I have been following the CF Profile and CDL and wonder
if there has been consideration of several items relating to
instrument physics and mathematical representations
of that physics. So
1. Has there been a discussion of how to represent
Time Series data that has been transformed to Fourier
spectra or two or three dimensional fields that use a
spherical harmonics representation? It looks like the
current conventions consider only data sampled
directly in space (usually longitude, latitude, and altitude)
and time. Fourier or spherical harmonics expansion
coefficients depend on all the temporal or spatial points
used to compute them, although the weightings depend
on which coefficient is selected for examination.
2. Has there been a discussion as to whether the CF
profiles need to consider the interpolating functions
between the sampled data points? This concern
arises if one wants to compare two collections of data
that may have a point sampling, but which might use
different functions for interpolating between the samples.
There is a deeper concern that follows. One question
that arises is whether two data collections should be
considered identical if the interpolating functions are
different - even if the data values at the sampled points
are the same.
3. Has there been a discussion about how to deal
with the tradeoff between resolution and uncertainty
when such instruments as satellite-borne infrared
temperature sounders invert the measurements to
retrieve vertical temperature profiles or when airborne
or satellite-borne imagers do orthorectification that
needs to include the instrument Point Spread Function?
4. Has there been a discussion about what to do
with the different altitude axes (geometric altitude,
pressure altitude, and sigma coordinates)? If one chooses
equal spacing in one of these, then the others are not
going to produce equally sampled data points. For
example, if we assume that the atmosphere is in
hydrostatic equilibrium, then a grid with evenly spaced
coordinates in pressure altitude will not have evenly
spaced coordinates in geometric altitude if the
temperature field has spatial variabilities or such
phenomena as inversions. Furthermore,
to convert from one coordinate system to another, the
algorithms are going to need a vertical temperature
profile and perhaps a humidity profile. I assume that
these additional fields would need to be added to
the data representation in some of the arrays.
5. In representing radiances, does the CF profile
intend to treat the direction of propagation as having
at least two coordinates (zenith angle and azimuth)
as separate dimensions - and perhaps an additional
one for solar zenith?
6. Has there been discussion of the correlations
and dimensions involved in some of the stratospheric
sounding instruments (SAGE, MLS, for example),
in which there is a correlation invoked by the scanning
geometry?
7. Has there been discussion of how to handle
such complex scanning patterns as the MISR
stereoscopic camera or the CERES scanning
patterns that use directionality of radiances to
create stereoscopic movies or sample angular
patterns of emitted and reflected radiance?
8. Has there been a discussion as to whether or
not vertical "picket fence" sampling by satellite-borne
lidars and radars should be handled as "swaths"?
I'm aware that NASA's ECS project used that approach
in HDF-EOS - but I'm not sure that the patterns of
image rows (Landsat or MODIS) have similar
characteristics to such samples as Lidar shots.
As a note, I'm a retired federal scientist who has led
the ERBE and CERES science teams as PI and then
moved over into managing the LaRC DAAC. I've
taught courses on radiative transfer, including one
on remote sensing, which used the weighting functions
in inversions for temperature.
Bruce R. Barkstrom
e-mail: brbarkstrom@xxxxxxxxx
cell: (828) 337-7128
