Abstract
Correlations of tropical mean water vapor
with its surface values have been calculated for all
AMIP1 models and some of the AMIP2 models. The
previously noted discrepancy between the GFDL model
and rawinsonde data also exists for other models: the
interannual correlations of water vapor with its surface
values decline to smaller values in observations than in
the models. This discrepancy is reduced somewhat after
the data from the models were sampled at the same
locations where observational data were collected, but
remains significant even at low levels. Significant discrepancy
between models and observations also exists in
the regressionso f water vapor on temperature, suggesting
something other than differences in the noise level
as the main cause for the discrepancy in the vertical correlations
between model and rawinsonde observations.