Name Class Date

EARTH EXPLAINED · LESSON GUIDE

The World Series Earthquake

Loma Prieta, 1989: why the worst damage came 60 miles from the fault

Video 8:34Level Grades 6–12 · Introductory university geologyStandards MS-ESS2-2, MS-ESS3-2, HS-ESS2-1, HS-ESS3-1, HS-PS4-1, HS-ETS1-3
Play fromfor school networks that block YouTube

Why did an earthquake centered in the mountains near Santa Cruz do its deadliest damage about 60 miles away, in San Francisco and Oakland?

At 5:04 p.m. on October 17, 1989, minutes before Game 3 of the World Series, a magnitude 6.9 earthquake broke a buried section of the San Andreas fault system beneath the Santa Cruz Mountains. This lesson uses it to explain how a sliding plate boundary stores strain and releases it suddenly (the elastic rebound theory that grew out of the 1906 earthquake), how a 1988 forecast singled out the segment that broke, and why the worst damage came far from the fault, where soft mud and artificial fill amplified the shaking and turned loose ground almost to liquid. It ends with what the region changed afterward, from freeway retrofits to earthquake alerts on phones, and with the USGS estimate of a 72% chance of at least one magnitude 6.7 or larger earthquake in the San Francisco Bay region between 2014 and 2043.

Captions (.vtt)Captions (.srt)

Learning objectives After this lesson, students can:

  1. Describe the San Andreas fault system as a boundary where the Pacific plate slides northwest past the North American plate, and name the main faults of the Bay Area.
  2. Explain the elastic rebound theory, using the ground offsets measured after the 1906 earthquake as evidence.
  3. Distinguish an earthquake forecast (a probability over a period of time) from a prediction (a date, place and size), and interpret the 1988 and 2014 Bay Area forecasts correctly.
  4. Explain how soft soil and artificial fill amplify seismic waves and how liquefaction damages buildings, and use these ideas to explain why damage was concentrated in the Marina District and along the edge of the bay.
  5. Describe the failures of the Cypress Street Viaduct and the Bay Bridge, and evaluate the engineering responses that followed: retrofitting older structures and replacing the Bay Bridge's eastern span.
  6. Explain how earthquake early warning works, why it can give some places tens of seconds of warning, and why places closest to the fault may get none.
  7. Evaluate the present hazard from the Hayward fault and explain why practicing Drop, Cover and Hold On makes sense in a region with a 72% chance of a large earthquake in 30 years.

Before watching

Key vocabulary ★ extension terms for older students

Fault
A break in Earth's crust where the rock on either side moves past each other. Movement on a fault can be slow and steady, or sudden, in an earthquake.
Tectonic plate
One of the large, rigid pieces of Earth's outer shell. In California, the Pacific plate moves northwest past the North American plate at about 2 inches (5 cm) a year.
San Andreas fault system
The family of faults that takes up most of the motion between the Pacific and North American plates in California. Around San Francisco Bay it includes the San Andreas, Hayward and Calaveras faults.
Transform boundary★
A plate boundary where two plates slide sideways past each other, neither created nor destroyed. The San Andreas fault system is the best-known example on land.
Elastic rebound
H. F. Reid's 1910 explanation of earthquakes: the rock on either side of a locked fault bends slowly as the plates move, storing strain, until the fault slips and the rock snaps back. It is still the basic model of the earthquake cycle.
Magnitude
A measure of an earthquake's size. Each whole step on the moment magnitude scale releases about 32 times more energy.
Hypocenter and epicenter★
The hypocenter (focus) is the point underground where the rupture begins; the epicenter is the point on the surface directly above it. Loma Prieta's hypocenter was about 11 miles (18 km) down.
Earthquake forecast
A statement of the probability that an earthquake of a given size will occur in a given area within a given time, such as '30% in 30 years'. It is not a prediction of a date.
Artificial fill
Land made by people by dumping or pumping sand, soil and debris into water or low ground. Much of the shoreline of San Francisco Bay, including the Marina District, is built on fill.
Liquefaction
The loss of strength of loose, water-saturated sand during shaking, so that it behaves like a liquid. Buildings on it sink or tip, and sand and water spurt up as sand boils.
Site amplification
The strengthening of shaking where seismic waves pass from stiff rock into soft soil: the waves slow down and their amplitude grows. In 1989, soft ground around the bay shook two to five times harder than nearby rock.
Fault creep
Slow, steady slip on a fault without earthquakes. Parts of the Hayward fault creep a few millimeters a year, bending curbs and walls that cross it.
Earthquake early warning
A system in which sensors detect the first, faster seismic waves near an earthquake and send an alert that can arrive seconds before the strongest shaking. California's system is built on the USGS ShakeAlert network.
P waves and S waves★
The two kinds of seismic waves that travel through the Earth's interior. P (primary) waves are faster and usually weaker; S (secondary) waves arrive later and do much of the damage.
Unreinforced masonry★
Older brick or stone construction without steel reinforcement. Such buildings, like those on Santa Cruz's Pacific Garden Mall, are among the most likely to collapse in strong shaking.
Seismic retrofit
Strengthening an existing structure so it can withstand earthquakes, for example by wrapping concrete columns in steel jackets and tying bridge spans together.

Chapter outline Select a time to play from there

0:005:04 PM

October 17, 1989: more than 60,000 fans are at Candlestick Park for Game 3 of the World Series between the San Francisco Giants and the Oakland A's when, at 5:04 p.m., the stadium starts to shake and the live broadcast cuts to static.

