Why are repeating earthquakes so infrequent: partial ruptures or slowly-slipping fault strands?

Repeating earthquakes are earthquakes that repeatedly rupture particular patches of a fault. One might therefore expect that we could estimate the recurrence time of a repeater by dividing the slip per earthquake by the long-term fault slip rate. However, repeaters in Parkfield and globally recur much less frequently that one would expect given their slip. Smaller repeating earthquakes recur particularly infrequently (relative to the expectation). Here we attempt to assess two explanations for repeaters' long recurrence intervals.First, we assess whether repeating earthquakes have long recurrence intervals because smaller, partial ruptures occur between the characteristic ruptures, accommodating much of the slip (Cattania and Segall, 2019). We find that many repeating earthquakes do have partial ruptures. The existence of the smaller ruptures suggests that many repeaters are significantly larger than their nucleation size. However, the partial ruptures accommodate only about 20% of the slip on the repeating patch. The slip in partial ruptures cannot explain why many recurrence intervals can be 5to 50 times longer than one would expect given seismic observations.Next, then, we consider a different explanation of repeaters' long recurrence times: that repeaters occur on small, slowly slipping fault strands that make up the large-scale fault zone, whose slip rate is the sum of the strands' slip rates. In this model, smaller earthquakes occur particularlyinfrequently because they occur on smaller fault strands, which slip more slowly (Williams et al, 2019). Here we further develop this model by determining the earthquake and fault sizes required to reproduce repeaters' observed recurrence intervals. We find that we can reproduce the recurrence interval-moment scaling if earthquakes and fault strands have power-law size distributions, with similar power-law coefficients, and we estimate the size distribution of faults likely to host variously earthquakes. Finally, we look for these sizes by examining clustering of earthquakes around M1, M2, and M3 earthquakes. Larger earthquakes appear to be surrounded by more spread-out earthquake clusters, which could represent larger fault areas. However, with the data used so far, the change in fault size is only marginally significant.