“Interesting question,” Dragert replied. “We have no idea.”
In a paper published in Science on May 25, 2001, Dragert and Wang released their mysterious findings to the world. “In the summer of 1999, a cluster of seven sites briefly reversed their direction of motion,” they wrote. “No seismicity was associated with this event.” Meaning there were none of the normal seismic shockwaves one would expect from something that looked in every other respect like an earthquake.
Could two plates really slip without producing detectable seismic waves? Dragert and Wang noted that all the GPS tracking stations that moved backward were some distance to landward of the locked part of the zone. So whatever was causing the backward movement had to be happening way down deep, where the rocks were hotter, softer, and less likely to stick together for long periods. They calculated that if an area 30 by 190 miles (50 by 300 km) were to move backward less than an inch (2 cm), the fault would still be generating the energy equivalent of a magnitude 6.7 earthquake. But where was that energy going?
They concluded that these slip events were probably transferring stress “uphill” to the shallower part of the locked zone in “discrete pulses.” So even though nobody could feel them at the surface, each time one of these bizarre reversals happened, it was probably pushing the fault one notch closer to failure—a giant earthquake.
The mystery of “silent slip” took another unexpected turn when Herb Dragert traveled to a science conference in New Zealand, where he learned from Kazushige Obara that something very similar was happening in Japan. “He’s the one that discovered tremors,” said Dragert, “but he didn’t know what they were. He had no idea there was crustal displacement—crustal motion—involved with these. He just kind of said, ‘Hey, these are weird signals that aren’t earthquakes, but they’re not volcanic.’ So he called them deep, non-volcanic seismic tremors.”
After quizzing Obara about the details, Dragert returned to the Pacific Geoscience Centre and started hunting for a connection between tremor and slip. Garry Rogers, by now one of Canada’s top seismologists, had an office just upstairs and down the hall. Dragert provided Rogers with dates when the zigzag patterns had showed up on the GPS. Rogers then dug out boxes of seismograph records from the PGC archives and both were stunned to find a near-perfect match-up.
“I opened the box,” said Rogers, recalling the search, “and there was the tremor!” He showed me the seismogram and pointed to a squiggle of tremor noise that coincided with one of the backward jumps on the GPS. “Herb gave me the next date—I opened the next box—and there was the tremor event. And boy, the hairs just stood up on the backs of our necks.”
“Oh yeah,” Dragert said, beaming. “Yeah, that was exciting. Every time a slip event occurred, there was a huge increase in this background noise, a huge increase. And so it was at that point the eureka came through. We said, ‘Hey, these things are intimately related.’ We knew we had something.”
Condensing it all into a neat little sequence, Rogers explained that the deepest part of the fault—way down where it’s hot and gooey—fails, or slips loose, every fourteen months. When it slips—for a period of about ten days—the GPS antennas on the surface record a backward jump. The land actually recoils as the fault slips, and the seismographs record a silent tremor. Fourteen months’worth of deep tectonic stress is transferred upward into the colder, harder rocks of the part of the fault that has remained stuck. Then, with the stress transferred, the lower part of the fault locks up again and the cycle repeats itself. “It’s a very unique phenomenon,” Rogers said. “We called it episodic tremor and slip, or ETS for short.”
The most fascinating revelation in all of this was that the upper part of the fault—the part that’s been locked in place and building stress ever since the last great Cascadia earthquake more than three hundred years ago—gets another increment of stress added to its load on an incredibly and mysteriously regular basis. Almost like clockwork. In other words, the stress doesn’t just add up gradually until the fault ruptures; it comes in discrete little jolts every fourteen months.
“It’s like adding straws to a camel’s back,” Rogers suggested. “Probably when one of these straws are added it will break the camel’s back.” And we’ll have the megathrust earthquake we’ve all been waiting for. But maybe—just maybe—with this bizarrely regular timing of the ETS events, there might be a way to anticipate that quake.
