What happened last March 11 wasn’t supposed to be possible. The seismic hazard maps didn’t entertain the idea of a 9.0 magnitude earthquake off the Tohoku coast of Japan.

But the Earth paid no heed to scientific orthodoxy. A massive slab of the planet’s crust lurched 180 feet to the east. It rose about 15 feet, lifted the ocean and tipped the Pacific’s waters onto the Japanese coast.

The quake and the resulting tsunami killed about 20,000 people, wiped out entire towns and triggered power outages and then meltdowns at the Fukushima nuclear power plant.

It also humbled the scientific community.

Since 2004, earthquake scientists have been caught off guard, or to some extent consternated, by huge killer earthquakes in the Indian Ocean, Haiti, China, Japan and New Zealand.

Now the geologists are in a state of soul-searching. They want to do better, get smarter and help save lives on a shaky planet. But they feel chastened by what happened in Japan and are reexamining their basic assumptions about earthquakes.

Humans can be gifted at perceiving patterns in nature. We can also imagine patterns that do not exist. We can focus our attention on too narrow a frame. It is the special challenge of earthquake scientists that they must contend with terrestrial forces that exist outside the frame of human lifetimes, or even the lifetimes of entire civilizations. Some geologic faults may endure thousands of years of strain before a catastrophic rupture.

“This is a humbling field. If you want to be smug, don’t be an Earth scientist, and certainly don’t be an earthquake researcher,” said Ross Stein of the U.S. Geological Survey.

Seismic hazard maps typically show where earthquakes are most likely to occur over a certain period of time, and the expected maximum intensity. But critics say these maps merely describe what has happened before and have virtually no predictive value. They call it “Texas sharpshooting” — shooting the side of a barn and then drawing a bull’s-eye around the bullet hole.

Defenders of the maps argue that they are better than nothing. Policymakers have to decide where to put resources. Which locations have older buildings that are most in need of seismic retrofitting? How high should a tsunami wall be?

Public officials may say, in effect, we know this map is probably wrong, but we still need it for planning purposes.

If there’s one obvious change in attitude among geophysicists since Japan’s 3/11 disaster, it’s a recognition that huge earthquakes can potentially happen on any subduction zone — any of the places where one tectonic plate is diving beneath another. For many decades, the presumption had been that some subduction zones were significantly more likely to generate a great quake.

As Thorne Lay, a seismologist at the University of California at Santa Cruz, noted in a commentary just published in the journal Nature: “[W]e must allow for the possibility of larger earthquakes in regions where we thought that potential did not exist. That is one difficult lesson that we can consider learned.”

Lay and Stein have both worked on GEM, the Global Earthquake Model, a new project based in Italy set up to create a set of detailed, more precise seismic hazard maps for people around the world, particularly in poor countries.

Critics have their doubts.

“It’s almost impossible to make a sensible earthquake hazard map,” argues Northwestern University geophysicist Seth Stein (no relation to Ross). “The onus is on Ross to prove that his complicated maps work better than the old maps — and that they work better than random.”

Robert Geller, a University of Tokyo geophysicist, said the standard maps “are simply wrong” and are based on the false premise that earthquakes repeat themselves at more or less regular intervals.

“We call this the ‘whack-a-mole model’ of earthquake hazard mapping. The mole will come up the same hole that it went down,” Seth Stein said. And that’s rarely the case.

Geller and Seth Stein contend that the seismic hazard maps ­haven’t shown themselves to offer information about potential earthquake location and intensity that’s better than a random guess.

The U.S. government disagrees. David Applegate, associate director for natural hazards at the USGS, said the hazard maps in this country are incorporating data going back thousands of years in some cases. And Art Frankel, a USGS geophysicist who led the National Seismic Hazard maps program from 1993 to 2004, said the maps are useful for designing building codes.

“I don’t buy this idea that we don’t know anything and every place is the same hazard. We know a lot,” Frankel said.

Since the development of the theory of plate tectonics in the 1960s, scientists have a better understanding of why earthquakes occur along plate boundaries. The 1906 San Francisco earthquake, for example, released the strain that had built up as the North American plate and Pacific plate fitfully tried to slide past each other along a plate boundary, the San Andreas Fault. Scientists have tried to monitor the motion of tectonic plates and create maps showing where strain is building at a locked plate boundary.

But earthquakes remain fundamentally unpredictable and eccentric. Scientists were surprised by the location of the 9.1 magnitude earthquake in the Indian Ocean on Dec. 26, 2004, which generated a huge tsunami and took 230,000 lives. That portion of the subduction zone near Sumatra had been considered an unlikely source of a great earthquake.

Since then, more surprises. The 2008 earthquake in China’s Sichuan province, which killed 68,000 people, was in an area that “did not look like a very active region,” American geologist Peter Molnar said afterward. The 2010 earthquake in Haiti occurred along a fault generally considered less dangerous than another fault to the north. And New Zealand has had two significant earthquakes on unmapped faults in the past year and a half.

Surprises may be the norm for the seismic future, even in places considered hazard-free. The Virginia earthquake in August, which damaged the Washington Monument and the National Cathedral, is a reminder that the East Coast could be more vulnerable to tremors than most people realize.

“All of the region plausibly has within it faults and residual strains and stress accumulations from the history that has been experienced by the crusts,” Lay said Friday.

One area in the United States that is receiving increased attention is the Pacific Northwest. A number of popular beach towns, such as Seaside, Ore., are in the line of fire of the Cascadia subduction zone, said Patrick Lynett, a University of Southern California professor of civil engineering who is an expert on tsunami hazards.

The West Coast does not have evacuation towers, and residents facing an incoming tsunami would have to evacuate on foot to higher ground, he said, before the first waves arrived in 15 to 30 minutes. Lynett noted that Seaside has summer festivals in which thousands of people crowd the beach. “You could have a lot of really bad factors coming together,” he said.

The Cascadia subduction zone last ruptured in the year 1700. When is the next Big One? The answer, unfortunately, is somewhere off the coast, in ancient crust beneath the sea.