Evidence against segmented earthquake rupture behavior

Jeanne Hardebeck USGS, Menlo Park, CA Segmentation as a useful concept for predicting fault rupture endpoints: a. The evidence for or against fault ruptures arresting at fault segment boundaries b. Potential barriers to rupture, if any, along the faults identified near DCPP, including the Hosgri and Shoreline faults. What is fault “segmentation”? o Model assumes that faults are divided into segments, separated by boundaries that act as barriers to earthquake rupture. o Model assumes that earthquakes are restricted to a single segment, or to a limited set of “multi-segment” rupture scenarios. o Segment boundaries often defined to correspond to irregularities of the fault surface trace: • bends, • offsets, • intersections with other faults, • changes in slip rate, • other geometrical irregularities. o Segments often defined to correspond to rupture zones of past earthquakes. Segments often defined to correspond to the rupture zones of past earthquakes.

o Irrelevant at this time for DCPP because we don’t have a record of past earthquakes. o If you’re interested in what I might have to say about this, you can watch a debate on segmentation archived by SSA: http://www2.seismosoc.org/members/2012-presentations/eq-debate1.php Shoreline Fault, as currently segmented by PG&E: Boundaries at surface trace gaps, intersection with San Luis Bay fault. Ruptures can’t span Hosgri and Shoreline faults. • Highest probability in logic tree: Lmax = 16 km Mmax ≈ 6.2 • Maximum considered: Lmax = 23 km Mmax ≈ 6.5 • Shoreline + Hosgri: Zero Probability. PG&E Shoreline Fault Report, 2011 Mmax ≈ 7.2 Hosgri Fault, as currently segmented by PG&E: Boundaries at fault intersections, ruptures can’t span Hosgri and San Simeon faults. • Highest probability in logic tree: Lmax = 45 km Mmax ≈ 6.8 • Maximum considered: Lmax = 110 km Mmax ≈ 7.1 • Hosgri + San Simeon: Zero Probability.

PG&E LTSP Report, 1988 Mmax ≈ 7.3 (UCERF2) Numbers from PG&E Shoreline Fault Report, 2011 Key Questions:

1)Are fault offsets or gaps barriers to earthquake rupture?

2)Are fault intersections barriers to earthquake rupture?

3)Can earthquakes begin/end somewhere other than at a fault trace irregularity?

4)Are multi-fault ruptures possible?

5)Is segmentation a reasonable simplifying assumption for hazard assessment?

6)If not, is there a good alternative? 1. Are fault offsets or gaps barriers to earthquake rupture?

Surface trace gaps and offsets: basis for segmentation of Shoreline Fault.

gap: ~3 km offset: ~1 km (?) (Hans’ talk Tuesday)

offset: ~0.35 km

gap: ~2.5 km

PG&E Shoreline Fault Report, 2011 1. Are fault offsets or gaps barriers to earthquake rupture?

Not the small surface trace gaps and offsets of the Shoreline Fault.

Shoreline gaps (compare with dots and triangles)

Shoreline offset (compare with rectangles and diamonds) Wesnousky, Nature, 2006. 1. Are fault offsets or gaps barriers to earthquake rupture?

Offsets at seismogenic depths: none identified, but given location uncertainty, hypothetical offsets of <1-2 km might go undetected. 1. Are fault offsets or gaps barriers to earthquake rupture?

Dynamic rupture modeling shows that 1-2 km offsets in vertical strike-slip faults are not barriers to rupture propagation.

1999 Izmit, Turkey, Harris et al., BSSA, 2002 2. Are fault intersections barriers to earthquake rupture? • Basis for all segmentation of Hosgri Fault south of San Simeon fault.

• Basis for highest-weighted logic tree branch for Shoreline Fault (barrier at intersection with San Luis Bay fault.)

PG&E LTSP Report, 1988 PG&E Shoreline Fault Report, 2011 2. Are fault intersections barriers to earthquake rupture? Not even the intersection with a tectonic plate boundary poses a barrier to earthquake rupture.

