Millions of years ago, the Pacific tectonic plate did not flank California. Back then, the San Andreas Fault did not exist.

Instead, in the midst of an older version of the Pacific Ocean, a mid-ocean ridge gave rise to the Pacific plate on its west side and the Farallon plate on the east. The Farallon plate did not slide past western North America, but instead plunged below. This subducting plate drove blasts of volcanoes that dotted the original Sierra Nevada. Today, the eroded roots of this once explosive continental arc constitute the towering mountains.

Most of the Farallon has been swept beneath North America. When the mid-ocean ridge that begat the Farallon impinged on the continent around 30 million years ago near modern-day Los Angeles, the San Andreas fault zone was born. Since then, the strike-slip plate boundary has zippered its way north and south. As the margin morphed from subduction to strike-slip, the Sierra Nevada’s arc volcanism sequentially ceased. Now, the Pacific plate, which moves to the northwest, laterally grinds against North America along much of California’s length.

Remnants of the once mighty Farallon plate live on, like its northern leftover, the Juan de Fuca plate. The Juan de Fuca plate is subdivided into smaller sections that sink beneath the northwestern coast of the U.S. This is the Cascadia subduction zone, which sits off the shores of Northern California, Oregon, and Washington, reaching north into Canada. The Cascade Range — an arc that includes Mount Saint Helens — remains a string of active volcanoes.

These disparate plate boundaries meet at the Mendocino Triple Junction, which, on a map, often looks like a single point marking intersecting faults: the right-lateral San Andreas Fault, the Cascadia subduction zone, and a right-lateral transform fault that connects to the mid-ocean ridge that continuously creates fresh Juan de Fuca plate.

But the triple junction is far more complex than a simple intersection of three lines. Faults splice the area in convoluted ways, seismicity rates are high, and the zone of deformation is broad.

In a recent Science paper, USGS seismologist David Shelly and coauthors use peculiar seismic signals — tectonic tremor and low frequency earthquakes — to peer into the complex region. They suggest that a snippet of the Farallon slab did not sink into the mantle with the rest of the plate’s pieces, but rather stuck itself to the Pacific plate. This bit of former Farallon may now be scraping its way northward — beneath North America, as opposed to next to it. This has major implications for the supposed slab window, considered an important feature of dying subduction zones. Moreover, if this hypothesis is correct, the hidden bits of tectonic plate could pose a seismic hazard that has not been accounted for in assessments.

What’s a slab window?

At the latitudes where the Farallon plate has fully disappeared beneath North America, the mid-ocean ridge entered the trench, halting subduction and forcing a switch to strike-slip motion. Where the ridge went down, the slab window hypothesis posits that because no plate followed, a window (or gap) formed between the newly minted strike-slip margin and the piece of Farallon continuing its downward trajectory. Hot mantle should flow into the gap, filling it.

This is a testable hypothesis; volcanism would change. Indeed, volcanic rocks that aren’t typical of continental arcs, but instead have signatures closer to hot, fresh mantle exist throughout California. These rocks signal that a slab window (or several slab windows) appeared as the Farallon fell away.

Moreover, the age of these volcanic rocks should follow the formation of the San Andreas. And in the northward direction, that is largely the case.

At the Mendocino Triple Junction, there’s no subducting mid-ocean ridge, but instead a subducting strike-slip fault. So what could be happening at this curious juncture? One hypothesis proposes that beneath North America, south of the southern edge of the subducted Gorda plate, a slab window exists where mantle wells up from below.

However, another model suggests that instead of falling away, a subducted fragment of former Farallon — the purported Pioneer Fragment — is stuck to the Pacific plate just southeast of the triple junction. If this is the case, at this location, there’s no mantle upwelling, no window in the slab.

One way to begin to differentiate between these possibilities is to look at the abundant seismic data from the region.

