Comparing Adjoint Waveform Tomography Models of California Using Different Starting Models

Adjoint waveform tomography (AWT) sits at the cutting edge of seismic tomography on local, regional, and global scales. However, the choice in starting model may have a significant impact on the final inversion results. In this paper, we present 3 AWT models of California that are based on different starting models. We chose three models that were inverted at different scales: SPiRaL, a global travel-time tomography model (Simmons et al., 2021, 10.1093/gji/ggab277), CSEM_NA, a regional adjoint tomography model of North America and the North Atlantic (Krischer et al., 2018, 10.1029/2017JB015289), and WUS256, a regional adjoint tomography model of the western US (Rodgers et al., 2022, https://doi.org/10.1029/2022JB024549). We then inverted three AWT models using the same source and receiver set. We ran each model over three period bands: 30-100 s, 25-100 s, and 20-80 s. Once the iterations were finalized, we used five methods of testing model similarity in both the model and data space. We conclude that the choice of starting model has a minimal impact on long wavelength models if an appropriate multi-scale inversion approach is used.Plain Language Summary Seismic tomography uses earthquake records to resolve three-dimensional models of seismic wavespeeds below Earth's surface. Full waveform inversion seeks to model the physics of wave propagation to create models that can most accurately reproduce the waveform data that are observed from recorded events. In this paper, we sought to study the effect of the choice of starting model on final inversion results in California using AWT, a type of Full waveform inversion technique; to our knowledge, this type of experiment has not been carried out before with real data at this scale. We began with three starting models that were created on different scales with different datasets. We then used five methods to compare the final model results. We focused on comparisons of the visual similarity of the models, the tectonic structures the models resolve, and how well the synthetic waveforms match the observed data. We find that the choice of starting model does not have a large impact when the model domain has azimuthally distributed events and stations over long length-scales.