0:42The Fault Next Door

California sits on the boundary where the Pacific plate slides northwest past the North American plate at about 2 inches (5 cm) a year, mostly along the San Andreas fault system, which splits around the bay into the San Andreas, Hayward and Calaveras faults. The faults stay locked for decades or centuries, then release their stored strain all at once.

1:261906

The magnitude 7.9 earthquake of April 18, 1906 broke about 296 miles (477 km) of the San Andreas fault; fire followed, about 28,000 buildings were destroyed and more than 3,000 people died. Offsets of up to about 21 feet (6.4 m) led H. F. Reid to the elastic rebound theory, and the filled shoreline that became the Marina District set the stage for 1989.

2:40The Forecast

In 1988 the Working Group on California Earthquake Probabilities gave the southern Santa Cruz Mountains segment of the San Andreas fault a 30% chance of a magnitude 6.5 to 7 earthquake within 30 years. The earthquake came the very next year.

3:14Fifteen Seconds

A magnitude 6.9 rupture began about 11 miles (18 km) beneath the Santa Cruz Mountains, near Loma Prieta peak; the strongest shaking lasted about 15 seconds and the rupture never reached the surface. Old brick buildings fell in Santa Cruz, the Marina District's fill liquefied, part of the Bay Bridge's upper deck fell, and the collapse of Oakland's double-decker Cypress Street Viaduct caused most of the deaths.

5:29Aftermath

The toll: 63 deaths, 3,757 injuries, more than 18,000 homes damaged, about 12,000 people left homeless and about $6 billion in damage. The stadium held, and the World Series resumed ten days later.

5:57What Changed

San Francisco tore down the damaged Embarcadero Freeway in 1991 and reopened its waterfront; the Bay Bridge's eastern span was replaced with a new one designed for a major earthquake (opened in 2013); older bridges and freeways across California were retrofitted; and on the 30th anniversary, in 2019, the state began sending early-warning alerts to phones.

6:56The Next One

The Hayward fault runs directly beneath the East Bay's cities, creeps a few millimeters a year, and last ruptured in a magnitude 6.8 earthquake in 1868. The USGS puts the chance of at least one magnitude 6.7 or larger earthquake in the Bay Area between 2014 and 2043 at 72%, which is why millions practice Drop, Cover and Hold On every October. Three things to remember close the episode.

Common misconceptions For the teacher

MisconceptionThe closer you are to an earthquake's epicenter, the worse the damage will always be.

What the evidence showsDistance matters, but so does the ground. In 1989 the deadliest damage was in Oakland and San Francisco, about 60 miles from the fault, on soft bay mud and artificial fill that amplified the shaking and, in places, liquefied. Engineered buildings on firm ground near the epicenter generally performed well.

MisconceptionScientists predicted the Loma Prieta earthquake.

What the evidence showsThey forecast it, which is different. The 1988 Working Group gave the southern Santa Cruz Mountains segment a 30% chance of a magnitude 6.5 to 7 earthquake in 30 years: a probability, not a date. No one can yet predict the day of an earthquake, and the 1989 rupture was itself unusual, with nearly as much up-and-over (reverse) slip as sideways slip, on a fault plane dipping steeply beneath the mountains.

MisconceptionA creeping fault such as the Hayward fault is safe, because it lets off its strain gently.

What the evidence showsCreep on the Hayward fault is largely confined to the top few miles. Below about 3 miles (5 km) the fault is locked and storing strain, and it produced a magnitude 6.8 earthquake in 1868. The USGS gives the Hayward–Rodgers Creek fault system a 33% chance of a magnitude 6.7 or larger earthquake between 2014 and 2043.

MisconceptionEarthquake early warning predicts earthquakes before they start.

What the evidence showsEarly warning detects an earthquake that has already begun. Sensors near the source catch the faster P waves, and an electronic alert outruns the slower, damaging waves. Places far from the source can get many seconds of warning; places closest to it may get none.

MisconceptionA magnitude 7.9 earthquake is only a little bigger than a magnitude 6.9.

What the evidence showsEach whole step on the magnitude scale is about 32 times more energy, so the 1906 earthquake released roughly 32 times the energy of Loma Prieta, and it broke a fault about ten times longer.

Discussion questions

  1. The epicenter was in the Santa Cruz Mountains, yet many of the most destructive failures were about 60 miles away. Using the episode, explain why. What does this suggest about how a city should decide where to strengthen buildings first?
  2. The 1988 forecast gave the Santa Cruz Mountains segment a 30% chance of a large earthquake in 30 years, and it came the next year. Was the forecast a success? What would it have meant if no earthquake had happened by 2018?
  3. Much of the land that failed in the Marina District was made by people, by filling in a cove. Who should be responsible for knowing about, and telling buyers about, the ground a home is built on? Are there places in your own region built on fill?
  4. After 1989, San Francisco chose to demolish the Embarcadero Freeway rather than repair it, while the Bay Bridge's eastern span was replaced at a cost of billions. Compare these decisions: what criteria (cost, safety, traffic, the look and life of the city) do you think mattered most in each?
  5. Traffic on the Cypress Street Viaduct was lighter than usual because many people had gone home early to watch the World Series. What does that tell you about how the timing of an earthquake shapes its toll? How might the toll differ at 8 a.m. on a weekday, or 3 a.m. on a Sunday?
  6. An early-warning alert might give you anywhere from no time to perhaps 20 seconds before strong shaking. Make a list of what you, your school, a train operator and a hospital could usefully do with five seconds, and with twenty.

