Using a landscape fingerprint to identify changes in fault-slip behavior

Large strike-slip faults rarely exhibit a consistent and constant slip rate along their entire length. Dense GPS and global navigation satellite system (GNSS) arrays and interferometric synthetic aperture radar (InSAR) analyses reveal that many major strike-slip faults contain both locked segments and creeping segments as well as gradients in slip rates that can be hard to identify due to relatively short time scales of observation and relatively long seismic cycles (Harris, 2017). Determining the past slip behavior of strike-slip faults over thousands of years or longer is particularly difficult because the ultimate result of aseismic slip and seismic slip over time is the same: displacement. Paleoseismic activity, a proxy for past slip behavior, can be identified using trenching but is limited to a geographic snapshot at a single point on the surface (Liu-Zeng et al., 2015). Identifying where faults change from locked slip to creeping slip is an important component in seismic hazard estimates, especially along faults that are not densely instrumented. By understanding that landforms may form in tandem with a slip transition, we can reinterpret landforms near faults around the world that exhibit slip rate changes, including the San Andreas (California, USA), North Anatolian (Turkey), Philippine, and Chaman (Afghanistan-Pakistan) faults.

One of the most dramatic changes in slip style occurs along the San Andreas fault near the town of Parkfield, California. To the northwest, fault creep is relatively rapid at 25–30 mm/yr, approaching the expected geologic slip rate (Titus et al., 2006). To the southeast, the fault is locked. Titus et al. (2011) documented regions of extension and contraction calculated from the GPS velocity field on either side of the slip transition that coincide with geologic evidence of deformation, including recent basin development to the southwest and young fold growth to the northeast. The change in slip behavior also coincides geographically with the Gabilan mesa, a 110-km-long by 25-km-wide landform on the Pacific plate that tilts gently away from the fault toward the Salinas valley (Fig. 1) (Durham, 1974; Richardson et al., 2020). In a wide deforming region with folds flanking either side of the San Andreas fault, the mesa stands out as a relatively undeformed feature. The landform is located on granitic Salinian basement and overlain by young sedimentary rocks, which are only moderately consolidated (Dohrenwend, 1978). Formation of the mesa has been attributed to the small convergent component of plate motion in central California (Montgomery, 1993; Richardson et al., 2020).

The Gabilan mesa has been used by geomorphologists as a natural laboratory due to its uniform erosion processes and tectonic forcing as well as simple boundary conditions (Perron et al., 2008a, 2008b). The ridge tops are accordant and define a paleosurface that slopes ~3° away from the fault (Dohrenwend, 1978). The trend of beheaded drainages that dissect the mesa varies systematically from westward draining in the northwest to southward draining in the southeast, resulting in a fanning pattern (Fig. 1). This drainage pattern suggests a subtle half-dome shape that is difficult to detect from field measurements due to the shallow tilt of the mesa. Because the drainages are beheaded across much of the mesa, quantitative geomorphological characteristics such as longitudinal profiles and drainage sizes cannot be meaningfully compared.

We use geologic and geomorphic data to bracket the timing of mesa formation. There is evidence of several phases of uplift across central California (Page et al., 1998) since plate motion became transpressional sometime between 3.5 and 8 Ma (Cox and Engebretson, 1985; Atwater and Stock, 1998). Uplift of the Gabilan mesa must have occurred after deposition of the ca. 4 Ma Paso Robles Formation (Addicott and Galehouse, 1973; Sarna-Wojcicki et al., 1991) and the Pliocene Pancho Rico Formation (Durham and Addicott, 1965), which are now gently dipping away from the fault. The minimum age of uplift is constrained to 400 ka. This age, which coincides with a major range-building phase that followed a period of relative quiescence (Page et al., 1998), is based on stream incision rates of 0.42–0.32 mm/yr that are consistent with the rate required to tilt a fluvial strath over 400 k.y. to its present angle of 0.66° (0.38 mm/yr; García and Mahan, 2012).

