Clustering Interseismic Strain Rates Near the Mendocino Triple Junction

Abstract:

The Mendocino Triple Junction, where the Pacific, North American, and Gorda plates meet, is surrounded by high levels of active deformation with long-term uplift rates reaching 4mm/year. This region is also densely forested and experiences high weathering rates, leading to large uncertainties about the locations of, and activity on modern fault traces. To assess deformation in the poorly-constrained region, we look to GPS velocity-derived strain rates. However, the multidimensionality of strain rate tensors is difficult to map spatially: the typical approach involves computing many invariant components from the tensor field. This results in maps that are challenging to interpret without knowledge of previously-mapped fault systems, leading to interpretation bias in poorly-mapped regions, where exact locations, locking depths, and fault mechanisms may be unknown.

By calculating and grouping strain rates, we can assess the implications of active deformation and interpret the relative stability of various regions. Using five distinct methods, we computed interseismic strain rate tensors across the region from GPS velocity vectors, and compared these methods to find areas of high and low consistency between methods. We used three-dimensional k-means clustering to group these strain rate tensor fields into domains of similarity to assess the boundaries of distinct deformation regimes. We performed this clustering across a spectrum of k-values to establish how many strain types were represented in the region. Gathering strain rates into clusters will help associate local deformation in specific regions with one or more fault processes. These a priori clusters are also easier to visualize and interpret, with less bias than previous strain-rate mapping methods.