several ways to do this, but the most straightforward is to acquire an accurate orthometric height by GPS at a previously geoidally determined benchmark.

Uncertainty derived from the variation between ellipsoidal and orthometric height is small over short distances, and estimation of the geoidal undulation modeled by GEOID03 (National Geodetic Survey) yields a horizontal deviation of ±2 ppm of the baseline distance. For baselines up to 20 km this equates to a vertical uncertainty of ∼0.5 mm/km. As such, the geoidal uncertainty adds essentially no statistically significant error to elevation determinations across most of the study area. For differential elevations measured across a fault, the geoidal uncertainty contributes nothing to the measurement error budget. Over the entire distance of the study area, however, a small error is introduced and is propagated in calculations. In our work, we set a lower limit of 2 km as the cutoff for inclusion of the geoidal uncertainty in our propagated error budget.

The final concern is the contribution of isostatic rebound due to water removal and the impact on the differential elevation of ancient shorelines. We estimated the potential isostatic displacement using a simple viscoelastic model (Turcotte and Schubert, 2002). With the removal of ∼90 m of water since 11 ka in a basin 15 km wide, the vertical deflection of the center of the basin would be ∼0.8 m; the margins of the basin would be ∼0.6 m. Fortunately, the contribution to elevation uncertainty is not the total isostatic uplift of the shoreline, but rather the differential uplift of sites as the distance from the axis of greatest water depth changes. Although we used a symmetric model in our calculations (D. Harry, 2006, written commun.) that does not replicate the actual lake geometry, it represents a fair approximation. The ancient lakebed west of Alvord and Serrano Points is relatively high topographically and represents a bay that did not accommodate significant water depth. At the latitude of our study, the northeast- southwest axis of deep water coincides with the basin separating Alvord Point and Little Sand Gap and did not contribute significant vertical deviation between the western and eastern ends of the transect. As such, rebound for the Alvord basin is not a major contributor to our calculated displacement budget.

As part of our work in the Alvord basin, we developed and tested a system for the acquisition, reduction, and analysis of the morphology and elevation of wave-cut terraces (Singleton, 2004). In the following sections, we summarize methods and provide examples of the error analysis applied to the observations.

Laboratory reconnaissance of the field area used aerial photography, DEM coverage, and digital orthophoto quadrangle (DOQ) and digital raster graphic (DRG) data to carry out first-order regional correlation of lake shorelines and to identify specific target areas for detailed study. DOQ images of 1 m resolution, DEM data with a horizontal resolution of 10 m, and DRG data of 1:24,000-scale U.S. Geological Survey topographic quads were used for mapping, basic orientation, and navigation in the field. All digital elements were input into a geographic information system (GIS) framework using ESRI (Environmental Systems Research Institute) ArcGIS 9.0 and integrated to produce 2-D and 3-D images.

Geomorphic surfaces are imaged with TLS that produces dense point clouds of XYZ data. Individual sites have dimensions of as much as 500 × 700 m and are scanned from at least four look directions to produce a 3-D image. Data are acquired at a rate of 1000 points/s and individual images commonly contain as many as 3.0 × 10⁷ points. We use a Riegl LPM-800HA reflectorless laser (Fig. 10) that provides point positions located with an accuracy of 1.5 cm ± 1 ppm of baseline length. Laser beam dispersion increases with baseline length at a rate of 13.0 cm per 100 m, resulting in greater surface averaging with distance. Point position and point-cloud precision is better than 5 cm and surface averaging does not exceed 90 cm² and is typically >30 cm² for most images.

All phases of data collection were spatially registered via two Leica SR530 dual frequency GPS receivers coupled with Leica AT502 L1/L2 antennas. A base station, located over a monument drilled and set into bedrock, and the roving receiver were operated in rapid static or real-time kinematic mode. Observations were recorded at 5 s intervals and the base station was operated continuously for up to 15 h each day. We used the same base station for sites within 20 km or less and combined the base station reoccupations to produce a robust solution used to adjust the network. All laser scanner positions are located with differential GPS together with as many as three to four reflector reference sites within the scan area.

Scans of geologic structures of interest, including paleolake terraces and faults, were taken from several look directions to maximize surface coverage and improve surface model resolutions. The scans ranged in area from ∼400 m × 400 m to as much as 500 m × 700 m. The total number of scans taken at any one particular geographic area ranged from three to seven, depending upon local topographic conditions.

Data reduction requires access to high-end personal computers (PC) and proprietary commercial software. The work flow employed during this study is summarized here.

The first step is to produce GPS positioning control for all points. Differential dual-frequency GPS provides rover positions with precisions of 2–3 cm (Featherstone et al., 1998), and base station locations are referenced to the National Geodetic Survey Continuous Observing Reference System (CORS) using the On-line Positioning User Service (OPUS). GPS positions provide the absolute frame with which to compare terrace features at different locations.

Base station locations for 16 GPS days were returned from OPUS, and all positions had an average root mean square error (1σ) of >0.014 m. Two base station positions were found to be centimeter-scale outliers to the group solution and were rejected using the Chauvenet criterion (Taylor, 1997). The best-fit solution for the remaining 13 positions yielded an uncertainty of ±0.0113 m provided the base station value was used in the network adjustment for all relative positions.

Alignment of multiple laser scans (different look directions) for a single site into a single 3-D point cloud utilized Innovmetrics, Poly-works 9.0.2. All look directions are transformed to a common reference frame and combined as a single point cloud, often containing from 10⁷ to 10⁸ irregularly spaced points. Alignment requires individual images to meet two criteria: (1) each image must share some redundant information, or overlap, with adjacent images, and (2) each image should contain some shape variation of the subject. A shape-based alignment uses collocated, distinctive three-dimensional shapes that contribute to lock individual look direction images into a common reference using a least squares iterative algorithm that minimizes the distance between surface overlaps in a set of 3-D images. This technique distributes the alignment error over all the image-transformation matrices and ensures a well-balanced alignment in a set of 3-D images within a local frame of reference.

