COPALIS BEACH, WASH. — When Japan issued its first-ever “megaquake” warning last week, Harold Tobin, Washington state’s seismologist, was watching carefully.

The advisory came after a 7.1-magnitude earthquake struck the southern island of Kyushu. Although that shaking caused little major damage — the biggest tsunami wave it produced would have risen up to your knee — it wasn’t the main worry.

Rather, seismologists were concerned that the quake would create stress that could trigger a bomb ticking offshore: Japan’s Nankai trough, likely the country’s most dangerous fault. The subduction zone has the potential to generate 100-foot-tall tsunami waves and kill nearly a third of a million people, according to Japanese government estimates.

Did the smaller quake mean that the “big one” was on the doorstep? No one could say for sure, but the odds were suddenly higher — if only by a few percentage points.

“Exactly what might keep me up at night,” Tobin said, if it were happening on the U.S. West Coast.

In Japan, the advisory prompted officials to close beaches, cancel fireworks celebrations and slow trains. People rushed to stock up on emergency supplies.

In the U.S., Tobin said, “we don’t have such a protocol.”

We do, however, have a similarly dangerous fault: the Cascadia subduction zone.

A magnitude-9.0 earthquake on the Cascadia fault and the resulting tsunami would kill an estimated 14,000 people in Oregon and Washington, according to the Federal Emergency Management Agency.

But if a smaller quake like the one Japan just saw happened near Cascadia, seismologists would have to decide on the fly whether and how to alert the public.

It’s the scenario Tobin has been thinking about for years: If he finds clues that a devastating earthquake is more likely, even just slightly, what warrants sounding the alarm? If the odds say you would be crying wolf — should you?

“You don’t want a mass evacuation panic that’s not warranted, but you want people not to go on their merry ways,” Tobin said.

His quandary is, in part, the product of this strange time in Tobin’s field: Researchers think they are homing in on the triggers or precursors of earthquakes in the world’s most dangerous seismic regions, but the science is far from settled. And even when the likelihood of an earthquake could be higher, the chances remain small. That leaves high-stakes questions about when to issue a warning.

On a chilly summer day in Washington state, Tobin and a dozen other scientists canoed up the Copalis River to a graveyard of cedar trees killed 324 years ago.

A kingfisher chittered and the wind sent shivers through tall, golden grass. It’s a peaceful place about a mile from the Pacific shore that tells the story of a violent day.

On Jan. 26, 1700, an earthquake on the Cascadia fault caused the forest to lurch downward by more than 3 feet. Soon after, a tsunami perhaps 100 feet high barreled through at 20 or 30 mph.

The scientists were visiting the forest to view the geologic evidence of the Cascadia quake in person. Occasionally, they’d hop out of their canoes, dig through the muck and pull out a 300-year-old pine cone as evidence.

Experts know the earthquake was at least a magnitude-8.7, because that’s how powerful it had to be to send the wave across the world that was documented in Japan.

“Some of the very best written records of our tsunami in 1700 come from Nankai,” said Brian Atwater, a USGS geologist emeritus who led the canoe flotilla. Atwater has used those Japanese records, along with plants buried in tsunami-deposited sand and dates from the rings of the Washington cedar trees, to piece together that tsunami’s story.

Research by USGS geophysicist Danny Brothers indicates there have likely been at least 30 large earthquakes over the last 14,200 years in sections of the Cascadia subduction zone, which runs along the U.S. West Coast from Northern California to northern Vancouver Island. A large earthquake there can be expected at least once every 450-500 years, on average.

But for years, Cascadia has remained quiet; some scientists say that’s because much of it is “locked” and building stress. When it rips, a chunk of the seafloor will lurch forward — perhaps by dozens of feet or more. The vertical displacement of the seafloor will send a tsunami toward shore.

“It’s going to be the worst natural disaster in our country’s history,” said Robert Ezelle, the director of Washington state’s emergency management division.

