The Repository Hook Echo

1906 San Francisco Earthquake: A Road Moved 20 Feet

October 27, 2026 · Runtime 15:13 · Watch on YouTube

The San Francisco earthquake of 18 April 1906 offset a road near Point Reyes Station by 20 feet. A year later, the Coast and Geodetic Survey re-measured its network of hilltop stations and found the ground had moved miles from the fault too: the Farallon lighthouse, 23 miles west of it, by 5.8 feet. From those surveys Harry Fielding Reid of Johns Hopkins worked out the theory of elastic rebound, still the model of the earthquake cycle. We redid the arithmetic from the 1908 tables: the survey and the tape measure on the fences give the same slip.

Chapters

  1. 0:00 Market Street, after the fire
  2. 0:46 Four days before
  3. 1:01 5:12 AM, 18 April 1906
  4. 1:42 Three days of fire
  5. 2:47 The seismographs go off the scale
  6. 3:53 The commission and the fences
  7. 5:43 The survey between hilltops
  8. 6:36 The lighthouse that moved
  9. 7:46 Our measurement: survey against tape
  10. 9:30 Harry Fielding Reid
  11. 10:31 Elastic rebound
  12. 11:30 A line of piers
  13. 12:12 Where the 7.9 comes from
  14. 12:54 The same island, by GPS
  15. 13:57 What the record supports
  16. 14:52 Next: the EF scale

Transcript

Market Street, after the fire 0:00

The San Francisco earthquake of 1906 moved a road 20 feet. It struck at 5:12 in the morning of April 18th, and the fire that followed burned for three days.

We read the survey that measured it, published in 1908, and we redid the arithmetic ourselves, from the surveyors' own tables. On an island twenty-three miles west of the fault, a lighthouse had moved almost six feet.

No one on that island could have seen it move. The surveyors found it a year later, in the angles between hilltops. This episode follows those measurements: the seismographs that were thrown off the scale, the fences, the survey, and the geologist who read them and explained what an earthquake is. His explanation is the one we still use.

Four days before 0:46

Four days earlier, on April 14th, the Miles brothers filmed this. Market Street, from the front of a cable car, heading for the Ferry Building. They left the city with the film. Their studio burned in the fire.

5:12 AM, 18 April 1906 1:01

At almost exactly 5:12, a foreshock was felt across the Bay Area. Some twenty to twenty-five seconds later, the great earthquake broke loose, with its epicenter near San Francisco. The strong shaking lasted forty-five seconds to a minute. It was felt from southern Oregon to south of Los Angeles, and inland as far as central Nevada.

The fault broke for two hundred and ninety-six miles. It was the northernmost part of the San Andreas, from northwest of San Juan Bautista to the triple junction at Cape Mendocino. San Francisco sits near the middle of that line, and the shaking was just as severe in many places along it.

Three days of fire 1:42

In the city, the fire did more damage than the shaking. Soldiers and firefighters blew up buildings with dynamite to make firebreaks, with mixed results. It burned for three days, across four point seven square miles.

About twenty-eight thousand buildings were lost. Of roughly four hundred thousand people in San Francisco, two hundred and twenty-five thousand were left homeless. The damage was put at four hundred million dollars, in 1906 dollars, and only eighty million of it was from the earthquake alone.

The Army's relief report of 1906 counted four hundred and ninety-eight deaths in San Francisco. Later research by Gladys Hansen and Emmet Condon estimated more than three thousand. The Geological Survey now says the old figure of about seven hundred underestimated the loss of life by a factor of three or four.

In New York, the American printed that half of San Francisco was now in ashes. That is the story the world remembers. This episode is about a different record, one made with instruments.

The seismographs go off the scale 2:47

The shaking was so strong that, in the words of the 1910 report, practically all seismographs, in any part of the world, recorded it. In Göttingen, in Germany, about nine thousand one hundred kilometers away, the record shows the first waves, then the slower shear waves. When the surface waves arrived, the instrument went off the scale.

