Readout · Series one · Japan
Readout: four Japanese earthquakes compared
- Context
- My own data story
- My role
- Designed and built it alone: the research, the data, the charts and the writing
- What I built
- Five earthquakes compared, with live charts and a dashboard, on US Geological Survey data
- Status
- Live. The data is a fixed snapshot, saved with how it was fetched
The events
Four earthquakes, five shocks
Japan has an earthquake of magnitude 6 or above about once every four weeks. (Magnitude measures how much energy an earthquake releases where the rock breaks.) Most pass with little notice. A few are remembered across the country, and I picked four of those to compare side by side. Kumamoto struck twice in two days, a foreshock and a mainshock, and the catalog records both, so five shocks sit in the charts.
They are Kobe in January 1995, Tohoku in March 2011, the two Kumamoto shocks 28 hours apart in April 2016, and Noto on the afternoon of New Year’s Day 2024.
They happened within thirty years of each other in one country, but they are very different events. The largest released about twenty thousand times the energy of the smallest, and their depths differ by a factor of three. What each one did to the people above it doesn’t follow neatly from either number. The rest of this page works out why.
What this means: The five are spread along the length of the country. Kumamoto is on Kyushu in the south west and Tohoku is off the north east coast, about 1,200 kilometers apart. The two Kumamoto shocks were 5 kilometers apart, so they share one pin. Kobe and Kumamoto broke directly beneath cities, while Tohoku broke well out to sea. That difference explains a lot in the charts below. The faint marks behind them are every other event of magnitude 6 and above in the same fifty years.
Size
How big each one was
Magnitude is easy to misread because each whole step up the scale means about 32 times more energy. So magnitude 6.2 and magnitude 9.1 look only three apart, but in energy they are more than twenty thousand times apart.
Tohoku is in a class of its own. It released about 190 times more energy than the other four put together. But as the next sections show, it doesn’t come top on every measure.
What this means: Each half ring is one earthquake, drawn out from the same starting point. The size of each ring uses a logarithmic scale, which squeezes big differences down so they fit on one chart. Four of the five sit close together. Tohoku sits far outside them, at more than twenty thousand times the energy of the smallest, and the gap is still that wide even after the scale has squeezed it.
Depth
How deep they were
People feel an earthquake at the surface, not where the rock breaks. The deeper it starts, the more its energy spreads out before it reaches them. So a moderate earthquake a few kilometers under a city can shake that city harder than a much bigger one forty kilometers down and far out to sea.
That explains much of the difference between Kobe and Tohoku. Kobe broke in the earth’s crust, 21 kilometers down and almost directly under a dense city. Tohoku was a subduction earthquake offshore, where one tectonic plate slides beneath another.
What this means: Kumamoto and Noto are both listed at exactly 10 kilometers. That is the default value the catalog (the USGS list of recorded earthquakes this story is built on) enters when it can’t work out the real depth. So two of the five depths here are not measurements, and any comparison that relies on them is relying on a placeholder.
Shaking
How hard the ground shook
Shaking intensity measures what happened at the surface, not at the source. It is the closest figure in this data to what someone standing in the street actually felt.
Plot it against magnitude for every event in the study area around Japan, and the link is weak. The points form a cloud, not a line.
The two strongest shakings in fifty years came from a magnitude 6.8 and a magnitude 6.9, both shallow and almost directly under cities. The magnitude 9.1 isn’t in the top five.
What this means: The left side ranks the five by size. The right side ranks them by how hard the ground actually moved, and a line joins each earthquake across. If size decided shaking, every line would be flat. None of them is. Tohoku is the largest earthquake in the whole record but fourth of five for shaking. Kobe is fourth for size and first for shaking. The All events switch asks the same question of every event in the study area that has an intensity value.
Significance
How the catalog ranks them
The catalog gives each event a significance score. It combines magnitude, shaking intensity, reports from people who felt it, and estimated impact. That makes it the closest thing in the data to a ranking by how much an event mattered.
On that score, Kobe comes last of the five.
The reason is in the inputs, not the earthquake. In January 1995 the public had no way to file felt reports online, and the alert product behind the impact estimate didn’t exist yet. So two of the four inputs are blank. A blank is not a zero, but a score made by adding up its inputs can’t tell them apart.
This affects any comparison across decades. Older events score lower because less about them was recorded.
What this means: Kobe scores lowest of the five, even though it caused the second hardest shaking in the whole fifty year record. Two of the four inputs to its score were never recorded.
Summary
What the comparison shows
Size, depth, shaking and significance put these five shocks in four different orders. Tohoku is first by energy but third by significance. Kobe is last by significance but second for shaking across the whole fifty year record. And two of the five depths aren’t measurements at all.
None of that is a fault in the catalog. Each of those four columns records something different, so each gives a different answer about the same event.
The study area holds 629 events in total, and the dashboard behind this story shows every one of them.
Sources
Where the data comes from
All the data comes from two USGS catalogs, downloaded from the USGS FDSN event service (a standard web service for earthquake data). Each covers fixed start and end dates and was fetched once, on 3 September 2026. The query URLs, the time of the fetch and the rules for handling missing values are saved beside the data. Nothing here updates live, on purpose: a number from a live feed is only true for the moment it was read, and nobody can check it later.
Magnitude 6.0 and above, 1 January 1976 to 3 September 2026. 7,182 events, of which 7,128 are earthquakes. The largest is magnitude 9.10, the Tohoku earthquake, and the deepest is 688.0 km. 534 of the earthquakes are in Japan, and 29 of those carry the tsunami warning flag.
Every recorded event, 4 August to 3 September 2026. 11,247 events, of which 10,993 are earthquakes.
The five named events. Choosing which five to feature was my call, and the choice is written down in the script that builds the data. Everything else about them comes from the catalog. Each one is looked up by its USGS event ID, and its magnitude, depth, location and time are copied from there, so a label on a chart can never drift from the data under it. Eight facts the text states are checked before the data file is written. Two examples: the local time in Japan was 05:46 for Kobe and 14:46 for Tohoku, and the tsunami column reads zero for Tohoku. The short note beside each event is context I wrote by hand, and the file marks it as having no source.
Precision. Depth and magnitude are kept exactly as the catalog publishes them. Rounding depth to one decimal place once moved four events onto exactly 10.0 km and changed the number of events at that depth from 1,189 to 1,193: a small difference, but a wrong one. So the build that produces this page’s data now checks fourteen counts against the source files and stops if any of them disagree.
More case studies
Three more projects about understanding a complex system quickly and correctly: finding the cause of an outage, interview scores you can question, and checking a website against accessibility law.