Preprint: Tremor migration patterns and the collective behavior of deep asperities mediated by creep

Abstract

Slow-slip events (SSE) and non-volcanic tremors have revealed a broad spectrum of earthquake behavior, involving entangled seismic and aseismic slip, and offer a unique window into fault mechanics at the bottom of seismogenic zones. A hierarchy of migration patterns of tremors has been observed in the Cascadia subduction zone, including large-scale along-strike tremor propagation and Rapid Tremor Reversals (RTR) migrating in opposite directions with much higher propagation speeds. Here we show that these tremor migration patterns can be reproduced by two end-member models of a fault with heterogeneous mechanical properties, composed of competent asperities embedded in a more frictionally stable, incompetent matrix. In the SSE-driven-tremor model, SSEs are spontaneously generated by the matrix, even in absence of seismic asperities, and drive tremor. In the tremor-driven-SSE model the matrix is stable, it slips steadily in absence of asperities, and SSEs result from the collective behavior of tremor asperities interacting via transient creep in the form of local afterslip fronts. We study these two end-member models through 2D quasi-dynamic multi-cycle simulations of faults governed by rate-and-state friction with heterogeneous frictional properties and effective normal stress, using the earthquake simulation software QDYN . In both models, tremor migration patterns emerge from interactions between asperities mediated by creep transients. The models successfully reproduce forward tremor propagation and RTRs, as well as various other observed tremor migration patterns, without the need to finely tune model parameters. Our modeling results suggest that, in contrast to a common view, SSE could be a result of tremor activity. Also, the hierarchical pattern of tremor migrations provides general constraints on fault zone rheology, and the location of RTRs and other tremor patterns might shed light on the finer scale spatial variability of fault properties. We also find that, despite important interactions between asperities, tremor activity rates are proportional to the underlying aseismic slip rate, supporting an approach to estimate SSE properties with high spatial-temporal resolutions via tremor activity.

Variations in slow slip moment rate associated with rapid tremor reversals in Cascadia

“…The rapid slip rates in RTRs could be facilitated not just by time‐dependent variations in fault healing, but also by spatial variations in fault properties, as suggested in the models of Ando et al . [], Luo and Ampuero [], and Ando et al . [].…”

“…Indeed, if the regions prone to RTRs do have lower normal stress and lower resistance to rupture, it is puzzling that these regions do not simply slip more as the main front passes through, and thus release all of the available stress drop before RTRs could occur. One option is that slip at asperities in the RTR‐prone regions take longer to nucleate than slip on the rest of the fault [ Luo and Ampuero , ], so that the stress drop occurs later. However, it is not clear whether such a delayed‐slip model can reproduce RTRs that repeatedly rupture a single location or RTR regions that host significant tremor before and between RTRs, as are suggested by the data [ Obara , ; Houston et al ., ; Royer et al ., ; Y. Peng et al ., ].…”

“…Tremor migration patterns like those in RTRs are often interpreted and modeled as indicators of changing aseismic slip rates [ Ando et al ., ; Luo and Ampuero , ; Rubin , ; Ando et al ., ; Hawthorne and Rubin , ]. In this context, RTRs may be thought of as subevents that nucleate and spread quickly within the larger slow slip event.…”

High‐resolution images of tremor migrations beneath the Olympic Peninsula from stacked array of arrays seismic data

“…This seems to make their model RTRs poor analogs for those observed. The observed behavior seems more consistent with the simulations of Luo and Ampuero [], where modeled tremor reversals are triggered by the failure of relatively large brittle asperities just behind the main slow slip front. When their background fault is velocity‐strengthening at steady state, these reversals can propagate much faster than the average speed of the main front, as observed.…”

“…If correct, this implies that tremor may not be merely a passive indicator of the underlying creep, but could sometimes significantly influence the propagation of slow slip, even though the seismic moment release in the frequency range of tremor is only about 0.1% of the total geodetic moment of the contemporaneous slow slip [ Kao et al ., ]. Nonetheless, it is not yet clear whether the spatially heterogeneous but frictionally simple model of Luo and Ampuero [] can account for the repeated occurrence of RTRs in the same source region.…”