Can we use AI to incorporate geological data and predict seismic susceptability?
How can we use geological data from the subsurface to inform our understanding of induced seismicity?
Presently, consensus on the incorporation of induced earthquakes into seismic hazard has yet to 22 be established. For example, the non-stationary, spatiotemporal nature of induced earthquakes is not 23 well understood. Specific to the Western Canada Sedimentary Basin, geological bias in seismogenic 24 activation potential have been suggested to control the spatial distribution of induced earthquakes 25 regionally. In this paper, we train a machine learning algorithm to systemically evaluate tectonic, 26 geomechanical, and hydrological proxies suspected to control induced seismicity. Feature importance 27 suggests that proximity to basement, in situ stress, proximity to fossil reef margins, lithium 28 concentration, and rate of natural seismicity are among the strongest model predictors. Our derived 29 seismogenic potential map faithfully reproduces the current distribution of induced seismicity and is 30 suggestive of other regions which may be prone to induced earthquakes. The refinement of induced 31 seismicity geological susceptibility may become an important technique to identify significant 32 underlying geological features and address induced seismic hazard forecasting issues.
http://onlinelibrary.wiley.com/doi/10.1002/2017GL076100/full