-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathREADME
101 lines (86 loc) · 4.34 KB
/
README
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
Overview
---------
ccast is software to take level zero data from the Cross-track
Infrared Sounder (CrIS), a Fourier transform spectrometer on the
Suomi NPP and JPSS weather satellites, and produce high-quality
calibrated radiances. It is written primarily in Matlab, allowing
for easy interaction, modification, and data visualization. Input
is a CCSDS packet stream, typically with an HDF5 wrapper, and the
output is Matlab-format HDF5 SDR files.
Features
---------
Transparent reference truth. The UMBC calibration algorithm uses
"flat" reference truth, with a simple sinc basis at the user-grid
OPD. This gives residuals similar to the NOAA and UW algorithms,
which use reference truth convolved with instrument responsivity.
But changing the definition of reference truth in that way amounts
to adjusting the ruler to fit the measured quantity.
Periodic sinc at the decimated sensor grid for the ILS. This is
in contrast with simple sinc and an extended transform matrix, as
used by UW and NOAA. Periodic sinc gives the correct basis, and
worked better than simple sinc (including sinc with various matrix
extensions) in our tests. This is mostly significant for the SW
band.
A "ratio first" calibration algorithm as the default. This takes
the ratio (ES - SP) / (IT - SP) before doing the self-apodization
correction. ccast can also do "SA (self apodization correction)
first," as an option. Ratio first allows for better cancellation
of instrument artifacts. See doc/ccast_eqns for more details.
Processing at the full sensor-grid decimated OPD, the "extended
resolution" point set. Points between the user- and sensor-grid
OPD are given a cosine apodization. This in contrast with the UW
and NOAA algorithms, which approximate a circular shift followed
by truncation.
Spectral interpolation to the low (aka "standard") res user grid
vs truncating interferograms to the original low-res sensor grid
spec. Interpolation is better for quality and consistency with
high res, while truncation may be slightly better for consistency
with obs before the switch to high res mode.
Cleaner and simpler resampling, from noticing that several of
the proposed resampling algorithms converged as n -> inf and were
being evaluated at "large n" (the undecimated point set), a value
large enough that the differences were not significant.
Simplicity and speed. The package consists of less that 10,000
lines of code, with about 5,000 lines of that for the main source
directory. Translation from RDR to L1a and from L1a to L1b/SDR
each take about 1 second/scan on a single core.
Limitations
------------
ccast does not do polarization or Doppler corrections, although we
have tested Doppler corrections as a post-processing step. These
corrections may give the UW processing a small advantage in the SW,
but this should be be balanced against questionable reference truth
and use of simple rather than periodic sinc for the ILS.
Top level directories
----------------------
doc - documentation
source - main source directory
test - test and development
motmsc - a few assorted utilities
batch - batch processing scripts
davet - Dave Tobin/UW components
readers - MIT reader with local mods
inst_data - instrument-specific data
npp_MWx - hacks for the NPP MW dropout
Documentation
--------------
doc/sdr_format.txt - ccast SDR field definitions
doc/ccast_eqns - overview of ILS and calibration equations
doc/nonlin_notes - notes on the ccast nonlinearity correction
doc/nedn_notes - notes on the ccast NEdN measurement
doc/finterp - notes on double-Fourier interpolation
source/README - short descriptions of most functions
Focal plane fitting
--------------------
The ccast focal plane fitting is done in another repo, cris_test.
See the directories focal_fit_npp and focal_fit_j1 there for more
info. The SA inverse matrices produced there are copied to the
ccast directory inst_data and specified in the various opts files
in the ccast/source directory.
To Do
------
- write up
- ccast is not currently under active development
- it might be useful to add doppler and polarization corrections
- since psinc is used for the ILS, it would be interesting to try
it for reference truth, as well