CHAPTER 19
Turbidite Timeline: Cascadia’s Long and Violent History
“Initially, my colleague Hans Nelson and I didn’t believe it,” said Chris Goldfinger with a smile. “We thought it was probably wrong. It was way too simple. It can’t be right. So we wrote a proposal to the NSF [National Science Foundation] to go prove John Adams wrong.”
This was Goldfinger holding court in the ship’s lounge off the coast of Thailand en route to Sumatra to collect piston cores of ocean mud from landslides triggered by the catastrophic quake and tsunami of Boxing Day 2004. He was setting the scene for those who had not yet experienced the intrigue and frustration of trying to study deadly subduction zones that lie hidden beneath the sea. For Goldfinger and his research partner at Oregon State University, Hans Nelson, the story had begun back in 1985 when John Adams, at the Geological Survey of Canada, wrote a controversial paper based on some old OSU core samples.
The original work in question had been done in 1968 by graduate student Gary Griggs and his thesis advisor, LaVerne Kulm, who (along with Bob Yeats) later became Goldfinger’s thesis advisor as well. The Griggs and Kulm piston cores revealed a series of undersea landslides that had traveled hundreds of miles downhill from the edge of the continental shelf into deeper ocean water along a network of seafloor canyons and channels off the coasts of Washington, Oregon, and northern California. The question in 1970, when the original data were published, was what had caused the landslides.
The core samples had been gouged with steel tubes and plastic pipes from a web of deep-sea channels many miles apart, each showing evidence of thirteen or more landslides. Griggs and Kulm over beers after work one night came up with several possibilities. Either the sediment flows “self-triggered” once the seafloor mud had piled up deep enough to collapse under its own weight. Or big storms with very deep waves might have done it. Or perhaps it was big earthquakes.
Griggs and Kulm knew from looking at the core samples and measuring how thick the layers were and how far they were from the Mazama ash layer that each of the thirteen slides had happened within minutes of each other—all up and down the coast. How could so many turbid flows happen in so many canyons so far apart all at once? If it were mere coincidence, the same coincidence had happened thirteen times, once every six hundred years. In one of a series of papers on the subject, Chris Goldfinger would later refer to this as “coincidence beyond credibility.”
The self-triggering idea seemed the least plausible because it was unlikely that sediment would accumulate at exactly the same rate along hundreds of miles of the continental shelf. Rivers of different size and volume dump differing amounts of debris in different places along the coast at different rates—so how would piles of sand so far apart all collapse into turbidity currents at exactly the same moment? Thirteen times?
Storms with waves big enough to trigger deep-sea landslides also seemed a tad improbable, especially on such a wide geographic scale. They were pretty sure that turbulence from waves generated by winter’s worst storms generally did not reach that far below the surface, so how could they disturb heaps of sand at the heads of steep canyons (where most of these landslides began) that were anywhere from 500 to 1,300 feet (150–400 m) deep? If storm-triggered landslides had happened, the odds were higher that they would have occurred at different times in different places—not all at once, as the core samples showed.
Seismic shockwaves also seemed unlikely culprits because in 1970 there was no historical record of megathrust ruptures in the Pacific Northwest. When one of the grad stude
nts casually suggested the quake hypothesis that night over beers at OSU, Professor Kulm legendarily replied, “Nobody would believe it.” Thus the earthquake story entered a state of limbo, with no obvious way to prove or disprove it.
Goldfinger can still recall the doubt and disbelief that flashed through the corridors of the marine geology department when the first John Adams paper appeared in the scientific literature saying, in essence, that Griggs and Kulm should have ignored the potential skeptics. Without going to sea, without collecting a single new sample himself, Adams, the outsider, this transplanted New Zealander who had moved to Canada via Cornell University, wrote a paper based on the Griggs and Kulm cores and concluded that seismic jolts were the best—indeed the only logical—explanation for the thirteen turbidites.
It must have seemed as though a foreign spy had raided the OSU lab and stolen the thunder of the home team’s original discovery. As it turned out, not a bit of skulduggery was involved. Adams simply wrote to OSU officials asking permission to look at some of the core logs that had been sitting in storage since 1968. Picking up the storyline more than three decades later, Chris Goldfinger told his colleagues aboard the Roger Revelle off the coast of Sumatra how an amazing feat of deductive reasoning had come about.