Quiz Choose an answer; it checks itself

1.Which description best fits the plate boundary beneath the San Francisco Bay Area?

2.What did H. F. Reid conclude from the ground offsets measured after the 1906 earthquake?

3.What lies beneath the part of the Marina District that was badly damaged in 1989?

4.In 1988, what did the Working Group on California Earthquake Probabilities say about the southern Santa Cruz Mountains segment of the San Andreas fault?

5.Why did buildings in the Marina District sink and tip during the 1989 earthquake?

6.Why did parts of San Francisco and Oakland, about 60 miles from the epicenter, shake harder than closer sites on rock?

7.Which set of figures matches the USGS summary of the Loma Prieta earthquake?

8.How can an earthquake early warning system alert people before strong shaking reaches them?

9.Why do many scientists consider the Hayward fault the most dangerous fault in the Bay Area?

10.What does the USGS forecast for 2014 to 2043 actually say?

Written response

1. Using the idea of elastic rebound, explain how plates that move only about 2 inches (5 cm) a year can produce sudden earthquakes. Use evidence from 1906 in your answer.

What a strong answer includes

The plates move steadily, but the faults between them are locked by friction, so the rock on either side bends instead of slipping, storing strain for decades or centuries. When the stress exceeds the strength of the fault, it slips suddenly and the bent rock snaps back. In 1906 roads and fences crossing the San Andreas fault were offset by as much as about 21 feet (6.4 m), representing well over a century of plate motion released in under a minute. Reid used those measurements to formulate the theory in 1910.

2. The Marina District is about 60 miles from the epicenter, yet seven of its buildings collapsed and four people died. Explain why, using the history of the land and the science of how ground responds to shaking.

What a strong answer includes

The damaged part of the Marina is built on a filled-in cove: loose sand pumped in for the 1915 Panama-Pacific International Exposition, together with older fill that may include some 1906 debris. In soft, water-saturated sediment, seismic waves slow down and grow larger, so the shaking was amplified well above that on the rock of nearby Fort Mason. The shaking then made the loose sand liquefy: it lost strength, buildings sank and tipped, sand boils erupted and lateral spreading broke gas and water lines. Fires broke out, and because the water mains had failed, a fireboat had to pump water from the bay. Buildings with weak, open garage floors were especially vulnerable.

3. A friend says: 'Scientists predicted Loma Prieta, and now they say the next big one will hit by 2043.' Explain what is wrong with both halves of this statement.

What a strong answer includes

Neither statement is a prediction. In 1988 scientists gave the southern Santa Cruz Mountains segment a 30% probability of a magnitude 6.5 to 7 earthquake within 30 years; that it came the next year was not a forecast of the date. The 2014 to 2043 USGS forecast is a 72% probability of at least one magnitude 6.7 or larger earthquake somewhere in the Bay Area in those 30 years. That means a large earthquake is more likely than not, but it is not certain, it could come on any fault in the region, and it could come any day, which is why preparing now matters.

Going deeper University and advanced high school

How big was Loma Prieta?

Moment magnitude is defined from the seismic moment M₀ = μ·A·D: the rigidity of the rock (μ), times the area of fault that slipped (A), times the average slip (D). Hanks and Kanamori (1979) set Mw = (2/3)(log₁₀M₀ − 9.1), with M₀ in newton-meters, so M₀ = 101.5 Mw + 9.1. The radiated energy grows by about 101.5 ≈ 32 for each whole step in magnitude. USGS Circular 1045 reports that the 1989 rupture involved slip on about 190 square miles (about 490 km²) of buried fault surface, and that geodetic surveys indicate about 6 feet of right-lateral slip and 4 feet of reverse slip.

Problem(a) How many times more energy did the 1906 earthquake (M7.9) release than Loma Prieta (M6.9)? How does Loma Prieta compare with the 1868 Hayward fault earthquake (M6.8)? (b) Calculate the seismic moment of an Mw 6.9 earthquake. (c) Taking μ = 3.0 × 1010 Pa, estimate the average slip and compare it with the geodetic estimate.

Worked answer

(a) 101.5 × 1.0 ≈ 32 times. Loma Prieta released 101.5 × 0.1 ≈ 1.4 times the energy of the 1868 earthquake, so the two are similar in size, but the Hayward fault runs beneath the East Bay's cities. (b) M₀ = 101.5 × 6.9 + 9.1 = 1019.45 ≈ 2.8 × 1019 N·m. (c) A ≈ 4.9 × 108 m², so D = M₀ ÷ (μA) ≈ 2.8 × 1019 ÷ (3.0 × 1010 × 4.9 × 108) ≈ 1.9 m, about 6 feet. The geodetic slip vector is √(6² + 4²) ≈ 7.2 feet (2.2 m): agreement within about 15%, which is good given the uncertainty in μ and in the rupture area.

The strain budget of the Bay Area's faults

Across the San Francisco Bay region the plates move about 1.6 inches (4 cm) a year relative to each other (USGS Fact Sheet 2016-3020). The USGS estimates that 1906-type earthquakes, with offsets of about 20 feet, recur roughly every 200 years. On the Hayward fault, dated stream deposits give a long-term slip rate of about 8 mm (0.3 inch) a year (Lienkaemper and Borchardt, 1996). The fault creeps at the surface at roughly 4 to 5 mm a year, but the creep is mostly confined to the top 3 miles (5 km); below that the fault is locked. Historical surveys suggest about 6 feet (2 m) of horizontal slip in the 1868 earthquake (USGS Fact Sheet 2008-3019).

