Regionalization of a California-wide seismic tomography model

Laurent G. Montesi, & Arfan Rashid

Submitted August 30, 2026, SCEC Contribution #15452, 2026 SCEC Annual Meeting Poster #083

Due to its long geological history, the continental crust is composed of many heterogeneous fragments separated by sharp faults and boundaries. Better understanding this heterogeneity and its effect on the stress field can help refine estimates of seismic hazard. However, it is challenging to fully understand the make-up of the continental crust at depth. Geological studies can define lithotectonic blocks but are subject to numerous assumptions and extrapolations. Seismic tomography provides a three-dimensional image of the crust but cannot uniquely constrain rock types. Moreover, tomography typically assumes smooth variations in material properties, which contrasts with the sharp contact of geological units and lithotectonic blocks. Combining these two approaches can help determine which heterogeneities are significant. Eymold and Jordan (2019) pioneered the approach of clustering a seismic tomography model, in their case SCEC’s CVM‐S4.26, and show that the geographical distribution of its clusters resembles that of SCEC’s Southern California Geological Framework (GFM). Here, we present clustering results of the Statewide uwlinca model (Lin et al., 2010). We extracted velocity profiles from the original model grid without interpolation and retained only those for which the derivative weight sum (DWS) is greater than 300 at every depth. The resulting dataset includes 1476 profiles with 8 depth horizons. We tested several clustering algorithms and found statistically favorable solutions with 4 or 5 clusters using K-means or Agglomerative clustering. Density-based clustering algorithms failed to return a valid solution. We attribute the low number of clusters to the smoothing inherent in seismic tomography, which results in gradual transitions from one block to another, even if their cores have markedly different velocity profiles. Nevertheless, the major geological divisions of California appear in our clustering solutions. The Sierra Nevada and the Great Valley stand out most prominently, the first with a thick low-velocity lower crust, the other with a slow upper crust. Some solutions divide the Great Valley into Western and Eastern domains, likely reflecting a gradual change. Other regions feature a sharp velocity jump at depth, likely the Moho, but systematically shallower in the Coast Ranger than in Southern California. Other geological divisions cannot be confidently inferred from the regionalization of uwlinca.

Key Words
Clustering, tomography, blocks

Citation
Montesi, L. G., & Rashid, A. (2026, 08). Regionalization of a California-wide seismic tomography model. Poster Presentation at 2026 SCEC Annual Meeting.

Related Projects & Working Groups
Community Earth Models (CEM)