The local reference frame for the 3-D images was transformed to Universal Transverse Mercator coordinates via a seven-parameter Helmert transformation by using GPS measured positions located within the image area. Transformation residuals are a measure of uncertainty and for our work typically are on the order of 0.055 m with the worst case reaching 0.49 m. Once transformed to a geospatially referenced frame, the data were cropped to reduce file size, which typically is several hundreds of megabytes, and output as ASCII (American Standard Code for Information Interchange) point clouds.

The georectified point clouds were gridded from analysis in ArcGIS using minimum curvature (Golden Software Corporation, Surfer 8.0). Grid spacing was set to 20–30 cm, depending upon the corresponding length scales used during Polyworks alignment. We used a relaxation factor of 1 for internal and boundary tension during gridding. Faceting typical of modeled surfaces was undetected, probably due to the high density of data points within the scans. The gridded surfaces were transformed for compatibility with ESRI software using Geospatial Designs Corporation Grid Convert.

Our high-resolution DEMs (0.3 m) were combined with 10 m DEM coverage supplied by the U.S. Geological Survey to complete coverage of the wave-cut terraces and fault scarps in the study area. Although vital for establishing continuity of terraces, the lower resolution topographic data have limited utility in assessing characteristic elevations of terraces and vertical displacements across faults. DEMs are based on gridded data used to establish a mean elevation for the centroid of a pixel, the dimension of which is established by the grid spacing. An elevation taken from a 10 m grid node may vary in actual units on the ground by as much as ±3.5 m. In practice, we found that elevation estimated for the same point from both 10.0 m and our 0.30 m data sets in our study area had a variance of ±1 m, consistent with elevation uncertainty reported for the 10 m DEM by the U.S. Geological Survey.

The riser crest of terraces, which represents one of the most stable components of the lake-terrace geomorphic surface (Hare et al., 2001), was used to map vertical offset of geomorphic surfaces across extensional faults. In our study we were not particularly interested in the best representation of water depth, but rather selected the riser crest as a geomorphically stable and easily recognizable feature with which to establish a vertical datum. To locate and map terrace riser crests accurately in the Alvord study area, slope analysis of 0.30 m and 10.0 m DEM data was used to produce slope-gradient images.

A variety of spatially rectified surface coverage information is employed in surface analysis. Hillshade analyses of the DEM data at 10.0 m and 0.3 m resolutions was combined with DOQ coverage of 1 m resolution to establish the relative locations of surface features. The primary tool for terrace riser crest mapping, however, was based on slope analysis. Initial slope analysis of high-resolution DEMs was performed and exported with a grid value of 0.30 m. This yielded a slope image contaminated by large boulders and vegetation that interfered with the recognition of large-amplitude geomorphic features of interest (Fig. 11A). To suppress the short wavelength noise and enhance longer wavelength features, slope images at pixel dimensions of 1.0 m to 5.0 m were developed (Figs. 11B and 11C). Through inspection and comparison to 1.0 m DOQ images and 0.30 m hillshades, we discovered that slope-image outputs of 1.5 m to 2.0 m cell size best characterized the riser crests (Fig. 11C).

During slope analysis, the tops of high-slope gradients corresponding to riser crests were mapped using line coverage within the ArcMap application of ArcGIS. The placement of the interpreted line was checked visually against the DOQ and hillshades, using 2-D and 3-D visualization in ESRI ArcMap and ArcScene applications, respectively (Fig. 11D). From these interpretations, the elevations of the riser crests were calculated. Calculations were based on elevation samples taken from the 0.3 m DEM data along lines of terrace riser-crest location. Sample elevations were carefully selected based on interpretation of slope, hillshade, and DOQ data sets in both two and three dimensions. During sample elevation collection, care was taken to select elevations from areas that did not appear to be modeling local vegetation or erosional features, such as streambeds and washes, that may intersect terrace riser crests. This was accomplished by cross-checking locations of sampling with slope gradient, hillshade, and DOQ data sets in both two and three dimensions. Elevation samples were selected and elevation data were entered into an attribute table for each individual terrace riser crest.

Based on our mapping, each geomorphic surface was assigned a series of elevations. For 0.30 m surface grids, characteristic elevation values are based on statistical analysis of individual elevation samples by the statistics feature in the Spatial Analyst tool of ESRI ArcMap application. The calculated standard error in elevation (Taylor, 1997) stems from the point-cloud alignment process and uncertainty in GPS positioning. Error values were calculated for the characteristic elevation of individual geomorphic features and were propagated using methods outlines by Taylor (1997) for each geomorphic surface.

In some locations, correlation between geomorphic features imaged using high-resolution scansrequiredassessmentofcharacteristicelevations through areas covered by lower resolution 10.0 m DEMs. All 10.0 m data were acquired from the 1:24,000 DEM for the Andrews, Oregon, U.S. Geological Survey Quadrangle. The U.S. Geological Survey protocol for assessing error values in each DEM is based on assessment of 28 points, 8 at the edges, and 20 throughout the coverage, that are compared to control values of higher spatial accuracy. The Andrews, Oregon, DEM has a reported elevation uncertainty of ±1.0 m that was used in combination with measurement errors calculated in a manner similar to that for 0.30 m data. The propagated error resulted in a standard error of 1.0 m for the 10 m coverage ²⁰.

