UNIVERSITY OF CALIFORNIA, SAN DIEGO
UC SAN DIEGO LIBRARY
3 1822 04429 7455
OPTICAL SYSTEMS GROUP
TECHNICAL NOTE NO. 231
October 1991
Offsite
(Annex-Jo
rnals)
QC
974.5
. T43
no. 231
WSI Ratio and Cloud Decision
Processing Summary
T. L. Koehler
UNIVERSITY
OF
CALIFORNIA
SAN DIEGO
The material contained in this note is to be considered
proprietary in nature and is not authorized for distribution
without the prior consent of the Marine Physical Laboratory
and the Geophysics Laboratory
Contract Monitor, Mr. B. Kunkel
Atmospheric Sciences Division
ulitiiliui:
FORNIA
38.
Prepared for
The Geophysics Directorate of the Phillips Laboratory
Air Force Systems Command, United States Air Force
Hanscom AFB, Massachusetts 01731
under contract No. F19628-88-K-0005
SCRIPPS
INSTITUTION
OF
OCEANOGRAPHY
MARINE PHYSICAL LAB San Diego, CA 92152-6400
UNIVERSITY OF CALIFORNIA, SAN DIEGO
III
III
11
11
III
III
1
1
1111
III
1
TIT
III
I
3 1822 04429 7455
TABLE OF CONTENTS
List of Figures ...........
List of Tables
1.0 Introduction .......
o
Data Tape Quality Assessment ....
2.1 Enter the Field Tape into the Data Base ......
2.2 The TAPEQC Program ...............
2.2.1 Checks Performed by TAPEQC ..
2.2.2 TAPEQC Program Execution ....
2.2.3 Files Created by TAPEQC .....
2.3 The DGN Program .........
2.4 TAPEQC and DGN File Screening Procedures .......
2.5 Post TAPEQC Procedures. .
3.0
Calibration Data Reduction ....................
3.1 Keeping Track of the Various Calibrations .......
3.1.1 Field Unit Calibration Log .........
3.1.2 List of Acquired Calibrations ....
3.1.3 Hardware Log .....
3.2 Calibration Data Reduction Algorithms ....
3.2.1 Program LINCALC ....
3.2.2 Program FILTER ..........
3.2.3 Program ABSCALC ....
3.3 Absolute Radiance and Ratio Calibration ....
3.4 Preliminary Ratio Calibration and Hardware Version Files ...
4.0
4.1
Ratio Input File Preparation ..............
Image Size and Offscale Threshold Determination ............
4.2
Sensor Azimuth Offset Determination ...
4.3 LAN Image Evaluation and Associated Ratio Input File Entries ...
4.3.1 Running CHECKLAN ...........
4.3.2 DATES.LST Specification ............
4.3.3 OCCs.DAT Specification .....
4.3.4 TIMES.COR Specification ....
4.3.5 VERSION.LOG Quality Indicators ...
4.4 . Obstacle Mask Definition .......
4.5 Final CALsVvw.RAT Adjustments ....
5.0
Red/Blue Ratio Image Construction ........
5.1 Description of TAPRATPL ......
5.2 Running TAPRATPL....
5.3 TAPRATPL Output Files ..................
6.0
Fixed Threshold Cloud Decisions ....
6.1 Ratio Threshold Determination ...
6.2 Program CLDDECM ..
6.2.1 Running CLDDECM.
6.2.2 Format of the Cloud Cover Summary File ......
6.3 Program CLDDECT..
7.0
Opaque Threshold and Clear Sky Background Specification ........
7.1 Opaque Threshold Determinations ..........
7.1.1 Running RATREF ...
7.1.2 Defining Opaque Thresholds with RATREF .
7.1.3 CMPDEC.INP File Entries ...
Clear Sky Normalized Background Determination ...
7.2.1 OBSBETA - The Normalized Ratio Extraction Program
7.2.2 Combining the Sample Files Using COMBNORM .....
7.3 Clear Sky Reference Value Specification ..
7.3.1 Estimate the Reference Parameter for Each Image
7.2
97
99
107
107
7.3.2
Using GETREF erence Parame
7.3.2
Select Representative Clear Sky Reference Values .....
110
8.0
112
Composite Cloud Decision Image Construction ......
8.1. Description of CMPDECTP..
8.2 Running CMPDECTP ....
ö
119
121
9.0 Decision Tape Stacking Procedures ........
9.1 The Tape Copying Program - EXCOPY ..
9.2 Stacked Decision Tape Quality Evaluation ..
Acknowledgements
References
126
LIST OF FIGURES
Figure
Number
1-1
WSI image processing steps ............................
1-2
Symbol legend for WSI procedure conceptual flow charts .........
Tape quality assessment procedure flow diagram. .....
2-1
sses
3-1
Calibration measurement and file preparation sequences ..........
3-2
Preliminary calibration and hardware version file
preparation sequence .................................
3-3
Spectral filter curves provided by Oriel for the filters
in hardware version 7V1 (Columbia, MO initial hardware
version)..........................................
4-1
Ratio input file preparations ........
4-2
Final ratio calibration file entry determinations ...
4-3
Sample sun position plots before and after azimuth .....
correction
4-4
Dark red (Quad 4) image from Columbia, with and without
the obstacle mask superimposed. Note the radar tower and
the tree in the original image. .......
Ratio processing conceptual flow diagram .
5-1
6-1
Fixed threshold determination and cloud decision
processing steps .....
7-1
Clear sky background distribution and composite decision .....
input file preparation steps
7-2
OBSBETA sample points ...........................
7-3
8-1
Haze loading parameter definition flow diagram ............
Composite cloud decision algorithm input and output
specification .......................................
..
Decision tape stacking procedure .........
113
9-1
LIST OF TABLES
Table
Number
2-1
Description of diagnostic error summary entries .........
3-1
Hardware component list from initial Columbia, MO
installation ........................................
19
3-2
Sample linearity calibration log for LINO22 from the
Columbia, MO version 7V1 camera
3-3
1
Sample calibration log from a Type 1 (Abs. vs Filter)
absolute calibration for the 7V1 sensor fielded at
Columbia on 8/FEB/89.
4-1
Revised CAL7V01.RAT image size table .
4-2
.....
4-3
Sample CHECKLAN summary list...............
Sample CHECK.OUT contents .........
TAPRATPL log file entry description .......
.
5-1
7-1
7-2
Sample RATREF processing notes ......
Sample entry into file COL018G.OBB .....
Columbia, MO ratio images processed by OBSBETA in the
clear sky background determination ...
7-3
7-4
7-5
Sample output from COLCOMB.OUT ........
COLSKY.DAT (COLCOMB.REV) sample table...
Composite decision categories ...
8-1
............
9-1
Sample stacked 1-min cloud decision tape summary for
March 1989 at Columbia, MO .....
1.0
Introduction
The Whole Sky Imager (WSI) is one of several video based imaging systems that have
been developed and fielded by the Optical Systems Group of the Marine Physical
Laboratory. A WSI system consists of two major components: a weatherproof sensor
package that normally was placed on the top of a building to provide as unobstructed a
view of the entire sky dome as possible, and a control and data archival computer ideally
placed in an office environment that was connected to the sensor by a set of cables. The
system was designed to run unattended for a week, then it would eject the tape containing
the radiance imagery, a blank tape would be inserted, and the system would resume
collecting more data. The resulting weekly data tape contains up to 1.2 gigabytes of data
from 5544 images taken in the red and blue parts of the visible spectrum. As many as 7
WSI imagers were in the field from the time of the deployment of the first system in late
March of 1988 to the end of data collection on 31 December 1990. See Tech. Note 226,
"WSI Data Base Summary" for more detailed information concerning data format and
system deployment schedules. In all about 4500 data days were collected.
The primary purpose of collecting this data was to provide a high spatial and temporal
resolution set of cloud cover data from several different sites across the country. These
data would then be used to validate the statistical parameters used in models of the cloud
free line of site and cloud free arc developed by the Air Force. Before this data analysis
could be accomplished, the raw radiance information collected in the field had to undergo
data quality control assessments, information from the red and blue image pairs had to be
combined and sensor calibration information applied to form one red/blue ratio image, and
finally, the discrimination of the presence of cloud at each ratio pixel had to be made. This
note summarizes the procedures used to perform the data processing.
Fig. 1-1 outlines the basic steps in the ratio and cloud decision processing. Numbers
in the lower right hand corners of the boxes refer to the section of the report describing that
procedure. The programs and instructions for the raw radiance archival and calibration data
collection procedures from the top two boxes in Fig. 1-1 are not included in this note. All
subsequent steps are presented. Little has changed from earlier data handling overviews,
such as Tech. Memo AV88-065t, "Cloud Program Data Handling Overview" from
November 1988. Considerable effort was expended in developing the software tools and
procedures needed to accomplish the ratio and cloud decision tasks. The algorithms were
being upgraded periodically during the data collection and processing. Most of what is
presented here refers to the current programs, as they would be employed today.
An attempt is made in the following sections to describe to a potential user what
function each step in the procedure performs, how to run the computer programs and
perform the procedures, and what products do the procedures generate, including disk file,
ExaByte tape and handwritten outputs. Descriptions of what appears on the computer
screen during execution of a given procedure, or in the associated disk file outputs is
presented in a different font that the rest of the text. User responses to program prompts
are highlighted in boldface italics (see for example page 7 in section 2).
Conceptual flow diagrams outlining the procedural steps and the associated input
requirements and output products are provided in each section. The legend for these
diagrams is presented in Fig. 1-2. The organization of each flow diagram is tailored to the
interelationships of the input and output requirements of the individual steps for each
procedure. The starting and ending points for each flow diagram can be ascertained from
the flow direction arrows.

Radiance Data
Acquisition
Field Unit
Calibrations
Data Tape Quality
Assessment
Calibration Data
Reduction
2
Ratio Input File
Preparation
Ratio Image
Construction
1-min data
10-min data
Fixed Thin and Opaque
Threshold Determination
- - - - - - - -
Fixed Threshold
Cloud Decision
Opaque Threshold and
Clear Sky Background
Determination
cision
6
Variable Threshold
Cloud Decision
18
Cloud Decision
Tape Stacking ---
Stacked Tape Quality
Assessment ( 1-min only) a
Fig. 1-1. WSI image processing steps

ExaByte Data Tape

Processing Procedure
If processing is performed
by a computer program, the
name of that program appears
in brackets.
[Program]

Input or Output Disk Files
If a description of the file
is given, the file name
appears in parentheses.
(File Name )

Handwritten Material
Procedure logs or notes.
Processing Flow Indicator.
Fig. 1-2. Flow diagram legend.
2.0
Data Tape Quality Assessment
Since the WSI data were being collected from several sites around the country, and
with limited on-site supervision, it was important to develop a procedure to perform
diagnostic checks on the field tapes when they arrived at MPL. This enabled us to assess
data quality and system performance, so that the necessary corrective action (if any) could
be performed in a timely manner.
Figure 2-1 provides a summary of the quality assessment operations and resulting disk
files and logs available for subsequent steps in the data processing procedure. The only
input to this step is a radiance image data tape. The quality control procedure was evolving
as the WSI data was being collected. The following discussion outlines the steps that
would be performed if a data tape were received today. Many of the steps have remained
unchanged since Tech. Memo AV88-046t describing the tape check-in procedures was
written in the fall of 1988.
2.1
Enter the Field Tape into the Data Base
A tape was received from the stations approximately once a week. It contained data
from seven days unless something unusual happened; i.e. power outages or mechanical
problems.
Each station has been assigned a 3 letter Station ID as well as a 1 digit Station number.
These are given in the following table:
Station
Station
Number
ID
Location
MPL
WSC
WSH
KAA
CL4
MAG
BAR
COL
CMP
POR
Marine Physical Laboratory, San Diego, CA
C-Station, White Sands Missile Range, NM
HELSTF, White Sands Missile Range, NM
Kirtland AFB, Albuquerque, NM
China Lake NWC, CA
Malmstrom AFB, Great Falls, MT
Malabar Tracking Facility, FL
National Weather Service Office, Columbia, MO
Composite System, San Diego, CA
Portable System, various locations
A log book is maintained for each station, containing a summary of the pertinent
information for each tape. The tape sequence numbers for each station began with 001.
When a tape arrived from a station, it was assigned a tape identifier (LLL###), consisting
of the 3 letter Station ID (LLL) and the next available sequence number (###) for that
station. Tapes were assigned sequential numbers even if they were received out of date
order.

Enter Field Tape
into Data Base
Diagnostic Files for Entire
Tape (LLL###.DGN)
Radiance
Image Field
Tape
( Master)
Tape Quality Assessment
Data Extraction ( TAPEQC)
Diagnostic File Error Summary
and Day Separation (DGN ]
Daily LAN Images
(LLL###C.LAN)
LAN Image Histograms
(LLL#**C.HST)
Tape Hourly Dark Signal
Values (LLL###.DRK)
Daily Diagnostic Files
(LLL###c'.DGN)
Field Tape
Backup
(EXCOPY]
Manual Screening of TAPEQC
and DGN Output and Overall
Tape Quality Assessment
Tape Catalog Information
(LLL###.EXT)
Tape Hourly Temperatures
( LLL###.HOT)
Radiance
Image Field
WSI Database
Catalog Entry
Tape
TAPEQC
and DGN
File Archive
Tapes
(Copy)
| Station Log Entries
Fig. 2-1. Tape quality assessment procedure
conceptual flow diagram.
2.2
The TAPEQC Program
The TAPEQC program (and its predecessor, READTPQC) provides the diagnostic
information needed for data quality and system performance assessments. Tech. Memo
AV88-043t provides a brief description of the READTPQC program from the fall of 1988.
Several features have been added to the current version, TĀPEQC. The current program
version is presented because it would be applied to any new field tapes.
TAPEQC loads WSI images from the field tapes onto the FG-100 board where the
image assessments are made. A variety of diagnostic tools are generated, including flags
indicating non-standard performance, average pixel values in certain portions of selected
images, and a sample red image from local apparent noon (LAN).
2.2.1 Checks Performed by TAPEQC
Information from every image on a FIELD tape is examined by TAPEQC. The
frequency and extent of the checking procedures applied to a given image depend on the
image type.
ment and ents. Time nanges bet
TAPEQC examines the header information on each of the 1-minute, subset resolution
images. Two types of checks are made: those indicating timing inconsistencies, or those
pertaining to filter wheel movement and flux control procedures. The program expects time
to proceed forward in 1-minute increments. Time skips ahead of greater than one minute
and backward time skips are noted. Time source changes between BIOS and WWV clocks
are also flagged. TAPEQC also verifies the spectral filter sequence for the four quadrants.
If the sequence is out of order, the program reports a spectral error. TAPEQC senses the
IRIS character string for a special character combinations that indicate any of the following:
the spectral filter wheel could not be placed in the proper position, the neutral density filter
wheel could not be placed in the proper position, or the flux control routine in the FIELD
program was unable to converge on a solution in the time allotted.
All the checks described above for the 1-minute images are also performed on each 10-
minute, full resolution image. TAPEQC also checks if the iris is fully closed when neutral
density buffering is activated. If not, the iris may have to be repaired. For every sixth 10-
minute image encountered (hourly under normal operating conditions), TAPEQC performs
dark signal and onscale checks, and records the temperature inside the camera housing, if
available. The dark scale checks are made by averaging 6x6 pixel arrays from the three
corners of the bright blue (quadrant 1) image not containing header strings. The averages
are saved, and warnings written when certain values are exceeded, indicating deterioration
of the video signal. Differences between the corner values are also evaluated to identify
vertical or horizontal ramping. The onscale checks evaluate the average brightness level of a
6x6 pixel array near the northern horizon in all four images. Warnings are given if
quadrants 1 and 2 are too dark, or quadrants 3 and 4 too bright. This check is often
activated on images taken before sunrise or after sunrise, when quadrants 1 and 2 are off
scale dark.
The hourly TAPEQC checks are performed on LAN and LAN + 3 hour images. The 3
dark scale pixel arrays are also saved. At LAN for each day on the tape, the bright red
(quadrant 2) image and its corresponding brightness level histogram are saved on separate
output files.
2.2.2 TAPEQC Program Execution
Select an available processing computer with at least one ExaByte drive. Each
processing computer has a TAPEQC directory with subdirectories for each station denoted
by the station identifier (LLL). Place the radiance image tape into the read ExaByte (drive
0). The following example uses tape COL051 from Columbia, MO.
To Execute: Enter TAPEQC at the prompt, i.e.
D:/TAPEQC/COL>TAPEQC
Screen Output:
TAPEQC--program version MOD--8.3. 11/6/89
Time/Date: TIME=12:20:03 DATE=01/FEB/90
Station=COL
Please Enter Tape Number >> 051 (3 digit sequence number assigned to
the tape.)
Station + Tape ID=COL051
(If the station ID cannot be read from
amLAN=15:20
the first image header then the program
LAN=18:20
will request it before the tape
pMLAN=21:20
sequence number.)
DATE=01/FEB/90
Software Version= 2V2S
Hardware Version= 2V6H
TIME=
--------------TEN
From here on the program will continue to run unassisted. Information will continue
to scroll on the screen. The time it takes to write to the screen will contribute to the overall
run time. To inhibit the screen output depress the SPACE BAR once. To resume screen
output depress the SPACE BAR again. When TAPEQC is finished, the screen output
should look similar to the sample output below:
-------ONE
Quad Time Iris Occultor ND SP
23:20 160
164 0 0
#2
23:20 160
164 0 1
23:20 160 164 0 2
23:20 160
164 0 3
Data search in progress
NNNN
:
ON
mmm
AW
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
•
•
•
•
•
..
..
..
...
..
..
..
...
..
...
....
...
....
...
..
....
...
..
...
...
... Finished
23:20:01 Ten Minute Image count=
One Minute Image count=
Stop - Program terminated.
72
720
D:/TAPEQC/COL>
2.2.3 Files Created by TAPEQC
TAPEQC creates three types of files: diagnostic files, LAN files and histogram files.
The diagnostic file contains the output from the data checking procedures presented earlier
in the TAPEQC program description for all days on the tape. Its filename is in the form
LLL###.DGN. Two files are also produced for each LAN on the tape. The quadrant 2 red
image pixel values are saved in a file named LLL###c.LAN, where c is a character (A, B,
C, ...,G) that indicates the sequence of the LAN image on the tape. A brightness level
histogram file for this image is saved under the name LLL###C.HST. The TAPEQC run
for COL051 produced the following files:
COL051.DGN -
for the entire tape
COL051A.LAN & COL051A.HST - for the first LAN on tape
COL051B.LAN & COL051B.HST . for the second LAN on tape
COL051C.LAN & COL051C.HST for the third LAN on tape
COL051D.LAN & COL051D.HST - for the fourth LAN on tape
COLO51E.LAN & COLO51E.HST - for the fifth LAN on tape
COL051F.LAN & COL051F.HST . for the sixth LAN on tape
COL051G.LAN & COLO51G.HST - for the seventh LAN on tape
2.3
The DGN Program
The DGN program takes the diagnostic file created for the entire data tape by TAPEQC
and divides it into separate files (LLL###c.DGN) for each day on the tape. It also creates
three special use files (.EXT, .DRK and .HOT). The contents of the special use files are
described below:
LLL.EXT
contains the catalog information: date, start/stop time,
hardware/software versions and individual DGN file names.
RENAME this file so that the tape sequence number follows the
station identifier. These files were used in putting together the WSI
Data Base Catalog (Tech. Note 228).
LLL###.DRK - intended to contain the hourly dark signal values for all the days.
However, a mistake in the DGN program omitted the dark scale
print to this file. At present these files contain no useful
information.
LLL###.HOT - contains the hourly temperatures for all the days. The files are used
in the tape quality evaluations.
Continuing with the example for tape COL051:
To Execute: Enter DGN followed by the TAPEQC diagnostic file name, i.e.
D:/TAPEQC/COL>DGN COL051. DGN
A diagnostic error summary is appended to the end of each day's DGN output file. Table
2-1 describes the errors listed in the diagnostic summary. The files created by the DGN
program for our sample-are:
COL.EXT - special use file
COL051.DRK - special use file
COL051.HOT - special use file
COL051A.DGN - the first day on the tape
COL051B.DGN - the second day on the tape
COL051C.DGN - the third day on the tape
COLO51D.DGN - the fourth day on the tape
COL051E.DGN .. the fifth day on the tape
COL051F.DGN - the sixth day on the tape
COL051G.DGN - the seventh day on the tape
Table 2.1 Description of Diagnostic Error Summary Entries
- the spectral filters are out of sequence.
1) Spectral Errors
2) Occultor Errors
- the occultor moves more or less than 15 degrees per hour.
If the images are missing this error could be flagged.
3) Signal Errors
OIS
- the dark signal exceeds a certain threshold in three regions
outside the image in Quadrant 1.
4) Image Errors
- the signal within the image in a given quadrant is too high
or low.
- the minute skips to far ahead or back.
5) Minute Errors
6) Hour Errors
- the hour skips to far ahead or back.
7)
Iris & N.D. not set
Correctly Errors
- the IRIS reading is greater than 020 and the selected
Neutral Density is not equal to zero.
8) No Time for Flux Ctrl - not enough time to do flux control.
9) Can't put filter in
Position
- the spectral filter could not be put in the proper position.
10) N.D. Runaway
11) Not on WWV time
12) Missing Ten IMG
13) One Minute Errors
14) Ten Minute Errors
- the neutral density filter could not be put in the position.
- number of times system used BIOS clock.
- total number of missing ten minute images.
- READ and COUNT values differ for one minute images*
- READ and COUNT values differ for ten minute images*
READ is the image number from the header and COUNT is a counter that increments
when each READ occurs. If they differ, then COUNT resets to the READ value and
is flagged. COUNT resets after the first image of the day is read.
count gets to the release increments
10
2.4
TAPEQC and DGN File Screening Procedures
Certain features of the image quality and system performance must be noted in the
individual station log books for each tape received. This information is obtained by
examining the files produced by TAPEQC and DGN for each day on the tape. The
appropriate entries outlined below are then made to the station log.
Step 1: Sort the TAPEQC and DGN output files for the tape.
Use NORTON Utilities as shown in Tech. Note 230. This step makes any
inconsistencies in the sequence labelling between the LLL###c.DGN files and the
LLL###c.LAN and .HST files readily apparent. In our example the day and LAN
sequence characters match, i.e. the first LAN occurs on the first day, the second LAN
on the second day, etc. This is not necessarily the case for every tape. If the data
collection began after LAN on the first day of the tape, then the LLL###A.LAN (the
first LAN on the tape) will actually be from the second day, creating an offset in the
.LAN and .DGN character sequences. Offsets can also arise from missing LAN
images on subsequent days.
Step 2: Examine the LAN images, their associated histogram (HST) and diagnostic
(DGN) files for each day.
Use an image viewing program such as SHOIMG, VIEWIMG or SAMPIMG to
examine LAN files, and the DOS LIST Utility to view .HST and .DGN files.
a)
LAN image entries:
- Date (Header)
- Occultor position (Header)
Number of the occultor arm used (Bands on the arm)
- Temperature (Header)
Degree of cloudiness (i.e. 10% ... 100%)
Cloud type (if you cannot determine, note whether opaque or thin)
WWV or BIOS time used (Header)
Other things to note:
- Is the occultor off-center or tilted?
- Is stray light present? (Occultor shadow)
Does a bright halo appear in the dark area of the image? (This may indicate
an inappropriate flux control threshold in the FIELD program input file.)
- Any obstructions on the dome? (birds, dirt, etc.)
- Is there noise in the dark area? (grainy or speckly appearance)
Are the images too bright? If so, then note the IRIS setting and the
Neutral Density (ND) used from the header.
Does the temperature on the image differ from the average LAN
temperature for all days on the tape by more than £ 5° C?
11
b) Histogram (HST) entries:
Are the values distributed properly? (There should be a larger
concentration at zero)
Is signal compression evident? This occurs when the highest pixel values
fall below 240. Note, the character string readable on the image are
assigned a value of 255, so the histogram will always register some pixels
at that level.
- Is the dark signal rising?
c)
Diagnostic (DGN) entries:
- Abnormal start and stop times.
- Are there Image Signal errors? (If so, when, location and values.) In the
winter these errors occur at the beginning and end of the day due to
darkness.
- Time gaps (including the missing ten minute midnight image.)
- Excessive "NO TIME FOR FLUX CONTROL" errors (more than 100
occurrences)
- Note the time of "CAN'T PUT FILTER IN POSITION" errors if any.
- If the LAN images are too bright there may be "IRIS AND ND NOT SET
CORRECTLY" errors. If so, note the time and quadrant where they
occur.
- Note any Dark Signal errors over 15.0.
All noteworthy items for the tape are summarized in a memo to the group.
IA
2.5
Post TAPEQC Procedures
All of the files created by TAPEQC and DGN are then backed up to ExaByte tape for
use in downstream processing. Each station has numerous backup tapes, with each tape
having no more than 10 TAPEQC runs appended to allow for timely future retrieval. The
backup procedure is described in detail in Tech. Note 230.
If the quality assessment has proceeded normally, a copy of the raw data tape is then
made using the program EXCOPY as described in Tech. Note 230. Both the master and the
copy are then appropriately labelled. The master and copy tapes are stored in different
locations.
3.0
Calibration Data Reduction
A significant effort is made by our group to acquire the calibration data needed for
converting the raw WSI measurements to useful, radiometrically correct quantities. This
section describes the types of calibration information collected, and the procedures used to
organize the calibration information into a limited number of input files for the ratio routine.
C
Figure 3-1 illustrates the four types of calibration data collected and the output
products generated by the calibration procedure. While the three computer programs used
in the calibration sequence are shown in the figure, instructions for running these programs
are the subject of other documents. Descriptions of the four calibration types are provided
below.
Linearity Calibration - documents the nonlinear response of the camera to changes in
the input flux as measured by the distance of a standard lamp to the camera chip.
A technique can then be devised to compensate for the nonlinear response.
Absolute Calibration - provides the measurements needed to relate the byte value
grabbed from the video signal to the irradiance viewed by the sensor for given iris
settings, and different spectral and neutral density filter combinations.
Rooftop Calibration - determines the image size in different filter combinations and the
acceptable dark pixel level for flux control in the FIELD program. Also saves
images with the geometric dome in place to aid in the geometric calibration. The
image size information is the most important of these parameters for ratio
processing. These calibrations are normally performed both before the sensor
leaves MPL and after it arrives in the field.
Spectral Filter Response - most applications use the half power points provided by the
manufacturer. Selected high spectral resolution calibrations have been performed.
As shown in Fig. 3-1, results from the calibrations come in three forms: calibration
image files, handwritten log entries, and computer text files. A later section will outline the
transfer of the handwritten log information to computer text files for input into the data
reduction algorithms.
13

