B1147
Title: Multivariate Stochastic Climate Generator
Authors: Matthew Edwards - Newcastle University (United Kingdom) [presenting]
Abstract: In order to understand and quantify the uncertainties in climate model projections, a collection of climate simulations (an ensemble) is typically used. Given the high-dimensionality of the input space of a climate model, as well as the complex, non-linear relationships between the climate variables, a large ensemble is often required to accurately assess these uncertainties. If only a small number of climate variables are of interest at a specified spatial and temporal scale, the computational and storage expenses can be substantially reduced by training a statistical model on a small ensemble. The statistical model then acts as a stochastic generator able to simulate a large ensemble, given a small training ensemble. Previous work on stochastic generators has focused on modeling and simulating individual climate variables (e.g.~surface temperature, wind speed) independently. Here, we introduce a stochastic generator (trivariate stochastic model) that jointly simulates three key climate variables. The parameters of this nonstationary global model are estimated with a sequence of marginal likelihood functions using large-scale parallelisation across many processors for more than 80 million data points. We demonstrate the feasibility of jointly simulating climate variables by training the stochastic generator on five ensemble members from a large ensemble project, and assess the stochastic generator simulations by comparing them to the ensemble members not used in training.
