Quantification of Model Parameterization Uncertainty in the NASA GEOS GCM

Parameterization of processes that occur on sub-grid length scales is a key source of uncertainty in global climate models. This study investigates the relative importance of a number of parameters used in NASA GEOS Single Column Model simulations, concentrating on cloud, convection, and boundary layer parameterizations. Latin Hypercube Sampling is used to generate a few hundred realizations of sets of 14 input parameters. A Gaussian process model is then used to create an emulator for the simulation code, and a response surface is computed. A more complete measure of relative parameter sensitivity is then determined by sampling the model response surface for a very large number of parameter values. Parameter sensitivity experiments are repeated for different geographic locations and seasons, and optimal parameter values are obtained for each experiment via comparison with available observations of cloud content and radiative fluxes. The results lend insight as to which parameters exert the greatest influence on simulation output, and whether parameter sensitivity and/or optimal values change with location and time.