For seismologists, the key question now is how to forecast this future violence. Fast-developing research is hinting that faults like Cascadia and Nankai might send out warning signals: a smaller quake as a foreshock, or a subtle groan only detectable by sensors, which scientists call a slow-slip event.

In Tobin’s nightmare scenario, the Cascadia fault suddenly issues that type of groan. Then — what to do?

If a major Cascadia quake were to hit, more than 100,000 people would be injured, projections say — assuming the quake hits when few people are at the beach. The shaking would last five minutes. Tsunami waves would batter the coast for 10 hours.

Inland hillsides would liquify, taking out roads and bridges. Some 620,000 buildings would be critically damaged or collapse, including an estimated 100 hospitals and 2,000 schools.

“We’re unprepared,” Ezelle said frankly.

Washington state advises residents that they would likely have to fend for themselves and against the elements for two weeks.

“It’s going to be neighbors taking care of neighbors,” Ezelle said.

A map of the Pacific Ring of Fire — where tectonic plates converge to form subduction zones and volcanoes — leaves Ezelle particularly uneasy.

“Over the last 50 to 60 years, and you will see that every subduction zone fault has had a major rupture — with the exception of Cascadia,” he said.

Japan ended its “megaquake” advisory on Thursday, after no unusual activity was detected on the Nankai trough.

In a similar situation in New Zealand in 2016, things played out a little differently.

That November, the magnitude-7.8 Kaikoura earthquake rumbled off the east side of New Zealand’s South Island, killing two and causing more than a billion dollars in damage.

A day later, scientists noticed a few centimeters of movement near the shore of the North Island via satellite monitoring. Subtle vibrations were emanating from the Hikurangi Margin, a subduction zone and the country’s largest fault, which is directly under the capital city of Wellington.

It was a slow-slip earthquake, the sloth of the seismic world, kicked off by the Kaikoura shaking. Such quakes release their energy slowly over weeks or months and don’t cause perceptible shaking. Scientists first recognized their existence about two decades ago, thanks to advances in GPS technology.

Some scientists, like Tobin and geophysicist Laura Wallace, think these slow-slip events might sometimes precede big subduction zone quakes. Scientists recorded a slow-slip event in 2011 before the magnitude-9 Tohoku earthquake and tsunami in Japan, which killed more than 18,000 people and touched off the Fukushima nuclear disaster. A similar pattern played out in 2014, before a magnitude-8.1 earthquake in Chile.

Wallace, who was working for the New Zealand research institute GNS Science at the time of the 2016 quake, spent her waking hours scrambling to track the quake’s every movement, model risk and answer questions from the government.

“I don’t think I’ve ever felt such an immense load of responsibility,” Wallace said. “I was taking my dog with me to the office because if we had a big earthquake, I didn’t want to be separated from my dog.”

Wallace and her colleagues determined that the probability of a major earthquake was elevated as much as 18 times, and that the risk within a year was 0.6% to 7%. But the big one never materialized.

“Which of these slow-slip events are going to essentially trigger the next big one?” Wallace said. “It’s one of the most important problems we’re trying to understand.”

For the Cascadia subduction zone, gaining a better understanding of the warning signs requires more data on slow-slip events, improved mapping of the fault zone and an enhanced capability of monitoring faults on the seafloor.

Tobin was part of a team that recently mapped the Cascadia subduction zone in the greatest detail yet. They found the fault is separated into four sections, which could rupture all at once or individually in succession. The individual segments are capable of producing a magnitude-8 earthquake or higher.

Meanwhile, researchers are trying to bolster the offshore monitoring network for Cascadia.

Japan has a sophisticated array of seafloor sensors, but it’s “one of the few places that have those instruments,” said David Schmidt, a geophysicist at the University of Washington.

The U.S. lags on seafloor monitoring, but Schmidt and Tobin are part of a group that received $10.6 million in federal funding to add seismic sensors and seafloor pressure gauges to a fiber optic cable off the Oregon coast.

The devices will help keep tabs on Cascadia. If the data can help researchers learn about what’s normal for the fault, they might also be able to determine when it’s time to worry.