Local and Global Sensitivity for Computationally Expensive Groundwater Simulation Models with Optimization and Surrogate Models

Many important problems in engineering and science require optimization and sensitivity analysis s of computationally expensive (costly) functions.  With costly functions (like groundwater models with nonlinear systems of partial differential equations), this optimization is made difficult by the limited number of model simulations that can be done because each simulation takes a long time (e.g. 10 minutes to many hours  for one simulation).   The sensitivity problem is even more difficult if it the objective function surface is multi-modal.

I will discuss application of our new global and local multivariate sensitivity methods and their application to nonlinear groundwater (PDE) models for f(x).  We can show a very large decrease in computational effort for global sensitivity.    The method employs Latin Hypercube sampling in addition to optimization and a surrogate model.