(1) Graphics showing the location of clear sondes according to
Mitch's "pre-level 2" clear
flag, color coded with relevant variables as noted below.
Sonde locations versus scan angle.
Sonde locations versus amsu distance.
Sonde locations versus land fraction.
(2) Uploaded RTP files containing point profiles from these sondes and
the biased radiances. Note that these RTP files are (a) point profile
RTP files, and (b) are missing level parameters needed to produce a usable
layer RTP file. Consequently they cannot be used by SARTA until missing levels
have been added and the resulting file has been run through a layering
program (klayers). Their main use at the moment is to let people test
their RTP readers.
These RTP files are located on weather at
where xx=00,06,12,18 to match the NOAA prepqc files. This directory
also contains Matlab files of these data with the same name but with
a .mat instead of a .hdf extension (ask if you need help with these).
(1) Found 455 sonde matchups with Mitch's intial clear flag
(fov_clear_flag in L2Sup) for unbiased data, and 579 matchups
for biased data. Evidently the bias was enough to significantly
change the effects of fixed limit switches in the algorithm and let more data
through as clear. Yes, Mitch, I will try to implement your new algorithm today!
(2) I have been comparing the brightness temperature between 900
and 2616 cm-1 in the sonde matchups. Both of these are fairly clear
window channels. I could correct small differences between these
channels for water vapor (using the sonde data) and for emissivity
(using know models) to fine tune this comparison, will do later. Right
now I have just subset the sonde matchups to only look at ocean,
night, and either 1- or 3-hour matchups. I *think* Planck is helping
to see cloud contamination. If both channels had an emissivity of
0.98, then you should see BT2616 cm-1 about 0.8K warmer. I think
these channels should have very similar emissivities, and they are
known quite accurately (I think).
First plot is scatter plot of the two channel B(T)s. Blue dots go with
the subset of ocean/night/1-hr match.
BT 900 vs 2616 cm-1 for ocean/night/1h-match
Next plot is histogram of the BT differences for the ocean/night/1-hr match.
Histogram of BT900-BT2616 cm-1 for ocean/night/1h-match
Final plot is same as previous (histogram), but relaxing to a 3-hour match.
Histogram of BT900-BT2616 cm-1 for ocean/night/3h-match
We are now working to take these sonde matches and put the data into RTP *layer*
format (using NCEP model to fill in), and run SARTA to compare radiances.
See Friday's Notes
(now updated with the truth - June 7, 2001.)
- Post-Exercise Results:
A global map showing L1B FOVs we found to be clear,
which also shows the truth, but for a L2 golfball.
I have also provided plots of these
results for individual granules. Just click on the filename in the
directory listing. The red circles in these plots denote a clear L2 golfball (truth),
while the blue dots are plotted whenever we determine the center FOV of this golfball
to be clear (we only tested on the center L1B FOVs for clear). It is expected
that we will find clear L1B FOVs in regions where there are no red circles for
L2 clear since L2 clear requires all 9 golfball L1 FOVs to be clear.
The clear flag methodology is described here.