March 11, 2011, and the voice that stayed at the microphone

I want to tell you about March 11, 2011, but I have to be honest about where I was standing when it happened, which was nowhere near it. I was twenty-two, in Missouri, living an ordinary American life. The earthquake hit the other side of the planet, and I learned about it the way nearly everyone outside Japan did, through a screen, watching water do something water is not supposed to be able to do. It has stayed with me for fifteen years, and it is a real part of why Japan eventually became one of the central things in my life. So I want to get it right.

What actually happened

At 2:46 in the afternoon, about 130 kilometers off the coast of Sendai, a slab of the Pacific seafloor that had been quietly straining for centuries finally let go. This is what a subduction zone does: one tectonic plate grinds down underneath another, sticks, loads up like a bent ruler for hundreds of years, and then releases all at once. The release that day was a magnitude 9.0, the most powerful earthquake ever recorded in Japan and one of the most powerful measured anywhere since we have had the instruments to measure. The seabed lurched, in places by tens of meters, and shoved an unfathomable volume of ocean upward and outward.

The shaking itself, as violent as it was, is not what killed most of the people who died. The water did that. The displaced ocean spread out as a tsunami, and when it reached the shallow coast it stood up into walls of water that, in the worst spots up in Iwate, ran nearly 40 meters up the land, higher than a ten-story building. By the time it was over, somewhere around 18,500 people were dead or never found, whole towns were gone, and three reactors at the Fukushima Daiichi plant had melted down. It is one of the worst days, by almost any measure, in modern Japanese history.

Here is the detail that matters most for the rest of this story. Japan has the best earthquake and tsunami warning system in the world, and that day it was not enough. The first tsunami warning went out fast, generated automatically in about three minutes from the quake’s estimated size and location. But that first automated estimate was far too small. It told some stretches of coast to brace for something around three meters. What came for them was many times that. The system, and every person trusting it, started from a number that was already wrong, and the real wave was faster than anyone could recompute.

The voice over the loudspeaker

Minamisanriku was a fishing town of about 17,000 people in Miyagi Prefecture. Like a lot of coastal towns, it had a public address system, loudspeakers wired across the town so the local government could reach everyone at once. And it had a young woman named Miki Endo, 24 years old, who worked in the town’s Crisis Management Department. Part of her job was to be the voice on those speakers in an emergency.

When the warning came, she got on the microphone in the disaster-prevention building, a three-story structure near the water, and she started telling her town to run. Get to high ground. A large tsunami is coming. She said it over and over, for about thirty minutes, as the situation went from bad to unimaginable. She kept going as the water arrived. The wave that reached Minamisanriku was around 15.5 meters. The building she was broadcasting from had a roof about 12 meters up, and the water went over it. People who had climbed to that roof clung to the antenna tower as the sea swallowed everything below. Of the roughly thirty who made it up there, only about ten lived. Forty-three people died at that building. Miki Endo was one of them. Her body was not found for weeks.

More than half of Minamisanriku evacuated and survived, and her voice is a large part of why. The skeletal red frame of that building, stripped to its steel by the water, became one of the defining images of the whole disaster, and she became a national symbol of a very specific kind of courage: the person who stays at the post so that everyone else can leave theirs.

”Why didn’t they just automate it?”

When I have talked about Miki Endo online over the years, the question that comes up, every single time, is some version of this: it is 2011, not 1911, so why was a human being standing at a microphone at all? Why not automate the warnings and let her go home?

It is a fair question with a frustrating answer, and the answer is the whole point.

Japan sits on top of several plate boundaries and feels earthquakes constantly, a few noticeable ones on a typical day, and any of them could in theory start a tsunami. The country’s early-warning system, rolled out to the public in 2007, is genuinely a marvel. It works by racing physics against itself. An earthquake sends out fast, weak P-waves first and slower, destructive S-waves a few seconds behind. A dense national network of seismometers catches those first P-waves, the computers estimate where and how big in a second or two, and a warning goes out ahead of the hard shaking. Sometimes that is only a handful of seconds. Sometimes those seconds are a life.

But it is not magic, and it is not perfect. It gets fooled. Lightning, construction blasts, two earthquakes happening at once, all of it can trip a false alarm, and a few times a year, even now, the system tells a region to brace for a big one that never comes. Tsunami warnings are automated too, but as 2011 showed in the cruelest way, the automation can tell you a wave is probably coming without being able to tell you, in the moment that counts, exactly how big this particular one will be. A computer that cries wolf gets ignored. A computer that under-calls a once-in-a-millennium wave gets people killed.

