Seismotectonic Implications Of Clustered Regional GPS Velocities In The San Francisco Bay Region, California

We have used a hierarchical agglomerative clustering algorithm with Euclidean distance and centroid linkage, applied to continuous GPS observations for the Bay region available from the U.S. Geological Survey website. This analysis reveals 4 robust, spatially coherent clusters that coincide with 4 first-order structural blocks separated by 3 major fault systems: San Andreas (SA), Southern/Central Calaveras-Hayward-Rodgers Creek-Maacama (HAY), and Northern Calaveras-Concord-Green Valley-Berryessa-Bartlett Springs (NCAL). Because observations seaward of the San Gregorio (SG) fault are few in number, the cluster to the west of SA may actually contain 2 major structural blocks not adequately resolved: the Pacific plate to the west of the northern SA and a Peninsula block between the Peninsula SA and the SG fault. The average inter-block velocities are 11, 10, and 9 mm/yr across SA, HAY, and NCAL respectively. There appears to be a significant component of fault-normal compression across NCAL, whereas SA and HAY faults appear to be, on regional average, purely strike-slip. The velocities for the Sierra Nevada - Great Valley (SNGV) block to the west of NCAL are impressive in their similarity. The cluster of these velocities in a velocity plot forms a tighter grouping compared with the groupings for the other cluster blocks, suggesting a more rigid behavior for this block than the others. We note that for 4 clusters, none of the 3 cluster boundaries illuminate geologic structures other than north-northwest trending dominantly strike-slip faults, so plate motion is not accommodated by large-scale fault-parallel compression or extension in the region or by significant plastic deformation , at least over the time span of the GPS observations. Complexities of interseismic deformation of the upper crust do not allow simple application of inter-block velocities as long-term slip rates on bounding faults. However, 2D dislocation models using inter-block velocities and typical depths to locking on the faults, with the addition of assigning 6-7 mm/yr slip to the San Gregorio Fault (based on the geologic rate), fit the available GPS data as well or better than similar models using presently accepted long-term slip rates (Wills et al., 2008). The main differences between the two suites of long-term slip rates are that rates derived from inter-block velocities place significantly less slip on SA (about 11 vs. 17 mm/yr) and more on NCAL (about 9 vs. 6 mm/yr), and that the total slip rate is somewhat less (about 37 vs. 41 mm/yr). Continuing the clustering process beyond 4 clusters yields spatially coherent groupings of velocities that reveal the elastic interseismic strain of blocks between faults, as well as highlighting some secondary intra-block faults such as the West Napa Fault: The sub-blocks separated by the West Napa Fault have an average inter-block velocity of about 4 mm/yr. The Greenville Fault is not well delineated by cluster analysis, suggesting that its long-term slip rate is low and/or its locking depth is anomalously deep.