Parametric uncertainty quantification of large complex dynamical system models

Quantifying parametric uncertainty is a critical step in developing and improving any dynamical system model. For a large complex dynamical system model such as Weather Research and Forecasting (WRF) model, it becomes an extremely difficult task because of model complexity and computational demand in running such a model. This talk presents a methodology that allows us to quantify parametric uncertainty and identify optimal parameter values of large complex dynamical system models. This methodology consists of two key steps: parameter screening and surrogate modeling. We will illustrate these two steps with a case study using WRF model to improve the 5-day precipitation forecasting in the Greater Beijing Area. The methodology we present should be applicable to any other large complex dynamical system models.