Statistical analysis of the elevation sampling for each terrace was calculated, and for our preliminary work yielded standard deviations of 0.060–0.559 m. Standard error estimates for characteristic elevations ranged from 0.013 m to 0.084 m with a mean standard error for all elevation measurements of 0.041 m (Singleton, 2004). All sources of uncertainty were propagated (Taylor, 1997) and included in an error budget for the characteristic elevations of geomorphic surfaces. Contributions from GPS positioning, geoidal uncertainty, TLS uncertainty, coordinate transformation, and surface sampling were combined. Where elevations are determined using 0.3 m data, standard errors typically are between 0.06 and 0.08 m with a total range of 0.06–0.4 m.

The utility of mapping terraces in three dimensions using the surface analysis methods outlined above is illustrated by comparing quantitatively determined riser-crest locations to an approach using topographic profiles to locate the same geomorphic surfaces. For the comparison, we extracted terrace profiles at Alvord Point and at Little Sand Gap from our 0.3 m resolution topographic data. The riser-crest locations shown on the profiles (Fig. 12) are, in most cases, recognizable for prominent terraces but difficult to resolve for more subtle surfaces. As established by Hare et al. (2001), who used lower resolution topographic data than those produced in this study, the lateral variation in the morphology of the shorelines typically overwhelms the sought after signal. Use of topographic profiles is of limited use unless measured at close intervals and stacked to suppress the natural variability of the geomorphic surfaces (Hare et al., 2001).

The vertical displacements of Serrano and Alvord terraces across exposed and inferred faults in the central Alvord basin provide the means to unravel the spatial and temporal pattern of fault displacement. The vertical component of the cumulative deformation is determined by summation of the vertical offset of shorelines accommodated by synthetic and antithetic faults across the basin. When combined with estimated ages of shorelines and fault activity, deformation rates and cyclicity in fault activity are estimated for the entire fault system.

Measurement of fault displacement is based on differential elevations of riser crests ²⁰ between correlated shorelines offset by observed and inferred faults along our transect. Rise-crest elevations are based on 0.3 m data in the best cases, but 10 m data were used locally to provide full spatial coverage of all of the geomorphic surfaces along the transect. The offset of individual shorelines across observed faults provides the measure of cumulative displacement on the structure since formation of the geomorphic surfaces ⁰³, and provides the basis for calculating average displacement rates on individual structures. A more useful measure is the history of fault displacement determined for intervals between successive shorelines, referred to here as the interval displacement for a given structure ⁰⁴. The interval displacement is determined by two methods that provide a cross check of internal consistency in the measurement. The difference between the cumulative displacements recorded by two shorelines cut by a fault provides one measure of interval displacement. In the other, the difference in spacing between adjacent shorelines measured on opposing sides of a fault provides an independent estimate of interval displacement. In instances where fault displacement occurred between periods of water-level drop, the riser-crest spacing in the hanging wall decreases by the amount of cumulative displacement accumulated during the time between successive shoreline development. In our estimates of interval displacements for faults, both measures are used and any discrepancies reconciled by averaging. We note that where differences occurred in interval displacement estimates using the different methods, the discrepancies were all within measurement uncertainty. The determined interval displacements ⁰⁴ are an aggregate estimate of displacement over a period of time (between shorelines) and may correspond to one or more earthquake surface ruptures.

The spatial and temporal patterns of displacement on faults within the central Alvord Desert are determined from terrace offset across exposed faults at the western end of the transect and by reconstruction of successive shorelines to horizontal across the basin. The pattern of shoreline displacement recorded across observed faults at Serrano and Alvord Points (Fig. 5) documents that lack of reversals in the reduction of cumulative displacement with decreasing age and, more importantly, consistently shows that shoreline spacing decreases in the hanging wall of faults active between formation of successive shorelines. These relations, taken together with the necessity of restoring offset terraces to the paleohorizontal for each shoreline interval, allow reconstruction of the geomorphic surfaces across the basin. In the reconstruction, we successively restore the displacement of lower, younger surfaces to a horizontal datum by reversing the normal displacement on observed or inferred faults. In the restored frame, the pattern or fault slip and displacement of older shorelines is more readily identified.

To capture the total fault displacement budget across the basin, the contribution of all synthetic and antithetic faults must be determined. Although this geometric necessity is unattainable within an extensional system where some faults are concealed beneath basin fill, the flights of areally extensive shorelines around the basin provide the means of establishing a good minimum estimate of cumulative displacement on concealed structures. The component of displacement accommodated symmetrically by antithetic and synthetic structures buried beneath the basin fill cannot be recovered for any one shoreline. However, in the likelihood that displacement is not symmetrically disposed for all shorelines composing the flights of eight surfaces, the asymmetric component of the displacement accommodated by concealed synthetic and antithetic structures can be determined by differences in shoreline elevations on opposing sides of the basin.

Analysis of shoreline altitudes across known and inferred faults illustrates a pattern of displacement that varied spatially and temporally ⁰⁴. Where shoreline flights are cut by faults, differences in the cumulative displacement recorded by progressively lower (younger) surfaces indicate a complex interplay between water-level drop and fault slip. Several faults have displacement histories recording multiple periods of displacement as the lake level regressed. Comparison of fault displacement histories illustrates a pattern of distributed displacement, with some faults exhibiting greater sustained displacement than others.

Faults exposed in the low hills at the western end of the transect (Fig. 5) document spatial and temporal variations in fault slip and lake level. The Serrano fault underwent displacement during most periods of shoreline development, and contrasts with the Alvord and West Alvord faults, the activities of which were more restricted.

The Serrano fault (Fig. 5) records nearly continuous displacement over time scales very similar to those of water-level variation. Six of seven intervals among Serrano and Alvord shorelines ⁰⁴ record differential displacement, with cumulative displacement on the fault systematically decreasing with younger shorelines (Fig. 13). The highest terrace, S1, records a cumulative displacement of ∼29 m, whereas the highest Alvord terrace (A1) is offset by ∼21 m (⁰³ and ⁰⁴). Cumulative displacement systematically decreases downward to ∼8 m for the youngest terrace (A5). Taken together with differences in cumulative offset, the decreased spacing between riser crests in the hanging wall indicates that the fault underwent six periods of displacement during progressive formation of the terraces. Total offsets of ∼3 to ∼4 m are recorded between S1 and S2, A1 and A2, A2 and A3, and A4 and A5, and small decimeter displacements occurred between S2 and S3 and S3 and A1.