Problem(a) What average slip rate on the northern San Andreas fault is implied by 20 feet of slip every 200 years? What fraction of the regional plate motion is that, and where does the rest go? (b) How much slip deficit has accumulated on the locked part of the Hayward fault between 1868 and 2026? Compare it with the slip of 1868. (c) Why does surface creep not remove the Hayward fault's hazard?

Worked answer

(a) 20 ft ÷ 200 yr = 0.1 ft a year, about 1.2 inches (3 cm) a year, roughly three quarters of the 1.6 inches a year across the region; the remainder is carried by the Hayward, Rodgers Creek, Calaveras and other faults of the system. (b) 158 years × 8 mm a year ≈ 1.3 m, about 4 feet: already a large fraction of the roughly 2 m of 1868, consistent with the USGS view that the fault can produce a magnitude 6.8 to 7.0 earthquake at any time. (c) Creep relieves strain only near the surface, and even there it is slower than the long-term rate; the deeper, locked part of the fault keeps storing strain until it ruptures.

Why soft ground resonates

A layer of soft sediment of thickness H over much stiffer rock traps shear waves and resonates. Its fundamental (quarter-wave) period is T = 4H/Vs, where Vs is the layer's shear-wave speed. In the Marina District, a borehole showed about 80 m (260 ft) of sediment, mostly older bay mud with Vs of about 300 m/s, over bedrock (Boatwright and others, in USGS Professional Paper 1551-F). Aftershock recordings showed the Marina amplifying the shaking by a factor of 7 to 10 near 1 hertz relative to the rock knoll at Fort Mason. A common rule of thumb gives a building's own natural period as about 0.1 s per story.

Problem(a) Calculate the fundamental period and frequency of the Marina sediment column and compare it with the observed peak. (b) Which building heights would be most at risk from resonance on such a site? (c) During the main shock, the Board of Inquiry reports amplification of only about 2 to 4 times on soft sites, much less than in the small aftershocks. Suggest why.

Worked answer

(a) T = 4 × 80 ÷ 300 ≈ 1.07 s, a frequency of about 0.94 Hz, matching the observed amplification peak near 1 Hz. (b) By the rule of thumb, buildings of roughly 8 to 12 stories (natural periods near 1 s) would resonate with the ground; low wood-frame houses have shorter periods, and in the Marina it was liquefaction and weak ground floors, rather than resonance alone, that brought them down. (c) Soft soils behave nonlinearly in strong shaking: they soften and dissipate more energy, which lowers the amplification and shifts the resonance to longer periods. Weak-motion amplification factors therefore overstate the amplification in a damaging earthquake, although the shaking on soft ground is still much stronger than on rock.

What a 72% forecast means each year

If large earthquakes in a region occurred randomly in time at an average rate λ per year (a Poisson process), the probability of at least one in t years would be P = 1 − exp(−λt). The USGS forecast (UCERF3, Fact Sheet 2016-3020) gives, for 2014 to 2043 in the Bay Area: 98% for at least one magnitude 6.0 or larger, 72% for magnitude 6.7 or larger and 51% for magnitude 7.0 or larger; for single faults, 33% for the Hayward or Rodgers Creek faults and 22% for the San Andreas fault, all for magnitude 6.7 or larger. The real forecast also accounts for the time since each fault's last earthquake, so the Poisson model is an approximation.

Problem(a) Convert the 72% into an equivalent annual rate, an average interval and a probability for a single year. Why is 72% ÷ 30 = 2.4% a year wrong? (b) Do the same for the 1988 forecast (30% in 30 years). What was the chance that the earthquake would come in the very next year? (c) If the Hayward–Rodgers Creek and San Andreas faults were independent, what would be the chance of at least one magnitude 6.7 earthquake on either of them? What accounts for the rest of the 72%?

Worked answer

(a) λ = −ln(1 − 0.72) ÷ 30 = −ln(0.28) ÷ 30 ≈ 1.273 ÷ 30 ≈ 0.042 per year, an average interval of about 24 years and a single-year probability of 1 − exp(−0.042) ≈ 4.1%. Dividing by 30 assumes that probabilities add, but chances of 'at least one' compound: a 2.4% chance each year would give only 1 − 0.976^30 ≈ 52% over 30 years, not 72%. (b) λ = −ln(0.70) ÷ 30 ≈ 0.012 per year, so about a 1.2% chance in any one year: the forecast identified the segment correctly, but the timing in 1989 was not part of it. (c) 1 − (1 − 0.33)(1 − 0.22) = 1 − 0.67 × 0.78 ≈ 0.48, or 48%. The rest comes from the Calaveras and other major faults, and from lesser-known and unknown faults (13% on their own).

How much warning is possible?

Earthquake early warning races electronic alerts against seismic waves. Take typical upper-crustal speeds of 6.0 km/s for P waves and 3.5 km/s for S waves. The Loma Prieta rupture began about 18 km (11 miles) deep, beneath an epicenter about 9 miles (14.5 km) from Santa Cruz and 60 miles (97 km) from San Francisco. Assume the system issues its alert 8 seconds after the rupture begins, which allows time for the P waves to reach several nearby stations and for processing; delivery to phones can add more.

