Fig. 3: Slip timing estimations, patterns identified by our model, and comparison to shear experiments in the laboratory.
From: An exponential build-up in seismic energy suggests a months-long nucleation of slow slip in Cascadia
a ML estimates in testing of time to failure (in blue), and ground truth time to failure (in red) from the PNSN tremor logs. For most events, estimates are close to zero for several weeks preceding the rupture. b The most important feature identified by our model plotted against time, for the best stations (B001), for time window intervals of 3 months (black curve). This energy-based feature shows clear patterns with respect to the time remaining before the next slow slip event (left axis). c Shows the best statistical feature found in laboratory slow slip experiments for comparison (acoustic power, right-hand vertical axis). The best features in the laboratory c and Cascadia b are related to the energy of the seismic waves, and appear to follow similar patterns, with a progressive increase in amplitude that peaks towards the end of each slow slip event.
