Time-Dependent Earthquake Forecasts on Oceanic Transform Faults

Abstract:

Earthquakes on oceanic transform faults (OTFs) exhibit many of the most systematic and predictable behaviors known in seismology. Many moderate to fast-slipping oceanic transform faults have small maximum magnitudes (Mw ~5.5-6.6) and short seismic cycles (4-20 years), making these faults ideal for investigating whether time-dependent models, which concentrate hazard towards the end of a seismic cycle, are appropriate for geometrically simple plate boundary faults. Using a surface wave cross-correlation relocation technique, I identify 19 fully-coupled rupture patches on six East Pacific Rise and Juan de Fuca Ridge OTFs with multiple earthquakes between 1990-2020. In the best-studied case, Gofar Transform Fault, the largest earthquakes (5.7 ≤ M_W ≤ 6.2) repeatedly ruptured the same portions of the fault, while the intervening fault segments hosted swarms of microearthquakes. These patterns were successfully used to capture two large earthquakes with ocean bottom seismic experiments: the 2008 Mw 6.0 on western Gofar and the 2020 Mw 6.1 on eastern Gofar that was recorded by our currently active deployment.

Here I build time-dependent earthquake recurrence models to quantify the predictability of OTF earthquakes. I create synthetic event catalogs of background seismicity and repeating large earthquakes. The background seismicity is governed by a Poissonian process in time and a tapered Gutenberg-Richter magnitude distribution. The large earthquakes are time-dependent and are drawn from a Gaussian magnitude distribution peaked at the median of the observed event magnitudes. The total moment rate on each rupture patch is required to match that observed in the Global Centroid Moment Tensor catalog. I compute the probability that the parameters explain the observed inter-event times from the Bayesian combination of three time-dependent models: the Brownian Passage Time (BPT), Weibull, and lognormal (LogN) distributions. Initial results show a best fitting model with weights of 17% BPT, 61% Weibull, and 22% LogN, with a mean recurrence interval of 5.5 years and a coefficient of variation (COV) of 0.2. The low COV and stable earthquake patterns indicate quasi-periodic OTF seismic cycles, with the rupture patch locations controlled by along-strike variations in fault zone properties.