Linearity Image
Files (LINeee.#)
Linearity
Archive
Tape
Linearity Calibration
Measurements (JLIN4 )
Linearity
Calibration Logs
Linearity Input
File ( LINIII.# )

Absolute Image
Files (ABSaaa.#)
Absolute
Archive
Tape
Absolute Calibration
Measurements (JABS4)
Absolute
Calibration Logs
Absolute Input
File (ABSaaa. INP)

Rooftop
Calibration Logs
Rooftop Calibration
Measurement
( at MPL or on-site )
(FLDCAL)
Image Edges, Offsets
and Magnifications
(OFFSET.FLD)
Rooftop
Archive
Tape
Rooftop Image Files
- ABSaaa or RTPrrr.EDG,
.FLX and .GEO

Filter Bandwidth
Characterization
Filter Input File
(FILSVV.INP)
Fig. 3-1. Calibration measurement and file preparation sequences.
14
3.1 Keeping Track of the Various Calibrations
With as many as 9 WSI systems in the field at a given time, and systems being
replaced due to intermittent failures, documenting the hardware changes and recording the
associated calibration information became a substantial effort. In order to process a given
raw radiance data tape through the ratio and cloud decision procedures, a means of
identifying the hardware fielded at site on the date specified must be devised, and reference
to the proper calibration information must be made. Several calibration logs have been
maintained to facilitate this process. These logs have been described in previous Tech.
Memos, such as AV90-026t. Excerpts for the Columbia, MO site are described here.
3.1.1 Field Unit Calibration Status Log
This first log subdivides the calibration information according to the installation site of
the hardware, where each site is assigned a specific field unit number (see Section 2.1).
The field unit calibration is stored under the file name CALUNIT.LOG. Contents of the
file include dates that hardware changes were made at the site, comments describing the
changes, and the linearity, absolute and rooftop calibration sequence numbers associated
with that hardware version. A sample from first WSI system (version 1) installed at
Columbia (field unit 7) follows.
FIELD UNIT CALIBRATION STATUS LOG
Updated:
2 April 90
FIELD UNIT # 1 C-STATION
Version la, 16 Feb 88:
FIELD UNIT #7 COLUMBIA, MO
Version 1A, 8 February 89:
Full calibration, looks normal. LINO22, ABS120-127. Name
ABS120 used for both rooftop and first abs calib. Also brought
home field calibrations and local apparent noon check from field,
COL001-003. Flux threshold 29000.
Version 1B, 23 March 89:
New antenna and occultor arms, use same calib as 1A
Version 1c, 22 June 89:
New FG100, use same calib as 1A
Note that many of the hardware changes, such as computer changes and occultor
repairs, will not change the radiometric calibration information. Normally, the version
number change will accompany new radiometric calibrations. For example, Unit 7
Versions 1A, 1B and 1C may use the same radiometric calibration, but Unit 7 Version 2
implies a camera change, and thus a new radiometric calibration.
15
3.1.2 List of Acquired Calibrations
This log file documents all calibrations acquired. The format of the file is outlined below.
LIST OF ACQUIRED CALIBRATIONS
Updated:
2 Apr 90
LINEARITIES
CALIB
DATE
COMMENTS
TAPE
ID
CAMERA
ID
FIELD
UNIT
#
LIN001 1/15/88
315
638-8671
1 v1
Designated 019 on files
LINO20 11/08/88 348
LIN021 1/06/89 001
LINO22 1/26/89 003
LIN023 3/02/89 008
LINO24 3/24/89011
716-87059
730-87345
638-8669
635-8640
1465-8852
6 v1
1 v4
7 v1
6 v2
1 v5
Calib as 8 vl
ABSOLUTES
CALIB
DATE
COMMENTS
ТАРЕ
ID
CAMERA
ID
FILT
CHG
FIELD CALIB
UNIT TYPE
ABS001 1/20/88 315
638-8671
FC3
1 v1
ABS118 1/11/89001
ABS119 1/12/89 002
730-87345
730-87345
FC51
FC5
v4
1 v4
6
7-9
WSC119 on
tape
7,8,9 Both calibs
1 called 120
7 vl
7 v1
7 v1
7 v1
7 v1
7 v1
warmer temp
ABS120 1/27/89 003
ABS120 2/01/89004
ABS121 2/01/89 004
ABS122 2/01/89004
ABS123 2/01/89004
ABS124 2/01/89004
ABS125 2/01/89 004
ABS126 2/01/89 004
ABS127 2/01/89004
COL001 2/08/89 006
COL002 2/08/89 005
COL003 2/09/89 007
638-8669
638-8669
638-8669
638-8669
638-8669
638-8669
638-8669
638-8669
638-8669
638-8669
638-8669
638-8669
FC8
FC8
FC8
FC8
FC8
FC8
FC8
FC8
FC8
FC8
FC8
FC8
cra AwNNN
7 vl
124 cont
7 v1
7 v1
7 v1
vl
7 v1
7-9
7-9
Grabs at LAN
Ck in field
Repeat ck
7-9
ABS128A 3/15/89 009
ABS129A 3/16/89 010
635-8640
635-8640
FC9
FC9
6 v2
6 v2
ROOFTOPS (early rooftops were designated absolutes)
CALIB
DATE
TAPE
CAMERA
FILT
FIELD CALIB
COMMENTS
#
ID
ID
CHG
UNIT TYPE
RTP001 8/11/89 023
0364-8918
FC11
5 v2
7-9
ABSOLUTE TYPES
Starting 6/12/89:
ABS VS FILTER AT APER =
AT APER=120, ND=1
ABS VS FILTER AT APER = 40 AT APER=160, ND=1
ABS VS FILTER AT APER = 120 AT APER=160, ND=2
ABS VS ND AT APER = 120
ABS VS IRIS AT ND = 0
OFF AXIS ATTENUATION CALIB
GEOMETRIC CALIB
IMAGE SIZE (FOR FLUX CONTROL)
EDGE CHECK, ALL FILTERS
FIELD CAL, INCORPORATES EDGE, THEN FLUX, THEN GEO
SPECIAL
O vous WN
7-9
10
:
The first six types are performed during the absolute calibration, while types 7 through
9 are the rooftop calibrations.
3.1.3 Hardware Log
The hardware log summarizes the hardware changes and their status, and accompanies
the field unit summary memos. The entry from AV90-070t appears below.
HARDWARE LOG
Updated:
1 May 90
This log documents the general changes to the Eo Camera field
units. The changes in specific components are listed in the
"Field Unit Major Components" tables kept on file.
* = New entry in log
FIELD UNIT # 1 C-STA
Version 1A, 16 Feb 88:
Memo AV88-0050
FIELD UNIT # 7 COLUMBIA, MO
Version 1A, 8 February 89: Memo AV89-028t
Fielded 8 February, abnormal clock antenna, otherwise
system appears normal.
Version 1B, 24 March 89: Memo AV89-0450
A new clock antenna and new occultor arms were installed.
Version 1c, 22 June 89: Memo AV89-0684
A new computer with new FG100 was installed, due to failure
of the Exabyte. This computer included a new modem.
This file also provides the cross-listing of the version number with the field status
summary memo containing its hardware inventory. A copy of the Unit 7 Version 1A
component list (Table 3-1) from AV89-028t is attached.
FIELD UNIT #7 MAJOR COMPONENTS
Fielded 8 February 89, Columbia
Date: 3 Mar 39
AV89-0230
FIELD UNIT #7 (conc.)
Date: 3 Mar 39
AV89-0230
Sensor Components
Item
Ident
Item #
(Table A)
Changed From
Previous Version
Item
Ident
Item #
(Table A)
Changed From
Previous Version
OCC-08
M-51010-12
M-51010-13
WLH-07
FL-07
FC 08
B-57530-15
R-57610-15
B-57530-17
Boards (Slot #)
Video Color Card (1)
Hayes Modem (7)
DIO 96 (2)
CPU & Co-proc (6)
FG 100 (4)
Hd Disk Control (8)
Excamal Drive (9)
SCSI Host (10)
Com Port (5)
Keyboard
External Disk
Sony Monitor
Cock Radio
SPSS
Sun Occultor
Filter. 4.0 ND
Filter. 4.0 ND (Spare)
Wx- Proof Lens Housing
Fisheye Lens
Filter Changer
Spea I (blue)
Spec. 2 (red)
Spec. 3 (blue)
Spec. 3 grim (0.5)
Spec. 4 (red)
Spec 4 rim (0.5)
ND 1 - donc
ND 2 (0.3)
ND 3 (0.7)
ND 4 (1.0)
Wx- Proof Cam Housing
Video Camera
GC 018536
A05400963598
86A1406
1755
KKM 1298
968644
W7994V-0
3211
20331352-0
518932
26-5045662
2031948
CLK-02
1
NI
19
R-57610-16
NI
NR
N-50810-09
N-50821-08
N-50830-08
WCH-07 .
638-8669
Peripheral Components
Item
Item #
(Table d)
Changed From
Previous Version
Ident
NI
Conmol Components
N
NI
Item
Ident
Item #
(Table A)
Changed From
Previous Version
Stand and Shroud
Cable Sec
19" Rack
Rack Slides
Pedestal
Air Conditioner
Heater
NO
NI
NI
Accessory Coagol Panel
TM Computer
Seagate Hard Disk
Exabyte Streamer
ACP-08
TMI 2039
147136
4972
NO
*NI = No Ident applied, NR = Not Recorded
Table 3-1
Hardware component list from initial Columbia, MO, installation
3.2
Calibration Data Reduction Algorithms
Three programs have been written to aid in the calibration data reduction: LINCALC,
FILTER and ABSCALC. Figure 3-2 shows how these programs fit in the overall ratio
calibration file preparation sequence.
3.2.1 Program LINCALC
The LINCALC program computes the linearity look-up conversion table (LUT) that
will be loaded onto the FG-100 board during ratio processing. The method for creating
this table is described in Tech. Memo AV89-056t, Software Documentation: Linearity
Processing. The following discussion explains how to run LINCALC.
The first step is to create the LINI11.INP file required by the program, where Ill is the
linearity sequence number. As indicated in the calibration logs, ill for the Columbia
example is 022. A copy of the handwritten linearity calibration log is attached (Table 3-2).
The corresponding LINCALC input file (LINO22.INP) appears below.
LINO22 Data
Log Signal
0.3
o
o o
0.6
255.0
253.2
218.6
175.9
140.2
111.1
85.5
· 66.9
51.8
39.1
28.9
21.7
14.4
10.4
7.0
3.9
2.4
1.6
0.0
1.5
1.6
1.7
1.8
1.9
2.0
4.0
The first and last entries are set to 255 and 0 respectively, while the remaining 17
entries are made from the linearity log. In this example, the values from the RUN ONE
column of the log were chosen.
20

Filter Input File
(FILSVV.INP)
Effective Filter
Irradiance Computations
[FILTER
Filter Irradiances
(FILSVV.OUT)
Absolute Input
Files ( ABSaaa.INP)
Mid Scale Radiance (MSR)
Computations (ABSCALC)
MSR Values
(ABSaaa. OUT)
Decimal Linearity
LUT ( LINIII.DEC)
Linearity Input File
(LINeee.INP)
Linearity LUT
Determination
[LINCALC)
Compute Absolute
Calibration Entries
for Ratio
Calibration File
21
Integer Linearity
LUT ( LINIII.INT)
Image Offset and
Magnification File
( OFFSET.FLD)
Preliminary
Ratio Calibration
File ( CALsVWW.RAT)
NDR, SPR and
Absolute MSR
Tables

Field Unit Version
Hardware Log
Preliminary
Hardware Version
Log ( VERSION.LOG)
Fig. 3-2. Preliminary calibration and hardware
version file preparation sequence.
RUN TWO
AVG
STD
SUB ASSEMBLY ASSIGNMENTS
E/O CAM: LIN CAL DATA SHEET
Camera Serial No. 6:38+8669
Tape Ident. No. LINOO
Filler Assembly Serial No. Done
Lens Type & Serlal No..
505 Serial No.
Unit 7ul
Special Filler
TMI: 2039 FG! NKAN 406
RUN ONE
Lamp
SPECIAL REMARKS
Position
N.O.
N.O.
(Log Unils)
| AVG STO Lamp Volls
Lamp Currents
0.0
6255, 255io
17.51 . 1.001
0.1
2550 255io
0.2
IsF .01255:0
0.3
255.: o: 1255 io
0.4
1253.24 2.5 254.9: 014
0.5 218.6 3.8 1226,4 4,6
0.6
11759: 3.21782.21 3,5
0.7
775.3
1140.27 2.81245.0;3,0
145,0
8 Illi 26 115.4 2.5
Assisi 24 1 88.91 215
.1.0
166.9 2.6 70,0!2.5
1.1 Isl.& 12.6 53.712.4 Szil's 53.7 se
1.2
139.112.91410:34
1.3
28.9.2.9. 1360;3.2
1.
4 21,73.0 73.0 13.4
1.5
14.4 i 3.0.1.313.0
1.6
10.7 3.
0 11.5 3.0
177.01 0.
97.5 3.11
1.8
3.912,814,1 17,8
1.9 Tal a l 123 1
- 2.0 TIG aro 11.6 11.9
016
DATE_182/89 START TIME 1430% END TIME / 500 OPERATOR JRV

