Using the numerical Discrete Element Method (DEM), I investigate potential correlations of pre-existing forearc structural features and mechanical properties to the megathrust earthquake sizes in subduction zones. My three projects provide good complements to field observation and laboratory experiments. Moreover, my results help establish a pool of candidate features that can be used for future seismic hazard assessments. My first project seeks to reproduce stress changes that occur after a large main shock, and explores the conditions that could cause stress switching both on- and offshore Tohoku. My simulations demonstrate that rapid fault weakening can produce stress change and predominant normal-fault earthquake mechanisms in the upper plate involved in the Tohoku-Oki earthquake. Several specific conditions seem to favor such stress-switching; the megathrust fault must have been frictionally strong before the main shock, and comparable values of internal and basal friction are necessary to cause the formation of the widespread normal faults within the wedge. This study also confirms that numerical simulations that incorporate dynamic upper plate extension are more appropriate for investigating stress changes associated with rapid fault weakening in the Tohoku area, than are steady-state models based on Critical Coulomb Wedge (CCW) models. Two adjacent segments of the South-Central (SC) Chile margin exhibit significant differences in earthquake magnitudes and rupture extent, during the 1960 Valdivia earthquake and 2010 Maule earthquake. My second project, informed by interpretations of the structure across these two segments, simulates the upper plate as wedges overlying megathrust faults that are partitioned into two frictional domains, modeled after dynamic Coulomb wedge models. The velocity strengthening outer wedge width strongly influences megathrust rupture extents, whereas the frictional conditions beneath the fixed strength or velocity-strengthening outer wedge can affect the size of megathrust earthquakes. My preferred models yield reasonable fits to published slip distributions, in particular, for the 2010 Maule rupture. The simulated slip distribution for the Valdivia earthquake, suggests that the margin probably experienced its highest slip close to the trench, which differs from published models. My final project focuses on how the activation of megasplay faults within the wedge can affect earthquake sizes and tsunami genesis. I model the upper plate as a wedge that is partitioned into inner (velocity-weakening) and outer (velocity-strengthening) wedges, combined with a splay fault rooting from the decollement. I examine the effects of the dip and friction along the splay fault, along with the width of outer (velocity-strengthening) wedge during seismic cycles. The splay faults can accommodate coseismic slip thus facilitating the generation of tsunamis. However, the presence of a velocity-strengthening outer wedge constrains the rupture size and tsunami generation. Moreover, my selected model, which best fits the derived slip distribution for the 2011 Tohoku earthquake, shows the reactivation of a megasplay fault moderately affects earthquake coseismic rupture and tsunami potential. In contrast, my selected model for the 2010 Maule earthquake, reasonably matches the published slip distribution for that event, suggesting that the activation of megasplay fault has minimal impact on earthquake size and tsunami potential along the SC Chile Margin.