Space-Time Modeling of Vertical Ozone Profiles
Jaeyong Lee and James O. Berger
Ozonesondes collect data relevant to ozone level at various altitudes.
Modeling this 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.
Postscript File (2915 kB)