Seismic Hazards 2 - Prediction

Introduction

  • Welcome to Seismic 2, focusing on going beyond earthquake prediction.
  • Objective: Examine early prediction attempts, case studies, and modern approaches.
  • Trajectory: Transition to probabilistic forecasting (covered in subsequent lectures).

Review of Introductory Seismic Lecture

  • Topics covered: Seismicity, P-waves, S-waves, seismic velocities, energy, wave trains, seismometers, intensity, magnitude distribution, and fault rupture.
  • Fault Rupture:
    • Average slip (e.g., 10 meters) is spatially variable.
    • Some fault parts may seize, while others slip freely, influencing aftershock patterns.
  • Rupture models propagate at 2-3 kilometers per second.

Historical Attempts at Earthquake Prediction

  • Early Attempts (1950s-1980s):
    • Focus on predicting earthquakes.
  • Modern Approaches (1990s-Present):
    • Shift towards improving forecasting with probabilistic models.

Defining Earthquake Prediction

  • Involves a short time window.
  • Includes a specific spatial window (area or zone).
  • Requires an estimate of magnitude and potential impact.
  • Needs knowledge of the system state, historical fault slip, and conditions conducive to the next event.

Requirements for Earthquake Prediction

  • Recognition of precursory signs indicating impending fault rupture.
  • Example with a ruler: Bending a ruler until it's about to break, with creaking and grinding serving as precursors.

Types of Prediction

  • Imminent Event Prediction:
    • Recognition of precursors in days, hours, or minutes before an event.
    • Time scales can be short-term or intermediate.
  • Decadal or Centennial Approach:
    • Falls into the realm of forecasting due to the extended time scale.

Spatial Scale

  • ll: Length of the source zone, dependent on earthquake size.
    • Magnitude 4: ~1 kilometer.
    • Magnitude 9: ~1,000 kilometers.
  • Uncertainty can range from specific fault segments to hundreds of fault lengths.

Early Research Programs

  • Russia, China, and Japan (1950s-1960s):
    • Aimed to save lives and lessen economic impact.
    • Countries with histories of disruptive earthquakes.
    • Investment in training seismologists to identify precursory signs.

Promising Case Studies and Increased Research

  • 1970s-1980s: Expansion into lab work to simulate larger-scale events.
  • Scale Problem: Bridging the gap between lab-scale seismicity and real-world events.
  • Optimism among scientists, decision-makers, and politicians regarding reliable earthquake prediction.

Parkfield, California Experiment

  • Objective: Monitor a seemingly predictable segment of a strike-slip fault.
  • Set up in the 1980s, anticipating an event based on previous patterns.
    • 1850s, 1880s, 1900, 1922, 1934, 1966 events
  • Recurrence Interval:
    • Mean of 22 years.
    • Standard deviation of 5 years.

Instrumentation and Monitoring

  • High-Resolution Seismograph Network (HRSN).
  • Continuous GPS for tracking ground elevation changes.
  • Water wells, tensor strains, and dilatometers for measuring groundwater changes.
  • San Andreas Fault Observatory at Depth (SAFOD): Drilled close to the fault plane for monitoring at depth.

Outcome of Parkfield Experiment

  • Anticipated event did not occur as expected (by the 1990s).
  • The fault slipped in 2004 (later than expected).
  • Magnitude 6 event with no precursors.
  • Significant setback for earthquake prediction.
  • USGS website (earthquake.usgs.gov) provides information on the Parkfield experiment.

Chris Shultz's Prediction (1970s)

  • Based on rock physics, predicting stages of geophysical changes before catastrophic failure.
  • Elastic Strain Buildup: Rock recovers to its original state if stress is removed.
  • Dilatency: Transition to the plastic domain, causing cracks and volume increase; water flows into cracks.
  • Seismic Velocity Changes: Damaged rock and cracks may alter seismic wave propagation, especially the ratio of P-wave to S-wave velocities.
  • Volume Change: Surface uplift may occur due to dilatency.
  • Influx of Water: Cracks join, creating thoroughfares for water movement, affecting well levels.
  • Slip Event: Followed by aftershocks (Omori's Law).

Lab Studies vs. Real-Life Geology

  • Lab studies show micro-cracking accelerating towards final failure.
  • Scaling up to real geology (kilometers) is challenging.

Potential Changes During Loading

  • Gas emissions (radon gas from radioactive decay) linked to pathway openings.
  • Electromagnetic radiation.
  • Groundwater level changes and deformation.
  • Anomalous animal behavior.

Follow-Up Studies After Parkfield

  • Jordan et al.: No causal connection established between radon and earthquake preparation.
  • Electrical precursors were not observed at Parkfield.

