Abstract:
|
Change-of-resolution models yield optimal and computationally feasible spatial smoothers of massive spatial data with non-stationary behavior (Huang et al., 2002). Our goal in this research is to develop dynamic spatial change-of-resolution statistical models, for space-time data, that yield the same computational feasibility. In this case, past data can provide valuable prior information about the current status of the space-time process; the status of the process is revised according to Bayesian theory, yielding the current posterior for the process. The temporal dynamic also allows one to forecast the process in the future and to derive the posterior of the process at any given time point in the past, based on all available data. An application will be given to Total Column Ozone (TCO), sampled remotely by a satellite over the entire globe. Initially, the large-scale spatial trend is removed, leaving the residual TCO process to be modeled. The temporal dynamic developed for the model honors the physics of motion and the mass-balance of TCO.
|