Farther to the east, the two splays of the Alvord fault system record displacement between two of the five Alvord shorelines that are cut by the structure ⁰⁴. The periodicity of fault slip is more sporadic than that of the Serrano fault, but progressive slip occurred as water level varied. The main strand of the fault system has a scarp height of ∼10 m and the scarp of the eastern strand is 1 m high. The splays merge ∼0.5 km south of Alvord Point, where the preserved scarp is ∼2 m high (Hemphill-Haley, 1987; Hemphill-Haley et al., 1999). At Alvord Point, the eastern splay (East Alvord fault) cuts across the height of land just above the present-day playa and only intersects the lowest Alvord terrace (A5). The main splay of the Alvord fault crosses higher on the topography and offsets all of the Alvord terraces (A1–A5). Serrano terraces (S1–S3) are exposed in the footwall of the Alvord fault, but the hanging-wall elevation is too low and Serrano terraces were not formed (Fig. 5). A total displacement of ∼5 m is observed on terrace A1 and the lowest terrace (A5) records ∼4.5 m of vertical offset. Riser-crest spacing between A1 and A2 decreases across the fault but is constant for all lower terraces, indicating that only one displacement event was captured during formation of the Alvord shorelines. Approximately 5 m of the scarp along the Alvord fault predates formation of the Alvord shorelines and is attributed to displacement during the S3–A1 interval ⁰⁴.

The West Alvord fault system is between the Serrano and Alvord faults and is a down-to-the-west structure that follows the northwest-ward trend of the western flank of Alvord Point (Fig. 5). The fault is a single structure in the north, but bifurcates into two splays to the south. Displacement on the fault system of ∼10 m is recorded on shorelines of the Serrano series ⁰⁴. In the south, the eastern splay ceased activity before formation of the late Pleistocene highstand (A1) of Lake Alvord, and the western splay continued activity, with a ∼2 m offset of shoreline A1, but ceased before formation of shoreline A2. The two splays of the West Alvord fault are overprinted by the higher Alvord terraces (Fig. 5), and have not been active since the late Pleistocene.

A comparable history of displacement during lake-level drop can be inferred across the Alvord playa by comparing shorelines on opposing sides of the basin. All members of the Serrano and Alvord terraces are exposed 13 km east of Alvord Point at Little Sand Gap (Fig. 6) along the Tule Springs Rim. The altitudes of the shorelines differ across the basin, as do the riser-crest spacings between each of the surfaces ²⁰. These relations are summarized graphically in Figure 13.

When the riser-crest spacing for shorelines at Little Sand Gap are compared to coeval surfaces at Alvord Point, they are clearly not the same, or systematically greater or less (²⁰; Fig. 13). The shorelines showing decreased riser-crest spacing alternate in descending order from one side of the basin to the other. At Little Sand Gap, the spacing of S1–S2 is substantially greater than that at Alvord Point, as is the spacing of the underlying S2–S3 interval. In contrast, the spacing for S3–A1 and A1–A2 is greater at Alvord Point than at Little Sand Gap. The spacings of intervals A2–A3 and A3–A4 are, within uncertainty, the same for both areas, and the spacing of A4–A5 is greater at Little Sand Gap. Taken together with the observation that shoreline spacing decreases in hanging-wall exposures across the Alvord and Serrano faults, the reversal in interval spacing across the basin is most reasonably explained as successive displacement on intervening west-facing and east-facing normal faults within the basin.

Along our transect, the east-facing Alvord fault and west-facing Tule Springs Rim fault (Fig. 3) mark the opposing sides of the basin and are separated by the featureless expanse of the Alvord playa (Fig. 4). The playa is underlain by a basin that, on the basis of gravity modeling (Whipple and Oldow, 2004), achieves depths to bedrock of >2.5 km. The subsurface basin floor is not smooth and is cut by several east-facing and west-facing normal faults that offset the bedrock–basin-fill interface by hundreds of meters. The structures within the basin have no surface expression, but high-resolution seismic reflection profiles across the basin several kilometers to the south (Bradford et al., 2006) show offset of stratigraphic markers to within 1 m or less of the surface. The implication is that even though the structures beneath the playa have no surface expression, the faults are active. Their lack of surface morphology is explained by periods of erosion when the playa is filled with water. According to local ranchers, water depths in the playa intermittently reach several tens of centimeters during the winter and spring.

The total cumulative displacement across the basin accommodated on individual concealed faults or by structures buried beneath the center of the basin cannot be determined from terrace spacing and altitude observations. What can be deduced, however, is the relative contribution of the asymmetric component of the total displacement budget accommodated by concealed or inferred east- and west-facing faults along the transect. Using the reversals in riser-crest spacing as a means of identifying the downthrown block (Fig. 13), we reconstruct the displacement history across the basin by restoring successively older terrace offsets to the horizontal. To maintain consistency in cumulative displacements and riser-crest spacing in hanging-wall blocks, displacements are restored on known or inferred east- and west-facing normal faults. For the extensional basin, restoration of each shoreline is referenced to the highest elevation attained by that surface after the previous restoration step. To preserve the paleohorizontal for each of the faults with east-side-down and west-side-downdisplacementarerequiredwithin the basin and along the western and eastern ends of the transect, respectively. Within the basin, the west-side-down fault displacement is attributed to the Tule Springs Rim fault system (Fig. 13), located along the eastern flank of the basin by steep gradients in gravity. The east-side-down displacement is attributed to a hypothetical fault (XEF) buried beneath the basin floor (Fig. 13), but as developed in the following paragraphs, displacement probably was accommodated on several structures. Shoreline restoration requires displacement on an east-facing fault west of our transect, in the Wild Horse Canyon segment of the Steens fault system (WHCF), and on west-facing normal faults (ETSRF) within the Tule Springs Rim physiographic high (Fig. 4).

