Efficient ensemble methods for simulating groundwater-surface flows

Speaker: Ying Li (OSU)
Dates: 2023/09/15
Location: MW154
Abstract: We propose and analyze a series of efficient, unconditionally stable, ensemble methods for stimulating groundwater-surface flows governed by the Stokes-Darcy model. For systems like the surface-groundwater flows, predictive simulations must account for the inherent uncertainties in the initial conditions, forcing terms and other physical parameters. Consequently, ensemble simulations that conduct repeated computations of the same model with different parameter sets become necessary. The efficient ensemble methods can greatly reduce the computation costs for ensemble simulations. The ensemble algorithm results in a common coefficient matrix for all realizations and consequently allows the use of efficient direct or iterative block solvers to reduce the computational cost.  Numerical experiments are also provided to illustrate the convergence, unconditional stability and effectiveness of the ensemble methods.