Tectonic tremor and other oddities

Typically, when the topic of earthquakes along the San Andreas Fault or the Cascadia subduction zone arises, major quakes come to mind. But these fault zones are always incrementally shifting in ways unknowable to humans without sensitive seismic instrumentation to measure the movement. Two unexpected signals in particular were only discovered within the past twenty years or so — tectonic tremor and low-frequency earthquakes.

“Tremor and low-frequency earthquakes are usually detected by the same networks, just using somewhat different techniques,” Shelly said. “They tell us about the same processes.”

Tectonic tremor is a low-amplitude seismic signal that lasts a long time — anywhere from minutes to hours to weeks. It emanates from deep within subduction zones, below the region that produces the more familiar megathrust earthquakes. This signal is associated with slow slip.

Low-frequency earthquakes are very small earthquakes that lack high-frequency waves, similar to tectonic tremor. But unlike tectonic tremor, low-frequency earthquakes have distinct seismic wave arrival times, so they can be identified as individual events. Plus, they often occur in patches at regular intervals. Low-frequency earthquakes in these “families” can be combined to improve the accuracy of their source location. What’s more, these low-frequency earthquakes comprise the longer duration tremor. Throughout the world, tectonic tremor and low-frequency earthquakes mark many major plate boundaries, most often deep within the subduction zone interface.

Cascadia anomalies

The Cascadia subduction zone has a band of tectonic tremor that stretches north to south along most of the margin, at a depth of more than 30 kilometers. However, at the southern end of the subduction zone, a stretch of tremor offset from the main band was identified in previous work. This bunch sits west of the main band by about 50 to 100 kilometers. Within this cluster are at least 27 families of low-frequency earthquakes that spark to life every two days or so. Taken together, the families dip to the northeast between depths of about 22 to 29 kilometers.

By studying the families of low-frequency earthquakes, scientists can figure out how they move, and in what direction the fault zone is oriented. That information can illuminate whether there’s a slab window or a Pioneer Fragment.

Shelly and colleagues came up with three different scenarios to test. “The thought experiment was really, ‘ok, we have slip on what may be a plate-bounding fault. We know the approximate orientation of the fault [west-northwest, east-southeast], but not the orientation of the slip.” With the three proximal tectonic plates, the team considered the question: What are plausible plate interactions?

In the first scenario they came up with, the structure would separate the Gorda plate to the north, and the North American lower crust to the south. If the families demonstrate normal motion between the two plates, that indicator of stretching or tearing would support the presence of a slab window. In this scenario, the Pioneer plate, which would have been south of the Gorda plate, would have either become attached to the North American plate or have been completely subducted, Shelly said.

The second scenario also has the Gorda plate to the north, but the Pacific plate would sit to the south. For this scenario to be supported, the families must move with right-lateral strike-slip motion. But the Pacific plate doesn’t subduct, so there must be something present south of the subducted Gorda plate. That “something” would be the Pioneer Fragment, necessarily captured by the Pacific plate so that it’s moving in the same manner as the Pacific—with right-lateral strike-slip motion—but under the surface. A corollary: The strike-slip boundary would extend beneath North America.

In the third scenario, Shelly and colleagues considered North American lower crust to the north and Pacific plus Pioneer to the south. This scenario would require oblique reverse motion, in which the captured Pioneer Fragment would encounter North American lower crust on its northern boundary. The Pioneer Fragment, moving northward with the Pacific plate, would get stuck and produce reverse faulting. The Gorda plate, Shelly said, would be either farther north or deeper.

Focal mechanisms, tides and sense of slip

To distinguish between the three scenarios, Shelly and colleagues first determined a composite focal mechanism (also known as a beachball diagram) for the families of low-frequency earthquakes. A focal mechanism can help conceptualize how a fault has moved, as well as potential orientations of the fault. The composite focal mechanism in this study involves combining the low-frequency earthquakes to combat the low signal-to-noise ratio of individual events. 

