Space-time modeling of vertical ozone profiles

Ozonesondes collect data relevant to ozone level at various altitudes. Modeling these data involves a combination of spatial and temporal modeling. The spatial component can be conveniently modeled as a four component mixture of normal distributions. The (relatively few) parameters of this mixture can then be modeled in a time-dependent fashion, via a dynamic linear model. Computations are carried out via Markov chain Monte Carlo methods. Copyright (C) 2003 John Wiley Sons, Ltd.