Close to the fault, the instruments failed. At Berkeley, the recording plate of one seismograph was lifted off its bearings and stopped turning. The other one there hit its stops, and recorded only eleven millimeters of ground motion, far less than was actually experienced.

At the Lick Observatory on Mount Hamilton, the seismograph was started by the violent shock itself. Its vertical pendulum was so deranged that it caught, and could no longer swing. Its record, the report says, is simply a circle on the plate.

So the seismographs could say when, and roughly where. They could not say how far the ground had moved. For that, the scientists went out to the fault with tape measures.

The commission and the fences 3:53

Three days after the earthquake, Governor George Pardee appointed a State Earthquake Investigation Commission. Its chairman was Andrew Lawson, head of geology at the University of California. Its report, published in 1908 and 1910, is still the main record of the earthquake.

One of its members was Grove Karl Gilbert, a geologist of the U.S. Geological Survey. Near Point Reyes Station, north of San Francisco, the road running southwest across the valley had been offset 20 feet. This is his photograph of it.

Near Olema, on the Skinner ranch, Gilbert measured four things the fault had crossed. The fence south of the barn was offset fifteen and a half feet. A garden path, fifteen. A row of raspberry bushes, fourteen and a half. And the barn itself had been dragged sixteen feet.

Near Woodville, he photographed a fence that was broken and offset eight and a half feet. He lined up the camera with the straight part of the fence beyond the break, so that the bend would show as well as the offset. Counting the bend, the straight parts of that fence had moved about eleven feet.

Along the whole fault, the offsets measured on fences, roads and other structures ranged from a foot or less up to 20 or 21 feet. Near Point Arena, at the north, two fences on one ranch had moved sixteen feet each. In the southern quarter of the fault, the largest offset was about eight feet.

The report also said what a fence cannot measure. A fence records the surface, across a strip a few feet or a few hundred feet wide. The movement of the firm rock below, it said, must be at least that large. And no tape measure could tell how far the ground had moved a mile away from the fault.

The survey between hilltops 5:43

That question had an answer, because of an older instrument. Since 1851, the Coast and Geodetic Survey had been measuring the angles between hilltops and lighthouses around San Francisco Bay. Each set of angles fixes where every station stands.

From July 1906 to July 1907, John Hayford and A. L. Baldwin of the Survey measured the network again, from Monterey County north to Fort Ross. The area was a hundred and seventy miles long. They recovered sixty-one of the old stations, and worked out where each one stood now.

To say how far a point had moved, they needed points that had not moved. They chose two, both thirty-three miles east of the fault: Mount Diablo, and a peak called Mocho, near Mount Hamilton. Every displacement in their tables is measured from those two.

The lighthouse that moved 6:36

This is their table, as printed in 1908. The Farallon lighthouse, on an island twenty-three miles west of the fault, moved one point seven eight meters, five point eight feet, toward the west-northwest. Mount Tamalpais, four miles east of the fault, moved one point nine feet, toward the southeast.

Hayford and Baldwin summed it up in two findings. Points on opposite sides of the fault moved in opposite directions: to the east of it, southward, and to the west of it, northward. And the farther a point was from the fault, the less it had moved.

On the west side, twelve points about a mile from the fault moved nine point seven feet, on average. Seven points, at three and a half miles, moved seven point eight. The lighthouse, at twenty-three miles, five point eight.

On the east side, ten points at about a mile moved five point one feet. Three points, at two and a half miles, two point eight. Mount Tamalpais, at four miles, one point nine. Beyond that, to the east, they could not detect any movement with certainty.

Our measurement: survey against tape 7:46

Those numbers carry an assumption. If Mount Diablo and Mocho had moved too, every number in the table would shift. Lawson himself, in the same report, wrote that it would be better to assume the two sides had shared the movement about equally.

So we used the tables in a way that does not depend on that choice. Take one station west of the fault and one east of it, and subtract their movements. Whatever the fixed points did, the difference is how far the two sides slid past each other between those two stations.

