Was the First Magnitude Wrong, or Just Early?
Japan's national weather agency, the Japan Meteorological Agency (JMA), is the office that issues tsunami warnings there. A tsunami can reach the coast soon after an earthquake, so a late warning may give people little time to act. JMA says its first tsunami warning must go out about two to three minutes after the earthquake [1].
To make that deadline, the first warning rests on a magnitude calculated quickly with JMA's own formula. JMA calls it the JMA magnitude. Later, within roughly 10 to 20 minutes, it can estimate the moment magnitude, a measure of how much the fault moved [1].
Reports from that day say the first figure was 7.9, and that the number then rose to 8.4 and to 8.8 later the same day [5]. On March 13, JMA announced that after a detailed analysis it was changing the magnitude from 8.8 to 9.0, and it noted that this 9.0 was a moment magnitude [2].
So the numbers were not corrections of careless errors. They were steps: a quick estimate, then better ones as more evidence was analyzed. The rest of the article asks why the quick estimate came out so low.
Why Is a Fast Magnitude Hard to Get Right for a Huge Earthquake?
JMA states plainly that for an earthquake above magnitude 8 it cannot quickly work out the exact size [1]. Part of the reason is the time limit above. The moment magnitude uses long-period waves, which rise and fall more slowly than the first sharp shaking, and those records take longer to collect and analyze [1].
Imagine a very big drum: its first boom arrives quickly, while the longer rumble gives you more clues about its size. If you had to guess after the first boom, you might guess too small. (An earthquake is not a drum, but the comparison shows why the first signal may not tell the whole story.)
On March 11, 2011, a second problem made the guess even harder: some instruments could not record the strongest shaking at all.
What Does It Mean When a Seismometer's Record Is "Clipped"?
A seismometer and its recording system can handle shaking only up to a certain size. If the shaking is stronger than that, the line on the record hits the top and goes flat. This is called clipping, and the part that was cut off is lost. You cannot tell from the record how much stronger the shaking really was.
JMA later explained what happened in one of its own research offices. On March 11 that team tried a method designed for earthquakes far from Japan, and this time used Japan's own data to estimate the size of the earthquake. Their figures were M9.0 about 10 minutes after the earthquake and M9.1 about 13 minutes after. But the Japanese waveforms were clipped, so the team judged the values unreliable and did not pass them on to JMA's headquarters. JMA states that these values came from clipped data and are not correct values [3].
It is tempting to say, "They had the right answer after ten minutes and nobody listened." That reading is wrong. A number calculated from a clipped record is not trustworthy even if it happens to land near the final answer. To know the true size, someone needed records that had not maxed out.
How Did Waves Recorded Far Away Show the True Size?
On March 13, JMA studied foreign observation data again to see how the break in the Earth's crust had unfolded. It found that three huge ruptures had followed one another, in a more complicated pattern than usual. Re-analyzing with that in mind gave a magnitude of 9.0 [2].
According to a Japanese science news report from the day, what triggered the review was seismic waves recorded in Australia. After the first wave, which the 8.8 had been based on, another wave appeared. The report said the ruptures ran from off the Sanriku coast to off Fukushima and off Ibaraki [4]. Put simply, one break was not the end: the fault kept breaking in sections, and adding them up made the earthquake bigger than the first estimate.
That is the answer to the title's question for this earthquake. The number changed because the quick method had a time limit, the closest instruments clipped, and the full story only showed up in complete records from farther away. The next question is whether this is just a Japanese problem.
Do Agencies in Other Countries Also Change Magnitudes After Announcing Them?
Yes. The United States Geological Survey (USGS) says it often updates a magnitude in the hours, and sometimes days, after an earthquake. For a large earthquake outside the United States, its first estimate comes within about 20 minutes, using real-time data from the closest stations [6].
The reason sounds familiar. Some of the waves used in magnitude analysis can take more than an hour to travel around the Earth to stations far from the epicenter, so they cannot be included in the first number. Some stations also send their data late, and a few data arrive days afterward. As data come in, the magnitude is refined [6].
So a changing number is not a sign that scientists are careless. It shows that more evidence has arrived. A number labeled "preliminary" is a promise to keep checking.
How Did Japan Change Its Warnings After 2011?
JMA changed both the measuring and the wording. On the measuring side, it said it would use strong-motion seismometers that do not clip, with the aim of fixing the moment magnitude about 15 minutes after an earthquake [3].
On the wording side, JMA describes a rule in place since March 7, 2013. If the earthquake might be too large for the quick method to measure, the first warning uses the largest magnitude that could be expected in that part of the sea. Until a reliable magnitude is known, JMA uses a word such as "huge" instead of a number [1].
In other words, the system now plans for the first number to be uncertain, and puts the danger first. That seems a sensible way to handle a number that may change.
How Can You Spot a Preliminary Magnitude in the News Yourself?
Watch the words
Next time a large earthquake is in the news, look for words such as "preliminary", "initial", or "revised". Compare two reports from different hours and see whether the magnitude, depth, or time changed. If you are in the United States, you can compare an early news story with the USGS event page a day later.
Try the arithmetic
A whole-number increase in magnitude means much more energy is released. The sources above do not give the multiplier; look for a USGS or JMA explainer if you want to find it. Then ask why even a small change, like 8.8 to 9.0, is worth announcing.
Open the sources
The JMA announcements of March 13 and September 10, 2011 are short and explain the changes in plain steps. One question we could not answer from these sources: how soon can agencies give a reliable magnitude for a very large earthquake today? The sources here give targets, not results.