Fragility Modeling of Precariously Balanced Rocks: Calibration, Benchmarking, and Sensitivity

Precariously balanced rocks (PBRs) provide natural geological indicators for constraining the upper bounds of earthquake ground motions over long timescales. However, translating these constraints into fragility models remains challenging because computationally expensive simulations limit inverse analysis and calibration of contact physics, reducing confidence in the fragility model predictions. To address the challenges, we present a simulated shake-table platform built on a physics engine. Using physical large-scale shake-table experiments on a natural PBR and 582 recorded earthquake displacement histories, we calibrate the contact parameters and benchmark overturning predictions against experimental results and a state-of-the-art discrete element method (DEM). The physics-engine approach reproduces overturning with predictive reliability comparable to DEM, while reducing wall-clock cost by approximately 102 to 105 times. This efficiency enables large ensemble analyses and allows us to evaluate how uncertainty in contact parameters propagates into inferred fragility boundaries. Among the contact parameters, lateral friction exerted the strongest influence on PBR fragility, whereas restitution and spinning friction had comparatively minor effects, with contact damping and stiffness exhibiting more complex behavior. Our study establish a practical pathway for using PBRs as quantitative constraints in seismic hazard assessment.