“He saw the same thing [that Griggs and Kulm had seen], that there were thirteen turbidites above the Mazama ash. So he tried to come up with a method to prove or test the origin of these things. And what he did was this: he noticed that there’s a confluence of two channel systems right through here.”
Goldfinger pointed to the map beamed from his computer via an overhead projector to a screen at the front of the ship’s lounge. “Well, it turns out that all of these cores have thirteen turbidites.” He pointed to sampling sites on both main channel systems. “So here’s his little test, right here. He said, ‘Well, okay—if you have two channels or two canyon systems, 200 kilometers [125 miles] apart, and one has thirteen turbidites and the other has thirteen turbidites—how could you possibly pass the confluence and go downstream and not have twenty-six turbidites?”
Goldfinger scanned the faces around the room to see fascinated smiles and raised eyebrows. “You should have twenty-six turbidites here, right?” Landslides send thirteen turbid currents of sand sloshing down two channels that run together, so there ought to be a total of twenty-six turbidite layers in the mud downstream from the confluence.
“And there are only two ways you could not have twenty-six,” he continued. “One is that it just coincidentally dropped out thirteen of them [for unknown reasons, thirteen of the landslides didn’t make it past the confluence], which didn’t seem very likely. And the other way is that they arrived at the confluence at exactly the same time, plus or minus about five minutes—and merged.” In other words, if a big quake triggered a landslide at the same moment at the head of each of the major canyons along hundreds of miles of the continental shelf, then all the mud flows would cascade downhill synchronously and would arrive at the downstream confluence where all the offshore sea channels meet—at the same moment.
Below the confluence there would be a single, merged turbidity flow. No matter how many small tributaries fed into the main channel from the steep slopes above, the total number of turbidites below the confluence would always be the same: one for each coastwide landslide. The only reasonable explanation for so many landslides happening synchronously along so many miles of coastline was large subduction earthquakes.
“And that’s what John argued,” said Goldfinger. And then he paid one of the highest compliments one scientist can offer another. “This was done purely with thinking power. And that’s out of fashion these days.” The praise seems all the more significant given that Goldfinger and Nelson had initially doubted the simplicity and elegance of the Adams hypothesis enough to ask the National Science Foundation for research money to prove him wrong.
Goldfinger had gone to sea in 1992 and discovered the underwater Elvis, a man-with-a-guitar–shaped mound of folded and fractured ocean sediment, along with eight other new faults in the upper plate along the continental shelf. He wrote that the rough edges of these fractures may limit the size of Cascadia’s big subduction quakes by inhibiting the build-up of strain energy and concluded that Cascadia may be the type of subduction zone at which magnitude 9 events “do not occur.” Without magnitude 9 ruptures, the Adams hypothesis had to be wrong. Smaller jolts just wouldn’t do what the hypothesis demanded.
But a subsequent research voyage in 1999 turned things around for Goldfinger and Nelson. The evidence in favor of big landslides was very obvious in the offshore mud when examined close up. Not only did they confirm the same thirteen turbidites along 375 miles (600 km) of coastline, but they ventured farther north to the Nootka fault, at the upper end of the Juan de Fuca plate off Vancouver Island, and farther south all the way down to Cape Mendocino in California. Along the way they collected nearly a hundred new cores and added several discoveries of their own, extending the count from thirteen to eighteen events—presumably large quakes—and extending the timeline back to the end of the last ice age, roughly ten thousand years ago. They saw these dark, sandy landslide scars on the ocean floor as “earthquake proxies,” the telltale markers of Cascadia’s long and violent past.
