Title: Seismological evidence for hidden crustal fluid diffusion transients in California
Earthquake swarms are primarily a consequence of aseismic driving processes that occur transiently, but the details of these processes are still unclear. We use deep learning algorithms to produce a significantly enhanced catalog of seismicity for Southern California and better resolve the details of swarms. Near the San Jacinto fault zone, a major strike-slip fault system, we identify nearly one hundred swarms that have durations in the range 0.5-6 years. These swarms are not directly driven by tectonic processes yet constitute about 1/4 of the total seismicity. We find that about half of the swarms exhibit ultra-slow diffusive migration patterns, along with propagating backfronts, in which a shut-down of seismicity propagates away from the source region. These observations are all consistent with expectations for natural fluid injection processes. Furthermore, we find evidence that permeability along the fault zone increases exponentially with time elapsed. The chronology of the swarms indicates that these aseismic driving processes were active in some part of the region throughout 2008–2020. The observations challenge common views about the nature of swarms, which would characterize any one of these sequences as anomalous. The regional prevalence of these sequences suggests that transient fluid injection processes play a key role in crustal fluid transport.