A suite of 4 proposed in-situ experiments will be performed in the BedrettoLab, a unique underground laboratory located in the Bedretto tunnel in the Swiss Alps. The Bedretto tunnel is a 5 km long side-track of the Furka railway tunnel leading through the crystalline Gotthard massive, with a local overburden of over 1000 m. To directly observe earthquake nucleation and propagation behaviour, four experiments at a scale of 50 m will be performed along a single fault, reactivated by high-pressure fluid injections (i.e. hydraulic stimulations). Stress changes artificially imposed onto the fault (i.e. stress preconditioning) will be used to investigate the role of stress heterogeneities and criticality on rupture nucleation and propagation behaviour. To accommodate the four experiments, we will excavate a 200 m long tunnel parallel to the target fault. The rock volume around the fault will be characterized in terms of rock mass properties (geological micro- and macrostructure, lithology, geochemistry, fault zone structure), hydrogeological conditions (in-situ pressure, transmissivity structure within and around the fault), as well as stress field heterogeneities along and across the fault. Monitoring instrumentation will be installed in the tunnels and in boreholes accessing the fault from the tunnels, including fiber-optics-based strain sensing, stress cells, tiltmeters, 3D dislocation sensors, pressure sensors, acoustic emission sensors, and high frequency accelerometers. Each experiment will test a unique set of pre-stress conditions. The exact experimental strategy and schedule will depend on the progress of the lab build out, but the following experimental phases are currently planned:
In addition to the various quasi-static stress and strain observations, the FEAR experiments will generate a rich database of continuous seismic records over a large frequency and energy range, from nano-earthquakes and micro-creep events (dimension 10cm-1m) to ruptures larger than 10m. This seismic data will be analyzed in real-time (or near-real time?) with state-of-the-art methodologies (e.g. template matching, deep learning powered seismic detection and characterization methods, waveform-based absolute and relative relocation methods, etc.). The resulting earthquake catalogues as well as other geophysical observations (strain, pressures, etc.) will flow as inputs into a novel test-bench for generating real-time, data-driven earthquake forecasts, which are in turn inputs to an adaptive traffic-light risk mitigation and control system for the safe operation of the tunnel.
The real-time data will drive a range of seismicity forecast models that are based on statistical, physics-based and hybrid modelling approaches (also integrating feedback and learnings from WP2 and WP3). We will develop a first-of-its kind closely-coupled induced seismicity experiment-simulator setup with a high-performance computational framework for analysis and modelling of fluid-rock interaction in near-real time, which will allow us to test and validate earthquake forecasting models and mitigation strategies.
An essential component of the experiments will be a systematic search for precursory signals, observed at laboratory scales. Such precursors systematically observed prior to ruptures of 10-50m length would be the most tantalizing discovery towards possible future applications in earthquake predictions, with potential to transform operational earthquake forecasting.