How Does the San Andreas Fault Extend into the Deep Crust? Testing Wavespeed Models of Southern California Using Gravity Modeling

SCEC Community Velocity Models CVM-H and CVM-S have large discrepancies with one another and with other published models e.g. Barak et al. (2015, G-cubed, hereafter BKL-15). Each model has different implications for the development of the San Andreas Fault System (SAFS). The 'propeller' model of a NE-dipping SAF south of the San Gorgonio Pass is strongly supported by BKL-15 that purports to show eastward-directed underthrusting of a lithospheric-scale wedge of the Peninsular Ranges Batholith (PRB) and the SAFS dipping NE to or through the base of the crust. CVM-H and CVM-S offer no strong support for this hypothesis.

Here we test which model (BKL-15, CVM-H15.1, or CVM-S4.26) best predicts the observed regional gravity field. We sample all three 3D models at the same voxel size, 1 km vertical, 0.05° lat./long., and extracted 2D cross sections with ~10-km cross-strike smoothing. We apply the Brocher (2005, BSSA) wavespeed-to-density conversion then use freeware FastGrav to calculate predicted gravity along each transect. Since the purely tomographic models have poor resolution at shallow depths, and all the models have degraded resolution below the Moho, we only model the gravity field due to wavespeed variations between 5-45 km depth. We remove DC offsets from the model and use the mean rms deviation between wavespeed-predicted gravity and observed gravity as our metric for goodness of fit.

Along three transects across the Peninsula Ranges Batholith, spanning a strike-length of >200 km, BKL-15 (rms=7 mGal) is marginally better than CVM-H (rms=17 mGal), but CVM-S has a substantially higher rms misfit of 42 mGal, equivalent to an rms error in a 10-km thick layer of 0.1 g/cm3, or a change from Vp=6.0 to 6.5km/s. The wavespeed model that best fits observed gravity also supports eastward underthrusting of the PRB, perhaps controlling the propeller shape of the San Andreas Fault.