Report 13: Statistical
Intercomparison of Global Climate Models: A Common Principal Component
Approach
Sengupta, Sailes K. and James S. Boyle
November 1993, 44 pp.
Variables describing atmospheric circulation and other climate parameters
derived from various GCMs and obtained from observations can be represented
on a spatio-temporal grid (lattice) structure. The primary objective of
this paper is to explore existing as well as new statistical methods to
analyze such data structures for the purpose of model diagnostics and intercomparison
from a statistical perspective. Among the several statistical methods considered
here, a new method based on common principal components appears most promising
for the purpose of intercomparison of spatio-temporal data structures arising
in the task of model/model and model/data intercomparison. A strategy for
such an intercomparison is outlined in two steps: first, the commonality
of spatial structures in two (or more) fields is captured in the common
principal vectors, and second, the corresponding principal components obtained
as time series are then compared on the basis of similarities in their
temporal evolution. (pdf file)
UCRL-MI-123395