Evaluating effects of fluid pressure on megathrust mechanics and earthquake dynamics with numerical models of the 2004 Sumatra-Andaman earthquake

Pore fluid pressure is one mechanism with promise for explaining multiple observations of subduction zone megathrust behavior, but its coseismic state and influence on rupture dynamics are poorly constrained. Here, we analyze the influence of fluid pressure ranging from hydrostatic to lithostatic on fault mechanics and first-order rupture characteristics using a suite of numerical models based on the 2004 Mw 9.1 Sumatra-Andaman earthquake. These 3D, geometrically complex models are run with SeisSol, a software package based on an Arbitrary high-order DERivate Discontinuous Galerkin scheme that solves for dynamic fault rupture and seismic wave propagation. We find that as fluid pressure increases, fault strength, traction, moment magnitude, cumulative slip, peak slip rate, stress drop and rupture velocity decrease. The two preferred scenarios support the presence of very high coseismic pore fluid pressure (97 percent of the lithostatic loading) with the megathrust subject to mean initial shear traction of 4-5 MPa and mean effective normal traction of 22 MPa. The mean stress drops for these scenario earthquakes are both 3 MPa and the mean rupture velocities are 2400 m/s and 2600 m/s, similar to observations of the 2004 earthquake. These scenarios have ratios of average stress drop to initial deviatoric stress magnitude of 0.6. Along the central rupture, compressive stress rotations in the hanging wall are 36° toward trench-parallel on average, resulting in a modeled post-earthquake stress field that agrees with numerous observed strike-slip aftershocks. We also compare the influence on earthquake characteristics of depth-dependent versus constant initial normal stress on the megathrust. Constant effective normal stress moves peak slip and peak slip rate up-dip and produces a more constant stress drop across the megathrust. From these models alone, it is difficult to conclude the likelihood of either condition. However, very high pore fluid pressures promote conditions of constant normal stress. If these conditions are present in-situ and coseismically along megathrusts, then constant normal stress should be used in place of depth-dependent normal stress (from lithostatic loading) to estimate stress and friction coefficients in subduction zones.