The reactivation of glacially induced faults is linked to the increase and decrease of ice mass. But, whether faults are reactivated by glacially induced stresses depends to a large degree on the crustal stress field, fault properties and fluid pressures. The background (tectonic and lithostatic) stress field has a major effect on the potential for reactivation, as the varying stresses induced by the ice sheet affects the state of stress around the fault, bringing the fault to more stable or more unstable conditions. Here, we describe the effect of glacially induced stresses on fault reactivation under three potential background stress regimes of normal, strike-slip and thrust/reverse faulting. The Mohr diagram is used to illustrate how glacially induced stresses affect the location and the size of the Mohr circle. We review these different cases by applying an analysis of the stress state at different time points in the glacial cycle. In addition, we present an overview of fault properties that affect the reactivation of glacially induced faults, such as pore-fluid pressure and coefficient of friction.

To model glacial triggering of earthquakes, it is necessary to obtain the spatio-temporal variation of glacial isostatic adjustment-induced stress during a glacial cycled. This can be computed efficiently using commercial Finite Element codes with appropriate modifications to include the important effects of ‘pre-stress advection’, ‘internal buoyancy’ and ‘self-gravity’. The modifications described in Wu (2004) are reviewed for incompressible and so-called materially compressible flat-earths. When the glacial isostatic adjustment-induced stress is superimposed on the background tectonic stress and overburden pressure, the time variation of earthquake potential at various locations in the Earth can be evaluated for any fault orientation. To model more complex slip and fault behavior over time, the three-stage Finite Element model approach of Steffen et al. (2014) is reviewed. Finally, selected numerical examples and their results from both modelling approaches are shown.