2007 M8.1 Solomon Islands earthquake, Furlong et al., Science, 2009 3. Can earthquakes begin/end somewhere other than at a fault trace irregularity?

YES, this happens frequently.

Wesnousky, Nature, 2006. 4. Are multi-fault ruptures (branches/bends) possible?

Shoreline-Hosgri joint ruptures could occur if:

• rupture on a smaller fault can continue onto a larger fault, and/or

• rupture on a main fault can branch onto a smaller fault, and/or

• rupture can branch in an unfavorable direction, and/or

• ruptures can take a ~30° bend. 4. Are multi-fault ruptures (branches/bends) possible?

Examples of multi-fault strike-slip earthquakes: - Bends to either side (e.g. Landers, Denali). - Begin on smaller faults and propagate onto larger faults (e.g. Denali). - Slip on branches off of main fault (e.g. Denali) or on multiple branches (e.g. Hector Mine).

1992 Landers, California, Sieh et al., Science, 1993

1999 Hector Mine, CA Ji et al., BSSA, 2002

2002 Denali, Alaska, Eberhart- Phillips et al., Science, 2003 4. Are multi-fault ruptures (branches/bends) possible? Four faults, including two unfavorable 90° bends.

2012 M8.6 Sumatra earthquake, Meng et al., Science, 2012 4. Are multi-fault ruptures (branches/bends) possible? Hypothetical Hosgri-Shoreline rupture looks quite tame in comparison to some recent strike-slip earthquakes. 5. Is segmentation a reasonable simplifying assumption for hazard assessment? May neglect important possible earthquakes, e.g. Tohoku.

National Seismic Hazard Maps for Japan (2005), Headquarters for Earthquake Research Promotion 5. Is segmentation a reasonable simplifying assumption for hazard assessment?

Segmentation puts bulls-eyes of higher hazard around segment boundaries. Because of the two earthquakes on either side of the boundary, they get “hit twice”.

Segmentation reduces hazard at locations that aren’t chosen as segment boundaries. Model says these locations can’t be earthquake end points and therefore can’t get “hit twice”. Boundary of Mojave & San Hazard Map with NSHMP 2002 Model Bernardino SAF segments

Boundary of two SJF segments (SB & SJ) From Ned Field 5. Is segmentation a reasonable simplifying assumption for hazard assessment?

Segmentation puts bulls-eyes of higher hazard around segment boundaries. Because of the two earthquakes on either side of the boundary, they get “hit twice”.

Segmentation reduces hazard at locations that aren’t chosen as segment boundaries. Model says these locations can’t be earthquake end points and therefore can’t get “hit twice”.

Using “floating” earthquakes of “characteristic” size is an internally inconsistent solution – if segment boundaries don’t correspond to rupture terminations, then they can’t control earthquake size. 6. Alternative to segmentation – model rate of all plausible earthquakes, similar to UCERF3 “grand inversion”.

Field and Page, BSSA, 2011 6. Alternative to segmentation – model rate of all plausible earthquakes, similar to UCERF3 “grand inversion”.

• Model for DCPP area would be much simpler and easier to understand than UCERF3 model. Many fewer relevant faults, many fewer plausible earthquakes, easier to visualize all of them.

• Could capture soft segmentation of slip rate changes at fault intersections. Recommendations:

- Shoreline and Hosgri Faults should not be segmented: o Any offsets or gaps in the Shoreline Fault are too small to be barriers to earthquake rupture. o Fault intersections are not barriers to earthquake rupture, so shouldn’t be used as hard segment boundaries.

- Multi-fault ruptures should be considered: o Shoreline-Hosgri joint rupture should not be ruled out. o Hosgri and San Simeon faults are a single continuous fault (see Sam Johnson’s Tuesday talk) and should be treated as such.

- A model similar to UCERF3 would be an appropriate alternative: o Could capture soft segmentation of slip rate changes at fault intersections. o Not as complex as UCERF3 model because there are fewer faults and fewer plausible ruptures.