Events

Presentation by Dr Men-Andrin Meier.

Despite vast progress across virtually all domains of earthquake seismology, we have not been able to substantially improve levels of earthquake predictability in the past decades. While existing methods for quantifying future earthquake probabilities such as hazard estimates and seismicity forecasts are highly useful, they usually do not provide the probability gains needed to respond to ongoing earthquake sequences in real-time. In this talk, I discuss the current status and prospect of two of the main avenues for improved levels of earthquake predictability: i) diagnostic precursory signals, and ii) next-generation rupture simulators that assimilate real-time data of ongoing earthquake sequences. I highlight how one of the best-monitored earthquake sequences to date contains a surprisingly clear signature of an impending M_w 6.5 earthquake. If this observation is not a chance occurrence, it may imply that diagnostic precursors may indeed precede large earthquake ruptures, but are usually too subtle to be systematically detected under standard monitoring conditions.

At the same time, our ability to detect and characterise subtle faulting processes is dramatically increasing, with advanced AI-powered processing methods and - in some cases - with direct on-fault instrumentation of natural faults. With multi-domain monitoring infrastructure built not just near but directly inside active fault zones, we can monitor also the key non-seismogenic aspects of faulting processes, including slow deformation, stress and pressure fields, and fluid hydro-bio-chemistry. Such on-fault observations will allow us to build and calibrate a new generation of earthquake sequence simulators, which dynamically assimilate data from an ongoing seismic sequence, in order to constrain both the evolving fault network itself, as well as the hydromechanic and fault deformation processes that take place across the network. I discuss how we can use such simulators to establish and push back the limits of earthquake sequence predictability. Building on these developments, I argue that the earthquake sequences of the future will not be analysed with the modular approaches of today, but with powerful digital twin systems that integrate raw data inference and process modelling.