The timing and evolution of the Gabilan mesa is further constrained by a series of wind gaps near the San Andreas fault (Fig. 1). These geomorphic features form when a stream channel is cut off from its upstream source and the abandoned channel becomes elevated relative to the present-day drainage network due to erosion and/or uplift. Four wind gaps formed as San Lorenzo Creek captured drainages that had previously flowed away from the fault toward the southwest, redirecting the flow of water in a fault-parallel, northwesterly direction away from the center of the mesa (García and Mahan, 2007). The wind gaps have a higher elevation in the northwest and are lower in the center of the mesa; there are no wind gaps in the southeast (Fig. S1 in the Supplemental Material¹). If we assume that the rate of uplift has been constant across these closely spaced ridges, but the uplift has progressed steadily southeastward with time, this pattern suggests that uplift has been occurring for longer in the northwest (adjacent to the creeping section) than the southeast (adjacent to the locked fault).

We use a physical model of a fault previously described in Ross et al. (2022) to isolate the influence of the slip behavior on deformation without the confounding variables present in nature such as varying lithology, fault geometry, oblique plate motion, or prior deformation. Our simple silicone model contains a dextral strike-slip fault that “creeps” along part of its extent and is “locked” along the rest (Fig. 2). We consider deformation during the interseismic period, when the locked fault section can be treated as an intact block. We track vertical displacements and hence the topographic development of the silicone using photogrammetry.

In the experiments (Fig. 2B), contraction dominates on the side where the creeping material “runs into” the locked material (i in Fig. 2B) while extension dominates on the other side where locked material “pulls away” from creeping material (ii). These prominent regions of dilatation in the model correlate to regions of geologic and geodetic contraction and extension observed in central California (Titus et al., 2011) and are described elsewhere (Ross et al., 2022).

Here, we focus on the secondary topographic feature that develops near the slip transition, which is caused by an elevation gradient near the slip transition (iii). Initially, this gradient takes the shape of a half-dome centered at the point of slip transition. Over time, the dome-like structure is offset from the transition point by right-lateral displacement. The displaced portion thus becomes a record of earlier tilting while the structure continues to develop near the slip transition, a process illustrated by the panels in Figure 2D.

The physical model limits our analysis to interseismic deformation, given that the linear viscous rheology of silicone does not allow for slip or fracture along the locked portion of the fault. While elastic rebound after an earthquake would effectively undo a portion of the interseismic strain accumulation in nature, previous work has demonstrated that not all interseismic landscape deformation is recovered elastically and instead accumulates to create significant landscape deformation over geologic time scales (Avouac, 2003; Baden et al., 2022). Additional research is required to reconcile the different approaches used to model elastic and inelastic contributions to off-fault deformation, particularly in poorly understood transition zones.

In her review paper, Harris (2017) describes 18 creeping segments of faults around the world. Many were identified using InSAR because the faults are in regions where GPS-GNSS coverage is limited or earthquake records are not sufficient to determine creeping versus locked behavior. Constraints of InSAR, including the direction of satellite orbits relative to fault motion, spatial and temporal resolution, vegetation, and atmospheric noise, make it difficult to identify the precise location of a slip transition (Fukushima et al., 2019). We examine all 18 faults, excluding those with significant fault-perpendicular contraction and also those where the terrain is too mountainous to allow for the identification of a subtle landscape feature akin to the Gabilan mesa. There are three faults, however, where we observe a tilted, fanning drainage pattern and half-dome shape in the topography of a comparable scale and fault-relative location to the Gabilan mesa (Fig. 3).

The simplest method to visualize the landform topography in each system is by comparing contours at an interval of 200 m (Fig. 3, left) to the millimeter-scale topography of the physical model (Fig. 2C). We also plot points in a 2 km grid across the area of the suspected feature and apply the ArcGIS “Trace Downstream” tool to identify the azimuth of flow from each point to the edge of the suspected feature. This tool uses a digital elevation model to trace downhill flow direction from each selected point. Flow-direction azimuths are color coded relative to fault strike (Fig. 3, right): Those within 5° of fault-normal in either direction are white, while azimuths >5° from fault-normal are yellow if flow is toward the creeping side or red if flow is toward locked side of the fault. The flow-direction patterns in conjunction with topographic maps confirm half-dome landforms centered near the slip-behavior transitions along each strike-slip fault.

Unlike our physical experiment in which confounding variables can be controlled, these three faults must each be considered in the context of their unique tectonic and geologic settings. Several factors can influence the direction of streamflow: rock type, underlying geologic structures, the effect of microclimate on valley walls, tectonic uplift, prior deformation history, and even human interventions. In the Gabilan mesa example, these variables are generally constant across the mesa and their effect on flow direction is well understood, but on other faults, it is important to consider the potential enhancing or obscuring role of these competing controls on streamflow direction patterns.