L440
DIA
Table 3-2
Sample linearity calibration log for LINO22 from the
Columbia, MO, version 701 camera
22
To Execute: Enter the LINCALC at the prompt, i.e.
D:\CALIB>LINCALC
Screen Output:
Enter the name of the linearity you wish to access
Sample name is LIN004. This will access
a file named LIN004. INP
ENTER LIN NAME:
lin022
THE INPUT VALUES ARE:
255.0
253.2
218.6
175.9
140.2
111.1
85.5
1.0 66.9
1.1 51.8
1.2 39.1
1.3 28.9
1.4
21.7
1.5 14.4.
1.6 10.4
1.7 7.0
1.8
3.9
1.9
2.4
2.0
4.0
THE LOG VALUE CORRESPONDING TO 128 IS .7419
in'ncininis initsioonis
ܩ ܘ ܩ ܩ ܗܘ
INTEGER F
S BE
AND
IT IN
INT
NOW THE INTEGER FORMAT LUT IS BEING GENERATED AND PUT IN FILE lin022.INT
NOW THE DECIMAL TABLE IS BEING GENERATED AND PUT IN FILE lin022.DEC
Stop - Program terminated.
D:\CALIB>
Integer and decimal LUTs are written to the two output files, LINIII.INT and
LINIII.DĒC, respectively. The decimal LUT has an entry for each tenth of input signal.
This file is not used in the downstream processing. The integer LUT is copied into the
ratio calibration file, and looks like the following example.
Integer Output Data for Linearity lin022
Inp sig
Out Sig
ONM
Econ
235
236
237
238
249
251
253
254
255
255
255
(The entire file for this sample is part of the CALZV01.RAT file presented in a later section
of this report.)
3.2.2 Program FILTER
Program FILTER computes the effective irradiances of the calibration lamp used in the
absolute calibrations as viewed through the 4 spectral filters placed in the spectral filter
wheel. Tech. Memo AV89-057t describes an earlier version of the routine. While some of
the input parameters changed, the basic methodology did not. FILTER requires an input
file named FILsVv.INP. The contents of FIL7V1.INP for Columbia are listed below.
Unit 7 v1
Lamp:
F12L
Filter
Ident
Lambdl Lambd2
O
AWNhd
57530-15
57610-15
57530-17
57610-16
430
615
430
620
480
695
480
690
O
O
The lamp identification is taken from the absolute calibration logs that will be presented in
the next section. The identification numbers for the 4 filters are taken from the component
list reproduced in Table 3-1. The final two entries, Lambd1 and Lambd2, are the filter half
power points read from the strip charts provided with each filter. The strip charts for the 4
filters in this example are shown in Fig. 3-3. FILTER assumes that the spectral
transmittance curve can be approximated by a Gaussian distribution fit to the half power
points (see Tech. Memo AV89-074t). This curve is then combined with the lamp
irradiance and nominal camera response curves to determine the effective filter irradiances.
24

WAVELENGNI
HEIMAH
BANDWIDTH
RAWIDU
0.18 YETENGTH
WAVELENGTE
Al met
.LOCKIES
Hochiile
ilmu
STANDuis
issot
Nb:1, 13
0.6
-
25
:
c
0.2
IIIIIII: TTTTTTTTTT! : ITTTTTTTTTI.IT
UP
0.0
Ulin
Fig. 3-3. Spectral filter curves provided by Oriel
for the filters in hardware version 711
(Columbia, MO, initial hardware version)
To Execute: Enter FILTER at the prompt, i.e.
D:\CALIB>FILTER
Screen Output:
Enter the station or unit number,
and then the version number,
for the instrument with the filters you
wish to process. For example, entering
i for both unit and version will cause a
file named FIL1V1.inp to be accessed.
ENTER UNIT (STATION) NUMBER:
ENTER VERSION NUMBER:
FILTER FILE FIL7V1. INP
Input Data:
Lamp: F12L
Filter
Ident
Lambdl Lambd2
ܝܕ ܢ ܚ
57530-15
57610-15
57530-17
57610-16
430
615
430
620
480
695
480
ܩ
690
Output Data:
Filter
Lambda Delta
Trans Cam Sens
Lamp
SxTXE
SxT
10
O
10
O
O
.00
.00
.03
.11
.26
.50
.78
.97
.97
O
O
380
390
400
410
420
430
440
450
460
470
480
490
500
510
520
O
.0987
.0987
.0987
.1017
.0972
.0965
.1047
. 1085
. 1130
.1152
.1137
.1122
. 1145
.1182
.1197
1.510
1.829
2.163
2.528
2.932
3.362
3.825
4.304
4.823
5.373
5.941
6.545
7.153
7.753
8.366
.003
.017
.075
.272
.733
1.622
3.120
4.542
5.301
4.823
3.377
1.887
.867
.320
.092
.002
.009
.034
.108
.250
.483
.816
1.055
1.099
.898
.569
.288
.121
.041
.011
O
O
M بر لبه ل
10
O
O
O
10
10
ON
O ON
O
10
FILT 1 RESULTS
Average Irradiance (UW CM-2 NM-1):
Average Irradiance (W M-2 UM-1):
4.677
46.77
26
590
10
.16
.1444
12.190
2.823
.232
730
10
.09
.1354
19.360
2.292
.118
FILT 2 RESULTS
Average Irradiance (UW CM-2 NM-1):
Average Irradiance. (W M-2 UM-1):
Avel
H
a
15.770
157.70
380
10
.00
.0987
1.510
.003
.002
W
.
520
10
.00
.1197
8.366
.092
.092
.011
w
FILT 3 RESULTS
Average Irradiance (UW CM-2 NM-1) :
Average Irradiance (W M-2 UM-1):
4.677
46.77
590
10
.09
.1444
12.190
1.612
.132
730
10
.04
.1354
19.360
1.087
.056
FILT 4 RESULTS
Average Irradiance (UW CM-2 NM-1):
Average Irradiance (W M-2 UM-1):
Stop - Program terminated.
15.758
157.58
D:\CALIB>
The output file (FIL7V1.OUT) echoes the screen print, beginning with the line "FILTER
FILE FIL7V1.INP" up to the filter 4 results.
3.2.3 Program ABSCALC
Program ABSCALC computes the Mid Scale Radiance (MSR) values used to derive
the calibration parameters employed in ratio processing. (Tech. Memo AV89-058t
describes the calculations.) Data from absolute calibration types 1-5 from the List of
Acquired Calibration output presented earlier are used to derive the absolute calibration
information entered in the ratio calibration file. These data are extracted from the manual
logs recorded during the JABS4 runs, and entered into files named ABSaaa.INP for input
to ABSCALC, where aaa is the absolute calibration reference number.
Table 3-3 is the type 1 manual log from ABS120. The corresponding input file for
ABSCALC (ABS120.INP) is listed below. ABSCALC also uses the corresponding
LINIII.INP file, in this case, LINO22.INP.
27
Abs usfilter
at Aporao
Unit
70l
E/O CAM: ABS CAL DATA SHEET
ALS 120

Lamp SPECTRAL #1
LAMP POS Pos com
of (cm) +N.D.' AVG | STO
40 155 10.3. 28
40 Izss 13./
30 49.0 1.0
Ho N.D.
liso
100
75
SPECTRAL #3
TN.D. AVG
4O 4.2
30
5.9 L
SPECTRAL #2
| AVG STD
1 12.3 12.
39.2 l 2.0
l 4.4 l z.7
140.
21 Hol
I zas.a I 6.0
STD
2.6
2.2
70
SPECTRAL #4
bene
+ N.B. AVG STD
100 Gen r o
7 I loob 22
55 376
2.5
40 I 76.0
3.2.
30 1 133,6 I sob
45
5.7
28
120 30
| 7040
40 11 55
148.8
24.9
494
l 3-1
3.1
3.0
| 208. I 1 5.9
ss 114104 l uol
75,1 7.
100 58.7 2.2
1o Tizu 12.5
30 34.?
vo | .?
55 1 23.3
75 1 17.0
100 | 20
30 Lor?
40 2.9
Tz.
| ..:
| 2.2
l
2.
3
SUB ASSEMBLY ASSIGNMENTS
Camera Serial No. –
672-26 69
Filter Assembly Serial No. _ o
Lens Type & Serial No..
505 Serial No.
IME:
2039 FG: Humovab
CALIBRATION LAMP DATA
Lamp Serial No.E12L
Outputlal HCP) Colorkal
Rated V_ VOC Rated
Set V_ VOC
Read _
CALIBRATION TARGET DATA
Target Ident
Target Retiectance
Attenuation Ident
Allenuation Rell.
Target Location; dı = dy + Oy:by
m
9
DATE
1 Feb
START TIME 1130
- END TIME 1700
OPERATIQR 563GMA
Table 3-3
Sample calibration log from a Type 1 (Abs. vs Filter)
absolute calibration for the 7V1 sensor fielded at Columbia on 8/FEB/89
ABS120 Data
Spec
ND
Aper
Lamp
55
40
30
30
40
55
150
100
75
55
45
55
ܢܙ ܙ
75
100
AAAAAAAAAA WWW WNNNNNNNNNNNNNN r >
Sig
10.3
25.5
49.0
48.8
24.9
9.4
12.3
39.2
74.4
141.2
208.9
208.1
141.4
75.1
38.7
12.4
4.2
8.9
10.3
2.7
6.6
15.6
37.6
76.0
133.6
134.3
76.7
23.3
17.0
7.0
150
ܢܝ ܢܙ ܩ
O
OOOOOOOOOOOOOOOOOOOOOOOOOOO
O
40
30
30
40
100
75
55
ܣ
ح
O
30
O
O
O
55
75
100
To Execute: Enter the program name ABSCALC at the prompt, i.e.
D:\CALIB>ABSCALC
Screen Output:
Enter the name of the absolute you wish to access
Sample name is ABS014. This will access
a file named ABS014.INP
ENTER ABS NAME:
ABS120
Enter the name of the linearity you wish to use with these data
Sample name is LINO04. This will access
a file named LIN004. INP
ENTER LIN NAME:
LINO22
THE LINEARITY INPUT VALUES ARE:
.3 255.0
4.0
.0
29
Enter the absolute irradiance,
in watts/(m* 2 nm) at 50 cm
These are computed from the spectral curves of
the filters and the lamp irradiance curve
Sample values are:
Filter 1: 44.6
Filter 2: 159.0
Filter 3: 44.6
Filter 4: 162.0
1 IRRADIANCE:
2 IRRADIANCE:
FILTER
46.8
FILTER
157.7
FILTER
46.8
FILTER
157.6
3 IRRADIANCE:
4 IRRADIANCE:
Lamp
Spec
ND
Aper
Sig
MSR
55
10.3
25.5
49.0
48.8
24.9
9.4
87.61
91.93
97.31
97.66
93.71
93.12
150
100
45
O
45
12.3
39.2
74.4
141.2
208.9
208.1
141.4
75.1
· 38.7
12.4
35.34
35.25
36.10
36.68
36.66
36.82
36.63
35.79
35.63
35.13
75
a pa A A A A A wwww NNNNNNNNNN trn
100
150
4.2
8.9
10.3
2.7
255.00
323.75
294.47
313.48
6.6
15.6
37.6
76.0
133.6
134.3
76.7
23.3
17.0
7.0
114.96
120.53
120.68
124.34
130.64
129.92
123.26
175.67
115.32
111.60
30
40
55
75
100
Stop - Program terminated
D:\CALIB>
30
The output file, AB$120.OUT, begins by naming the input ABSaaa file, and then lists
the irradiance values entered. The remainder echoes the screen output starting with the
"Lamp Spec" line.
3.3
Absolute Radiance and Ratio Calibration Constant Determination
The conversion factors needed to transform the raw digital WSI signal values into
absolute radiance and radiance ratios can be determined from the MSR values computed in
ABSCALC. Tech. Memo AV90-049t, "Derivation of Calibration Constants", provides a
detailed step-by-step discussion of the ABSCALC data reduction procedure. Two of the
resulting factors are of interest in the ratio processing: the Spectral Ratio (SPR) and the
Neutral Density Ratio (NDR).
Each red filter and each blue filter exhibits some variability in its peak wavelength and
bandwidth relative to the nominal values. Thus, the measured signal through two different
spectral filters of the same color will vary for the same sky radiance distribution. Similarly,
the ratio between the signals from two filters of different color will vary from filter pair to
filter pair for the same sky radiances. The SPR converts a ratio of the blue and red signals
(corrected for the camera response) to a ratio between the input radiances. As in AV90-
049t, SPR is defined as:
SPR = MSR(1) / MSR(2)
where 1 represents the blue filter in position 1 on the spectral filter wheel, and 2 represents
the red filter in position 2.
Due to the limited dynamic range of the video camera, two sets of red and blue images
were needed, with the second set buffered from the first by a .5 log neutral density filters.
Considering the spectral variability of the color filter response, and the slight non-neutrality
of the offset filters, the spectral ratio of the second pair (SPR) will not necessarily match
that of the first pair. SPR' is defined as:
SPR' = MSR(3) / MSR(4)
where 3 and 4 refer to positions on the spectral filter wheel.
The ratio routine chooses the 3-4 filter pair if either of the 1-2 pair is off scale bright.
Consider two regions of the sky with the same radiance ratio, one region brighter than the
other. If the brighter region is off scale bright in the 1-2 pair and the darker region is not,
the 3-4 pair would be used in the brighter region ratio computation, while the 1-2 pair
would be used in the darker region. If SPR and SPR' differ, the corrected signal ratios in
the two regions would also differ despite both region having the same radiance ratio. The
NDR factor is designed to compensate for the apparent difference, where NDR is defined
as:
NDR = SPR /SPR' = [MSR(1)/MSR(2)] /[MSR(3)/MSR(4)]
= [MSR(1)*MSR(4)] /[MSR(2)*MSR(3)]
The practical effect of applying these constants is as follows. The NDR factor is used
to make the corrected signal ratio from the 3-4 pair appear as if it came from the 1-2 pair,
while the SPR factor is used to correct the 1-2 corrected signal ratio to a radiance ratio.
Since the filters in the neutral density wheel are not perfectly neutral, SPR and NDR ratios
are computed where possible for each of the 4 positions on the neutral density filter wheel.
The SPR and NDR values computed for the Columbia example are given below.
AWNEZ
SPR
2.51
2.43
3.34
3.34
NDR
.99
99
.92
.92
In this case, the absolute calibrations for ND4 were off scale dark, and the values from
ND3 were used for ND4.
3.4
Preliminary Ratio Calibration and Hardware Version Files
The next step in the procedure is to enter all the pertinent calibration data on the
particular hardware version into one common file, and make the appropriate entry into the
hardware version log used by many of the ratio and cloud decision processing programs.
The ratio processing calibration files are named CALsVvvv.RAT where s is the station or
site number, and vvv is the hardware version identifier. The preliminary version of the
CAL7V01.RAT file available at this stage of the procedure is given on the next 7 pages.
The file name appears first, followed by the date of the most recent file entry. Other
pertinent features of the hardware are then listed. The linearity LUTs from the
LINO22.INT file created by LINCALC are next. In most cases, the value of the output
LUT at the 0 byte value location must be manually changed to 0.
The important ratio correction factors appear in the table of numbers labeled "Absolute
Computed Correction Factors" on page 6 of the calibration file. The first column of
numbers are the ND filter wheel position. The transition thresholds are in the next column.
If the corrected signal in either spectral filter 1 or 2 exceeds this value (offscale bright in 1
or 2), the ratio from the 3-4 pair will be used for the pixel. If either of the spectral 3 or 4
corrected signals exceeds the value in the third column, the ratio pixel output will be
assigned a value indicating offscale data. The values 250 and 254 for these thresholds are
reasonable for normal cameras. The NDR values are taken directly from the calibration
computations. The final column contains the numerical value of 4/SPR. A brief
explanation of this factor follows.
TU
The ratio routine takes this factor and multiplies it by 64. The resulting factor is then
used to scale the red/blue corrected signal ratio to an 8-bit integer value (between 0 and
255). In mathematical terms,
Ratio Byte Value = 64 [4/SPR) (CorSig(2)/CorSig(1)]
= 128 [2] [MSR(2)/MSR(1)] (CorSig(2)/CorSig(1)]
where CorSig is the signal corrected for nonlinearity. Using the definition of MSR in
AV89-058t, the ratio byte value can be related to the radiance ratio as follows.
Ratio Byte Value = 128 [2] [Radiance(2)/Radiance(1)]
Our original intent was to have the ratio byte value of 128 correspond to a red/blue radiance
ratio of 1.0, which is close to the expected ratio for opaque clouds. In practice, the
empirical factor of two was needed to raise the ratios to near the desired level. The reason
for the factor of two discrepancy is still open to question. (See the related discussion in
AV90-049t.)
The next set of entries in the CALsVvvv.rat files are the MSR values and aperture
factors at the different ND, SP and aperture settings that would be needed for absolute
radiance computations. These values are produced in the post-ABSCALC computations,
but have yet to be used. The final series of entries are the image geometry factors taken
verbatim from the OFFSET.FLD file of the FLDCAL program. These include, the image
edges, horizontal and vertical offsets, magnifications for all the spectral filter position
combinations with the first two neutral density wheel positions. Images in ND positions
are usually too dark for edge determination, so the values from ND position 2 are assumed
to be valid for positions 3 and 4. The ratio routine uses the offset and magnification
information to coregister pixels from the 4 spectral filter positions during the ratio
computations.
The final step in the calibration reduction procedure is to make a preliminary entry of
the new hardware version into the VERSION.LOG file, as shown below.
File VERSION.LOG
Updated 10 Oct 91
List of hardware versions used in field, for access by ratio programs.
STATION
DATE
Mo/Da/Yr
HARDWARE DATA
Vers Qual
QUALITY SOFTWARE
Indicators Vers
COMMENTS
2
3
2
3
3
4
16 88
19 88
29 88
12 88
M س با ما
0000000000
0000000000
0000000000
0000000000
1.10
1.10
1.10
1.10
2A
2B
04 03 90
02 08 89
10 11 91
0000000000
0000000000
0000000000
2.60
1.71
2.70
The last two lines pertain to the entries needed for the first version (1A) fielded at
Columbia, MO. The DATE column indicates that the station began collecting data on
8/FEB/89 using FIELD software version 1.71. At this point, the data quality values are
given default values of 1 for the DATA quality, and ten O QUALITY indicators. The last
line was included for smooth exit from the ratio routine. More will be said in the next
section about the entries into this log.
33
File CAL7V01. RAT
Date: 10/OCT/91
This file contains calibration data for unit and version
indicated in file name, for input to ratio programs
CALIBRATION CONFIGURATION
Field Unit No.
Field Version No.
Filter Chg No.
Camera Type
Camera Serial No.
TMI Serial No.
FG100 Serial No.
FC-08
2710
638-8669
TMI 2039
HKM 406
FIELD CONFIGURATION
Field Unit No.
Field Version No.
Filter Chg No.
Camera Type
Camera Serial No.
TMI Serial No.
FG100 Serial No.
FC-08
2710
638-8669
TMI 2039
HKM 1298
LINEARITY LUT VALUES
Data Source:
File LINO22.INT, 08/18/89
Original Calibration file: LINO22, 01/26/89
una WWN O vun no WNE
File CAL7V01.RAT - Page
1
34
M ا
M
N
N
M
N
N
N
M
ܝܕ ܢ ܚ ܩ ܟ ܗ ܢ ܣ ܩ ܘ
ا
ܢ
ا
ܢ
ا
ܚ
ا
ܩ
ا با لبه
ܩ ܗ ܙ
ه
ܣ
ه
ܩ
ه ه ه
ܘ ܢܙ ܝ
ه
ܚ
ه
ܩ
ه
ه
ܗ
ه
ܙ
ه
ܤ
ه 0
ܘ ܘ ܢ ܢܨ ܚ ܚ ܟ ܗ ܢ ܣ ܩ ܘ
0
0
ܝܕ ܢ
0
ܚ
0
ܩ
0
ܟ
0
ܗ
0
ܢ
0
ܣ
0 للا
ܩ ܘ ܢ
ܢ
ܚ
ܩ
aus W WNH
50
63
66
70
73
74
75
File CAL7001.RAT - Page 2
35
75
76
78
79
80
81
77
78
79
80
82
81
0
O
0
O
0
88
85
86
87
89
90
88
0
89
90
0
91
0
von A WWNA
0
0
0
UAWN
Ova
98
ao
99
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
100
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
115
116
117
118
119
120
121
122
123
124
125
126
127
128
127
128
File CAL7701.RAT - Page
3
36
129
130
م
م
م
م
الا هر )
م
م
129
130
131
132
133
134
135
136
137
139
140
141
142
143
144
144
145
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
146
ه
م
ل
ه
147
148
ه
)
و14
لا لا لا ح م
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
)
و
161
168
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
169
170
171
172
173
لا
174
175
م
176
178
179
179
180
181
182
183
179
180
181
182
لا لا لا لا
همه مه وه
File CAL7001.RAT - Page
4
17
183
184
185
186
187
188
189
190
191
184
185
186
187
188
189
190
191
O
W
193
194
O
192
O
193
Ur
195
196
O
O
197
198
199
200
201
202
O
O
203
O
O
O
194
195
196
197
198
199
200
201
202
203
204.
205
206
207
208
209
210
211
212
O
7 WN
fond
ܢ
204
205
207
208
209
210
211
212
213
214
216
217
218
219
220
222
223
224
226
227
ܠ
N p
ܙ
ܗ
N
ܙ
N
ܣ
N
218
219
N
220
N
ܘ ܝܐ
229
221
222
223
224
225
226
لا لا لا لا
w
w
w
228
229
w
230
232
233
235
236
238
239
241
243
244
246
248
249
251
A
231
A
A
د تا با به
File CAL7001.RAT - Page 5
38
W
u
un
253
254
255
255
255
255
255
255
ол
л
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
ол
л
un un
n
255
255
255
255
ABSOLUTE CALIBRATION DATA
Data Source:
File: Manually derived from AB$120-126.OUT, 08/18/89
Original Calibration data: ABS120-126, 02/01/89
Absolute Computed Correction Factors
2/1 - 4/3 4/3
ND Filter Transition Offscale NDR
250
254 0.99
250
254 0.99
250
254 0.92
250
254 0.92
14/SPR]
SPR
1.59
1.65
1.20
1.20
.
Absolute Mid Scale Values
These parameters are not used unless we are
generating a tape of calibrated data, however
they were used to generate the above correction
factors.
Mid Scale Values for APER = 0
SP =
1
2
3
4
ND = 1
ND = 2
ND = 3
ND = 4
104
253
635
1860
41.5
104
190
556
342
832
1780
5210
137
347
652
1907
Absolute Aperture Factors MSR(aper=0) /MSR (aper)
File CAL7V01.RAT - Page
6
39
APER =
APER
0
1.00
40
2.79
80
5.12
120
9.63
160
16.5
Factor:
IMAGE EDGE INFORMATION
Data source: Field calib, 8 Feb 89, calib COL002
Selected edges for neutral density = 1
Filter #
Top
Bottom
OM
Left
Right
# Iter.
---------
----
------------
-----
--
W
--------------
i
a
as WNW is
15
440
465
468
73
443
ir w
468
73
472
72
441
445
W
Resulting magnification and offset values
ND
SP
MAG
XOFF
YOFF
لی
O
لی
1
1
2
3
4
.984
1.001
.996
1.008
P
0
M
O
لیا
and
008
Selected edges for neutral density =
2
from 8 Feb 89
Filter # Top Bottom Left Right # Iter.
Iter.
----------------- ------------------------
Standard - 10 470 71 439
1
m
را
i
1
71
با ب
68
cina O
471
473
474
478
446
448
446
451
ه
68
ه
Resulting magnification and offset values
ND
SP
MAG
XOFF
YOFF
لما لا
1
2
3
4
1.013
1.026
1.022
1.035
۱
مر
OBSTACLE INFORMATION
Number of Obstacles = 0
File CAL7V01.RAT - Page 7
4.0
Ratio Input File Preparation
The processing of the raw radiance image tapes collected in the field to the ratio stage
is the most involved part of the entire WSI tape processing procedure. It is at this stage that
many of the idiosyncrasies in the data collection are resolved, and data quality specified.
Five input files must be prepared before a field tape from one of the stations can be
processed to ratio. These files are:
VERSION.LOG
- defines for all of the sites, when system changes were made
either to the EO sensor, or to the control computer. Changes
in data quality are also recorded.
CALsVvvv.RAT - contains the calibration information for a particular hardware
version (site s, version vvv). As noted in Section 3, this
information includes the linearity LUT, the spectral pair
selection and offscale bright thresholds, the NDR and SPR
ratio adjustment parameters, absolute radiance MSR values,
image size parameters, and image boundaries of obstacles
blocking the field of view.
DATES.LST - lists for site s, the tape identifier (LLL###), the first date as it
appears on the tape, the number of days with data on the tape,
and the date correction factor, if any.
OCCs.DAT
- documents the dates when the occultor parameters (arm
number and/or angular offset) changed at site s.
TIMES.COR
- lists for site s the dates when the time correction changed, and
the magnitude of the correction.
Preliminary versions of VERSION.LOG and CALsVvvv.RAT are available at the end
of the calibration data reduction described earlier in Section 3. An initial version of the the
DATES.LST file can be prepared from the station log books described in Section 2. The
following presentation describes the steps taken to verify and complete the information in
these preliminary files, and to determine the entries to the remaining files (OCCs.DAT and
TIMES.COR). Figures 4-1 and 4-2 outline the steps in this procedure. As noted
previously, the procedure was evolving as tapes were being processed, and the summary
presented here describes how a new tape would be handled today.
Raw
Radiance
Field
Tapes
Verify Image Size
Parameters and
Offscale Thresholds
Preliminary
CALsVvW.RAT
Linearity
Calibration Logs
Preliminary
VERSION.LOG
Determine Azimuth
Offset of Sensor
( CHECKAZM)
Revised
CALsVvW.RAT
Sun Position File
(Ssyymmdd.SUN)
Site Azimuth
Offset List
Determine Occultor
Arm and Offset and
Time Corrections
(CHECKLAN)
UCHA