Problem(a) Calculate the straight-line distance from the hypocenter to San Francisco's Marina District and the P- and S-wave arrival times there. How much warning would the Marina have had before the S waves? (b) Repeat for Santa Cruz. (c) Estimate the radius of the 'blind zone' on the surface, inside which the S waves arrive before the alert. Why might real warning times be shorter still?

Worked answer

(a) R = √(97² + 18²) ≈ 98.7 km. P waves arrive after 98.7 ÷ 6.0 ≈ 16.4 s and S waves after 98.7 ÷ 3.5 ≈ 28.2 s, so an alert at 8 s gives about 20 seconds of warning. (b) R = √(14.5² + 18²) ≈ 23.1 km; the S waves arrive after about 6.6 s, before the alert, so Santa Cruz would have had no warning. (c) In 8 s the S waves travel 28 km; at the surface that reaches √(28² − 18²) ≈ 21 km (about 13 miles) from the epicenter. Real warnings can be shorter because the rupture spread about 20 km in each direction along the fault, so strong waves also came from parts of the fault closer to many towns, and because alerts take time to reach phones. A deep hypocenter, as here, actually helps by delaying the S waves at the surface.

Standards alignment

Next Generation Science Standards

MS-ESS2-2

Construct an explanation based on evidence for how geoscience processes have changed Earth's surface at varying time and spatial scales.

In this episode: Slow plate motion stored as strain and released in seconds; the 1906 offsets; the Hayward fault's creep bending curbs over decades.

MS-ESS3-2

Analyze and interpret data on natural hazards to forecast future catastrophic events and inform the development of technologies to mitigate their effects.

In this episode: The 1988 and 2014 earthquake forecasts, freeway retrofits, the new Bay Bridge span and earthquake early warning.

HS-ESS2-1

Develop a model to illustrate how Earth's internal and surface processes operate at different spatial and temporal scales to form continental and ocean-floor features.

In this episode: A model of the transform plate boundary and the elastic rebound cycle; the oblique 1989 slip that lifted parts of the Santa Cruz Mountains.

HS-ESS3-1

Construct an explanation based on evidence for how the availability of natural resources, occurrence of natural hazards, and changes in climate have influenced human activity.

In this episode: Building on fill and bay mud, the toll of 1906 and 1989, the reopened Embarcadero waterfront and the Bay Area's preparedness today.

HS-PS4-1

Use mathematical representations to support a claim regarding relationships among the frequency, wavelength, and speed of waves traveling in various media.

In this episode: Seismic waves slowing and growing in soft ground, the resonance of the Marina's sediments, and P- and S-wave speeds in early warning.

HS-ETS1-3

Evaluate a solution to a complex real-world problem based on prioritized criteria and trade-offs that account for a range of constraints, including cost, safety, reliability, and aesthetics as well as possible social, cultural, and environmental impacts.

In this episode: Retrofit, replace or remove: the Cypress Street Viaduct, the Bay Bridge's eastern span and the Embarcadero Freeway.

Advanced Placement

University courses

Physical geology

Transform plate boundaries; the San Andreas fault system; elastic rebound and the earthquake cycle; fault creep; oblique slip

Natural hazards

Earthquake forecasting and probability; site effects and liquefaction; lessons from 1906 and 1989; earthquake early warning; preparedness

Engineering seismology and geotechnical engineering

Seismic moment and magnitude; site amplification and resonance of soft soils; liquefaction of hydraulic fill; seismic retrofit of bridges and viaducts

Transcript The narration, word for word

0:005:04 PM

San Francisco, on October 17, 1989. More than 60,000 fans have packed Candlestick Park for game three of the World Series, between the San Francisco Giants and the Oakland A's.

0:14At four minutes past five, the television broadcast has just begun. Then the stadium starts to move, and the live picture, going out to millions of viewers, cuts to static.

0:26This is the story of the Loma Prieta earthquake. We'll look at the fault behind it, the warnings that came before it, and how a whole region rebuilt around its lessons.

0:42The Fault Next Door

California sits on the boundary between two tectonic plates. The Pacific plate is grinding northwest past the North American plate, at about 2 inches a year.

0:53Most of that motion takes place along the San Andreas fault, which runs for more than 700 miles through California. Around San Francisco Bay, it splits into a family of faults, including the San Andreas, the Hayward and the Calaveras.

1:11Like the plates beneath Japan, these faults do not slide smoothly. They lock together, the rock around them bends for decades or even centuries, and then the stored strain is released all at once.

1:261906

San Francisco learned that the hard way on April 18, 1906, when a magnitude 7.9 earthquake ripped open the San Andreas fault for nearly 300 miles.

1:39Broken gas lines and water mains turned the shaking into a firestorm that burned for days. About 28,000 buildings were destroyed, and more than 3,000 people died.

1:52North of the city, roads and fences that crossed the fault were torn apart and shifted sideways. Near the town of Olema, the ground on one side of the fault moved about 21 feet past the other.

2:05A scientist named Harry Fielding Reid studied those measurements and worked out why earthquakes happen. The rock on either side of a locked fault bends slowly, storing energy, until it snaps back into place. He called this elastic rebound, and it is still how geologists explain earthquakes.

2:23The city rebuilt quickly, and in 1915 it celebrated with a world's fair. To make room for the fairgrounds, a cove along the northern shore was filled in with sand pumped from the bay, along with some debris, and that new land later became a neighborhood called the Marina District.

