Stresses that drive the plates from below: Definitions, computational path, model optimization, and error analysis

We divide the torques on each surface plate into three parts: lithostatic pressure, side strength, and basal strength. We compute each part for each of 52 plates using a thin shell finite element model of the lithosphere with topography, variable heat flow, variable crust and lithosphere thicknesses from seismic data, transient geotherms, nonlinear rheology, and weak faults. We present an iterative method of adjusting boundary conditions that results in correct plate velocities without the need for models of deep mantle flow. Uncertainty remains because side strength torques, and therefore inferred basal strength torques, depend on the effective friction of faults. Therefore, we compute a two-parameter suite of models with differing trench resistance and differing fault friction, and we evaluate their misfits relative to seafloor spreading rates, geodetic velocities, intraplate stress directions, and azimuths of seismic anisotropy. The minimum misfit occurs at effective fault friction of 0.1 and trench resistance of 2 x 10(12) N m(-1). In this preferred model, computed values of mean basal strength traction systematically increase for smaller plates. We analyze error sources and find that the largest source is unmodeled variation in effective friction of plate boundary faults. Discounting highly uncertain results, we find mean basal shear tractions of no more than 1 MPa for the six largest slabless plates: Africa 0.2 MPa; Antarctica 0.1 MPa; North America 0.6 MPa; Eurasia 1.0 MPa; South America 1.0 MPa; Somalia 0.9 MPa. The directions of basal shear traction on these plates are generally forward, meaning subparallel to absolute velocity. Basal strength torques on plates with subducting slabs represent the sum of net slab pull and distributed basal shear traction; if these torques are attributed to net slab pull alone, net slab pull is generally toward the trench and of order 5 x 10(12) N m(-1). Thus, present plate motions on Earth appear to be driven primarily by deep mantle convection rather than by topography and associated lithostatic pressures.