Earthquake Forecasts for Gorkha Immediately Following the 25th April, M=7.8 Mainshock

Abstract:

The M-7.8 Gorkha (Nepal) earthquake on the 25^th April, 2015 has shaken the central Himalayan front and immediately raised concerns for the severity of future triggered earthquakes. Here, we implement standard and innovative forecast models to predict the spatio-temporal distribution of triggered events. Key challenges addressed are: 1) the limited information on early aftershocks, 2) the low-productivity aftershock sequence in the near-source area, 3) the off-fault (>250 km) triggered events exemplified by the M=5.4 Xegar event, 3 hrs after the mainshock. We apply short-term empirical/statistical ETAS and physical forecast models, the latter based on the combination of rate/state friction law and Coulomb stresses. Within the physics-based model implementation we seek to evaluate the uncertainty related with the rupture style of triggered events by considering: 1) the geometry of active structures, 2) optimally oriented for failure faults and 3) all-potential faults described by the total stress field. The latter is represented by the full stress tensor before and after the mainshock and our analysis suggests that the preseismic stress magnitudes are still sufficient to cause earthquakes even after modification by the mainshock. The above remark reveals that there are no “stress shadows” affecting the spatial distribution of near-field aftershocks. It is also noted that the method allows for an a-priori determination of the rupture plan of the M=7.3 event, within the limit of uncertainty (20˚). The results show that: (1) ETAS models underestimate the number of observed events, since they heavily base their good performance in small magnitude earthquakes, not available in the first few weeks after the mainshock, (2) far field triggered events are captured only by physics-based forecasts, and (3) the total stress method improves the predictability of larger magnitude events. We conclude that frontier regions benefit from the implementation of physics-based models since they have comparable to improved performance when compared with empirical/statistical models with the advantage of less demanding input parameters at real-time experiments.