Non-Stationary Locked-Boundary Inversions for the Main Himalayan Thrust: Creep-Front Propagation and Viscoelastic Stress Redistribution

The Main Himalayan Thrust (MHT) is a strongly coupled continental megathrust that accommodates India-Eurasia convergence and drives the largest seismic hazard across the Himalayan arc. Existing geodetic coupling models broadly agree that the shallow MHT is highly locked, but they make conflicting inferences about (i) the downdip extent and sharpness of the locking-creep transition and (ii) along-strike segmentation, differences that largely reflect assumed block kinematics, inversion regularization, and the frequent neglect of time-dependent lower-crustal and mantle deformation. Given these divergent inferences, key questions remain about which portions of the fault interface are truly locked and whether viscous flow beneath the Himalaya-southern Tibet systematically biases geodetic coupling estimates. We re-evaluate MHT interseismic coupling by inverting GNSS baseline length-change rates for the depths of the upper and lower locked boundaries, using a physically constrained, boundary-based inversion that permits non-stationary locking by gradual erosion of locked areas through creep-front propagation, represented by negative stressing rates (Johnson & Sherrill, 2026 in prep.). Using interseismic GNSS velocities from Lindsey et al. (2018) and a viscoelastic earthquake-cycle model, we invert for the locked-zone boundaries, spatially variable interseismic creep, and creep-front-driven stress-drop rates along the locked-zone edges. We couple this physics-regularized kinematic locking model to a viscoelastic earthquake-cycle framework to capture interseismic stress redistribution by Maxwell relaxation in the lower crust and upper mantle. Uncertainties and epistemic tradeoffs are quantified with Bayesian MCMC and a 20-model ensemble spanning published block-kinematic configurations and viscosity structures (10¹⁹-10²¹ Pa·s). Across the ensemble, coupling is consistently concentrated above mid-crustal ramp-flat transitions, with robust locking to ~15–20 km depth, most strongly between ~77° and 86°E, and limited evidence for significant locking below ~20 km. Lower viscosities favor shallower, narrower locked zones, whereas higher viscosities permit deeper and wider locking. The non-stationary creep-front models better reproduce observed baseline rates than a stationary locking model (reduced χ² ≈ 1.17 vs. 1.58) and predict peak creep rates near the downdip edge of locked asperities, where seismicity is concentrated. These results present a physically grounded interseismic coupling model with quantified uncertainties that refines Himalayan seismic moment budgets. The inferred locked zone accumulates moment at ~ 5-15*10¹⁹ N·m/yr, consistent with the long-term potential for an M_w>9 earthquake on a 1000-year recurrence interval, and delineates persistently locked segments, particularly in western Nepal, capable of hosting future great megathrust ruptures.