Successes and Failures in Earthquake Prediction

  • Haijeng, China (1975):
    • Evacuation based on foreshocks led to drastically decreased death toll after a magnitude 7.3 earthquake.
  • Guangdong, China: Similar attempt with evacuation; no earthquake occurred.
  • Tangshan, China (1976):
    • Magnitude 7.6 earthquake with no precursors.
    • Over a quarter of a million fatalities.

Chinese Earthquake Cataloging

  • Over 1,000 magnitude 5 events.
  • Precursors recognized in some cases.
  • Accurate prediction rate of only ~1%.

Global Shift Towards Probabilistic Forecasting

  • Rejection of many proposed precursory phenomena.
  • IASPI (International Association of Seismology and Physics of the Earth Interior).
  • Focus shifts to earthquake generation processes, physics, modeling, and monitoring for forecasting.

Animal Behavior

  • Skepticism regarding animal behavior as a reliable precursor.
  • Viral misinformation example: A video of animal behavior falsely linked to the Turkish earthquake.

Probabilistic Forecasting

  • Long-term approaches with wider time windows and statistical involvement.
  • Hazard assessment and mitigation.
  • Ongoing monitoring efforts in Japan and China.
  • International collaborations and data sharing.
  • Use of statistical models and pattern recognition.
    • M8/MATE algorithm: 70% success rate for magnitude 8+ events over specific areas.

Societal Response

  • Two extremes:
    • Acceptance of seismic events with limited mitigation.
    • Stringent protection with potential socioeconomic consequences.

Prediction Considerations

  • Time window of a few days ideal for evacuation.
  • Robust evidence and clear communication are critical.
  • Risk assessment: Balancing vulnerability and hazard level.
  • Public cooperation and understanding.
  • False alarms must not devastate local economies.
  • Uncertainty must be communicated effectively.

Twenty-First Century and Operational Forecasting

  • Jordan et al. study: Operational earthquake forecasting state of knowledge and guidelines.
  • Example from Italy:

Italian Case Study

  • Using seismic catalogs to identify hotspots for seismic events.
  • Ground acceleration with a 10% probability of exceedance in the next 50 years.
  • Higher ground acceleration translates to higher intensity (7-9), causing structural damage.
  • Authorities use these maps for long-term planning.

The L'Aquila Earthquake (2009)

  • A seismic event caused over 300 fatalities.
  • Lawsuit against seismologists for negligence.

Forecasting Models in Italy

  • Multiple teams with different forecast models competing.
  • Models tested against actual events to refine accuracy.

Question About Precursors

  • Precursors depend on the type of fault.
  • Hydrological context plays a key role.

Distribution of Earthquake Sizes

  • Gutenberg-Richter Law: Many small events, few large ones.
  • Log plot: Straight line, indicating the rarity of large earthquakes.
    logN=a−bMlog N= a - bM
  • NN = cumulative number of events
  • MM = magnitude
  • Gradient: bb (B-value)
  • aa: productivity (number of earthquakes)

B-Value: An Important Parameter

  • Inverse dependence on differential stress (loading on the fault).
  • Higher differential stress (fault close to failure), the b-value decreases.
  • Lower B-Value: Starts to have more larger events taking place (cracks coalescing).

Julia and Wima Paper (Nature)

  • Experiment: Loading lava, recording micro-earthquakes and the B-value change. B-value decreases when these cracks were just linking up.
  • B-value change before final failure.

B-Value and Foreshocks

  • If an earthquake is followed by a steeper slope (increase in B-value), it suggests a return to smaller cracks.
  • If an earthquake is followed by lower B values, (red), then after this earthquake has taken place, comparatively speaking now you've got more larger events. it is likely to lead to another event.

Tohoku Earthquake (2011) - B Value Case Study

  • Magnitude 7.3 foreshock, followed by a magnitude 9 main shock.
  • B-value decreased after the 7.3 event, indicating a larger event was likely.

Bad Prediction Example: New Madrid Seismic Zone

  • Meteorologist Iben Browning predicted a 50% chance of a magnitude 6+ earthquake around December 2-12, 1990 based on astronomical/astrological calculations.
  • Public panic, school closures, and a surge in emergency preparedness supplies.
  • No earthquake occurred; highlighting the dangers of bad communication.

Communication Issues and Lessons Learned

  • Panic can kill more people than the earthquake itself (not always true).
  • There are bad predictions that went viral. Example a tweet made before the Turkish earthquake got millions of views.

More Recent Paper Caution Needed

  • Science papers need caution in terms of any kind of attempts to predict.
  • There are very different from the longer term forecasting

Summary and Final Recommendations

  • B values and foreshocks are promising
  • Probabilistic models are where we are now.
  • Recommended reading: Jordan study and Parkfield information.