In the lounge aboard the Roger Revelle, Goldfinger explained the challenge of adding the five new turbidites to the series of thirteen already established. The problem with the new samples was that they came from offshore river channels that were not physically connected to the network of channels flowing primarily from the Columbia River and the Strait of Juan de Fuca up the coast. The new evidence was found in a completely separate, unrelated grid of outflow channels from Barclay Canyon, off Vancouver Island, and from the Rogue River Canyon, midway down the Oregon coast.
How would it be possible to know whether the five additional turbidite flows had happened all the way down the coast at roughly the same time, in the same kind of synchronous gushes that Adams noticed in the main Cascadia channel? The question could be answered using oilfield techniques well known to many of the faculty at OSU, where petroleum geology was a significant part of the academic program. As Chris Goldfinger likes to tell it, oil drillers have been doing this sort of thing for years.
Again he pointed to the overhead map, zeroing in on the offshore region near the Oregon–California border. “These channel systems don’t have the same sources and they’re even further apart [than the channels that Adams studied]. They don’t have anything in common,” he said. Oregon’s Rogue River, for example, flows directly from Crater Lake—the former volcano Mount Mazama—to the sea with no downstream connection to the Cascadia channel. There is no confluence of canyon heads and tributaries that would physically link the Rogue turbidites with the others farther north.
The stratigraphic patterns in all the samples, however, did look very nearly identical. The relative age, thickness, and spacing of the alternating bands of turbidite sand, silt, and gray-green ocean mud were the fingerprints of Cascadia’s history. Goldfinger and Nelson used a process known as wiggle-matching to make a detailed, layer-by-layer examination of all the minute gradations of muck that had been laid down on the ocean floor.
“Correlating the wiggles” in core samples from the entire length of the Cascadia Subduction Zone took quite a while, but the match-up was pretty convincing. “Even though this hadn’t been used before in paleoseismology,” Goldfinger said, “this is basic, subsurface oilfield geology. This is how oil deposits are tracked from one place to another, because turbidites make good oil reservoirs of sand. So correlating turbidites from place to place is something that hundreds of people do on a daily basis all over the world. We’re just taking that technique and applying it to a different purpose here.” Instead of chasing oil, they were chasing earthquakes.
By 2003, when Goldfinger and Nelson published another paper based on more turbidite data, the tide of opinion had turned. The number of doubters had dwindled. The onshore r
ecord of sunken marshes, drowned trees, and sheets of tsunami sand had been accepted by most as evidence of Cascadia’s past quakes. The evidence from offshore turbidites was still circumstantial, although the case was strengthened now that the work of John Adams had been redocumented, confirmed, and extended. Still, there was a lot to be done.
Goldfinger knew there were not enough data yet to establish absolute numerical ages for each of the offshore events—progress was slow because radiocarbon dating was difficult to do with so little plankton or other biotic material available in the deep-sea samples—so it was initially hard to correlate the turbidite record with the land-based data. There were enough similarities in the offshore core patterns, however, to establish “lateral continuity” of the turbidite layers. Meaning turbidite bed number three from a core taken off Vancouver Island was probably in the same regionwide stratigraphic layer as turbidite bed number three from a core taken near the California border.
Whatever triggered the offshore landslide up north presumably also triggered simultaneous landslides hundreds of miles to the south. The exact date may not be known, but in all likelihood the matching turbidites made a synchronized plunge downhill. And the only force strong enough to rattle the sea floor all the way from Vancouver Island to California would have to be a very large temblor.
As a control sample, to see what the ocean mud looked like beyond the end of Cascadia’s fault, Goldfinger and Nelson took their 1999 cruise ninety miles (150 km) south of Cape Mendocino, where they collected three more cores in the Noyo channel, an offshore canyon that drains the northern California continental margin adjacent to the San Andreas fault. They discovered a similar cyclical pattern of sandy turbidite flows interlaced with ocean mud, the main difference being that here the landslides seemed to happen more often. In the last ten thousand years there appeared to be thirty-one events—most likely caused by San Andreas ruptures big enough to trigger offshore landslides. It seemed that California’s most famous fault was causing the same kinds of offshore landslides as the Cascadia Subduction Zone.
Cascadia's Fault Page 24
