Mañana viernes día 4 de octubre a las 12:30 horas Noel Cressie (distinguished Professor and Director, Centre for Environmental Informatics, University of Wollongong, Australia) dará una conferencia en el Seminario Sixto Ríos de la Facultad de Matemáticas de la UCM titulada The spatio-temporal random effects model: An application to remote sensing.
In this talk, I propose a hierarchical spatio-temporal statistical model, where data dependencies are introduced through a latent dynamic spatio-temporal linear mixed-effects model. Big data, such as those obtained from satellite remote sensing of the environment require some form of dimension-reduction. Here, the spatial random effects model, which depends on a relatively small, fixed set of basis functions, is generalised to a spatio-temporal random effects model that is dynamic in form. By using a small number of basis functions, not only does the statistical model achieve dimension-reduction, it is also non-stationary.
Model parameters are estimated, and an empirical hierarchical model (EHM) is used to obtain the predictive distribution of the latent spatio-temporal process. A spatio-temporal algorithm that is akin to the Kalman Filter is presented, and it is applied to the remote sensing of Aerosol Optical Depth (AOD), which is a measurement taken by the MISR instrument on board NASA's Terra satellite.