We construct a 3D Vs model in San Gabriel and San Bernardino basins using ambient noise correlation between dense array nodal, broadband and accelerometer stations.• We separated the Rayleigh wave fundamental mode and first higher mode in dispersion analysis based on the Rayleigh wave particle motion.• Our Vs model predicts deeper and slower sedimentary basins than the SCEC CVMS model, yet is consistent with geological and drilling data in these basins.

Summary The initiation and development of subduction zones are associated with substantial stress changes both within plates and at plate boundaries. We formulate a simple analytical model based on the force balance equation of a subduction zone, and validate it with numerical calculations of highly non–linear, coupled thermo–mechanical system. With two kinds of boundary conditions with either fixed velocity or fixed force in the far field, we quantitatively analyze the role of each component in the force balance equation, including slab pull, inter–plate friction, plate bending, and basal traction, on the kinematics and stress state of a subducting plate. Based on the numerical and analytical models, we discuss the evolution of plate curvature, the role of plastic yielding and elasticity, and how different factors affect the timing of subduction initiation. We demonstrate with the presence of plastic yielding for a plate of thickness, H, that the bending force is proportional to H2, instead of H3 as previously thought. Although elasticity increases the force required to start nucleating subduction it does not substantially change the total work required to initiate a subduction zone when the yielding stress is small. The analytical model provides an excellent fit to the total work and time to initiate subduction and the force and velocity as a function of convergence and time. Plate convergence and weakeing rate during nucleation are the dominant factors influencing the force balance of the plate, and 200 km of plate convergence is typically required to bring a nascent subduction zone into a self-sustaining state. The closed–form solution now provides a framework to better interpret even more complex, time–dependent systems in three–dimensions.

We construct a 3D Vs model in San Gabriel and San Bernardino basins using ambient noise correlation between dense array nodal, broadband and accelerometer stations.• We separated the Rayleigh wave fundamental mode and first higher mode in dispersion analysis based on the Rayleigh wave particle motion.• Our Vs model predicts deeper and slower sedimentary basins than the SCEC CVMS model, yet is consistent with geological and drilling data in these basins.

Passive seismic and gravity measurements are combined to estimate the 3D depth of the sediment-basement interface.• The maximum depth in the San Gabriel basin is 4.5 km, and Chino and San Bernardino basins are less than 2 km deep.• The trace of the Raymond fault is delineated in the gravity anomaly of the San Gabriel basin.

The San Gabriel (SG) and San Bernardino (SB) basins are sedimentary basins northeast of the city of Los Angeles (Figure 1). The SG basin consists of two sub-basin structures: the Raymond basin on the west and the San Gabriel basin on the east, separated by the Raymond fault. The SB region, immediately to the east of SG, is composed of three sedimentary basins: the Chino, Rialto-Colton, and San Bernardino basins from west to east. Bounded by mountains both to the north and south, the sedimentary structures in the SG and SB area were part of the opening of the Los Angeles basin region in the Miocene.Understanding the velocity structure of SG and SB area is important for the accurate hazard assessment of the densely populated Los Angeles region because the low-velocity basins in the SG and SB area may function as a waveguide that channels earthquake energy from the San Andreas Fault (SAF) into the Los Angeles region (Olsen et al., 2006). Numerical simulations such as the ShakeOut Scenario (Jones et al., 2008) and CyberShake (Graves et al., 2011) show events on the southern SAF may cause large ground motions in downtown Los Angeles. A study using ambient noise correlation estimate (Denolle et al., 2014) found the ground motion could be four times larger than the simulation. This implies the current Southern California Earthquake Center (SCEC) Community Velocity Model (CVM) used in the ground motion simulations do not adequately account for the channeling effect of the northern sedimentary basins (Clayton et al., 2019). A recent study in the Los Angeles basin constrains the velocity model using dense industry arrays correlated with broadband stations (Jia & Clayton, 2021), and the new fine-scale velocity model's strong motion amplification performs similar to the CVMH model but better than the CVMS model, two popular community models used in seismic hazard estimates. Similarly, Ajala &