LLL###C.LAN
Files
Prepare Occultor
Parameter and Time
Correction Ratio
Input Files
.
Handwritten LAN
Observations
Output File
(CHECK.OUT)
Occultor Input
File (OCCS.DAT)
Data Quality
Specification
Prepare Tape ID
Date List and Check
for Date Slips
Time Correction
File (TIMES.COR)
.
.
Station Log
Date List File
(DATES.LST)
Cloud Cover Observation
- Form 10 or ETAC
Summaries
Revised
(VERSION.LOG)
-
Fig. 4-1. Ratio input file preparations.
A2

Revised
VERSION.LOG
DATES.LST
OCCS.DAT
Preliminary
Composite Ratio
Processing
(TAPRATPN)
Raw
Radiance
Field
Tapes
TIMES.COR
Revised
CALSVvvv.RAT
Sample Ratio
Image Files
Adjust NDR and
SPR parameters
Final
CALsVvvv.RAT
Specify Obstacle
Boundaries
Fig. 4-2. Final ratio calibration file entry determinations.
4.1
Image Size and Offscale Threshold Determination
The first step in the procedure is to verify the image size and placement information in
CALsVvvv.RAT originally obtained from the FLDCAL output file OFFSET.FLD. The
verifications are made with full resolution images from raw radiance field tapes. The steps
in the procedure are outlined below.
Step 1: Use IMFIND to extract several image sets of all four quadrants for both the
first and second positions on the ND wheel.
The best periods to pick are when the flux control algorithm is alternating in its choice
of ND filter between positions 1 and 2. It is also preferable to choose cases for ND 1
when the iris is fully closed (IRIS=0) providing the best focus at the image edge.
Choosing periods when the occultor is not obscuring the sky at the top, bottom, left
and right edges also helps. WARNING - Use the IMFIND version from 15 May
1989 to access the data instead of earlier versions of IMFIND, or programs RDIMGS
or WRIMGS, because the placement of the image by this version of IMFIND agrees
with the placement in the ratio program.
Continuing with a Columbia example, the following table lists some dates and times
when four sets of radiance images were extracted from the radiance data tapes to
illustrate this step.
Time
ND
Tape ID
001
001
Date
12/FEB/89
12/FEB/89
6/APR/89
6/APR/89
16:40
17:20
19:20
18:50
009
009
N-
Step 2: Locate the image edges.
Use a mouse routine (eg., VIEWIMG, SHOIMG, or SAMPIMG) to locate the cursor
at the images edges, and record these positions. My preference is to find the outer
edge of the visible sky dome, instead of the bottom of the four screws visible in the
images. The edges for the sample images are
ND = 1 Cases
ND = 2 Cases
SP
lan
16:40 12/FEB/89
Top Bottom Left Right
13 462 79 440
8 466 76 443
13 464 77 439
8 470 75 445
AWNA
17:20 12/FEB/89
Bottom Left Right
468 75 444
473 72 447
471 73 444
4 476 71 450
Awng Awn-
SP
19:20 6/APR/89
Top Bottom Left Right
462
438
10 465 76 442
13 464 78 438
10 469 76 444
18:50 6/APR/89
Top Bottom Left Right
468 74 443
5 472 71 447
8 470 72 443
4 476 71 449
2
3
4
44
Step 3: Examine the bright regions on the images with the mouse.
While the image edges are being evaluated, the nature of the offscale bright portions of
the images can be examined. In the case of the Columbia images above, the offscale
pixel values were 255. Thus, the offscale bright thresholds can remain at 254, and the
transition threshold can stay at 250. Other cameras have a more pronounced rolloff at
the high end. For example, the camera fielded with version 3 at Malmstrom (513)
exhibited a significant high end rolloff as shown in a portion of its linearity log sheet
(LINO29) copied below.
Lamp
Position
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Avg. Pixel
Value
212.1
216.5
221.3
230.5
235.5
230.5
204.8
168.2
Note how the average pixel value increases from 168.2 at 0.7 logs to 235.5 at 0.4
logs, then decreases to 212.1 at 0.0 logs. A raw pixel value between 212 and 235
could be associated with one of two absolute radiance values. If the peak entry from
the linearity log is used, then some regions that should be defined as offscale bright
will have ratios that are on scale, but meaningless. If a lower threshold is chosen, then
the previously erroneous regions will be identified correctly as offscale bright, but
other regions with meaningful ratios will also be designated offscale bright. The latter
more conservative strategy was chosen.
The other point to remember is that the thresholds entered into CALsVvvv.RAT will
be applied in the ratio routine after passing the raw imagery through the linearity LUT.
Let's look a little more at the Malmstrom example. The high end of the linearity LUT
appears below.
Raw
Value
210
Corrected
Value
223
225
227
230
232
234
Raw
Value
218
219
220
221
211
212
213
214
215
216
217
Corrected
Value
240
242
244
247
249
251
253
255
222
236
223
224
225
238
If we wanted to exclude the entire rolloff region (above 212 in raw value), the
corrected value of the offscale threshold should be set at 227, and the transition
threshold would be set 5 to 10 counts below that.
45
Step 4:
Compute the image magnification and center offsets
After examining the raw sample images with the mouse, the image magnification
factors and center position offsets must be computed for all spectral filter positions in
ND positions 1 and 2. First, the average edge positions are computed. For the
Columbia example,
SP
SP
1
ND = 1 Averages
Top Bottom Left Right
14 462 79439
466 76 443
13 464 78 439
9 470 76 445
ND = 2 Averages
Top Bottom Left Right
468 75 444
473 72 447
471 73 444
476 71 450
3
4
The offsets and magnifications are then computed using the following equations.
XOFF = 255 - (Left + Right)/2 YOFF = 240 - (Top + Bottom)/2
XMAG = (Right - Left)/368 YMAG = (Bottom - Top)/460
MAG = (XMAG + YMAG)/2
The values for the initial Columbia setup are:
ND = 1
SP XMAG YMAG MAG XOFF YOFF
10.978 0.974 0.976
2 0.997 0.994 0.995
3 0.981 0.980 0.981
4 1.003 1.002 1.003 -6 0
ND = 2
SP XMAG YMAG MAG XOFF OFF
1 1.003 0.998 1.000 - 5
2 1.019 1.017 1.018
3 1.008 1.007 1.007
4 1.030 1.026 1.028 -6 0
AWN
Step 5: Make changes to the CALsVvvv.RAT file.
Appropriate corrections can now be made to the ratio calibration file. In our Columbia
example, no changes would be made to the absolute calibration tables at this point in
the calibration procedure, because the camera did not exhibit any rolloff at the high
end. The image size information would be updated using the values above, as shown
in Table 4-1.
The image size parameters have to be checked whenever significant hardware changes
are made. These include: relocation of the sensor, disassembly of the sensor either at MPL
or on site, and replacement of the FG-100 board (including swap out of the entire
computer). An FG-100 swap can cause the image to shift in the X direction, while the Y
placement remains unchanged. For example, the FG-100 in Unit 1 Version 5A was
replaced at C-Station on 6 July 1989, changing the version to 5B. The Y offsets for
versions 115A and 115B were the same, but the images shifted 8 or 9 pixels to right,
decreasing the X offsets from 5 or 6 for version 1V5A to -2 or -3 for version 1V5B.
46
Table 4-1. Revised CAL7001.RAT Image Size Table
IMAGE EDGE INFORMATION
Data source: Field Tapes 001 and 009
Selected edges for neutral density = 1
Filter # Top Bottom Left Right
tom
# Iter.
tandard
O
10
470
439
I
14
462
466
464
470
76
78
439
443
439
13
76
445
Resulting magnification and offset values
ND
SP
MAG
XOFF
YOFF
1
1
اسم ال
1
2
.976
.995
.981
1.003
-4
-5
2
ه
4
Selected edges for neutral density = 2
Filter #
Top
Bottom
Left
Right
# Iter.
--------
--
101
standard
10
O
470
71
i
1
5
468
473
471
476
PWN in
444
447
444
450
Resulting magnification and offset values
ND
SP
MAG
XOFF
YOFF
2
2
-5
-5
1
2
3
4
1.000
1.018
1.007
1.028
sur un
1
HOC
OBSTACLE INFORMATION
Number of Obstacles = 0
Revised File CAL7001.RAT - Page 7
4.2 Sensor Azimuth Offset Determination
An important step during the installation of a WSI unit is determining a landmark
designating true north, and aligning the instrument along the north-south direction. Proper
alignment is important for two reasons: keeping the dome shaded from direct sunlight, and
accurate specification of a pixel's zenith and azimuth angles in the image products. Neither
the landmark selection or the alignment procedure is perfect. Fortunately, the sun provides
a well defined landmark within the field of view, and the azimuth offset of a camera setup
can be determined. This is done by tracking the sun's position in the WSI images on a
clear day, then plotting the observed positions against an anticipated track computed from
the solar ephemeris in a program called CHECKAZM. The procedure for running
CHECKAZM is described below.
Step 1:
Find a reasonably clear day
If practical, chose a day between the vernal and autumnal equinoxes, preferably in the
summer, when the sun is visible for as long as possible, and a greater length of solar
arc can be observed.
Step 2: Load the appropriate field tape.
Select the field tape containing the chosen date and load it onto the read ExaByte.
Space over EOFs to the beginning of the selected day using IMFIND, WRIMGS or
RDIMGS. I prefer IMFIND for its multiple EOF skip ability.
Step 3: Enter a directory containing the VERSION.LOG file and the appropriate
CALsVvvv.RAT file with the revised image size information. Then run the
CHECKAZM program. The following example is from the first hardware version at
Columbia on 20 April 1989.
To Execute: Enter CHECKAZM at the prompt, i.e.
D:\CALIB>CHECKAZM
Screen Output:
Welcome to CHECKAZM - Version 2.0
This program
Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
1991 - All Rights Reserved
Site # Location | Site # Location
----------------------------------------
MPI 1 6
Malabar
C-Station
Columbia
HELSTF
Composite 1
Kirtland
Portable 1
China Lake I
Malmstrom i
in A WNHO
48
-> Please enter the site number: 7
Input the start date of the tape as follows:
Format: Month Day Year
Example: 1 10 89
-> Please enter the starting date: 4 20 89
Calibration File CAL7001B.RAT opened successfully.*
Initial header found: Date = 20/APR/89
Sun file 578 90420.SUN opened successfully
Time X Y
# (Min) Pos Pos
Sun Present? (Y or N):
At this point, change the monitor to RGB, and respond to the prompts on the screen.
If the sun is visible, type Y, if not, type N. Actions taken by the program are
N - the program skips the next EOF, loads the next full resolution image, then
asks "Sun Present?" again.
Y - activates the mouse and writes the "Locate sun with mouse "command to the
screen.
Place the curser on the center point of the sun, then press any of the three mouse
buttons. The following question will then appear.
Is this point OK? (Y or N):
Responses: N - will return to the "Locate ... " command. Adjust mouse position
and press a mouse button.
Y • a data line will be printed to both the output file and the computer
screen. The next action line will then appear.
Get another? (Y or N):
Responses: Y - skips the next EOF, loads the next full resolution image, then asks
"Sun Present?" again.
N
- begins the azimuth offset determination.
CAUTION - exit to the azimuth offset determination is made only after a new sun
position has been defined. If no usable images appear at the end of the day, answer
yes at the next "Sun Present?" prompt, and choose a location far off the active image
area. Then respond no to "Get another?".
The offset determination begins with the nominal WSI active image being
highlighted on the RGB screen. The solar arc is then plotted in yellow, followed by
the observed sun positions in purple. An initial offset of 0.0° is assumed. The
program then prompts
Offset = 0.0
Try Another? (Y or N):
Responses: Y - requests a new azimuth offset with the prompt
Enter new azimuth offset:
A positive value will rotate the yellow sun arc counterclockwise,
while a negative value will rotate the sun arc clockwise. A value
of -1.5 gives the best fit for the Columbia example. The adjusted
arc will the be replotted followed by the prompts
Offset = -1.5
Try Another? (Y or N):
N - ends this CHECKAZM run.
The output file is assigned a name of the form Ssyymmdd.SUN, where s is the site
number, yy is the year, mm is the month, and dd the day. The output file for the Columbia
sample is listed below.
Sun file 57890420.SUN opened successfully
Time
X Y
# (Min) Pos Pos
1 790 416 246
2 850 391 216
3 910 361 193
4 970 327 176
5 1030 292 165
6 1090 255 161
7 1150 218 163
8 1210 182 172
9 1270 148 188
10 1390 91 239
М
о
un N v
Uл
Final Azimuth Offset =
-1.5
The sequence number of the sun observation is in the first column, followed by the time in
minutes GMT. The last two columns list the nominal image coordinates of the sun
position. The final, best fit azimuth appears on the last line. The top panels of Fig. 4-3
show the CHECKAZM output before and after the sun arc is adjusted. The most extreme
offset encountered in our data processing thus far was 7.5° for version 5 at C-Station. The
bottom panels of Fig. 4-3 show the effect of azimuth offset adjustment for this case.
50
Columbia, MO - Version 7v1

