|
Abstract:
|
Ensembles are generated from global climate models in order to study the state of the earth's system. These ensembles often require tens of Terabytes of storage and substantial effort to control and coordinate access. This project explores data compression of ensembles via multivariate global space-time models. The methodology extends an existing scalar multi-step conditional approach, for axially symmetric processes, to the multivariate setting. This is achieved through correlated spectral processes which provide computationally feasible algorithms in high-dimensions. Ultimately, this paradigm provides earth system investigators with the a ability to generate bespoke ensembles through conditional statistical model simulations.
|