Near Fort Ross, twelve stations sit within a few miles of the fault. We took every pair with both stations within two kilometers of it, one on each side. That gives twenty pairs. The slip between them comes out between eleven point two and fourteen feet.

Now the fences on the same stretch. On the Call ranch, a fence had moved twelve feet in total, across a zone four hundred and fifteen feet wide. On the Doda ranch, a fence was offset fifteen feet. A road, twelve feet three inches.

About thirty miles farther north, near Point Arena, two survey stations sit less than a mile from the fault, on opposite sides. Between them, we get fifteen point five to fifteen point seven feet. On the Fitch ranch, on the same stretch of the fault, two fences had moved sixteen feet each.

Those are two instruments with nothing in common. One is a tape stretched along a fence a few hundred feet long. The other is a set of angles read between hilltops miles apart. They agree to within about a foot. The crack in the fence was not a surface effect: the whole strip of ground had moved.

Harry Fielding Reid 9:30

The second volume of the report, published in 1910, was written by Harry Fielding Reid, a professor of geology at Johns Hopkins University. Reid asked what the surveys meant.

He had three surveys to compare, not two: 1851 to 1865, 1874 to 1892, and 1906 to 1907. And the older ones showed that the ground had started to move long before April 1906.

Between the old surveys, the land on the two sides of the fault had already been shifting, slowly, almost parallel to it. By the 1880s, Reid calculated, two thirds of the stress that caused the rupture had already accumulated.

Going back to the surveys of the 1850s and 1860s, nearly half of it was already there. It seems not improbable, he wrote, that the strain was accumulating for a hundred years. He added that there was no reason to suppose it had built up at a uniform rate.

Elastic rebound 10:31

His explanation was this. The crust on the two sides of the fault is being slowly displaced, and the rock along the fault holds, and bends. When the strain becomes greater than the rock can endure, it breaks, and the strained rock rebounds under its own elastic stresses. He called it elastic rebound.

This is his drawing of it, made from the survey's own averages. The displacements are drawn a thousand times larger than the distances. The movement is greatest at the fault, and it fades away on both sides, in opposite directions.

Reid also tested the idea on a bench. He set a sheet of stiff jelly, about two centimeters thick, between two pieces of wood, drew a straight line across it, and cut it along the middle. He moved one board a centimeter while holding the cut closed, and the line bent. When he let go, the two halves slipped, the line straightened, and the slip was equal to how far the board had moved.

A line of piers 11:30

Then Reid wrote down how the next one might be seen coming. Build a line of piers, about a kilometer apart, across the fault, and measure their distances and directions from time to time. If the ground becomes strained through an angle of about one part in two thousand, he wrote, we should expect a strong shock.

The commission did set concrete piers on the fault, at Olema and at Crystal Springs Lake, four at each place, two on each side. Their iron caps read: To measure earth movements, 1906. The piers at Olema are on the Skinner ranch, where Gilbert measured the fence. Reid noted they were placed to measure slips on the fault, not to predict the next earthquake.

Where the 7.9 comes from 12:12

Those survey angles still decide how big the earthquake was. The traditional magnitude, eight and a quarter, comes from Richter, in 1958. Re-reading the seismograms from twelve observatories gives seven point seven. Re-analyzing the survey data from before and after 1906 gives a moment magnitude of seven point nine, the one the Geological Survey uses today.

The same survey models put the slip at depth near Point Reyes at about twenty-four feet, and up to twenty-eight feet near Shelter Cove, at the northern end. Over the southern hundred miles, it averaged eight to twelve. The fences, as the report warned, showed less than the rock below.

The same island, by GPS 12:54

Reid's line of piers exists today, as GPS. A receiver on Southeast Farallon Island, about three hundred meters from the 1906 lighthouse station, has recorded its position since 1994. Another sits on Mount Diablo. From the Nevada Geodetic Laboratory's velocities, we get the island moving twenty-nine point nine millimeters a year relative to Mount Diablo, toward the north-northwest.