.
.
.
.
.
.
.
O
.
.
.
.
.
.
.
.
.
+
...
.
.
...
..
... ..
......
TIL
.
.
.
..
.
..
.
.
DO
..
.
..
.
.
.
.
6
.
0
.
.
.
.
.
.
.
.
.
..
.
.
6
...
.
.
.
100.
.
4
.
.
6..
.
..
.
.
.
....
.
....
.
.
....
..
.
.
.
.
O
1
.
.
7
...
.
... .........
...... .......
.
.
18
.
.
.
.
.
+
.
.
.
...
1
.
.
.
6.
.
.
.
.
.
.
.
.
-
.
.
.
...
.
.
.
.
.
.
...
.
.
.
•••
6.
....
.
.
.
...........
... ... ...
.
.
.
10.000
.
.
1
•
.
M
.
.
..
......
.
.
...
.
.
.
.
.
.
.
.
.
+
1
.
.
.
.
.
....
.
.
.
.
.
........
1
....
.
.
.
.....
.
.
.
.
DOU.
.
.
.
.
.....
5
.
....
.
.
.
.
.
....
.
01
.
.
.
.
.
.
.
.
.
.
.
.
....!
.
1
11 TL
.
.
.
.
.
.
.
TO
11111
.
.
.
.
.
.
.
.
.
.
.
.
......
.
!!
.
.......
..
.........
.
+
.
.
.
1
.
.
.
..
.
.
.
6 .
100
It
.
.
0
00
.
.
.
.
......
.. .. .
....
.
117
.
.
.
$O
.
.
...
. .
.
:
.
..
.
.
.
1
......
.
.
....
. . .
.
..
.
...06
.
.
.
....
. . . .
.
.
....
...
.
.
.
..
. .
.
.
.
.
.....
.
.
.
.
.
.
.
.
16.
-
.
.
.
.
.
....
.
...
.
.
.
..
.
1
.
.
.
..100€
.
.
.
.
..
..
.
.
0
.
.
.
.
.
...
.
...
.
.
1
.
....
.
O
.
.
.
.
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
De
.
.
.
. ....
.. ...
.
.
.
.
.
4
.
.
.
.
.
.
#
.
...
.....
.
.
.
.
.
..
.
.
.
.
.
.
.
1
.
.
.
DO
.
.
.
10.
.
.
.
. ..
...
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. .. .... .
....... ...
.
.
..
.
.
#
.
0
1
.
.
.
6
.
I
.
.
.
.
.
.
D
.
.
.
.
.
.
,
+
.
.
.
.
.
.
.
.
...
.
.
+
.
.
00
.
.
.
0
.
.
+
...
..
.
.
.
.
.
. .. .. ..
10.05 ..
... .
. . .
..
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
LO
.
.
.
.
.
.
..
. .
. . . .
....
. !
.
.
.
.
.
.
.
.
.
..
.
.
.
.
.
.
.
.
...
.
.
.
*
YOYOY
.
.
.
0
.
.
..
6
.
.
.
. .
.
..
.
.
..
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
..
....
.
.
.
.
.
.
.
TDI.
•
.
.
0
.
...
.
D
..
..
.
$
.
.
.
..
.
.
NO
.
.
$
.
6
:
AN
4
.
.
.
...
.
0
.
.
99.
00
...
.
.
A
.
.
.......
.
.
109
.
.
.
.
.. UDO.
136.
.
.
.
..
.
.
.
.
.
#
.
.
#
.
.
.
.
.
.
+
.....
.
.
90
..
.
.
.
.
.
.
..
.
.
0
.
.
.
..........
.
tot
.
.
.
.
0106
.
+
.
.
... .........
... . .. .
.
.
110
+ + O2
.
.
T
+
..
.
.
ATTO
. .
.
1
..
.
.
*
.
1
.
.
.
.
.
...
H
.
.
.
.
.
.
*
.
..
.0
. .
.
.
.
TO
!!.00....10
0 . 00 .
.
.
.
.
....
*
..
.
.
1 1 1
11
1
.
.
1.
.
.
.
0
.
.
10.000
.
10
.
d .
._001
2.
.
.
. .
2
2017
OTTO
C
...
..
.
0
.
TE
.
.
10.1
.
...
..
.
.
_
.
1
.
0
.
.....
1000,
14.
.
CT
.
0
.
.
t
.
TOT
1
PA
DO
.
I
.
.
.....
....
. .
.
.
.
.
.
..
..
.
:
..
.
+
+
+
.
.
+
.
0
1
.
.
.
.
Od
O
.
.
.
.
.
.
.
.
..
.
.
.
•
01.
.
.
.
..
.
..
.
..
.
00
.
.
.
1
_OO.
.
.
..
.
.
.
.
.
.
.
C
.
..
o
.
O
1.10
1
00
.
....
..
OOO
.
...
+ 11
.
.
1
116
100..
.
ZO
t.
..
.
.
.
.
.6
.
O
.
.
...
.
O IT
.
.000
.
+0.
12.0
.
. .
7. OO
.
. -
... AO
..10
. +00
1
•
.
.
.
..
.
.
.
O
.
.
.
.
.
.
On
+ 1
.
..
2000
......
T
.
.
. . . .
.!..
.
.
.
.
.
..
1
.
.
.
.
.
2011
.
..
.
+
ftnT
2
1
2017
+ . . . +
ET
11
12.
.
.
.
.
.
.
.
.
.
.
.
114
1
1
.
....
.
106..
.
.
i
.
.
1
.
0
.
.
..
..
.
.
.
i
.
11.
.....
.
.
.
.
.
.
.
.
.
.
.
.
1
.
.
.
1
.
06
. . .
1..000
.
....
.....
.
.
.
..
X
.
1
.
.
.
.
1
1
1
.
.
+
.
DOO
"
..
.
.
.
.
.
.
.
1
.
.
7
0
.
......
.
.
.
1
1
0
.
.
..
.
#
.
1
11
.
0.01
..
.
1
.
...
OP
#
1
111
0
#
.
.
.
.
.
#
.
.
.
.
.............
.
1
O
.
.
.
.
.
.
..
.
.
.
.
:
..
.
.
.
.
.
.
.....
.
..
.
.
.
..
....
....
..
.
..
..
.
...
.
.
.
.
.......
.
.
....
.
.
.
.....
.
.
14
.
..
DI
..
.
.
.
.
.
...
.
.
.
.
.
.
.
.
DDT
.
.
.
1.
.
.
.
01.
. . . .
t
.
&
N
......
.
1
-
.
10
.
.
.
..
1
.
..
.
.
.
.
1
1
0
1
.
10+
.
.
.
.
.
1
C
.
.
.
•
1
.
.2
.
.
.
.
.
.
.
.
.
PO
1
.
.
.
.
6
...
.
•OTO
.
.
.
....
.
06
.
4
.
.
.
.
.
001
.
.
10
1
.
SAN
Uncorrected
Corrected (-1.5 deg )
C-Station - Version 1v5A & B

.
.
.
111
+
:
.
..
.
..
.
.
.
..
0
1
.
.
.
1
1
.
•
.
.
_.
.
.
•
.
.
.
.
.
1
.
....
..
.
TOTO
.
.
1
.
1
.
.
0
1
1
1
1
1
1
1
1
.
1
. .
.
.
..
.
.
.
.
.
.
1
..
.
In Otto
1
1
1
.
.
..
.
1
• TO
Tet
.
.
0
1
0
1
1
1
OR
.
.
.
.
1
1
.
AN
:
.
.
+
.
.
0
.
1
.
.
.
2 . . .
do
TOT. +
+ 2q1
.
0
.
1
1
0
...
.
1
.
C
2
1
1
.
.
...
.. 11 2011
.
0
.
0
0
1
.
.
1
1121
+
.
.
000
1
.
.
.
1
.
1
0
.
1.
.
.
.
.
.
.
.
.
.
.
I .
11
CC
LO
.
.
2
1
.
1
1
.
1
1
0
0
0
0
0
0
0
1
.
1
1
.
0
.
1
-
.
.
.
.
. .. .
............
. . .
.
1
1
+
.
.
•••••
.
1
1
-
.
.
1
0
.
.
.
1
.
1
1
.
1
.
.
.
1
.
-
.
.
...
.
1
+
.
.
1
.
.
1
1
1
.
.
.
.
1
1
1
+
1
1
.
1
D
.
.
.
#TTO
.
1
.
+
.
.
1
.
16.01
.
.
1
-
.
.
2
1
.
.
1
+
1
+
$
.
.
1
.
0
1
+
+
+
.
•
.
.
+
1
-
.
.
.
F
._
..
.
..
...
..
+
...
1
..
+
..
+
.
.
2.
.
1
11 1
1
.
.
.
.
.
.
.
1
1
1
.
1
01
.
–
.
0
..
+
0
0
1
-
-
.
0.
►
.
1
.
C
.
.
.
2
.
.
.
.
.
1
"
C
10000.
. . . .
• • .
1 1 1
1 OD
1 1
10
1 1 1 1 1 1
1
1
1
+
.
2017
LA21__10_
LALOLU
1 1
1
LA VILL
20_LNC4L
LLLL
1
.700 O .
.
011
+ .
.
16
.
10 11 12
.
+
TO
DO
.
10_0
6
.
.
!
1
A
.
.
.
.
.
.
.
19
..
.
.
.
.
1
50
4
+
+
. .
t...!
.
.
!
.
.
.
.
.
.
.
.
5
.
.
20
#111
1
.
+
+
+
+
+
+
.
.
.
..
1
1
.
1
.
.
.
.
.
.
.
114
.
.
.
.
.
.
.
.
.
.
.
.
.
O
.
.
.
.
1
..
.
.
•
1
.
•
0
1
.
1
.
-
....
.
+
.
.
.
.
1
.
.
.
.
+
1
.
•
.
1
1
.
1
0.
00
.
.
1
4
1
0
0
.....
.
.
+
1
+
+
.
0
+
.
1
.
...
OVAT
.
1
1
.
.
.
.0000
.
+
.
.
1
.
.
+
.
.
.
I
.
.
+
1
0
-
11
-
.
.
1
6
.
:
.
0
+
U
.
.
•
+
1
09
+
.
1
.
.
-
.
.
.
.
+
.
1
.
.
1
R
1
. .
1.
IT TO
.
11.
.
.
.
1
.
.
.
1
.
.
.
1
0
0
0
.
0
0
0
0
.
0
0
.
.
.
.
.
.
.
.
1
.
.
1
.
.
.
.
.
.
.
.
.
.
.
.
00
.
.
.
.
.
.
.
.
.
.
1
1
1
.
.
.
.
.
.
+
.
+
+
+
.
.
.
......
.
.
.
....
.
.
.
I
11
.
?
.
.
.
.....
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . .
....
. . . .
.
.
.
•••
. .
.....
.
.
...
.
.
.
.
.
.
.
OTO.. . .
........
.
.
...
. .
11.
.
.
.
.
.
.
.
.
.
.
.
.00
.
....
A
.
A
.
.
D
9
LLLL
Uncorrected
Corrected ( 7.5 deg )
Fig. 4-3. Sample sun position (+) plots before and
after azimuth correction.
51
4.3 LAN Image Evaluation and Associated Ratio Input File Entries
The ratio input files are used to pass several different types of information to the ratio
program, including occultor parameters for the proper placement of the occultor mask, time
and date corrections, and image quality assessment indicators. The CHECKLAN program
was developed to provide a systematic means of evaluating when the occultor arms were
changed at a site, what the offset in occultor angle position was, and how far the clock was
off if the WWV clock was not operating normally. The LAN images being extracted from
the field tapes during the data quality assessment provided a handy set of once daily image
samples that could be used in determining the ratio input parameters. CHECKLAN loads a
specified LAN image, then draws the occultor mask. The operator can adjust the occultor
arm and occultor angle until the mask covers the occultor visible in the image. The final
arm chosen and the angle offset are then recorded to a file CHECK.OUT. If the sun can be
seen on the image, a track of the sun position at 10 minute intervals is superimposed on the
image, and the operator can change the time until a cursor is centered on the sun. The time
recorded on the image is then subtracted from this best fit time to yield the time correction
that can be applied by the ratio program on that date.
4.3.1 Running CHECKLAN
Download the LAN images to be evaluated onto an appropriate subdirectory containing
VERSION.LOG and the revised CALsVvvv.RAT for the images being evaluated. In the
example below, we'll assume that these files reside on the TAPRAT directory, where the
ratio runs will be made. A CHECKLAN sample run for the first 6 Columbia tapes will
now be examined.
To Execute: Enter CHECKLAN at the prompt, i.e.
D:\TAPRAT>CHECKLAN
Screen Output:
Welcome to CHECKLAN - Version 1.0
Checks the occultor and sun positions on LAN images
Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
1990 - All Rights Reserved
Site #
Location
1 Site #
Location
I
6
en A WNroi
MPL
C-Station
HELSTE
Kirtland
China Lake
Malmstrom
Malabar
Columbia
Composite 1
Portable 1
-> Please enter the site number: 7
-> Enter the field tape sequence number (eg, 001):001
Input the start date of the tape as follows:
Format: Month Day Year
Example: 1 10 89
-> Please enter the starting date: 2 9 89
Calibration File CAL7001. RAT opened successfully.
Enter azimuth offset for this site :-1.5
-> Enter the LAN indicator (such as A,B,C...) :A
At this point switch to the RGB screen and answer the prompts for each of the LAN images
being evaluated. See Tech. Note 230 for a more detailed description of the responses to the
prompts.
A summary sheet for the CHECKLAN also needs to be filled out, such as that shown
in Table 4-2. Record the tape ID number, the date displayed on the LAN image, a visual
estimate of the total and opaque cloud cover, and comments pertinent to the data quality
assignments.
The CHECK.OUT file for the sample run is presented in Table 4-3. The data columns
are self explanatory. If the sun was not visible at noon, such as on 13-15 Feb, a 999 is
entered into the time correction column. A "W" would appear after the time correction if
the WWV clock was accessed. The following subsections describe how the CHECKLAN
output is used to update the ratio input files.
4.3.2 DATEs.LST Specification
The first file to evaluate is the DATEs.LST file. The procedure to follow is outlined
below, using the first 6 Columbia tapes as the example.
Step 1:
Construct a preliminary version from the station logs.
The station logs contain information on the start date of a field tape, and the number of
days it contains. This information can be entered into the DATEs.LST file as follows.
FIELD TAPE DATE LIST
Site Number 7 - Columbia, Missouri
Last Update - 18 Oct 1991 by TLK
Tape
Identifier
COL001
COL002
COL003
COL004
COL005
COL006
Start
Date
Mo/Da/Yr Days Corr.
02 09 89 7 0
02 16 89
02 23 89
03 02 89
03 08 89
03 15 89
ooo
Run by TLK
16 OCT 91
Stallon:
Columbia, MO - 1989
DATELO, EST
TAPE
- EST. EST
O DATE 186 90
Сонцето
001 2/91 0
2/10 O O
2/11 O O
2/12 100 150
2113100 100 | rain drops
2114 100 100
12/15 100 100 | rain drops

