Marine Denolle

Dynamic characterizations of earthquakes focus on whole-event representations, that is whether the total radiation of seismic waves is more or less energetic. Denolle et al (2015) and Yin et al. (2018) suggest to use the source spectrogram in order to analyze the radiation during the rupture itself. Here, we take a retrospective view on these studies to better establish the methodology of source spectrogram, and highlight its strengths and limitations. We provide clear interpretation of the temporal evolution of the source spectrogram through time-variant high-frequency falloff rate and radiated energy rate using canonical kinematic and pseudo-dynamic examples. The radiated energy rate provides the amount of energy radiated through time and its integral is the total radiated energy. It is most sensitive to fault heterogeneities in the local slip-rate function and its peak, and in rupture velocity. The high-frequency falloff rate peaks at times of zero moment acceleration, but remains constant otherwise and theoretically equal to one. The M7.8 2015 Nepal earthquake exemplified the propagation of a slip pulse and is thus perfectly suited to demonstrate this approach. We use 3D empirical Greens functions to remove wave propagation effects and construct the P-wave source function. We then construct spectrograms and explore the variations in the radiated energy rate functions. We find that, as expected from unilateral dislocation models, the Nepal earthquake radiated seismic waves at the beginning and at the end of the rupture, but not during the phase of high moment release. Finally, we interpret our results in light of rupture dynamics, i.e. the earthquake initiation, propagation, and arrest.