In the first step of the reconstruction (Fig. 14), the youngest Alvord terrace (A5) is restored to the horizontal. Between Little Sand Gap and the east-dipping normal faults at Alvord Point, 14.7 m of west-side-down displacement is accommodated. In our simplified reconstruction, the total displacement is attributed to the west-facing Tule Springs Rim fault concealed beneath the eastern edge of the basin, but it is probable that some component of slip was taken up by faults with the same facing within the interior of the basin. Restoration of the A5 shoreline also require restorations of east-side-down displacements of 1.0 m on the East Alvord fault, 4.5 m on the Alvord fault, 7.7 m on the Serrano fault, and 3.2 m on the Wild Horse Canyon fault. With A5 restored to a horizontal datum, relative displacements of older terraces become apparent.

In the A5 reference frame (Fig. 14), no displacement is recorded between Alvord Point and Little Sand Gap on west-facing faults during intervals A2–A4 (which show no difference in riser-crest spacing) but down-to-the-west displacement of ∼19 m is required on faults east of the basin margin (ETSRF) and nearly 2 m on the western strand of the West Alvord fault during A1–A2. Down-to-the-east motion on faults within or bordering the basin is recorded in the riser-crest spacing for interval A4–A5 and during intervals A1–A2 and S3–A1. This displacement cannot be accommodated by the main strand of the east-facing Alvord fault system, but rather must be taken up either by motion on the East Alvord fault (Fig. 5) or a fault or faults concealed beneath the basin. The Alvord fault system bifurcates into the Alvord and East Alvord fault splays immediately south of Alvord Point, with the East Alvord fault forming the western margin of the Alvord Desert playa. The East Alvord fault is traced north for nearly 2.0 km before it is lost beneath the playa. The main splay of the system (Alvord fault) was involved in a small east-side-down displacement between A1 and A2, but was quiescent until after formation of A5 and cannot accommodate the differential displacement observed across the basin during the A4–A5 and A1–A2 intervals (Fig. 14). The history of motion on the East Alvord fault prior to A5 cannot be determined because the elevation of hanging wall is too low to develop older shorelines. In our reconstruction, we simplify the model by attributing all of the inferred slip to a concealed fault beneath the basin (XEF), being mindful that some or all of the slip may have occurred on the East Alvord fault. Displacements on the east-facing structure consist of cumulative displacements of 2.0 m on A4 and 4.4 m on A1. The difference in riser-crest spacing for A4–A5 across the fault is ∼2 m, and between A1 and A2 is nearly 4.5 m. No spacing differentials are noted between younger terraces A2–A3 and A3–A4, indicating that there was no additional displacement before formation of the Alvord terrace (A4). Farther to the west, the Serrano fault underwent successive episodes of east-side-down displacement during intervals A4–A5, A2–A3, and A1–A2 for a total of 13.6 m.

Restoration of the terraces to the A1 reference frame (Fig. 15) shows continued east-side-down displacement on the inferred fault system (XEF) of ∼4.2 m. Along the western margin of the basin, the main splay of the Alvord fault system underwent ∼5.0 m of displacement, and the two splays of the West Alvord fault produced west-side-down displacement of ∼5.5 m and 3.5 m, respectively, for a total of ∼9 m. In this frame, the Serrano fault records cumulative east-side-down displacement of >7 m.

In the S3 reference frame (Fig. 16), west-side-down displacement of ∼14 m is attributed to the Tule Springs Rim fault and is required to accommodate the 10.6 m and 3.4 m offset of S1 and S2 shorelines, respectively, exposed on opposite sides of the basin. The structures on the western flank accumulated ∼15 m of east-side-down displacement, the Serrano fault contributing nearly 7 m of the total and the remainder being accommodated by the Wild Horse Canyon fault farther west. The eastern strand of the West Alvord fault showed minor west-side-down displacement of just >1 m.

The record of fault displacement preserved in terrace offsets across the central Alvord basin illustrates the complex spatial and temporal pattern of fault interaction that evolved during late Pleistocene and Holocene extension. Since the late-middle Pleistocene, the Serrano fault (Fig. 5; ⁰⁴) underwent nearly continuous displacement until activity ceased between ca. 5 and 2 ka (Hemphill-Haley, 1987; Hemphill-Haley et al., 1999; Personius et al., 2006, 2007). During this time interval, other structures across the basin were only intermittently active.

Cumulative displacement across the basin is large. For the Serrano terraces, vertical displacements of at least 137 ± 3.6 m are recorded, of which ∼52 m is expressed by exposed faults and the remaining ∼86 m inferred for structures buried beneath basin fill or located at the eastern and western margins of the basin transect. Offset of the highstand of the younger Alvord terraces is less, and yields a total of 72.5 ± 2.8 m with ∼30 m occurring across exposed faults and another 43 m accommodated on inferred structures.

The spatial distribution of fault displacement within the Alvord basin is localized on structures along the western basin margin. In particular, the Serrano fault exhibited nearly continuous displacement but accommodates just >20% of the total displacement during the late Pleistocene and Holocene. When the West Alvord, Alvord, and Wild Horse Canyon faults are included in the displacement budget, ∼50% of the total displacement is localized along the western margin of the basin. The remainder was distributed on structures within (30%) and along the eastern margin of the basin (20%).