DATE TOT.CVRLOPAQUEL COMMENTS
005 3/8 o o locc. shadow
3/9 o o
3/101 10110 I may be ovc thin
3/11100100
3/12 100 100
3/13 100 1 100
|3/14oo
II
| 3/15 olo
|3/16| 100 | 100
3/171 90 130
|3/18 1100 1100
|3/19| 100 | 100 Rain drops.
|3/201 10110
13/11 olo
00212/16100 80 sum dimely
201 100 100
2/181100 1 60 Cellular AC
2/18 | 100 | 100 | sun visible
2/201100 11001 Raindrops
2/21 | 100100
2/22 0 0 Oeculler,
adoop
0032/23 Oo
z/z400
2/25 80 160
2/26 8070
2/27 0
2/281 0 0
3/1 100 100
Few
Cu
0041 32 100
313100 100
3/4 1100 100
315100 100
3161100100
| 3171ol ditteunt
Джлаа
? 13/7lolo loce, shadow
Table 4-2
Sample CHECKLAN summary list
54
Table 4-3. CHECK.OUT Contents
O
O
1
Odont
O
Cor
O
1
O
CO
O
= 999
OPPO
oooo
H
H
H
H
1
H
H
H
O
1
COL001A. LAN
COL001B. LAN
COL001C. LAN
COL001D. LAN
COL001E. LAN
COL001F. LAN
COL001G.LAN
COL002A. LAN
COL002B. LAN
COL002C. LAN
COL002D. LAN
COL002E. LAN
COL002F. LAN
COLO02G. LAN
COL003A. LAN
COL003B. LAN
COL003C. LAN
COL003D.LAN
COL003E. LAN
COL003F. LAN
COL003G. LAN
COLO04A. LAN
COLO04B.LAN
COLO04C. LAN
COLO04D.LAN
COLO04E.LAN
COLO04F. LAN
COLO04G. LAN
COL005A. LAN
COL005B.LAN
COLO05C. LAN
COLO05D. LAN
COL005E.LAN
COLO05F. LAN
COL005G. LAN
COLO06A. LAN
COLO06B. LAN
COL006C. LAN
006D. LAN
COLO06E. LAN
COLO06F . LAN
COL006G.LAN
9/FEB/89 ND=0 Occultor: Arm=2 Cor=
10/FEB/89 ND=0 Occultor: Arm=2 Cor=
11/FEB/89 ND=0 Occultor: Arm=2
12/FEB/89 ND=0 Occultor: Arm=2 Cor=
13/FEB/89 ND=1 Occultor: Arm=2 Cor=
14/FEB/89 ND=0 Occultor: Arm=2
15/FEB/89 ND=0 Occultor: Arm=2
16/FEB/89 ND=1 Occultor: Arm=2 CO
17/FEB/89 ND=1 Occultor: Arm=2 Cor
18/FEB/89 ND=1 Occultor: Arm=2 Cor=
19/FEB/89 ND=1 Occultor: Arm=2 Cor=
20/FEB/89 ND=0 Occultor: Arm=2 Cor=
21/FEB/89 ND=1 Occultor: Arm=2 Cor=
22/FEB/89 ND=0 Occultor: Arm=2
23/FEB/89 ND=0 Occultor: Arm=3 Cor=
24/FEB/89 ND=0 Occultor: Arm=3
25/FEB/89 ND=1 Occultor: Arm=3 CO
26/FEB/89 ND=1 Occultor: Arm=3 Co
27/FEB/89 ND=0 Occultor: Arm=3 Cor=
28/FEB/89 ND=0 Occultor: Arm=3
1/MAR/89 ND=0 Occultor: Arm=3 Cor=
2/MAR/89 ND=0 Occultor: Arm=3
3/MAR/89 ND=0 Occultor: Arm=3
4/MAR/89 ND=0 Occultor: Arm=3
5/MAR/ 89 ND=0 Occultor: Arm=3
6/MAR/89 ND=1 Occultor: Arm=3 Cor=
7/MAR/89 ND=0 Occultor: Arm=3 Cor=
7/MAR/89 ND-O Occultor: Arm=3
8/MAR/89 ND=0 Occultor: Arm=3 Cor=
9/MAR/89 ND=0 Occultor: Arm=3 Cor=
10/MAR/89 ND=0 Occultor: Arm=3 C
11/MAR/89 ND=1 Occultor: Arm=4
12/MAR/89 ND=1 Occultor: Arm=4
13/MAR/89 ND=0 Occultor: Arm=4 Cor=
14/MAR/89 ND=0 Occultor: Arm=4 Cor=
15/MAR/89 ND=0 Occultor: Arm=4 Cor=
16/MAR/89 ND=0 Occultor: Arm=4
17/MAR/89 ND=0 Occultor: Arm=4 Cor=
18/MAR/89 ND=1 Occultor: Arm=4 C
19/MAR/89 ND=0 Occultor: Arm=4 CO
20/MAR/89 ND=0 Occultor: Arm=4
21/MAR/89 ND=0 Occultor: Arm=4 cor
Time: Cor
Time: Cor= 1
Time: Cor
Time: CO
Time: Cor 999
Time: Cor 999
Time: Cor
Time: Cor
Time: CO
Time: Cor 0
Time: Cor= 1
Time: Cor
Time: Cor
Time: Cor
Time: Cor= 1
Time: Cor=
Time: Cor= 999
Time: Cor= 999
Time: Cor= 1
Time: Cor= 1
Time: Cor= 999
Time: Cor= -1
Time: Cor= 999
Time: Cor= 999
Time: Cor= 999
Time: Cor= 999
Time: Cor= -8
Time: Cor= -9
Time: Cor= -10
Time: Cor= -11
Time: Cor= 1
Time: Cor= 0
Time: Cor= 999
Time: Cor= 999
Time: Cor= 1
Time: Cor= 1
Time: Cor= 999
Time: Cor= 1
Time: Cor= 999
Time: Cor= 999
Time: Cor= 1
Time: Cor= 1
O
O
Cor
O
O
O
O
1
O
Old Oo
O
с
O
Co
O O OPPO
1
co
55
The start date entered here is not yet corrected for any date skips that might have
occurred. The date correction is initially set to 0.
Step 2: Evaluate CHECK.OUT and the station log for date skips.
Do the dates in CHECK.OUT progress in the normal manner? Do duplicate dates
occur, or are missing dates evident. In Tables 4-2 and 4-3 for the Columbia example,
a duplicate 7/MAR/89 appears. The comment in the handwritten CHECKLAN
summary indicates that the second 7/MAR LAN image is different from the first, and
that it is a noon image. Sometimes if the clock skips backward over LAN, TAPEQC
will produce a second LAN image on the same day, but the sun will not appear near
the 180° azimuth in the second LAN. Since this is not the case in the second 7/MAR
LAN image, it appears that the control computer did not change the date to 8/MAR
after midnight on 7/MAR. The cloud cover estimates in the CHECKLAN summary
can be used to verify this possibility. A comparison between the LAN cloud cover
estimates and the 1750 GMT weather observations made at Columbia (from the Form
10 sheets) is presented below.
LAN
Date
LAN
Total
Cloud Cover
(tenths)
Form 10
Date
Form 10
Total
Cloud Cover
(tenths)
9 (1 opaque)
2/MAR
3/
MAR
4 MAR
5/
MAR
OMAR
7MAR
7/
MAR
8 MAR
9MAR
10/MAR
11 MAR
12 MAR
13 MAR
14 MAR
15 MAR
16/
MAR
17/MAR
18 MAR
19/
MAR
20 MAR
21 MAR
2/
MAR
3/
MAR
4/
MAR
5/
MAR
6MAR
7/
MAR
8/MAR
9/
MAR
10/MAR
11MAR
12 MAR
13/MAR
14/MAR
15/MAR
16/
MAR
17/MAR
18/
MAR
19/
MAR
20 MAR
21/
MAR
22 MAR
10
10
This sequence is well matched, supporting the idea of a computer date slip on 8/MAR.
56
Missing dates in the LAN sequence are a more typical problem. Usually these occur at
a tape change, and are due to a late start or a premature stop of the FIELD program.
This can be verified against the station log. Otherwise, the cloud cover check can be
used.
Step 3: Update the date corrections, if necessary.
As determined in the above example, The computer date slipped on 8/MAR/89 at
Columbia. For all tapes started after that date, a date correction of +1 should be
entered, until evidence suggests that the problem was rectified. The revised
DATE7.LST entries then become:
FIELD TAPE DATE LIST
Site Number 7 - Columbia, Missouri
Last Update - 18 Oct 1991 by TLK
Tape
Identifier
COL001
COL002
COL003
COL004
COL005
COL006
Start
# Date
Mo/Da/Yr Days Corr.
02 09 89
02 16 89
02 23 89
03 02 89
03 08 89
03 15 89
WWOOOO
4.3.3 OCCs.DAT Specification
The entries for OCCs.DAT are straightforward. Whenever the arm or the correction
angle changes, an entry into the file is made. Based on the CHECK.OUT occultor
information for our sample, the initial OCC7.DAT entries should be:
U
O
O
O
O
OCCULTOR ARM INFORMATION - Last updated 18 Oct 1991
Site #7 - Columbia, MO
Beginning Date (* normal schedule, not verified)
Mo/Da/Yr Arm Used Offset Angle
02 09 89
02 14 89
02 23 89
02 24 89
03 12 89
03 13 89
03 22 89
03 29 89
04 16 89
05 07 89
08 06 89
08 27 89
09 14 89
10 01 89
10 19 89
11 09 89
O
O
NW Auravan As WW NN
O
O
*
*
O
*
*
*
57
Note that I have filled in the normal arm change schedule for the remainder of the year, so
that only departures need to be entered in the future. At least one such additional entry is
helpful to keep the file read in the ratio program from aborting.
CAUTION - the dates entered in this table must be corrected dates, reflecting the date
corrections made in DATES.LST. Thus, 1 day has been added to the CHECK.OUT dates
after 7/MAR/89, when occultor changes occurred. That is, entries for changes on
11/MAR, 12/
MAR and 21/MAR have been dated 12/
MAR, 13/MAR and 22/
MAR,
respectively in OCC7.DAT.
4.3.4 TIMES.COR Specification
The system clocks on the control computer exhibit drift. The FIELD program was
designed to periodically update the system clock from WWV. At times, WWV was not
available, and the system clock was only updated from the more accurate BIOS clock when
the system was rebooted. The ratio routine uses the TIMES.COR file to place a better time
stamp on the processed images. As seen in Table 4-3, the only significant time
discrepancies (greater than 2 minutes) from the Columbia sample occur from 3/MAR to
10/MAR (using corrected dates). The corresponding TIME7.COR entries are:
ܘ
ܢܝ
ܩ
TIME CORRECTION INFORMATION - Last updated 18 Oct 1991
Site #7 - Columbia, MO
Beginning Date Time
Mo/Da/Yr Offset
02 09 89
03 03 89
03 04 89
03 05 89
03 06 89
03 07 89
03 08 89
03 09 89
03 10 89
03 11 89
O
ܗ
ܢ
ܣ
ܘ
ܘ
ܢ
ܘ
58
4.3.5 VERSION.LOG Quality Indicators
The data quality entries to VERSION.LOG can be made at this stage in the procedure.
The entries for Columbia in 1989 are given below.
File VERSION. LOG
Updated 24 July 90
List of hardware versions used in field, for access by ratio
programs.
STATION
COMMENTS
DATE
Mo/Da/Yr
HARDWARE DATA
Vers Qual
QUALITY SOFTWARE
Indicators Vers
O
O
02 08 89
03 24 89
05 08 89
05 16 89
05 31 89
06 22 89
11 09 89
89
12 07 89
12 14 89
12 25 89
12 26 89
12 28 89
O HONW
ܢܕ ܢܙ ܢܙ ܢܙ ܝ ܝܙ ܢܙ ܝ ܢܕ ܢܝܙ ܢܙ ܢܙ ܢܙ
Cam UU UUUUUU
0000000000
0000000000
0000000000
0000000000
0000000000
0000000000
0000001000
0000000000
0000000000
0000001000
0000000000
0000001000
0000000000
HM HNW WWWW
1.71
1.71
1.71
1.71
1.71
2.70
2.70
2.70
2.70
2.70
2.70
2.70
2.70
17
♡
O
No additional entries for the 6 tape sample period were made. Some comments about
subsequent entries are pertinent. As noted in the hardware log (Section 3.1.3), a new
occultor and a new clock antenna were installed on on 24/MAR (change to version 1B) and
on 22/JUN the computer was swapped, with a new version of FIELD loaded. The general
data quality indicators (DATA Qual.) can have one of three values.
1 = Cloud data appear normal, but check the quality indicators
2 = Cloud data may be abnormal
3 = Data missing.
Interpreting, these entries in the VERSION.LOG sample, the WSI was either down or a
data tape was lost or unreadable for 3 periods in 1989: 8/MAY to 15/MAY, 31/MAY to
21/JUN, and 7/DEC to 13/DEC. The images from 14/DEC to 24/DEC are of questionable
quality. Examining the 10 character quality indicator string tells us why. The quality
indicator string describes problems with the imagery based on the automated and visual
evaluation of the LAN images. Zeroes indicate that no problems are expected. Non-zero
flags are explained below.
Position 1 - No occultor present if = 1.
Position 2 - Stray light.
1 = Wrong occultor arm may lead to stray light problems.
2 = Large time offset displaces occultor off sun producing significant stray
light.
3 = Other.
Position 3 - Split images, data unusable if = 1.
Position 4 - Bad input look up table on FG-100 board in field if = 1, resulting in
gaps in field brightness values. During ratio processing, affected pixels
are indicated either as "no data" or eliminated by smoothing.
Position 5 - Range truncation at top of range, if = 1. Ratio data may be slightly
distorted.
Position 6 - Range truncation at bottom of range, if = 1. Ratio data may be slightly
distorted.
Position 7 - Obstruction to vision on dome.
1 =
Condensation inside dome. Ratio data may be useful.
Other obstruction on dome or portions of dome.
2 =
3 =
Dirty dome.
Positions 8 thru 10 - Currently unused.
Thus, the poor data quality at Columbia in mid-December 1989 was due to heavy
condensation in the dome (a 1 in position 7).
The date entries in this file must be corrected dates when the information comes from
field tapes that needed date adjustment. The data quality specification requires not only the
CHECKLAN information, but may also include information from the TĀPEQC and DGN
output evaluations.
4.4 Obstacle Mask Definition
At many of the sites, structures and other obstacles obscure the field of view, making a
cloud decision in that portion of the sky impossible. Since these ol
incorrectly identified as cloud, we decided to mask them as no data regions.
4
The obstacle mask is defined using raw radiance images that clearly depict the
obstacle. One of the mouse routines is then used to define a polygon (with up to 19
vertices) outlining the image. Up to 10 obstacles can be defined. The top panel of Fig. 4-4
shows a dark red image from Columbia. Two significant obstacles are apparent: the radar
tower to the left (west) and the tree near the top of the image (just west of south). The
boundaries of the obstacle masks are shown in the bottom panel. The X and Y locations of
the points defining the polygons are listed below.
Radar Tower
Tree
Pixel
Location
Pixel
Location
X
Point
X
Point
150
167
91
127
128
137
144
146
154
155
135
126
179
180
196
213
Noooo van AWN
230
232
212
202
188
202
230
218
118
220
221
225
226
233
230
178
112
72
76
91
150
Repeating the first point as the last point closes the polygon. These image coordinates must
be converted to the nominal ratio image coordinates. If an image with a magnification near
1 is chosen, the conversion to nominal coordinates consists of adding the X and Y offsets
to the raw image coordinates. In the example, the magnification is 1.003, the X offset is -6
and the Y offset is 0. The obstacle list appended to the end of CAL7V01.RAT then
becomes:
.
OBSTACLE INFORMATION
Number of Obstacles =
2
Obstacle
Point
1 - 15 Points on the Perimeter
Column Row
85
150
121
122
131 180
167
179
138
139
148
196
213
218
220
221
149
ܩ ܩ ܝ
لا لا لا لا هم پسر هم
ܢܙ ܢܙ ܘ
225
129
120
112
106
NO
66
226
233
230
178
150
14
70
85
15
Obstacle
Point
von A WNH
2 - 7 points on the Perimeter
Column Row
224
13
226
206
196
182
196
224
wwwur unw

.
.
.
.
.
..
2
.
.
..
.
.
.
S
.
.
.
. .
...
.
.
.
.
.
A
.
.
..
.
20
.
.
.
20..
..
.
10000.00
0
.
....
. .
.
.
.
on.
000
.
.
.
.....
O
.
O
0
.
.
.
AU
20.
ON-
VO
.
A
.
.
.
.
.
.
.
.
*
22

.
.
O
I
.
.
.
!
1
.
.
.
.
.
.
..
1
.!
.
1
.
.
.
.
.
.
.
.
.
.
.
.
..
+
.
..
.....
.
...
.
0
...
...
.
..
...
...
...
.
...
.
...
..
0
.
.
.
1
S1
.
.
.
dan
.
.
.
Fig. 4-4. Dark red (Quad 4) image from Columbia,
with and without obstacle mask outline
superimposed. Note the radar tower and
tree in the original image.
63
4.5
Final CALsVvvv.RAT Adjustments
The final step in the ratio file preparation phase is to run an abridged version of the
ratio program (TAPRATPN) and evaluate the many parameters that have been defined,
particularly the NDR and SPR-related parameters from the absolute calibrations. Images
from both of the first 2 neutral density filter wheel positions should be evaluated. The
periods selected for the edge verification could also be applied here. TAPRATPN differs
from the operational version of the ratio routine (TAPRATPL) in that the resulting ratio
images are not archived to ExaByte. TAPRATPN and TAPRATPL have to be run on a
computer equipped with a PL-1250 accelerator board. The steps for the parameter
adjustment procedure are as follows. The example shown is for 12 Feb 89, one of the
edge checking cases.
Step 1: Load the raw radiance tape onto the read ExaByte.
Step 2: Move the tape to the desired starting point using IMFIND.
Step 3: Run TAPRATPN.
Before this program is started, make sure the necessary files reside on the TAPRAT
directory. Change directory to TAPRAT.
To Execute: enter TAPRATPN at the prompt, i.e.
D:TAPRAT>TAPRATPN
Screen Output:
Welcome to TAPRATPN - TAPRAT Version 2.5
Computes the red/blue ratios - does NOT save results to ExaByte
· Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
1991 - All Rights Reserved
-> Enter the tape number (such as 001) : 0010
...WSI Image Selection
Enter the letter corresponding to your
choice for WSI image processing.
T = Ten Minute Images Only
O = One Minute Images Only
-> Your Selection: T
Note: You have several options which may be changed during a run.
Options may be changed by entering the appropriate keystroke,
without pressing the ENTER key. All of these options take
extra time, and are intended only for occasional use during
a run to visually check the ratio images.
The Key enter options are:
S to change Show Color flag
C to Change Color
D to change Save Radiance flag
T to change Save Ratio flag
F to change Show Full flag (this shows full
10 min image when process 1 only has been chosen)
P to change Print Log flag
X exit after current image
? to print current values of flags
Any other key will cause the available options to be printed
-> Press ENTER to continue
Do you want to change the color from default?
Note: Default thresholds are thin=120, thick=130.
Type O if Ok, or enter new thresholds
-> ENTER O or THIN CLD THRESHOLD: 0
*
Site # Location | Site # Location
---------------------------- ---------
MPL
6 Malabar
C-Station
Columbia
HELSTF
Composite 1
Kirtland
Portable 1
China Lake
Malmstromi 11
und WNHOI
-> Please enter the site number: 7
Opened diagnostic file 17001D.DGN
Input the start date of the tape as follows:
Format: Month Day Year
Example: 1 10 89
-> Please enter the starting date: 2 9 89
Field Sequence Number = 001 Confirmed ..
Expected number of days on tape is 7.
* Results of the Calibration File Selection *
65
Data
Qual.
rs.
1
Seq.
Day Day-Mo-or
406 9/FEB/89
407 9/FEB/89
408 9/FEB/89
409 9/FEB/89
410 9/FEB/89
411 9/FEB/89
412
9/FEB/89
Quality Hard.
Indicators Ve
0000000000
0000000000
0000000000
0000000000
0000000000
0000000000
0000000000
Soft. Calibration
Vers. File Selected
1.71 CAL7V01. RAT
1.71
CAL7V01. RAT
1.71 CAL7V01.RAT
1.71 CAL7001. RAT
1.71 CAL7001.RAT
1.71 CAL7V01.RAT
1.71 CAL7001. RAT
1
-> Enter the starting day number: 4
-> Enter the starting image sequence number: 1
Input the expected start date as follows:
Format: Month Day Year
Example: 1 10 89
-> Enter the starting date: 2 12 89
Initial header found: Date = 12/FEB/89
Initial date confirmed.
Field tape format confirmed: FLDTYP = 2
Initial Image type = 1
Log file 17001D.LOG
opened successfully
New calibration file CAL7V01.RAT opened successfully.
Calibration file date is :10/OCT/91
Basic Time Information for 12/FEB/89:
Estimated Start Time = 1223 GMT
Computed Sunrise = 1300 GMT
Computed Sunset = 2347 GMT
Time Correction = 0 min
Occultor Arm 2 selected: Offset = 1
Adding the "D" to the tape number informed the program that it was not starting at the
beginning of the tape, making the inquiries after the calibration file selection necessary.
In this application of TAPRATPN, the starting image number is not important, so a
value of 1 was entered.
At this point in the program, the first image is being read. Hit the hotkey T to stop the
ratio sequence after smoothing is complete. Switch to the RGB screen and watch the
ratio computation. Of particular interest is the boundary between the 1-2 and 3-4
spectral pairs. If this boundary is clearly evident, the NDR input value needs to be
changed. If the boundary is not evident, change NDR by a couple of tenths and run
again to see if the boundary was present.
Step 4: Adjust the NDR value, if necessary.
The direction to adjust NDR depends on whether the ratios using the 3-4 pair (usually
the bright regions on the raw imagery) are brighter or darker than those from the 1-2
pair. Increasing NDR will darken the 3-4 region, and decreasing NDR will brighten
Step 5: Erase the TOO1D.DGN and .LOG files
Step 6: If making another try, reposition the input tape and start again at Step 3.
The NDR values for both of the lower ND filter settings should be checked, using several
different cases, both clear and with clouds. A single NDR value will not necessarily fit all
cases, due to the different color temperatures of varying sky backgrounds, so a
compromise value is usually chosen. After the NDR values are set, it may be useful to
save some sample ratios for both ND positions, and examine the cloud and clear sky ratio
levels. It may be desirable to adjust the SPR parameter to bring minimum opaque cloud
level to near 130 ratio pixel counts.
The results from this procedure for the Columbia case are summarized in the before
and after correction factor tables from page 6 of the CAL7V01.RAT file shown below.
BEFORE:
Absolute Computed correction Factors
2/1 - 4/3 4/3
ND Filter Transition Offscale NDR
250
254 0.99
250
254 0.99
254 0.92
250
0.92
(4/SPR]
SPR
1.59
1.65
1.20
1.20
LAWN
250
254
AFTER:
Absolute Computed Correction Factors
2/1 - 4/
3 4 /3
ND Filter Transition Offscale NDR
250
254 0.87
254 0.87
250
254 0.74
250
254 0.74
14/SPR)
SPR
1.41
1.46
1.04
1.04
250
ما M با هر
In this case, both the NDR and [4/SPR) factors were adjusted downward.
The analyst should also examine several other features of the ratio images during the
TAPRATPN runs. These include:
· Are the red and blue image well registered? This is most evident at cloud edges.
- Is the occultor mask working properly?
67
- Are the obstacle masks in the right place?
- Are anomalous ratio points being defined in what otherwise would be an offscale
bright area? This indicates that the bright end rolloff of the camera has not
properly been addressed, and the transition and offscale thresholds should be
lowered.
5.0
Red/Blue Ratio Image Construction
The data processing discussed heretofore has involved preparing input parameters for
the ratio processing program. This section describes the ratio processing procedures.
Conceptually, the red/blue radiance ratio computations could have been just one step in a
multistep cloud decision procedure. However, the development of the cloud decision
algorithms lagged the ratio computation development. Thus, the ratio computation became
a separate processing procedure, and an intermediate ratio data base was created. This also
provided more flexibility in developing the cloud decision algorithm, in that the time
consuming ratio computations did not have to be repeated when improvements to the cloud
decision technique were implemented.
We found early in the development of the ratio program (TAPRAT) that the procedure
required large amounts of computer time. TAPRAT was then modified to run using the
PL-1250 Pipeline processor board to accelerate the computations, allowing either the full
resolution (10 min) or subset resolution (1 min) images from a single field tape to be
processed in about 12 hours on our AT class computers. The accelerated program was
named TAPRATPL. The conceptual flow diagram for TAPRATPL runs is presented in
Fig. 5-1.
69

