Relaxing Segmentation: Does It Improve Characterization of Fault Rupture Behavior?

Most faults have not ruptured once historically, let alone repeatedly. Estimating future rupture length of an earthquake source has been a challenge since the 1970s when concepts of full and half fault lengths were employed. In the 1980s paleoseismic event timing and observations of slip, coupled with geometric and other physical fault changes, led to concepts of fault segmentation and it’s modeling for hazard. The Uniform California Earthquake Rupture Forecast 3 (UCERF 3, Field et al., 2014) relaxed segmentation, guided by rules in which a separation distance of ≤5km and orientation to Coulomb stress changes at fault junctions are prime factors for allowing fault-to-fault jumps. A set of ~350 fault sections produced ~250K ruptures ranging in length from 15 km-1200 km. An inversion provided the rates of these, which range from 10²-10⁸ years. Many of the long ruptures have exceedingly low individual rates within the UCERF 3 geologic model but are sufficient in number to release cumulative moment that brings the long-term (Myr) and historical (since 1850) MFDs for the California region into close agreement.

Does UCERF 3 have too many multi-fault ruptures? Since 1850 there have been ~260 surface ruptures worldwide in shallow continental crust. 77% are 0-49km; 6% exceed 150km, and the longest is 1906 San Francisco (435-470 km). In California since 1857 there have been 31 surface ruptures. 77% are shorter than 49 km. The longest are 1906, 1857 Fort Tejon (297km), and 1872 Owens Valley (108 km). Most long historical strike-slip ruptures are continuous and geomorphically well-defined traces with limited geometric changes. In contrast, UCERF3 modeling of the south Hayward, as an example, allows it to participate in ruptures that extend to the south ends of the San Andreas or San Jacinto faults (900 km). These include branching (Hayward-Calaveras, San Andreas-San Jacinto) and jumps (Calaveras-San Andreas) on creeping sections of these faults. 5km is the connectivity threshold in UCERF 3 but only 40% of historical ruptures have negotiated this distance. There are other controls of rupture propagation: frictional properties, rupture dynamics, creep, and, as shown by the 2002 Denali-Totschunda rupture, the timing of the prior event and level of stress accumulation on adjacent fault sections. Time will be the tester.