There were 26 periods of displacement that occurred on faults between the formation of successive shorelines across the basin ⁰⁴. Along the western margin of the basin, 14 periods of fault slip are recognized on exposed faults. The Serrano fault was active both during and after formation of Alvord and Serrano terraces, but activity on the West Alvord fault system largely stopped before the younger lake inundation and was completely quiescent before formation of the A2 shoreline of the Alvord terraces. The Serrano fault has a cumulative offset of nearly 29 m on the highest stillstand (S1), and shows incremental displacements between all subsequent terraces but A3–A4. In contrast, the West Alvord fault records displacement of nearly 10 m between the formation of S1 and the advent of the younger Alvord shorelines, as two periods of slip of 1 m (S2–S3) and 9 m (S3–A1). Farther to the east, the trace of the Alvord fault is topographically lower than the Serrano shorelines and does not preserve a record of displacement during the earlier lake highstand. Nevertheless, the fault was active before the formation of Alvord shorelines and developed a scarp of at least ∼5.0 m before the onset of the Alvord highstand. During and following the Alvord highstand, the fault records two displacements with a cumulative displacement of just over 5 m. The history of East Alvord fault is limited as the structure only cuts the lowest shoreline (A5) with a displacement of 1.0 m. For inferred structures and those concealed beneath the present-day playa, 12 periods of displacement are inferred. The Wild Horse Canyon fault underwent four periods of activity, and structures exposed in the highlands of the Tule Springs Rim east of the basin accommodated displacement during two time periods. Terrace offsets between Alvord Point and Little Sand Gap have a minimum cumulative displacement of ∼40 m taken up on synthetic and antithetic faults within the basin.

From the displacement history, we estimate earthquake recurrence intervals for observed faults and for the basin as a whole. Due to the limited spatial extent and poor age constraints for the Serrano terraces, we limit our calculations to the period during and following formation of the Alvord shorelines.

During the formation of the highest and lowest Alvord shorelines (A1 and A5), the Serrano fault records four periods of displacement with a cumulative vertical offset of nearly 14 m with an additional ∼8 m of offset following formation of the A5 shoreline. During the same time interval, the Alvord fault records one offset of ∼0.5 m between A1 and A5 and an additional 4.5 m of displacement after formation of A5. The East Alvord fault records a displacement of 1.0 m following formation of the lowest Alvord terrace (A5). Along the western basin margin, the Wild Horse Canyon fault records just over 3 m after the formation of shoreline A5, but was quiescent until before development of terrace A1. Structures of the Tule Springs Rim accommodated ∼19 m of displacement during A1–A2 but have been inactive since. Within the basin, displacement on concealed east-facing structures consists of two periods of offset with an aggregate displacement of ∼6.5 m between A1 and A5 and nearly 15 m of displacement on west-facing faults (attributed to the Tule Rim Springs fault).

Using the estimated ages of the Alvord series shorelines and the record of fault displacement during the latest Pleistocene lake cycle together with results of paleoseismologic studies, we calculate ranges for earthquake recurrence in the central Alvord basin. In our calculations, we assume that fault ruptures have a characteristic slip behavior (Schwartz and Coppersmith, 1984) and use the range of vertical displacement of 1.1–2.2 m determined from multiple ruptures recorded in trenches across the southern and central segments of the Steens Mountain fault system (Fig. 2). We use the age range of 20–15 ka for shoreline A1 together with the estimated 11 ka lower bound for A5 to establish a 9–4 k.y. duration for deformation during formation of the Alvord terraces. For faults that record displacement following the formation of Alvord shoreline A5, the 2–5 ka age of the last surface ruptures on faults within the Alvord basin (Hemphill-Haley, 1987; Hemphill-Haley et al., 1999; Lindberg, 1989; Personius et al., 2006, 2007) provides a 9–6 k.y. range of activity. We compute the upper and lower limits for earthquake recurrence for three time intervals: (1) during shoreline formation (A1–A5); (2) following formation of the lowest shoreline (A5) and cessation of surface ruptures; and (3) between the Alvord series highstand (A1) and the last surface ruptures ⁰⁵.

The estimated recurrence intervals for individual faults and for the basin as a whole indicate substantial variations in the temporal and spatial pattern of deformation. Individual faults show order-of-magnitude differences in calculated earthquake recurrence during the late Pleistocene and early Holocene, highlighting the complex spatial-temporal pattern of deformation within the basin. The recurrence intervals for the studied faults, depicted graphically in Figure 17, show a broad range of activity. The recurrence of earthquakes calculated in this study is consistent with the rates determined for a single structure in a paleoseismologic study of the southern Steens fault system in Bog Hot Valley (Personius et al., 2007).

The recurrence for individual faults ranges from hundreds of years to tens of thousands of years. The Serrano fault has a record of nearly continuous displacement during lake-level variation and yields an average recurrence interval of between 500 and 1850 yr. The average rate is similar to recurrence calculated over shorter time intervals (Fig. 17), indicating that strain release on the fault was relatively uniform during the 18–10 k.y. period of earthquake activity. The recurrence intervals for observed displacements on the Alvord and East Alvord faults are in the range of thousands to tens of thousands of years, and are similar to recurrence estimates of the southern segment of the Steens fault estimated by Personius et al. (2007). The Alvord fault has an average recurrence interval of 2000–7850 yr, and recurrence estimated over shorter time intervals is on the same order of magnitude (Fig. 17). For the East Alvord fault, only one surface rupture is recorded and provides an estimated recurrence of 11,000–36,000 yr. Other displacements attributed to the east-facing and west-facing faults within the basin (XEF and Tule Springs Rim fault) were probably accommodated by multiple structures and are not reliable indicators of recurrence of individual structures. Nevertheless, the calculated recurrence intervals for the structures provide insight into the average recurrence intervals for faults concealed beneath the basin and indicate that the rate of earthquake activity within the basin was comparable to that for the Serrano fault, albeit more distributed. The recurrence intervals estimated for the Wild Horse Canyon fault of 3450–12,000 yr and for the structures of the East Tule Springs Rim fault system of 600–2100 yr are within the limits established from individual structures.