So when people say the technology “wasn’t there yet,” this is what it actually means. Not that they lacked tape recorders. Of course a town could play a recording. It means that a fully automated system trustworthy enough that no human is needed in the loop did not exist for a small town in 2011, and arguably still does not. And a recording cannot do the thing Miki Endo did. A recording cannot read out the updated wave height. It cannot say it is worse than we told you, go higher, go now. And it cannot carry what a live human voice carries. The people who survived talked about that voice, a real person clearly telling them this was not a drill, and that is what got them moving instead of waiting to see. You can dismiss that as sentiment. I think it is just true.

If you want proof the system has come a long way since, look at the 2019 Yamagata earthquake, a magnitude 6.7 off the Sea of Japan coast. The automated system issued a tsunami advisory within minutes and helped move people to safety, and the actual wave turned out to be tiny, around ten centimeters. To a cynic that is the system being wrong. To me that is the system being exactly right: erring toward caution, buying the evacuation, costing nothing but some inconvenience. That is the world Japan built after Tohoku, with enormous money and study poured in specifically because of that day. It is not the world Miki Endo was working in.

The argument I keep having

I will be straight about something, because it is part of why this sticks with me. Every so often, when this story comes up, someone wants to turn it into an argument that Miki Endo was a sucker. The take goes something like: the “fear in her voice” business is nonsense, she should have hit play on a recording and walked out, and the whole idea of sacrificing yourself for the group is a scam that governments and companies push to keep people in line.

I find that take cynical, and honestly a little sad, and I do not think it survives contact with the facts.

We have no evidence anyone forced Miki Endo to die at that microphone. There was no order to stay until the water took her. What there was, was a 24-year-old with a job to deliver information to her town, who, like everyone around her, had no idea in the moment how enormous this was going to be. Nobody did. By the time it was clear, it was minutes away, not hours. And then she was left with the same choice everyone in that building faced: keep doing the one useful thing she could do, or run for the roof, where most of the people who ran still died. She chose to keep talking. A tape would not have made that choice, because a tape does not know what is happening. She did, and she stayed.

I do not know whether staying was the “right” call in some cold accounting, and neither does anyone else, and I think pretending otherwise is its own kind of disrespect. What I know is that she made a choice, in the worst few minutes of her life, to spend that time on other people. The decent response to that is to honor it, not to launder it into a lesson about how looking out for each other is for fools. This had nothing to do with governments or corporations. It was one person choosing her neighbors. That instinct, to put the group ahead of yourself, runs deep in Japan, and I have never once seen it as a con. It is one of the better things people are capable of.

This was a one-in-a-thousand-year event. On March 11, 2011, the dice rolled wrong for a lot of people. She is not a cautionary tale. She is the opposite.

Where it led me

That day planted something in me that did not fully grow for years. Over time my interest in Japan turned into a real and lasting part of my life. I started studying the language, and eventually I got to go there. But one specific thread of it traces straight back to this: a kind of obsession with how a country watches its own ground, in real time, and tries to give its people those few seconds.

That obsession pulled me into a small, scattered online community of people who follow Japan’s live earthquake data, the same feeds the warning system runs on, and who build tools to visualize it. The best of those tools is jQuake, a desktop program that pulls Japan’s real-time seismic data and early-warning alerts and shows you, the instant the network detects a quake, where it is and how hard it is about to hit. It was created and mostly built by one person, a developer who goes by François LN, and the real bulk of the work is his.

Here is the part I am quietly proud of. A handful of us, scattered across the US, Germany, Australia, and elsewhere, had each been tinkering on our own versions of this kind of software. We could easily have stayed rivals, each hunched over our own little project, splitting a tiny audience among ourselves. Instead we threw in behind Francois, who had built the best version of it, and we set out to make his the one that won. Everybody took a role. We helped with testing, with getting releases out the door, and, very much in character for me, with hosting and building the website. He gets the credit, and he should. My job was to help the thing succeed, doing the parts I happened to be good at, and I did. Being part of something whose entire reason to exist is to hand people those few seconds of warning, the very thing Miki Endo was trying to buy her town by hand, means more to me than almost anything else I have helped build.

Because that is the whole hope, in the end. We build the automation so that one day no one ever has to stand at a microphone as the water comes. We are not there. We may never fully get there. But every second of warning is a person who gets to run. Miki Endo bought her town thirty minutes with nothing but her own voice. The least the rest of us can do is keep building the thing that might, someday, do it for her.