2:40The Forecast

By the 1980s, scientists were estimating which parts of the San Andreas were most likely to break next. One of the stretches they flagged was in the southern Santa Cruz Mountains, south of San Francisco.

2:54In 1988, a national panel of earthquake experts gave that segment a 30% chance of a large earthquake within 30 years.

3:05The earthquake came the very next year.

3:14Fifteen Seconds

At four minutes past five, the fault broke about 11 miles below the Santa Cruz Mountains, near Loma Prieta peak. The earthquake measured magnitude 6.9.

3:25The strongest shaking lasted only about 15 seconds, and the rupture never reached the surface. Its seismic waves, however, spread across the whole region, and in some places the ground itself made them worse.

3:40Close to the epicenter, old brick buildings on Santa Cruz's Pacific Garden Mall crumbled into the street, and three people were killed there.

3:51About 60 miles away, in the Marina District, the shaking turned the loose, wet ground beneath the houses almost to liquid. Buildings sank, and some collapsed.

4:03In effect, the land made for the city's celebration of its recovery helped bring buildings down 74 years later. Fires broke out from broken gas lines, four people died in the Marina, and seven buildings collapsed.

4:17The Marina was not the only place at risk. All around the bay, buildings and roads stood on soft mud and fill, and in soft ground, seismic waves slow down and grow larger, like jelly shaking in a bowl. Those sites shook two to five times harder than solid rock nearby.

4:35On the Bay Bridge, a 50 foot section of the upper deck broke loose and fell onto the deck below. One person was killed, and the bridge was closed for about a month.

4:47The worst disaster came in Oakland. The Cypress Street Viaduct was a double-decker freeway, and its upper deck collapsed onto the lower one, onto cars in the evening commute. Forty-one people died there.

5:01Traffic that evening was lighter than usual, because many people had gone home early to watch the World Series, and experts believe that kept the death toll from being far worse.

5:13Rescuers searched the wreckage for days. Four days after the earthquake, after 90 hours beneath the fallen deck, they found a man still alive in his car. He survived the rescue, but died a few weeks later.

5:29Aftermath

In all, 63 people died and more than 3,700 were injured. More than 18,000 homes were damaged, some 12,000 people were left without one, and the damage came to about $6 billion.

5:44At Candlestick Park, the stadium held, and the fans walked out safely. The World Series resumed ten days later, and the A's went on to sweep it.

5:57What Changed

The earthquake also reshaped San Francisco. The damaged Embarcadero Freeway, a wall of concrete along the waterfront, was torn down in 1991, and the city's waterfront opened back up to the bay.

6:11The eastern half of the Bay Bridge was replaced entirely. The new span, built to withstand a major earthquake, opened in 2013, at a cost of about $6.4 billion.

6:25Across California, engineers returned to freeways and bridges built before modern building codes, wrapping old concrete columns in steel and tying their spans together.

6:37California also built an earthquake early warning system, which detects the first, faster waves and sends an alert ahead of the strong shaking. On Loma Prieta's thirtieth anniversary, in 2019, the state began sending those alerts to phones.

6:56The Next One

The plates have not stopped moving. Many scientists consider the Hayward fault the most dangerous in the region, because it runs directly beneath the cities on the east side of the bay, and it last broke in 1868.

7:10You can actually see the Hayward fault move. Parts of it creep along continuously, about a fifth of an inch a year, and in downtown Hayward, old curbs that cross the fault have been bent sideways, bit by bit, over the decades.

7:26Its last large earthquake, in 1868, wrecked almost every building in Hayward and killed about 30 people. Until 1906, it was the earthquake people called the great San Francisco earthquake.

7:40The U.S. Geological Survey puts the chance of at least one magnitude 6.7 or larger earthquake in the Bay Area, between 2014 and 2043, at 72%.

7:55That is why, every October, around the anniversary of Loma Prieta, millions of people practice three simple steps in the Great ShakeOut drill: drop to your hands and knees, take cover under a sturdy table, and hold on until the shaking stops.

8:12So, three things to remember. First, Loma Prieta was a magnitude 6.9 earthquake on the San Andreas fault system, and its strongest shaking lasted only about 15 seconds. Second, the worst damage came far from the fault, in places such as the Marina District and the Cypress freeway, where soft mud and fill made the shaking two to five times stronger. And third, the Bay Area is not finished: the chance of another earthquake of magnitude 6.7 or larger there, between 2014 and 2043, is 72%.

8:21Loma Prieta was not the big one, but it was a reminder. The faults beneath the bay are still storing up strain, about an inch at a time.