Raw
Radiance
Field
Tape
Revised
VERSION.LOG
Final
CALsVvvv.RAT
1-min
Ratio
Image
Tape
DATES.LST
Ratio Computation
and Composite
Image Construction
10-min or 1-min
Resolution
(TAPRATPL)
Os###c.DGN
OCCs.DAT
Os###c.LOG
TIMES.COR
1-minute runs
- - - - -
-
-|
-
-
-
7
-
-
-
10-min
Ratio
Image
Tape
15
Ts###c.DGN
--
Ts###C.LOG
-
-
-
10-minute runs
- - - - -
-
-
-
-
-
-
-
-
-
.
Fig. 5-1. Ratio processing conceptual flow diagram.
5.1 Description of TAPRATPL
The TAPRATPL program does more than just computing the red/blue ratio. The
following discussion outlines the many tasks performed in TAPRATPL. Steps 5, 6 and 7
employ the PL processor.
Step 1: Find the next image of the type being processed on the raw radiance field
tape.
The image is transferred to the FG-100 board with an offset in the column locations.
Step 2: Perform limited quality control checks on the radiance headers.
This includes checking for improper filter and aperture settings, time and date skips,
and mismatches between the time and occultor angle settings.
Step 3: Apply time and date corrections, if necessary, and check against the times of
sunrise and sunset.
The WSI was designed to collect data for a 12 hour period beginning 6 hours before
LAN. During the winter months, many images were collected either before sunrise or
after sunset, when low light levels made meaningful ratio computations impossible. To
avoid expending resources processing these dark images and retaining them in the data
base during subsequent processing procedures, images collected before sunrise and
after sunset (based on the corrected time and date) were skipped in the ratio processing.
Sunrise and sunset were defined to occur when the sun is 1.5° below the horizon.
Step 4: Pass all 4 images through the linearity LUT to correct for nonlinear sensor
response.
This procedure is performed directly on the FG-100 board using the digitize command
after loading the appropriate LUT from the CALsVvvv.RAT file. The corrected
images are shifted to back to their proper column locations during the LUT pass
through.
Step 5: Construct the composite ratio images.
This includes many more steps than just performing the ratio computation, including:
transferring the corrected signals from the FG-100 to the PL board, determining
collocated pixels in all 4 quadrants from the magnification and center position offset
inputs, checking whether brightness values are off scale bright or dark, selecting the
appropriate red/blue filter pair, scaling the resulting ratios, and sending the results back
to the FG-100 board.
71
.
Step 6: Smooth the ratio images to remove effects of random noise.
Two passes over the ratio data are made with the 3x3 smoothing kernel shown below,
Off scale dark or bright points are not included. If half or more of the total weight of
the kernel comes from off scale points, smoothing is not performed at that location.
Smoothing Weights
.0625 1250 .0625
.1250
.1250
.0625
2500
1250
.0625
Step 7: Derive the ratio histogram for the active image region.
Step 8: Construct the dummy DOS header and ratio header, and imbed the ratio
header into the ratio image.
Step 9: Transfer the ratio image with its corresponding dark red image (Quad 4) to the
output ExaByte drive.
TAPRATPL has several hotkey options that modify the image and data display
sequences, and allow sample radiance or ratio images to be saved to disk. These options
are not normally employed during routine processing.
Tech. Memo AV90-042t describes the structure of the ratio headers, and AV90-123t
describes the formats of the 1-min and 10-min ratio tapes.
5.2 Running TAPRATPL
After all the ratio input files described in Section 5 have been loaded into the TAPRAT
directory, a raw radiance tape has been loaded into the input ExaByte (unit 0), and a blank
tape has been loaded into the output ExaByte (unit 1), TAPRATPL can be executed.
To Execute: enter TAPRATPL at the prompt, i.e.
D:\TAPRAT>TAPRATPL
Screen Output:
Welcome to TAPRATPL - TAPRAT Version 2.5
This program computes the red/blue ratios and saves the resulting
images to ExaByte
Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
1991 - All Rights Reserved
-> Enter the tape number (such as 001) : 035
...WSI Image Selection
Enter the letter corresponding to your
choice for WSI image processing.
T = Ten Minute Images Only
O = One Minute Images Only
-> Your Selection: O
Note: You have several options which may be changed during a run.
Options may be changed by entering the appropriate keystroke,
without pressing the ENTER key. All of these options take
extra time, and are intended only for occasional use during
a run to visually check the ratio images.
The Key enter options are:
S to change Show Color flag
C to Change Color
D to change Save Radiance flag
T to change Save Ratio flag
F to change Show Full flag (this shows full
10 min image when process 1 only has been chosen)
P to change Print Log flag
X exit after current image
? to print current values of flags
Any other key will cause the available options to be printed
-> Press ENTER to continue
73
Do you want to change the color from default?
Note: Default thresholds are thin=120, thick=130.
Type O if Ok, or enter new thresholds
-> ENTER O or THIN CLD THRESHOLD: 0
LO
Site # Location | Site # Location
----- -------------------------------------
MPL
Malabar
C-Station
Columbia
HELSTE 1 8 Composite 1
Kirtland
Portable 1
China Lake
10
Malmstrom
11
-> Please enter the site number: 7
Opened diagnostic file 07035.DGN
Input the start date of the tape as follows:
Format: Month Day Year
Example: 1 10 89
-> Please enter the starting date: 10 19 89
Field Sequence Number = 035 Confirmed
Expected number of days on tape is
7.
* Results of the Calibration File Selection *
Data
Qual.
1с
Day Seq.
* Day Day-Mo-or
658 19/OCT/ 89
659 20/OCT/89
660 21/OCT/89
661 22/OCT/89
662 23/OCT/89
663 24/OCT/89
664 25/OCT/89
Quality Hard. Soft. Calibration
Indicators Vers. Vers. File Selected
0000000000
2.70 CAL7V01C.RAT
0000000000
2.70 CAL7001C.RAT
0000000000
CAL7V01C.RAT
0000000000
2.70 CAL7001C.RAT
0000000000 1C 2.70 CAL7001C.RAT
0000000000 1C 2.70 CAL7001C.RAT
0000000000 1C 2.70 CAL7001C.RAT
Initial header found: Date = 19/OCT/89.
Initial date confirmed.
Field tape format confirmed: FLDTYP = 3
Initial Image type = 1
Log file 07035.LOG
opened successfully
Writing initial EOF to target tape ***
New calibration file CAL7001C.RAT opened successfully.
Calibration file date is :11 / APR/90
Basic Time Information for 19/OCT/89 : :
Estimated Start Time = 1153 GMT
Computed Sunrise = 1219 GMT
Computed Sunset = 2329 GMT
Time Correction = 0 min
Occultor Arm 2 selected: Offset = 2
The example shown illustrates a 1-min TAPRATPL run. The only changes for a 10-
min run is to respond "T" to the WSI image selection request. The output files would then
be named T7035.DGN and T7035.LOG.
75
5.3 TAPRATPL Output Files
Two output files are generated for each TAPRATPL run. The first contains some of
the diagnostic information for the run and is named Os###c.DGN for 1-min runs or
Ts###c.DGN for 10-min runs. (The character c does not appear during normal
processing. See for example the TAPRATPN run in section 4.5.) The initial output is
echoed from the TAPRATPL computer screen output. The error summary tables only
appear here in the .DGN file. The sample shown below is from tape 035 at Columbia.
07035.DGN output
Starting Mo/Da/Yr on tape is 10 19 89
Field Sequence Number = 035 Confirmed
* Results of the Calibration File Selection *
Data
659
Day Seq.
| Day Day-Mo-or
658 19/OCT/89
20/OCT/89
660
21/OCT/89
661 22/OCT/89
5 662 23/OCT/89
663 24/OCT/89
7 664 25/OCT/89
Quality Hard. Soft. Calibration
Indicators Vers. Vers. File Selected
0000000000
2.70 CAL7V01C.RAT
0000000000
2.70 CAL7001C.RAT
0000000000
2.70
CAL7001C.RAT
0000000000 10 2.70 CAL7001C.RAT
0000000000
2.70 CAL7001C.RAT
0000000000 10 2.70 CAL7001C.RAT
0000000000
2.70 CAL7001C. RAT
10
10
Initial header found: Date = 19/OCT/89
Initial date confirmed.
FIELD tape format confirmed: FLDTYP = 3
Initial image type = 1
Log file 07035.LOG
opened successfully
New calibration file CAL7001C.RAT opened successfully.
Calibration file date is :11/APR/90
for 153 GM
Basic Time Information for 19/OCT/89 :
Estimated Start Time = 1153 GMT
Computed Sunrise = 1219 GMT
Computed Sunset = 2329 GMT
Time Correction = 0 min
Occultor Arm 2 selected: Offset = 2
Summary for 19/OCT/89
Number of 1 min images processed to ratio = 460
FREQUENCY
PROBLEM
Time skips ahead :
Time skips back :
Occultor out of range :
Occultor disagrees with time :
Neutral density inconsistencies :
76
0
Spectral filter inconsistencies :
Quadrant inconsistencies :
Basic Time Information for 20/OCT/89 :
Basic Time Information for 25/OCT/89 :
Estimated Start Time = 1153 GMT
Computed Sunrise = 1225 GMT
Computed Sunset = 2321 GMT
Time Correction = 0 min
Occultor Arm 2 selected: Offset = 2
Summary for 25/OCT/89
Number of 1 min images processed to ratio = 657
FREQUENCY
PROBLEM
Time skips ahead :
Time skips back :
Occultor out of range :
Occultor disagrees with time :
Neutral density inconsistencies :
Spectral filter inconsistencies :
Quadrant inconsistencies :
.
Ooo
The only substantial change in the output for the 10-min run (17035.DGN) is in the
line in the summary table indicating the number of images processed. The summary table
for the 10-min run on the first day of tape 035 begins:
Summary for 19/OCT/89
Number of 10 min images processed to ratio =
46
The second file generated by TAPRATPL is the log file containing numerical
information from each of the images processed, named Os###c.LOG for 1-min image, and
Ts###c.LOG for the 10-min images. The first 11 lines from the tape 035 1-min output are
reproduced below.
07035.LOG sample
panen pada
OOOOO
ooooo
7 8 91019 1 1550 891019 1550 100000
7 8 91019 2 1551 891019 1551 1 00000
7891019 3 1552 891019 1552 1 00000
7891019 4 1553 891019 1553 1 00000
7891019 5 1554 891019 1554 1 00000
7891019 6 1555 891019 1555 1 00000
7891019 7 1556 891019 1556 1 00000
7891019 8 1557 891019 1557 1 00000
7891019 9 1558 891019 1558 1 00000
7891019 10 1559 891019 IS 59 1 00000
7891019 11 1600 891019 1600 1 00000
0
0 0
0 0
Ojo
0 1 0
O 10
0 10
010
0 10
O 10
O 10
0 1 0
0
OOOOOOOOOOO
OOOOOOOOOOO
OOOOOOOOOOO
ooOOOOOOOOO
OOOOOOOOOOO
OOOOOOOOOOO
OOOOOOOOOOO
OOOOOOOOOOO
ooOOOOOOOOO
OOOOOOOOOOO
OOOOOOOOOOO
OOOOO
oooooo
NNNNNNNN
OOOOOOOOOOO
2 8 28 30 13 5
8 27 31 13 5
8 26 31 14 5
2 9 27 30 14 S
2 9 27 29 14 S
2 10 28 27 13 5
NNNNNNNNNNN
0 0
0 0
0 0
0 0
Oo
0 0
0 0
0 0
0 0
0 0
0 0
ooOOOOOOOOO
OOOOOOOOOOO
10 29 29 13
1 8 28 30 14 5
i 7 27 32 14
1 6 28 32 14 5
1 7 28 33 14 S
rostor
OOOOO
ooooo
to po p
Table 5-1 describes the entries into this table for the first line. The first 5 entries for the 10-
min run are given below.
T7035.LOG output
89109 1
00000_ܙ 1559
091019
1550
ܘܕ 0_0_6
ܘܙ ܕ3
33
3 _ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘܘ_ܘ
ܕ
ܕ 9
oooo0
1
1600
891009 ܘ160 ܐ891019
7
0
ܘ
ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ
ܘ
113
44
30 ܐ
891013
7
3610891009
3
161o 1 ooooos
0 ܘ1 ܘ 4
0 0 0 0 0 0 0 0 0 0 0 0 0
ܕ 10
38
35
3
8910194
7
0ܙ 6
412
00000 ܀ 1620
891009
1620
0 0 0 0 0 0 0 0 0 0 0 0 0
3ܐ 34
32
3
5
891049
7
ܐ 3
oo06ܘ ܐܘ163
891009
91630
ܘ
ܘ1
ܘ
ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ_ܘ
ܘ
ܙ1
37
29
6 ܙ
3
ܘܘܘܘܝܙ
ܕ
ܘ
ܘ_ܘ_ܘ_ܘ
ܘ1
ܘ
ܘ
1
ܘ
ܘ
ܘ
ܘ
ܘ
2 1 0 0 0 0
0 ܐ
0
0_0_0
ܘܘܘܝܙܘ
ܘܘܘܘܘ
ܘܘܘܘܘ
(Note: These examples are from the actual processing. The histogram results in the 1 and
10 min runs differ slightly for the same times, because small adjustments were made to the
scaling factors between the 1 and 10 min runs.)
Table 5-1. TAPRATPL Log File Entry Description
I New Seq New old old N Flags Off Sca Ratio Histgram (Max-105)
D YrModa # Time YrMoDa Time D TONSQ LU HU HL 0 1 2 3 4 5 6 7 8 9
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
7 891019 1 1550 891019 1550 0 00000 1 0 0 10 0 0 0 0 0 0 0 0 0
Ratio Histgram (Max-105)
10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 >24
++++++++++++++++++++++++++++++++++++++++++++++++
0 0 0 0 2 8 28 30 13 5 2 1 0 0 0 0
ID
- station number
CU
New YrMoDa - corrected date (Year Month Day)
Seq # - ratio image sequence number for this date
New Time - corrected time (GMT)
Old YrMoDa - original date
Old Time - original time
ND - neutral density indicator. Values 0 to 3 correspond to ND
filter positions 1 to 4
Flags TONSQ - ratio processing problem flags - 0 values indicate normal
operation
T - time check indicator
Values: 1 - time advanced more than 1 minute
2 - time unchanged or backtracked
8 - date change expected but not found
9 - date changed by more than 1 day
O - occultor check indicator
Values: 1 - position out of range (0 to 180)
2 - recorded position differs by more than
3° from that expected at Old Time
N - ND check: 1 - ND not identical for all 4 quadrants
S - spectral filter check: 1 - not in proper order
Q - quadrant check: 1 - not in proper order
Off Sca LU HU HL - % of active image off scale
LU - % offscale dark in 1-2 pair
HU - % offscale bright in 1-2 pair
HL - % offscale bright in 3-4 pair (before masking)
Ratio Histogram
(Max 10s)
- % of active image in 10 count ranges labeled by
10's digit of top of range. For example, the value
under 8 is the % of image with a ratio between 71 and
80, inclusive. O indicates % no ratio due offscale
dark, or in occultor or obstacle mask. >24 indicates %
off scale bright in the 3-4 pair (after masking).
6.0
Fixed Threshold Cloud Decisions
The first-generation cloud decision algorithm employed a relatively simple fixed
threshold technique. Two thresholds were defined, one at the clear/thin cloud transition,
and the other at the thin cloud/opaque cloud transition.
In January 1990, our group was asked to provide a preliminary set of 1-min cloud
decision data from 4 sites over a 14 month period. The data were sent to Dr. Charles
Medler of The Analytic Sciences Corporation (TASC) for evaluation. The only cloud
decision algorithm available at the time was the fixed threshold technique.
The CLDDECM program was written to process the TASC 1-min data set. The
CLDDECM procedure, outlined in Fig. 6-1, consists of two steps. First, the thin and
opaque thresholds must be defined from the ratio imagery. Then the cloud decision is
made.

