Posted by Randy Watts on November 09, 2000 at 21:50:49 (UTC):
In Reply to: abyssal eddies posted by Julie McClean on November 07, 2000 at 10:32:39:
: clarification regarding the form of the covariances in the
: IO scheme. It is stated that it is a function of the
: measurment locations and does not depend on the measurements
: themselves. I assume then that ``r_0'' (equation 3 in the
: appendix on the multivariate IO) is the first baroclinic
: rossby radius. Did you consider using length scales from
: the data? Can you please explain your choice?
Julie, The r_o was determined from data: by
least squares fitting a Gaussian curve to the
observed pressure correlations, NOT the first
BC Rossby radius at all. I'm pretty sure we
removed the mean field before calculating the
correlations.
The velocity correlations follow from the forms
listed in Table 1 on the .ps file that you
can access in the poster.
Note that the meanfield and the choice of
correlation field start to play a big role
when observations are sparce and barely
correlated. In SYNOP the sites were well
correlated, and the mapped fields were "robust"
in the sense of being insensitive to the
correlation length r_0. The Watts, Tracey, and Friedlander
(JGR, 1989) paper discusses these issues.
--Randy