The aggregated earthquake recurrence interval for all structures within the basin is on the order of 150–550 yr. Of observed structures, only the Serrano fault has a recurrence similar to that of the basin as a whole and for the most part, individual structures record earthquakes with a periodicity of several years to tens of thousands of years, as do aggregated recurrence estimates for faults inferred beneath and along the margins of the basin. Earthquake activity clearly is heterogeneously distributed across structures of the extensional fault system. The recurrence for earthquakes in the Alvord basin during the late Pleistocene and early Holocene is greatly in excess of the 2–5 k.y. of seismic quiescence recorded in the late Holocene (Fig. 17). The record of earthquake activity documents that seismic release in the Alvord basin is nonperiodic over a 10⁴ yr time scale and that the extensional basin experienced an earthquake cluster (Wallace, 1984, 1987) during the latest Pleistocene and early Holocene.

The cumulative displacement recorded by the highest shoreline (Serrano terrace S1) provides a minimum vertical displacement since the late-middle Pleistocene (200–130 ka). Summation of vertical offsets of S1 reconstructed in our transect indicates at least 137.5 ± 3.6 m of vertical displacement. Of this displacement, at least 52.0 m is directly measured across exposed faults, with the remainder determined from our reconstruction of shorelines in the central part of the basin. The poor age constraints established for the Serrano terraces (200–130 ka) provide for only a crude estimate of average vertical displacement rate of between 0.7 and 1.1 mm/yr.

For the Alvord terrace series, the cumulative offset of the highest stillstand (A1) from Serrano Point to Tule Spring Rim is 72.5 ± 2.8 m, with ∼29.5 m accommodated on observed faults exposed in the western part of the study area and the remainder estimated from terrace patterns from both sides of the Alvord playa. Using an age for the Alvord highstand (A1) of 20–15 ka provides a vertical displacement rate of 3.6–4.8 mm/yr. Displacement on A1, however, occurred over a shorter time interval of 18–10 k.y. (between 20–15 and 5–2 ka) bracketed between formation of the geomorphic surface and the last surface rupture on faults in the area. Using the restricted time interval, vertical displacements are between 4.0 and 7.3 mm/yr.

Given the dip of faults, the vertical displacement rate provides an estimate of horizontal displacement rate that can be compared to the geodetic rates across the Alvord basin. Potential field models of the Alvord basin (Whipple and Oldow, 2004) and seismic reflection profiles across the basin ∼12 km south of our transect (Bradford et al., 2006) indicate fault dips of ∼60°. For the uppermost Alvord terrace (A1), this yields an average horizontal displacement rate of 2.0–2.8 mm/yr. Within the uncertainty, this is in excess of the geodetic rate of 1.75 ± 0.3 mm/yr determined across the basin. If the time interval (18–10 k.y.) limited by the formation of the Alvord highstand and last surface rupture is used, the horizontal rate increases 2.3–4.2 mm/yr and is nearly double the geodetic rate. The average horizontal displacement rate for the Serrano terrace (S1) and its relation to the contemporary displacement field present a very different picture. For S1, the estimated horizontal displacement rate of 0.4–0.6 mm/yr is slower than the geodetic rate by at least a factor of three.

Active normal faults within the Alvord extensional basin disrupt flights of three and five shorelines of pluvial Lake Alvord formed during at least two periods of lake-level high-stand in the Pleistocene. Although recognized as a potentially complicating factor (Hemphill-Haley, 1987; Lindberg, 1999), differences in the altitude of individual shorelines caused by faulting far exceed earlier estimates (Carter et al., 2006) and provide a record of the complex interaction between coeval fault displacement and lake-level drop. During the late Pleistocene and Holocene, 26 periods of fault slip occurred on at least eight faults during and after water levels decreased.

The displacement history of faults in the Alvord basin illustrates the complexity in the spatial-temporal pattern of deformation within an extensional basin. The cumulative displacement of faults bounding the western margin of the basin takes up nearly 50% of the total offset recorded by lake shorelines. Exposed structures along part of the western margin, the Serrano, West Alvord, and Alvord faults, show dramatically different histories of activity. The Serrano fault was active nearly continuously and accommodated >20% of the total displacement across the basin since the late-middle Pleistocene. In contrast, the Alvord and West Alvord faults underwent long periods of quiescence and accounted for only ∼15% of the aggregate displacement, the remainder being accommodated by the Wild Horse Canyon fault farther to the west (Fig. 4). Nearly 30% of the aggregate displacement was taken up on structures concealed within the basin; the remaining 20% was accommodated by structures in the highlands bordering the eastern flank of the basin.

Deformation rates varied with time, and displacement was heterogeneously distributed on structures across the extensional basin. Displacement rates estimated over 10⁴ yr outstrip geodetic rates by two to three times and are greater than rates estimated over 10⁵ yr by a factor of five or ten. Surface rupture of basin faults ceased ca. 2–5 ka (Personius et al., 2007; Hemphill-Haley et al., 1999; Lindberg, 1989), but the basin fault system records an average earthquake recurrence of several hundred years during a late Pleistocene–early Holocene cluster of earthquake activity. During the earthquake cluster, recurrence on individual structures varied from hundreds of years on one fault to several thousand years to tens of thousands of years for the other structures within the basin. Strain release was characterized by substantial spatial and temporal heterogeneity within the extensional fault system.