Focal mechanisms have two nodal planes, one of which corresponds to the likely orientation of the fault. The composite focal mechanism determined by Shelly and colleagues contained a nodal plane that matched how the families were distributed at depth. The families’ hypocenters — their locations in three dimensions — line up along a plane that strikes west-northwest and dips to the northeast.

The team also attacked the orientation of slip on the fault via a second method — studying tidal modulation. Because the low-frequency earthquakes have such short recurrence intervals — every two days — they’re likely not driven by large, transient slow-slip events that operate on the scale of weeks to months to longer than a year. Instead, these low-frequency events are likely linked to Earth’s tides, as documented elsewhere. (Remember that in this context, frequency refers the number of cycles per second the seismic waves produced, not how often the earthquakes occur.)

The team demonstrated that these particular low-frequency earthquakes are strongly tidally modulated, and they explored what fault orientation would maximize tidal modulation. In other words, what is the optimal faulting geometry that would allow Earth’s tides to trigger low-frequency earthquakes every other day?

The tidal analysis is broadly consistent with either normal faulting or strike-slip faulting — that rules out the third scenario, which requires reverse faulting. Moreover, the strongest and most systematic tidal analysis results in a beachball that matches that of the composite focal mechanism. Both beachballs support strike-slip motion, which is inconsistent with the first scenario, a slab window. This analysis points to Pacific-like plate motion beneath the North American plate, supporting the hypothesis that the Pacific plate captured the Pioneer Fragment. And, it implies that the Mendocino transform — that third fault of the triple junction — continues below the surface. If the second scenario is right, the Pioneer fragment is scraping the underside of North America as it’s dragged northwestward by the Pacific.

Another puzzle piece

But that’s only the first bit of tectonic trickery explored by these authors. The second involves a largely aseismic low-velocity zone, which has also been imaged in tomography studies. This zone’s lower, southern boundary is marked by the families of low-frequency earthquakes we’ve been discussing. The gap persists until seismicity illuminates the main band of subducting Gorda plate.

The low-velocity zone could indicate the presence of melt or fluids, or even heat. “We don’t know for sure,” Shelly said. But melt isn’t a likely explanation because it wouldn’t typically support small, sudden slip that generates the low-frequency earthquakes, he explained. “Fluids are probably involved.” Their preferred hypothesis: the subduction of water-rich sediments.

These sediments, the authors argue, are those of the accretionary wedge initially stuck to the edge of the North American plate — a feature typical of subduction zones that have plenty of sediment shedding into the trench via rivers draining the overriding plate. The downgoing plate has its own layer of sediment that’s scraped off at the front edge of the overriding plate and into the accretionary wedge (or prism). Some of the accretionary wedge’s sediment might hitch a ride on the downgoing slab, heading into the subduction zone. The subducting sediment could then be underplated — stuck to the underside of the overriding plate — or continue deeper into the mantle.

In this case, Shelly and colleagues posit that some of the accretionary prism was indeed subducted, but currently remains on the downgoing Gorda plate. In other words, a squishy bit of sediment accreted to North America’s front edge may be attached to the Gorda plate somewhere below North America’s western edge.

“There’s a volume problem for what happens to the shallow [North American accretionary] wedge as the Pacific moves northward,” said Shelly. “Subducting part of the wedge would solve (or at least mitigate) that problem.”

Detachment

How do these pieces fit together? The upper surface of the Pioneer Fragment would be a large, low-angle detachment into which the near-vertical San Andreas soles. This detachment could slip seismically, and could be capable of generating rare — but large — earthquakes.

The families of low-frequency earthquakes trace the northern boundary of the Pioneer Fragment, which dips to the northeast. Above and to the north of this fault zone (but below the overriding North American plate) is the low-velocity zone of subducted sediment that moves with the Gorda slab.

If the hypothesis put forth by Shelly and colleagues is correct, then an unknown and unconsidered seismic hazard exists for the region around the Mendocino Triple Junction — one of the most seismically active places in the world.

This study utilized seismic data from the NSF National Geophysical Facility data archive.