Earthquake Probabilities in the San Francisco Bay Region, 2002–2031

Earthquake Probabilities in the San Francisco Bay Region, 2002–2031

Overview

• History and scope of the Working Group reports (and what is new in this one)

• Uncertainty, and what they mean by it

• The earthquake model, and ‘background events’

• Probability models

• Putting it all together

History of the WGCEP…

• 1988: SAF and HF slip rates & time predictable model, estimated 50% probability of M~7 in 30 years

• 1989: M 6.9 Loma Prieta earthquake

• 1990: Post-Loma Prieta recalculation, added RCF, stress changes, new rupture scenarios, p = 67%

• 1995: SoCal only, included geodetic slip rates, multiple segment ruptures, regional bounds, etc

• 1999: Included CF, SGF, GF, C-GVF, MtDT, 18 segments, 35 scenarios, more sophisticated data/methods as per 1995 report, p = 70%

…and what’s new?

• Improved 1906 stress shadow model

• Probabilities of different magnitude earthquakes included, as are different time intervals

• Incremental improvements to: slip rate estimates (geodetic/geologic), historical eq knowledge (locations/intervals/magnitudes), knowledge of creep, 1906 eq slip, regional strain budget, etc…

Area covered by report

Fault segments considered

A treatise on uncertainty

• Aleatory uncertainties – natural random variability, which is irreducible

• Epistemic uncertainties – owing to our lack of understanding of natural processes, use of incomplete models, measurement error, etc

• The study is devoted to reducing, and quantifying epistemic uncertainty

• A Monte Carlo approach is used

Monte Carlo treatment of uncertainty

The earthquake model

• Fault segments – their lengths, widths, slip rates and ‘seismogenic scaling factors’ (= % not creeping)

• Rupture sources – 35 combinations of segments that can rupture alone or in groups (+ ‘floaters’)

• Rupture scenarios – possible combinations of rupture sources in a single earthquake cycle

• Fault rupture models – weighted combined probabilities of the various scenarios occurring

• Regional model – all the above must satisfy GPS strain budget

Division of the plate motion budget

Probability models

• Poisson – simple, time invariant (based on mean recurrence)

• Empirical – new for 2002! Modulates Poisson recurr-ences by current seismicity rates (shadow effects)

• Brownian Passage Time – deterministic loading/ stress shadowing + stochastic element

• Time Predictable – using last earthquake rupture time/size and loading rate, how long until next one?

• Estimation of inherent randomness ‘remains a significant scientific challenge’

Strain accumulation and release

From rupture area to eq magnitude

• Earthquake scaling relations are used to estimate moment release from the fault segment area(s)

• 3 relations used – Wells & Coppersmith (1994), Ellsworth (1999) and Hanks & Bakun (2002)

• Largest source of uncertainty in the whole process – can get factor of 2 differences in M0 for different scaling relations

Putting it all together 1: recurrences

Putting it all together 2: probabilities