Sources

  1. U.S. Geological Survey. M 6.9 – October 17, 1989 Loma Prieta Earthquake. Earthquake Hazards Program. www.usgs.gov/programs/earthquake-hazards/science/m-69-october-17-1989-loma-prieta-earthquakeUsed for: Time, magnitude, epicenter, deaths, injuries, damage, Santa Cruz, Bay Bridge, liquefaction
  2. Plafker, G., & Galloway, J. (Eds.) (1989). Lessons Learned from the Loma Prieta, California, Earthquake of October 17, 1989. U.S. Geological Survey Circular 1045. pubs.usgs.gov/publication/cir1045Used for: Hypocenter depth, rupture size and slip, buried rupture, amplification, Marina District, Cypress and Bay Bridge
  3. O'Rourke, T. D., et al. (1992). The Loma Prieta, California, Earthquake of October 17, 1989: Marina District. U.S. Geological Survey Professional Paper 1551-F. pubs.usgs.gov/publication/pp1551FUsed for: History of the Marina fill, liquefaction, collapsed buildings, site resonance and amplification
  4. Borcherdt, R. D., & Glassmoyer, G. (1992). On the characteristics of local geology and their influence on ground motions generated by the Loma Prieta earthquake in the San Francisco Bay region, California. Bulletin of the Seismological Society of America, 82(2), 603–641. doi.org/10.1785/BSSA0820020603Used for: Amplification of shaking on fill and bay mud
  5. Housner, G. W., & Thiel, C. C., Jr. (1990). Competing against time: Report of the Governor's Board of Inquiry on the 1989 Loma Prieta earthquake. Earthquake Spectra, 6(4), 681–711. doi.org/10.1193/1.1585592Used for: Soft-soil amplification, the Cypress Street Viaduct and Bay Bridge failures
  6. Working Group on California Earthquake Probabilities (1988). Probabilities of Large Earthquakes Occurring in California on the San Andreas Fault. U.S. Geological Survey Open-File Report 88-398. pubs.usgs.gov/publication/ofr88398Used for: The 1988 forecast
  7. U.S. Geological Survey. The Great 1906 San Francisco Earthquake. Earthquake Hazards Program. earthquake.usgs.gov/earthquakes/events/1906calif/18aprilUsed for: 1906 rupture length, casualties and buildings destroyed, Reid's elastic rebound theory, recurrence
  8. Brocher, T. M., et al. (2008). The Hayward Fault: Is It Due for a Repeat of the Powerful 1868 Earthquake? U.S. Geological Survey Fact Sheet 2008-3019. pubs.usgs.gov/fs/2008/3019Used for: The 1868 earthquake, Hayward fault creep, recurrence
  9. Lienkaemper, J. J., & Borchardt, G. (1996). Holocene slip rate of the Hayward fault at Union City, California. Journal of Geophysical Research: Solid Earth, 101(B3), 6099–6108. doi.org/10.1029/95JB01378Used for: Hayward fault long-term slip rate
  10. Aagaard, B. T., et al. (2016). Earthquake Outlook for the San Francisco Bay Region 2014–2043. U.S. Geological Survey Fact Sheet 2016-3020 (ver. 1.1). pubs.usgs.gov/publication/fs20163020Used for: 72% forecast, single-fault probabilities, regional plate motion, Drop, Cover and Hold On
  11. Burkett, E., Given, D. D., & Jones, L. M. (2014). ShakeAlert: An Earthquake Early Warning System for the United States West Coast. U.S. Geological Survey Fact Sheet 2014-3083. pubs.usgs.gov/publication/fs20143083Used for: How earthquake early warning works
  12. Berkeley News (2019, October 17). California rolls out first statewide earthquake early warning system. University of California, Berkeley. news.berkeley.edu/2019/10/17/california-rolls-out-first-statewide-earthquake-early-warning-systemUsed for: Phone alerts launched on Loma Prieta's 30th anniversary (MyShake)
  13. Metropolitan Transportation Commission. San Francisco–Oakland Bay Bridge. www.baybridgeinfo.orgUsed for: The new eastern span: opening and cost
  14. Hanks, T. C., & Kanamori, H. (1979). A moment magnitude scale. Journal of Geophysical Research, 84(B5), 2348–2350. doi.org/10.1029/JB084iB05p02348Used for: Moment magnitude and seismic moment

Images in the episode: USGS, NOAA and NASA (public domain), with maps from NASA/USGS Landsat via NASA GIBS; animations by EarthWatch 24-7. Full credits: earthwatch247.com/credits.html.

Cite this lesson

EarthWatch 24-7. (2026). The World Series Earthquake: Loma Prieta, 1989: why the worst damage came 60 miles from the fault [Video lesson]. Earth Explained, episode 2. https://earthwatch247.com/learn/loma-1989/

Answer key

  1. C. It is a transform boundary: the plates move sideways past each other at about 2 inches (5 cm) a year, and the motion is shared among the San Andreas, Hayward, Calaveras and related faults.
  2. A. This is elastic rebound (1910). Its key insight is that slip on the fault causes the earthquake, not the reverse.
  3. D. USGS studies trace the filling of the cove, which ended with hydraulic sand fill placed for the 1915 Panama-Pacific International Exposition. Loose, wet sand fill is exactly the kind of ground that liquefies.
  4. B. The forecast was a probability, not a prediction of a date. The 72% figure is the USGS forecast for the whole Bay Area for 2014 to 2043.
  5. A. This is liquefaction. Sand boils and cracks from lateral spreading appeared across the district, about 60 miles from the epicenter; the rupture itself never reached the surface.
  6. C. Site amplification: the Governor's Board of Inquiry found that soft ground around the bay amplified the shaking more than building codes had anticipated.
  7. B. 63 deaths, 3,757 injuries and about $6 billion in property damage. The first option describes the 1906 earthquake and fire.
  8. D. Alerts travel almost instantly, while seismic waves travel a few miles per second. The earthquake must already have started.
  9. C. Its 1868 earthquake (magnitude 6.8) killed about 30 people when the region was sparsely populated; today millions live along it.
  10. A. The 72% covers the whole region and all its faults over 30 years. For single faults the figures are lower: 33% for the Hayward or Rodgers Creek faults and 22% for the San Andreas.