The North Anatolian fault in Turkey is right lateral with both seismic and aseismic behavior. The 70-km-long Ismetpasa segment is creeping (Fig. 3B), which is observable via InSAR and wall displacements (Bilham et al., 2016). Creep rate estimates range from ~6 to 10 mm/yr, and creep has been observed in bursts rather than a continuous constant rate (Bilham et al., 2016). This segment of the North Anatolian fault has also experienced large earthquakes in the past, including a 7.2 M_w event in 1944 (Kondo, 2005). The creep rate has decreased since the behavior was first documented in the 1950s, indicating that the creep is a postseismic signature (Kaduri et al., 2017). Despite this, the region is comparable to the others because the fault segments on either end of the creeping section remain fully locked, creating a gradient in slip rates. The fanning drainages are somewhat complex here but overall show a similar pattern to those in the San Andreas fault system (Fig. 3A).

The Philippine fault accommodates left-lateral displacement within a complex tectonic regime, with subduction occurring from the east and west of the Philippines (Fig. 3C). Duquesnoy et al. (1994) inferred a creeping section on the Philippine fault using GPS on Leyte Island, which Fukushima et al. (2019) mapped in more detail using InSAR. Creep has also been validated using alignment arrays (Tsutsumi et al., 2013) and by monitoring concrete structures and roads (Perez et al., 2008), although anthropogenic subsidence complicates the interpretation of slip signals (Fukushima et al., 2019; Dianala et al., 2020). Creep rate estimates range from 21 to 36 mm/yr (Dianala et al., 2020). Along the Philippine fault, the topography dips west toward the Camotes Sea. The half-dome feature with fanning drainage patterns is centered slightly north of the transition in slip next to the locked portion of the fault, consistent with left-lateral displacement of the feature over time.

The Chaman fault near the border of Pakistan and Afghanistan defines a left-lateral boundary between the Indian and Eurasian plates with a slip rate between 10 and 30 mm/yr (Barnhart, 2017) (Fig. 3D). The proposed 125-km-long creeping segment borders an arc-shaped alluvial basin drained by highly incised, parallel streams (Ul-Hadi et al., 2013). The Chaman Basin southwest of the fault is considered the result of transpressional uplift of the crystalline bedrock material, enhanced by vertical motion along secondary thrust faults bordering the Roghani ridge thrust block. Similar to those at the Gabilan mesa, wind gaps along the high side of the uplifted ridge decrease in elevation toward the north, suggesting that uplift timing or rate is not uniform along the entire length of the uplifted section (Ul-Hadi et al., 2013) (Fig. S1). The topographic map of the basin demonstrates the subtle half-dome shape centered around the hypothesized change in slip behavior from Barnhart (2017), with flow changing directions near Roghani ridge.

Due to ever-improving InSAR coverage, geodetic monitoring, and seismic detection, present-day variations in slip velocity along major strike-slip faults are now identifiable. However, many locations still exist where these methods are not available at a resolution that is useful for estimating local seismic risk, particularly on remote or unmapped faults. Short-term geodetic and InSAR observations are insufficient for distinguishing locations of slip transitions that have persisted over geologic time scales. Landscape-scale features, on the other hand, form over hundreds of thousands of years and are evidence for non-elastic, permanent deformation sustained over repeated earthquake cycles. By investigating the topography and drainage patterns at suspected transitions in slip behavior on four strike-slip faults, we have identified a potential topographic signature for faults that shift between seismic and aseismic behavior along strike: a “half-dome” landform with fanning drainages developed. Our physical models of a fault with a change in slip behavior are useful for identifying this topographic “fingerprint” because they allow us to isolate the slip transition as a variable causing off-fault deformation. Because the landform is observed along the San Andreas fault, in physical models, and along three other faults, we suggest that a transition from locked slip to creeping slip can shape large-scale topography.

Ross, Reber, and Titus were supported by National Science Foundation grants EAR-1916970 and EAR-1917048. Authors have no competing interests. Thank you to Emery (Grasshopper) Anderson-Merritt, Sean F. Gallen, and three anonymous reviewers for thoughtful comments and suggestions on this manuscript. Thanks to Chris Harding for help with GIS analysis.