Research ArticleGEOPHYSICS

A physics-based earthquake simulator replicates seismic hazard statistics across California

See allHide authors and affiliations

Science Advances  22 Aug 2018:
Vol. 4, no. 8, eaau0688
DOI: 10.1126/sciadv.aau0688
  • Fig. 1 Hazard-relevant measures in the simulator compared with the UCERF3 hazard model.

    (A) Recurrence intervals in the earthquake simulator compared with the UCERF3 hazard model. Plot shows a two-dimensional (2D) pointwise histogram of 265k elements on fault for M ≥ 7. Color shows number in histogram. The diagonal solid black line shows what would be perfect agreement. The lower left part shows the fastest-moving areas of the fault system with the shortest recurrence intervals. The horizontal lines at the top show the finiteness of the catalog, with the top horizontal set being patches that have broken once during the catalog above the cutoff magnitude, the next horizontal set below having broken twice, and so on. (B) Pointwise shaking hazard in the simulator compared with UCERF3. This is for a standard hazard measure, 2% in 50 years peak ground acceleration (PGA). Plot shows 2D pointwise shaking hazard across California coming from M 6.5+ events for the simulator events compared with on-fault UCERF3 events. Note that the hazard measure agrees even better than the recurrence intervals.

  • Fig. 2 Maps of shaking hazard in the earthquake simulator compared with the UCERF3 hazard model and plots of differences.

    The immediate predecessor UCERF2 California hazard model is shown for comparison. Maps show PGA 2% in 50-year exceedance. Units are in fractions of the acceleration of gravity g. (A) UCERF2. (B) UCERF3. (C) RSQSim model. (D) Map of ln ratio of UCERF2/UCERF3 shaking hazard. (E) Map of ln ratio of simulator/UCERF3 shaking hazard. Note that the simulator is even closer to UCERF3 than UCERF3 is to UCERF2.

  • Fig. 3 Full hazard curves for some example cities.

    The first four cities are by largest population in California, and the next two are added as examples due to proximity to San Andreas and thrust faults. The horizontal axis is PGA as a fraction of gravity acceleration g. The vertical axis is annual probability of exceedance. Red lines show RSQSim results. Blue lines show UCERF3 results for on-fault events. Black lines show full UCERF3 hazard results for all events, including off-fault events, as reference to also show off-fault hazard that is not being included. Horizontal dashed line shows the 2500-year standard reference curve in the middle, the 1000-year curve on top, and the 10,000-year curve at the bottom. Cities are as follows: (A) Los Angeles, (B) San Diego, (C) San Jose, (D) San Francisco, (E) San Bernardino, and (F) Santa Barbara.

  • Fig. 4 Good agreement in long-term hazard.

    Averaging over the state, we plot mean absolute ln(RSQSim/UCERF3) as a function of probability level for different PSA(T). PSA(T) at different resonant period T seconds is pseudo-spectral acceleration, a measure of ground motion used by engineers relevant to building response. Different curves are for different spectral periods T, shown with PGA (red), at T = 0.2 s (yellow), T = 1 s (green), T = 5 s (blue), and T = 10 s (black). Note the very good correspondence for long-term hazard at annual probabilities of p = 10−3 year−1 and below, which are regions of central engineering importance. At time scales shorter than a few centuries, below the repeat times of large events where details concerning smaller events become important, the curves begin to diverge.

  • Fig. 5 Maps of PSA(1) 1-s spectral acceleration shaking hazard in the earthquake simulator compared with the UCERF3 hazard model, and plots of differences, for different return periods.

    Note the illumination of slower-moving faults at longer return periods and the good correspondence between the simulator and the hazard model across these changes.

  • Fig. 6 Dominant magnitude.

    Maps of dominant magnitude in the earthquake simulator compared with UCERF3, relationship with UCERF3 fault linkage rules, and impact on long-period hazard. (A) Simulator. (B) UCERF3. (C) UCERF3 fault system connectivity. Colors show connected clusters, with colors indicating cluster sizes in rank order from magenta (largest) to blue (smallest), with only the largest 11 clusters shown. The underlying colors are thus not continuous but a discrete rank ordering. Note that the dominant magnitude in UCERF3 correlates closely with the largest cluster. (D) Difference in dominant magnitude simulator–UCERF3. Note that, on the major faults, UCERF3 tends to have larger dominant magnitudes, whereas on the minor outlying faults, the reverse occurs. The color here shows magnitude difference. (E) Map of long-period long-term hazard difference, showing ln simulator/UCERF3 of PSA(10) at p = 10−4 year−1, a measure for which dominant magnitude differences are expected to have more of an impact. Note that the two areas with substantial dominant magnitude differences of order unity on major faults, around the San Francisco Bay Area and in the Southern San Andreas and San Jacinto fault area, are the two areas that show substantial hazard differences. Otherwise, the hazard differences are modest.

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/4/8/eaau0688/DC1

    Fig. S1. Map of difference in recurrence intervals in the earthquake simulator compared with UCERF3.

    Fig. S2. Parameter sensitivity study.

    Fig. S3. Parameter sensitivity study.

    Fig. S4. Parameter sensitivity example showing the spatial structure of changes in hazard.

    Fig. S5. Parameter sensitivity example showing the spatial structure of changes in hazard relative to UCERF3.

    Fig. S6. Weak magnitude dependence of GMMs.

    References (2529)

  • Supplementary Materials

    This PDF file includes:

    • Fig. S1. Map of difference in recurrence intervals in the earthquake simulator compared with UCERF3.
    • Fig. S2. Parameter sensitivity study.
    • Fig. S3. Parameter sensitivity study.
    • Fig. S4. Parameter sensitivity example showing the spatial structure of changes in hazard.
    • Fig. S5. Parameter sensitivity example showing the spatial structure of changes in hazard relative to UCERF3.
    • Fig. S6. Weak magnitude dependence of GMMs.
    • References (2529)

    Download PDF

    Files in this Data Supplement:

Navigate This Article