And the two points the 1906 survey held fixed? Today, Mount Diablo, and Mount Hamilton, eleven miles from Mocho, move together, within a quarter of a millimeter a year of each other. Relative to the interior of North America, both are moving about eighteen millimeters a year.

At today's pace, the island moves as far from Mount Diablo as it jumped in 1906 in about sixty years. Since 1906, the total is about three and a half meters. Not all of that is on the San Andreas: the Hayward fault and others lie between the island and the mountain.

What the record supports 13:57

Here is what the record supports. On April 18th, 1906, the ground on the two sides of the San Andreas slid past each other along two hundred and ninety-six miles, by up to twenty feet at the surface. Fences measured it at the crack, surveys measured it miles away, and the two agree. And the surveys from the decades before show the ground had been bending before the fault broke.

Here is what it cannot say. Whether the strain built up at a steady rate; Reid himself said there was no reason to think so. How many people died; the count is still an estimate. And when the next great earthquake on this part of the fault will come.

One question we could not close. The report gives one fence at San Andreas Lake as ten point nine feet in its chapter on the peninsula, and sixteen point nine feet in its summary. If you know which number is right, tell us in the comments.

Next: the EF scale 14:52

Next, a scale that turns damage into wind. The Fujita scale, from 1971, and the Enhanced Fujita scale that replaced it in 2007: twenty-eight kinds of damage, read as a wind speed that no instrument measured. That is the next episode. The EF scale.

Description and sources

The State Earthquake Investigation Commission (Lawson, ed., 1908; Reid, 1910) measured offsets on fences and roads along the San Andreas: 20 feet on the road southwest of Point Reyes Station, 15.5 feet on the Skinner ranch fence at Olema, 16 feet on two fences of the Fitch ranch near Point Arena. Hayford and Baldwin's re-triangulation of 1906-07 showed that the two sides moved in opposite directions, less the farther from the fault, measured from Mount Diablo and Mocho, assumed fixed. Our measurement from their Table 1: the slip between stations on opposite sides of the fault, which does not depend on that assumption, is 11.2-14.0 feet near Fort Ross (20 pairs; fences 12 and 15 feet) and 15.5-15.7 feet near Point Arena (fences 16 feet). Comparing the surveys of 1851-65, 1874-92 and 1906-07, Reid concluded the strain had been accumulating for decades and the rupture was an elastic rebound. The modern Mw 7.9 comes from the same geodetic data (Thatcher, Marshall and Lisowski, USGS). Today the GPS station on Southeast Farallon Island moves 29.9 mm a year relative to Mount Diablo (Nevada Geodetic Laboratory MIDAS velocities; our subtraction).

PRIMARY SOURCES

ARCHIVE PLATES — public-domain documents, with author, date, archive and licence.

  • Fault line that caused the San Francisco earthquake (Popular Science M | Popular Science Monthly (1906) | public-domain

... and 9 more plates, each credited on screen in the video and listed in this project's manifest.

Hook Echo reads a disaster through the instruments that recorded it: radar, sensor, survey and report, with the primary sources above.

About the material. Film: San Francisco in April 1906, from 1906 prints in the Prelinger Archives on archive.org (public domain in the United States by age), including the Miles Brothers' A Trip Down Market Street (14 April 1906) and an Edison newsreel. Report pages, plates and figures are reproduced from the Report of the State Earthquake Investigation Commission (Carnegie Institution of Washington, 1908 and 1910; public domain), including G. K. Gilbert's photographs. The Göttingen seismogram is from the U.S. Geological Survey; the portrait of Reid, from the Bibliothèque nationale de France. Our measurement uses the displacement table of Hayford and Baldwin (1908, Table 1, read from the page image) and the MIDAS GPS velocities of the Nevada Geodetic Laboratory (Blewitt et al., 2016). Every diagram is drawn in code from the sources listed. Nothing in this video is AI-generated imagery.

#HookEcho #SanFrancisco1906 #Earthquake