The fact that the 10⁴ yr deformation rate exceeds the geodetic rate in the Alvord basin differs from geologic estimates in many extensional basins where geodetic rates commonly are greater (Friedrich et al., 2003; Wesnousky et al., 2005). The lack of late Holocene earthquakes may explain the difference from basins where elevated geodetic rates are attributed to viscoelastic strain transients associated with ancient earthquakes (Wernicke et al., 2000; Dixon et al., 2003; Oskin and Iriondo, 2004). Alternatively, the difference in rates in the Alvord basin may reflect the ability to estimate aggregate displacement across the Alvord fault system via the use of high-resolution measurement of a regionally extensive vertical datum (shorelines). To adequately resolve this critical issue, a better understanding of the alongstrike variation in fault slip is required within the Alvord basin and must be combined with better age controls on the ages of shorelines. Nevertheless, the distribution of slip recorded on faults within the Alvord basin documented in our study across the center of the basin clearly shows that geologic deformation rates based on individual fault zones within extensional systems are susceptible to substantial underestimation.

The discrepancy between geodetic, 10⁴ yr, and 10⁵ yr deformation rates in the Alvord basin has important implications for physical models for the earthquake deformation cycle. Studies based on geodetically and geologically determined displacement rates typically assume that far-field displacements and associated strain accumulation rates remain constant (Reid, 1910; Savage and Burford, 1973; Shimazaki and Nakata, 1980; Scholz, 1990). In this context, well-constrained slip histories, determined over several earthquake cycles, should equal interseismic displacement rates and strain accumulation histories provided by tectonic geodesy, so long as short-term transient contributions to displacement fields are adequately compensated (Foulger et al., 1992; Hager et al., 1999; Kenner and Segall, 2000; Wernicke et al., 2000; Dixon et al., 2003; Meade and Hager, 2004; Oskin and Iriondo, 2004).

As a counterpoint, Friedrich et al. (2003) discussed the possibility that the assumption of constant strain loading during the earthquake cycle may not be valid and may vary over time scales of 10³ to 10⁵ yr. The spatial and temporal clustering of earthquakes (Wallace, 1984, 1987; Swan, 1988; Sieh et al., 1989; McCalpin and Nishenko, 1996; Grant and Sieh, 1994; Marco et al., 1996; Zreda and Noller, 1998; Rockwell et al., 2000) together with the lack of highly variable slip per earthquake along well- documented fault segments (Sieh and Jahns, 1984; Sieh, 1984; Schwartz and Coppersmith, 1984) are consistent with end-member interpretations: (1) that observed variation in strain release (earthquake clustering) reflects comparable changes in strain accumulation, or (2) that strain accumulation is constant and variations in strain release will average to strain accumulation rates over sufficient time intervals (10³ yr or greater).

Taken in the context that tectonic geodesy provides the only practical means of assessing strain accumulation rates on faults, the two scenarios have significant implications for the application of geodetically determined displacements and their relation to the earthquake cycle for individual fault systems. For the case where variable strain release during earthquake clusters tracks changes in strain accumulation, GPS surface velocities have limited predictive potential for characterizing the earthquake deformation cycle over intermediate time scales. In contrast, if strain accumulation rates remain constant and displacements arising from stress diffusion associated with earlier earthquakes can be compensated, GPS determined surface motion might represent a good proxy for earthquake slip rates integrated over 10³ to 10⁶ yr.

The ability to determine whether strain release on active faults directly reflects strain accumulation has been hampered by the investigation of extensional fault system where recent earthquakes may have contributed short-term displacement transients to the contemporary displacement field. Over intervals of several hundred years, the viscous or viscoelastic behavior of the substratum can contribute 50% of the velocities of sites within a few hundred kilometers of large earthquakes (Foulger et al., 1992; Hager et al., 1999; Kenner and Segall, 2000; Wernicke et al., 2000; Dixon et al., 2003; Friedrich et al., 2003) and may reach values of up to 95% (Oskin and Iriondo, 2004). Thus for seismically active regions, the contribution of a short-term transient to the velocity field contaminates the signal to a degree that, when combined with uncertainties in subsurface fault geometry and ages and displacement of Holocene and late Pleistocene geologic markers, differentiation between constant strain accumulation and variable strain accumulation cannot be made (Friedrich et al., 2003).

The lack of significant earthquake activity since 2–5 ka in the Alvord basin (Hemphill-Haley et al., 2000; Personius et al., 2007) suggests that the contemporary displacement field for the region (Miller et al., 2001; Hammond and Thatcher, 2005; Mazzotti et al., 2003) is not affected by short-term displacement transients arising from earlier earthquakes. As such, the rate discrepancy over different time scales observed between geodetic, 10⁴ yr, and 10⁵ yr deformation rates in the Alvord basin lends credence to the supposition by Friedrich et al. (2003) that changes in strain release and accumulation may be directly linked. Although an intriguing possibility, our work in the Alvord basin does not provide adequate spatial coverage to preclude the likelihood that the rate discrepancy reflects migration of the locus of strain release to other locations within the basin and/or to locales outside the basin fault system. The displacement magnitude and rate of heterogeneity recorded on faults within the transect over the 10⁴ yr time scale may extend to substantial variation in deformation along strike within the basin over a 10⁵ yr time scale or to the migration of deformation to other basins in the region. As pointed out by Dolan et al. (2007), the reciprocal nature of the temporal pattern of strain release in the Los Angeles basin and Eastern California shear zone does not require changes in far-field boundary conditions. This uncertainty reemphasizes the need to characterize deformation patterns in complex fault systems and a broad range of time and length scales.

This work was partially funded by National Science Foundation grants EPS-0132626 and EAR-0225421. We thank Luigi Ferranti for assistance in the field and for discussions about the interaction of shorelines and faults. We acknowledge the insightful and constructive reviews of the manuscript by Marith Reheis and Sotiris Kokkalas.