Evidence for and Implications of Rupture Velocity Acceleration During Earthquake Rupture Onsets

Much of what is known about earthquake rupture processes comes from the study of small earthquakes. While, for sufficiently large earthquakes, key source characteristics can directly be inferred from waveform observations, observational resolution limits preclude such direct estimations in the study of small earthquakes. As a consequence, some characteristics have to be assumed or extrapolated from studies of large earthquakes. In particular, rupture velocities ????r are typically assumed to reach their steady-state value quasi-instantaneously after rupture initiation. Here we show that this assumption is irreconcilable with the bulk of available seismic near-source observations. We use an extensive near-source record data set of shallow crustal earthquakes from a broad magnitude range (????2.0 ? ????7.6) to attain a maximally objective and parameter-free characterization of ground motion onsets. We find that shallow crustal earthquakes develop in two phases, each of which has distinct statistical properties: the events almost invariably start with an observable period of rapid potency rate growth, dP / dt ? t^?, with ? > 4 independent of event magnitude. These growth rates strongly exceed predictions of standard rupture models (?(pred) = 2). Small earthquakes reach peak P-phase amplitudes during this short initial period of 0.1-0.8 seconds. Large earthquakes subsequently transition to growing in the second phase with significantly lower growth rates, ?~1.5. We use scaling relations to demonstrate that the high growth rates in the first phase are likely caused by increasing rupture velocities ????r. Our observations suggest that small earthquake ruptures never reach their steady-state ????r. This implies that the common practice of assuming constant ????r in the study of small earthquakes source properties may introduce a strong bias and that some fundamental source characteristics, including spatial and temporal rupture dimensions and static stress-drops, have to be revised. Furthermore, the observed potency rate decrease after the first phase points to a fundamental, yet surprisingly obvious, scaling break in earthquake source properties.