Written response

  1. The plates move steadily, but the faults between them are locked by friction, so the rock on either side bends instead of slipping, storing strain for decades or centuries. When the stress exceeds the strength of the fault, it slips suddenly and the bent rock snaps back. In 1906 roads and fences crossing the San Andreas fault were offset by as much as about 21 feet (6.4 m), representing well over a century of plate motion released in under a minute. Reid used those measurements to formulate the theory in 1910.
  2. The damaged part of the Marina is built on a filled-in cove: loose sand pumped in for the 1915 Panama-Pacific International Exposition, together with older fill that may include some 1906 debris. In soft, water-saturated sediment, seismic waves slow down and grow larger, so the shaking was amplified well above that on the rock of nearby Fort Mason. The shaking then made the loose sand liquefy: it lost strength, buildings sank and tipped, sand boils erupted and lateral spreading broke gas and water lines. Fires broke out, and because the water mains had failed, a fireboat had to pump water from the bay. Buildings with weak, open garage floors were especially vulnerable.
  3. Neither statement is a prediction. In 1988 scientists gave the southern Santa Cruz Mountains segment a 30% probability of a magnitude 6.5 to 7 earthquake within 30 years; that it came the next year was not a forecast of the date. The 2014 to 2043 USGS forecast is a 72% probability of at least one magnitude 6.7 or larger earthquake somewhere in the Bay Area in those 30 years. That means a large earthquake is more likely than not, but it is not certain, it could come on any fault in the region, and it could come any day, which is why preparing now matters.

Going deeper

  1. How big was Loma Prieta?: (a) 101.5 × 1.0 ≈ 32 times. Loma Prieta released 101.5 × 0.1 ≈ 1.4 times the energy of the 1868 earthquake, so the two are similar in size, but the Hayward fault runs beneath the East Bay's cities. (b) M₀ = 101.5 × 6.9 + 9.1 = 1019.45 ≈ 2.8 × 1019 N·m. (c) A ≈ 4.9 × 108 m², so D = M₀ ÷ (μA) ≈ 2.8 × 1019 ÷ (3.0 × 1010 × 4.9 × 108) ≈ 1.9 m, about 6 feet. The geodetic slip vector is √(6² + 4²) ≈ 7.2 feet (2.2 m): agreement within about 15%, which is good given the uncertainty in μ and in the rupture area.
  2. The strain budget of the Bay Area's faults: (a) 20 ft ÷ 200 yr = 0.1 ft a year, about 1.2 inches (3 cm) a year, roughly three quarters of the 1.6 inches a year across the region; the remainder is carried by the Hayward, Rodgers Creek, Calaveras and other faults of the system. (b) 158 years × 8 mm a year ≈ 1.3 m, about 4 feet: already a large fraction of the roughly 2 m of 1868, consistent with the USGS view that the fault can produce a magnitude 6.8 to 7.0 earthquake at any time. (c) Creep relieves strain only near the surface, and even there it is slower than the long-term rate; the deeper, locked part of the fault keeps storing strain until it ruptures.
  3. Why soft ground resonates: (a) T = 4 × 80 ÷ 300 ≈ 1.07 s, a frequency of about 0.94 Hz, matching the observed amplification peak near 1 Hz. (b) By the rule of thumb, buildings of roughly 8 to 12 stories (natural periods near 1 s) would resonate with the ground; low wood-frame houses have shorter periods, and in the Marina it was liquefaction and weak ground floors, rather than resonance alone, that brought them down. (c) Soft soils behave nonlinearly in strong shaking: they soften and dissipate more energy, which lowers the amplification and shifts the resonance to longer periods. Weak-motion amplification factors therefore overstate the amplification in a damaging earthquake, although the shaking on soft ground is still much stronger than on rock.
  4. What a 72% forecast means each year: (a) λ = −ln(1 − 0.72) ÷ 30 = −ln(0.28) ÷ 30 ≈ 1.273 ÷ 30 ≈ 0.042 per year, an average interval of about 24 years and a single-year probability of 1 − exp(−0.042) ≈ 4.1%. Dividing by 30 assumes that probabilities add, but chances of 'at least one' compound: a 2.4% chance each year would give only 1 − 0.976^30 ≈ 52% over 30 years, not 72%. (b) λ = −ln(0.70) ÷ 30 ≈ 0.012 per year, so about a 1.2% chance in any one year: the forecast identified the segment correctly, but the timing in 1989 was not part of it. (c) 1 − (1 − 0.33)(1 − 0.22) = 1 − 0.67 × 0.78 ≈ 0.48, or 48%. The rest comes from the Calaveras and other major faults, and from lesser-known and unknown faults (13% on their own).
  5. How much warning is possible?: (a) R = √(97² + 18²) ≈ 98.7 km. P waves arrive after 98.7 ÷ 6.0 ≈ 16.4 s and S waves after 98.7 ÷ 3.5 ≈ 28.2 s, so an alert at 8 s gives about 20 seconds of warning. (b) R = √(14.5² + 18²) ≈ 23.1 km; the S waves arrive after about 6.6 s, before the alert, so Santa Cruz would have had no warning. (c) In 8 s the S waves travel 28 km; at the surface that reaches √(28² − 18²) ≈ 21 km (about 13 miles) from the epicenter. Real warnings can be shorter because the rupture spread about 20 km in each direction along the fault, so strong waves also came from parts of the fault closer to many towns, and because alerts take time to reach phones. A deep hypocenter, as here, actually helps by delaying the S waves at the surface.