Raw
Radiance
Field
Tapes
VERSION.LOG
CALsVvvv.RAT
DATES.LST
Ratio Computation
- No ExaByte
Archive
(TAPRATPN)
OCCs.DAT
TIMES.COR
Selected Ratio
Images
Thin and Opaque
Cloud Threshold
Selection
(DECVIEW)
Thin and Opaque
Threshold Summary
1-min
Ratio
Fixed Threshold
Cloud Decision
( CLDDECM )
1-min
Cloud
Decision
Tape
Tape
Cloud Cover
Summary File
(LLL###.CVR)
*.CVR
Archive
Tapes
Fig. 6-1. Fixed threshold determination and cloud
decision processing steps.
81
6.1 Ratio Threshold Determination
The TASC data set was processed in two month intervals. Anywhere from 3 to 6
cases per site were selected for the threshold determination. Most of the images were from
partly cloudy conditions, some with distinct clear/opaque boundaries, and others with both
ihin and opaque clouds present. Ratio images were produced for these cases using
TAPRATPN, as presented earlier in section 4.5. Full resolution, 10-min ratio images are
required for threshold evaluation. To start, load the raw radiance tape for one of the cases
into the read ExaByte, and move the tape to the proper position using IMFIND.
TAPRATPN is then executed using the "T" image selection option, and the "T" hotkey,
which halts the ratio processing after the composite ratio image has been smoothed, and the
header imbedded. The following prompt is then written to the computer screen.
If you wish to save the ratio, enter the file name (up to 12 characters).
Otherwise press ENTER
SAMPLE1.RAT
In the example above, the file SAMPLE1.RAT would be created, in which the current
512x480 ratio image would be copied. This procedure would then be repeated for the
remainder of the samples. If a hardware version change occurred during the period of
interest at a particular site, two ratio samples would be extracted from before and after the
change.
For the sake of argument, assume that 3 sample ratio images were extracted for one of
the 4 sites, named SAMPLE1.RAT, SAMPLE2.RAT, and SAMPLE3.RAT. The next step
in the procedure is to run a program called DECVIEW, that color enhances a ratio image
using thin and opaque thresholds defined through user input. A run stream from the
example is shown below.
To Execute: enter DECVIEW at the prompt, i.e.
D:\TAPRAT>DECVIEW
Screen Output:
Enter image filename: SAMPLE1. RAT
Define the thresholds:
-> ENTER O or THIN CLD THRESHOLD: 120
-> ENTER THK CLD THRESHOLD: 130
Shift to the RGB screen to view the color-enhanced ratio image. Ratio byte values of 0 and
1 are colored black, values from 2 to the thin cloud threshold are shaded blue indicating
clear, values from 1 plus the thin cloud threshold to the thick cloud threshold are shaded
yellow indicating thin cloud, and values above the thick cloud threshold are shaded gray
indicating opaque cloud. If some of the thin cloud region is being defined as clear sky,
lower the thin threshold. If clear regions are defined as thin, raise the thin threshold.
Likewise, if opaque clouds are defined as thin, lower the thick threshold, or if thin clouds
are defined as opaque, raise the thick threshold. Assume in our example that the clear
region overlapped the thin, while the thin overlapped the opaque for the 120-130 thresholds
selected. Thus, the thin should be lowered and the thick raised. This is done as follows:
82
Enter image filename:
Define the thresholds :
-> ENTER O or THIN CLD THRESHOLD: 115
-> ENTER THK CLD THRESHOLD: 140
Pressing the enter key at the "Enter image filename: " prompt causes no new file to be
loaded on the FG-100 board. The new threshold estimates are then entered, and the image
can be viewed again on RGB. Assume for the time being that these thresholds are
satisfactory for this image. Now load the next sample image.
Enter image filename: SAMPLE2.RAT
View the image in RGB. The previously defined thresholds are still active. Assume that a
minor modification of the opaque threshold is necessary. Return to the computer screen to
modify the thresholds again.
Define the thresholds:
-> ENTER O or THIN CLD THRESHOLD: 115
--> ENTER THK CLD THRESHOLD: 135
This sequence is repeated for all the sample images until a consensus set of thresholds
is selected. Usually, the thin threshold is the more difficult to define. If too low a value is
selected, portions of the clear sky near the horizon and the solar aureole will be incorrectly
defined as thin cloud, particularly when the sun is near the horizon. Too high a value leads
to underestimating thin cloud in the downsun portions of the sky. Our compromise values
tended toward the latter. Handwritten notes taken during the DECVIEW processing
provide the threshold values for input into the cloud decision routine.
83
6.2 Program CLDDECM
The fixed threshold program, CLDDECM is much faster than the ratio program, and
conceptually easier to understand. A LUT approach is employed, similar to that used in
TAPRATPL to apply the linearity correction. After identifying the input ratio tape, and
defining the thresholds, CLDDECM performs the following steps.
Step 1: Find the next 1-min ratio image on the ratio tape.
The image is transferred to the FG-100 board with a column position offset.
Step 2: Pass the ratio image through the cloud decision LUT.
The LUT is defined from the threshold input in a manner similar to that used in the
DECVIEW color enhancement process. The byte values assigned to the different sky
cover categories are outlined below.
CLDDECM Output Values
Cloud Decision
Category
Ratio Value
Value
No Data
Ratio = 0 or 1
0.
Clear
2 5 Ratio 5 Thin
100
Thin Cloud
Thin < Ratio 5 Thick
150
Opaque Cloud
Thick < Ratio 5 239
200
Offscale Bright
Ratio = 240
202
Step 3: Construct a histogram of the cloud decision image.
Sky cover percentages can be computed from the active pixel counts in each of the
above 5 categories.
Step 4: Imbed the cloud decision header information.
The format of the 1-min, fixed threshold header is defined in Tech. Memo AV90-042t.
Step 5: Copy the sky cover information to the LLL###.CVR file for every tenth
image.
Step 6: Transfer the cloud decision image to ExaByte.
The format of the cloud decision tapes is outlined in Tech. Memo AV89-082t.
84
6.2.1 Running CLDDECM
Having defined the thresholds for each of the sites in the preliminary data sample
period, we can now proceed to the fixed threshold cloud decision program, CLDDECM.
The 1-min ratio tape to be processed is loaded into the read ExaByte (drive 0), and a blank
tape is loaded into the write ExaByte (drive 1). A PL board is not needed for CLDDECM
processing. The example is for Columbia, Tape 035.
To Execute: enter CLDDECM at the prompt, i.e.
D:\CLDDECM>CLDDECM
Screen Output:
Marine Physical Laboratory - ptical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
CLDDECM
Version 2.1
27 Mar 90
Forms Cloud/No Cloud decision from TAPRATPL images.
(One minute only)
Location
O
1
Site # Location | Site #
---------------------
MPL
C-Station 1 7
HELSTE
1 8
Kirtland
China Lake 1 10
Malmstrom
Malabar
Columbia
Composite 1
Portable 1
-> Please enter the site number: 7
-> Enter the field tape sequence number (eg, 001): 035
-> Enter the Thin Cloud Threshold: 115
-> Enter the thick cloud threshold: 135
Note: You have several options which may be changed during a run.
Options may be changed by entering the appropriate keystroke,
without pressing the ENTER key. All of these options take
extra time, and are intended only for occasional use during
a run to visually check the ratio images.
The Key enter options are:
S to change Show Color flag
D to change info Display flag
I to change Image Save flag
X exit after current image
? to print current values of flags
Any other key will cause the available options to be printed
-> Press ENTER to continue
Opening Cloud Cover File COL035.CVR
85
cid Rat
Count Sea
Total Opaque Number
Date Time Cover Cover of Pts
460 images processed for this date.
0
0
15
Total
Opaque
CLOUD COVER SUMMARY
Cloud Cover (Tenths)
2 3 4 5 6
5 2 1 1 1
2 1 2 7 3
1
10
9
Nun N
7
2
3
8 9 10
3 19 416
5 62 351
***CLDDECM processing finished***
657 images processed for this date.
CLOUD COVER SUMMARY
Cloud Cover (Tenths)
0 1 2 3 4 5 6
Total 108 201 87 137 86 27
Opaque 240 156 152 103 4 2 0
10
owv
owo
Ono
Stop - program terminated.
D:\CLDDECM>
6.2.2 Format of the Cloud Cover Summary File
As noted earlier, CLDDECM creates a cloud cover summary file. The following
example of this file continues the Columbia Tape 035 case.
Cover File COL035.CVR
Selected thresholds: Thin = 115
Thick = 135
cid Rat
Count Seq Date
1 1 19/OCT/89
11 11 19/OCT/89
21 21 19/OCT/89
Total Opaque Number
Time Cover Cover of Pts
15:50 100.0% 99.0% 17783
16:00 99.9% 99.4% 17822
16:10 100.0% 99.6% 17794
441
451
441 19/OCT/89
451 19/OCT/89
23:10
23:20
62.8%
6.7%
7.9%
1.0%
18457
13144
460 images processed for this date.
CLOUD COVER SUMMARY
Cloud Cover (Tenths)
2 3 4 5 6
1
7
0
Total
Opaque 15
Mno
0
8 9 10
3 19 416
5 62 351
dolal
9
2
2
7
3
3
cid Rat
Count Seq Date
1 1 20/OCT/89
Total Opaque Number
Time Cover cover of Pts
12:20 35.9% 15.7% 18544
Output for 20 - 25 OCT
646 646 25/OCT/89 23:10 33.8%
656 656 25/OCT/89 23:20 82.5%
***CLDDECM processing finished***
17:4%
46.2%
19121
18902
657 images processed for this date.
CLOUD COVER SUMMARY
Cloud Cover (Tenths)
0 1 2 3 4 5 6
Total 108 201 87 137 86 27 4
Opaque 240 156 152 103 4 2 0
7
3
0
8
3
0
9
1
0
10
0
0
The "Number of Pts" column gives the total number of active pixels. While an
individual entry is made for only every tenth image, the cloud cover summary tables at the
end of each day include every 1-inin image processed.
6.3 Program CLDDECT
A 10-min version of the fixed threshold algorithm was also written, but never
employed in the standard data reduction procedure. This program was given the name
CLDDECT. It differs from CLDDECM in only a few ways. Before starting the run, a 10-
min ratio tape must be loaded into the read ExaByte, and a blank tape loaded into the write
Exabyte.
To Execute: enter CLDDECT at the prompt, i.e.
D:\CLDDECM>CLDDECT
Screen Output:
Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
CLDDECT
Version 1.0
23 Apr 90
Forms Cloud/No Cloud decision from TAPRATPL images.
(Ten minute only)
The remaining lines are almost identical to those from the CLDDECM sample, except for
the number of images processed. The output file name has the form LLL###T.CVR. The
format of the CLDDECT decision tapes is discussed in Tech. Memo AV90-057t. This
program was used to produce a sample of 10-min cloud image tapes for the preliminary
data release to NRL.
88
7.0
Opaque Threshold and Clear Sky Background Specification
The experience gained from applying the fixed threshold algorithm demonstrated
several features of the cloud decision problem that needed to be addressed in an improved
algorithm. Several of these points are outlined below.
Point 1:
Ratio values in clear sky situations exhibit some variation across the scene.
Generally speaking, the clear sky ratios are lower downsun than upsun, and
greater near the horizon than overhead.
Point 2:
A fixed threshold technique works reasonably well in identifying opaque
clouds, if the proper threshold is defined. A single threshold was used in
the TASC data set processing. Uncertainties in the SPR behavior as a
function of ND setting, and the frequent selection of the darker ND settings
at some sites suggest that different opaque thresholds at each ND setting
encountered would have improved the opaque cloud decisions.
Point 3:
The nature of the clear sky background, and thus the appropriate thin cloud
threshold, changes not only on a seasonal basis, but also from day to day
when weather patterns are changing significantly.
Point 4: Under certain conditions, particularly near sunrise and sunset, the red/blue
ratio in parts of the clear sky exceed that of opaque clouds, making a
meaningful cloud determination impossible. A means of identifying these
regions would be helpful.
Several contrail cases were studied to gain a better understanding of the behavior of
thin cloud red/blue ratios relative to their clear sky backgrounds. Carefully chosen contrail
cases are helpful in this endeavor because of their relatively uniform cloud optical depth
properties over significant distances. The primary result of these tests was that the
variation of the contrail red/blue ratio follows the variation in the clear sky red/blue ratio to
the extent that the ratio of these ratios remains relatively constant over considerable extents
of the sky dome.
.
A hybrid cloud discrimination method was then developed. The opaque cloud
discrimination is still performed using fixed thresholds that are specified individually for
each ND setting. The thin cloud discrimination is done by examining the observed/clear
sky ratios. Finally, clear sky ratio estimates are compared to the opaque thresholds to
identify those regions of the image where reasonable cloud decisions cannot be made, and
categorizing these regions as indeterminate. The success of this composite decision
algorithm (CMPDEC) hinges on our ability to define reasonable estimates of the clear sky
ratio field at specific locations and times.
Either a deterministic approach based on radiative transfer models such as FASCAT,
or an empirical approach based on the red/blue ratio imagery data base could be used in
determining the clear sky background fields. Several problems were encountered in
adapting the FASCAT model to this problem. Considering the time constraints imposed in
order to meet the project's schedule, the empirical approach was selected. Experience
gathered from studying many clear sky ratio images suggested that the clear sky ratio
distributions were driven by two major variables: the position of the sun in the sky,
particularly the solar zenith angle; and the haze layer characteristics. We discovered that the
haze effect on the clear sky ratio distribution at a given solar zenith angle, could be removed
to a reasonable degree of approximation by normalizing the ratio distribution by values
taken from specific look zenith angle-solar scattering angle combinations. As shown in
Koehler, et al. (1991), normalizing by the ratio value at the 45° look zenith angle and 45°
solar scattering angle is quite effective at removing the haze layer dependence.
The strategy for empirically modeling the clear sky background became a two step
process. First, site-specific normalized clear sky background distributions were extracted
from clear sky ratio samples. Then, values of the clear sky ratio at the 45° - 45° points for
individual images were estimated. Multiplying the normalized distribution for the observed
solar zenith angle by the reference value valid at the time of the ratio image yields a good
estimate of the clear sky background for that image.
In summary, the composite cloud decision algorithm needs three basic pieces of
information: opaque thresholds at each ND setting, normalized clear sky background
distributions at selected solar zenith angles, and clear sky reference values for the ratio tape
being processed. The following descriptions are organized around the three basic input
data sets.
7.1
Opaque Threshold Determinations
Fig. 7-1 shows the conceptual flow for determining the first two CMPDEC input
parameters. The opaque threshold definition is shown in the bottom half of the diagram.
The RATREF program is used in this step of the procedure. In its interactive form,
RATREF allows the user to color portions of the image below a certain threshold. The
selection of the threshold is performed by moving a vertical line across a ratio histogram
display and activating the mouse at the desired value. Comparisons of the color-enhanced
ratio image to its corresponding dark red radiance image are also possible.
A set of opaque thresholds must be defined for each hardware version being
processed. Examine the Ts###.LOG files to find a limited number of periods (at least 2)
when ND positions 2, 3 and 4 (1, 2 and 3 in the .LOG output) are in use. Of particular
interest are periods with frequent ND changes.
7.1.1 Running RATREF
Load one of the selected 10-min ratio tapes onto the read ExaByte (drive 0), and
position the tape to the period of interest. RATREF can then be started. The following
example is from Columbia, MO on 25/JUN/89.
To Execute: Enter RATREF at the prompt, i.e.
D:\CMPDEC>RATREF
Screen Output:
Welcome to RATREF - Version 1.2
This program makes the cloud/no cloud decision
after correcting for scattering and zenith angles. * This description
The corrected images are written to ExaByte
really applies
to CMPDECTP
(Ten Minute Full Resolution Images Only)
Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
1991 - All Rights Reserved
Site #
Location
|
Site #
Location
1
6
en A WNHO
MPL
C-Station
HELSTF
Kirtland
China Lake
Malmstrom
Malabar
Columbia
Composite 1
Portable: 1
i
10
11
-> Please enter the site number: 7
-> Enter the field tape sequence number (eg, 001): 017D
91

Normalized Ratio
Distributions
(LLL###c.OBB)
Combine Distributions
at Same Solar Zenith
I COMBNORM)
Combined Distibution
File (LLLCOMB.OUT)
Extract Clear Sky
Normalized Background
Distributions
(OBSBETA)
Manually Edit the
Combined File
10-min
Ratio
Tapes
Site Azimuth
Offset List
Normalized Clear Sky
Background Distributions
(LLLSKY.DAT)
Determine Opaque
Thresholds for All
ND Settings
( RATREFI
Composite Decision
Input File (CMPDEC.INP)
RATREF Processing
Notes
Fig. 7-1. Clear sky background distribution and composite
decision input file preparation steps
-> Do you want interactive processing? (Y or N): Y
Opening Cloud Cover File COL017D.RRE
Always choose interactive processing for threshold determination. At this point in the
program, switch to the RGB screen. You should see the image being loaded onto the FG-
100 board from the ExaByte, followed by a histogram plot of the percent of active image
area in 3 count ratio bins. A vertical line is also plotted that can be moved left and right to
some point of interest on the histogram using the mouse. An information block is written
in the top center of the image, giving the date, time, site ID and ND setting A mouse button
selection menu (see below) is written in the lower right corner of the screen.
BUTTON SELECTION
Center: view highlighted image
Right : continue to next image
If the right button is depressed, the program will load the next ratio image from the input
tape. Pressing the center button will cause the color-enhanced ratio image to appear, where
the entire region of the sky with ratios less than or equal to the threshold displayed on the
previous image is shaded blue. The information block now appears in the upper lefthand
corner, while the numerical value of the threshold is written in the upper right. The button
selection menu in the lower right corner lists the mouse options.
BUTTON SELECTION
Left : return to graph
Right: view red image
Pressing the mouse button on the left returns to the histogram plot with the previous button
selection menu, while the rightmost button brings the red image into view. The following
instruction appears on the red image.
Hit left button to return
When the left button is depressed here, the highlighted image and its button selection menu
reappears. To exit the program, return to the graph, enter Ctrl C, then press the right
button.
7.1.2 Defining Opaque Thresholds with RATREF
The ratio histogram distribution of an opaque overcast sky resembles a Guassian
distribution. The lower tail of the distribution is where the opaque threshold should be set.
For the highest ND settings, it is often difficult to identify a clear/opaque boundary, so the
following approach to using RATREF was developed. First find some cases at a lower
ND setting where reliable estimates of the opaque threshold can be made. Choose the most
representative of these values. Thresholds for the remaining ND values will be defined
relative to the chosen threshold using ratios of the center position of the opaque
distribution taken from images before and after a neutral density filter change. This
procedure is better illustrated in the RATREF processing notes taken during the Columbia,
MO opaque threshold determination (Table 7-1).
Ratio tape COL017 was loaded onto the ExaByte, and moved to the beginning of the
fourth day (25/JUN/89). A good opaque threshold estimate in ND 1 of 135 was obtained
93
on the 16:50 image. The center of the opaque distribution at that time was at 149. The ND
changed to 2 on the next image, with the opaque center moving to 164. The ratio between
the 17:00 and the 16:50 images was 165/149 = 1.101, as shown in the last column of Table
7-1. The transition ratios were also measured several times on the following day.
Averaging the six estimates of the ND2/ND1 transition ratios yields a value of 1.131.
Similarly the NDO/ND1 transition ratio was estimated from 27/JUN/89 and 1/SEP/89
images to be 1.079. Next, multiplying the ND 1 threshold of 135 by the two transition
ratios yields threshold estimates of 145.7 for ND 0 and 152.7 for ND 2. An independent
estimate of the ND O value from the ratio images on 27/JUN/89 was 144. A compromise
value of 145 for ND 0 was chosen, while the transition estimate of 153 for ND2 remained
unchanged. ND 3 was never reliably chosen for Unit 7V1.
7.1.3 CMPDEC.INP File Entries
The final step in the opaque threshold specification is to enter the thresholds into the
CMPDEC input table, as shown below for our example.
File Name : CMPDEC. INP
File Date: 30 Oct 1991
Calibration
File Name
CALIVO4B
Azimuth
Correction
3.5
ND=0
173.
Opaque Threshold
ND=
1 NDE2ND=3
165. 165.* 165.*
Thin Cloud
Accept Lul
1.20
CAL7001A
CAL7001B
CAL7V01C
-1.5
-1.5
-1.5
145.
145.
145.
135.
135.
135.
153.
153.
153.
153.
153. *
153.
.
1.20
1.20
1..20
The * after the ND=3 value indicates that this value is never used. The last column, the
thin cloud acceptance level, will be described later in the report.
94
Table 7-1 Sample RATREF Processing Notes
ND 2 Opaque Threshold Determination
Columbia, MO
2/AUG/91
Opaque
Center
ND2 / ND1
25/JUN/89
Opaque
Time
ND
Threshold Comments
16:40
Dark cloud problems
16:50
135
better threshold
17:00
17:10
149
1.101
164
172
26/ JUN/89
Opaque
Threshold
ND
Time
16:20
ND2 / ND1
Opaque
Center
140
Comments
Peak of clear sky
1.157
16:30
Peak of clear sky
162
17:40
1
Overcast
17:50
18:10
18:20
165
1.115
184
176
1.143
154
1.162
179
1.105
162
Average ND2 / ND1 = 1.131
18:30
2
18:40
1
27/JUN/89
ND
Opaque
Center
Comments
NDO / ND1
Time
13:40
Opaque
Threshold
144
good
19:50
157
1.076
20:00
0.
169
1/SEP/89
Opaque
Threshold
ND
Comments
Opaque
Center
160
Time
13:50
NDO , ND1
1.088
174
14:00
14:50
171
1.089
15:00
1
157
Table 7-1 Page 1
95
1/SEP/89 (Cont.)
Time
ND
Opaque
Threshold
Opaque
Center
Comments
ND2 / ND1
16:20
16:30
0
16:40
19:00
19:10
19:20
159
1.044
166
1.051
158
161
1.093
176
1.086
162
1.086
176
1.100
160
161
1.075
173
1.081
160
Average NDO , ND1 = 1.079
19:30
19:40
21:10
21:20
21:30
Columbia, MO - Version 7V1
Estimated Opaque Thresholds
ND
Threshold
0
145
1
135
2
153
Table 7-1 Page 2
96
7.2
Clear Sky Normalized Ratio Background Determination
O
The next step in the CMPDEC input file preparation is to build the normalized clear
sky ratio distribution tables, as shown in the top portion of Fig. 7-1. Using the Form 10's
or some other weather data, a set of possible clear days at a particular site is defined. The
normalized distributions are only a function of site, so the sample can be built from data
taken with different hardware versions. (Normalizing the data removes the primary
influence of hardware version, the SPR selection.) The sample should, as much as
possible, include days from different seasons and a broad spectrum of haze conditions.
Completely clear days near the summer solstice are often in short supply. Remember also
that cases near the horizon are only available between the autumnal and vernal equinoxes,
because data collection begins after sunrise and ends before sunset in the summer.
7.2.1 OBSBETA - The Normalized Ratio Extraction Program
The OBSBETA program was written to facilitate the normalized data extraction, and to
allow the operator to remove nonrepresentative data from the data base. After loading the
10-min ratio tape into the read ExaByte (drive 0), and positioning the tape at the beginning
of the day selected, the program can be started. The example below is for 29/JUN/89 at
Columbia, MO.
To Execute: Enter OBSBETA at the prompt, i.e.
D:\CMPDEC>OBSBETA
Screen Output:
Welcome to OBSBETA - Version 1.0
This program extracts the normalized ratio data from
ten minute full resolution data tapes
Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
1991 - All Rights Reserved
Site #
Location
Site #
Location
uns WNHO
MPL
C-Station
HELSTF
Kirtland
China Lake
Malmstromi
o ovo
Malabar
Columbia
Composite 1
. Portable 1
1
1
9
10
11
-> Please enter the site number: 7
-> Enter the field tape sequence number (eg, 001): 018G
Enter a new azimuth offset:-1.5
97
-
At this point in the program, switch to RGB output, and answer the appropriate screen
prompts as described below.
The RGB display now shows a 10-min ratio image, with the date, time and site ID
written in the upper left corner. A yellow * marks the estimated sun position, and the solar
zenith angle is written on the lower right of the screen. The option menu appears at the
bottom of the screen.
MENU: P-Process G-Get new img
S-Stop
Option Descriptions:
S-Stop
- stop processing for this data set. This closes the
LLL###c.OBB output file.
G-Get new img - causes the next ratio image to be read
P-Process
- continues the processing of this image. Normally this option
is chosen if the solar zenith angle displayed is within 1° of an
even 5° increment, and if the image is clear enough and of
reasonable quality to yield a representative sample.
Assume that the P option was selected by keying in P then Enter. The RGB image
changes. A blue 1 and possibly a blue 2 are written on the image to indicate the positions
of the 45°-45° points. The reference values for these points computed from 5x5 point
averages appear in the upper right corner. A new option menu is written at the bottom of
the screen. (Several versions of this menu can be written. The example shown here is for
the 2 reference point case.)
Enter
Option
1-#1 2-#2
B-Both
N-None R-view, red
Option Descriptions:
N-None - Discontinue processing of this image and read the next.
R-view red
- view the corresponding dark red image. Return to the "Enter
Option" menu by pressing Enter.
1-#1
2-#2
- normalize using the reference value at point 1.
- normalize using the reference value at point 2.
- normalize using the average of reference values 1 and 2.
B-Both
The selection of 1, 2 or B is based upon the visual assessment of whether the reference
values are reasonable, and not contaminated by dirt, clouds, or stray light. If 1,2 or B are
chosen, the image is normalized, i.e., the individual pixel values are divided by the
reference value, then multiplied by 100. After the computations are complete, the
normalized ratio sample is extracted from points spaced every 5° in zenith from 0° to 75°,
and every 15° in azimuth relative to the solar azimuth from 0° to 345º. The values extracted
98
are averages over the 3x3 arrays centered on these points. This distribution of points is
illustrated in Fig. 7-2. A + sign is then plotted on the screen, either in red or green. A red
+ indicates that all 9 points were off scale, while a green indicates that at least one point is
on scale. A blue arrow also appears somewhere on the screen. The user can now use the
mouse to position the arrow over green points that do not appear to be representative of the
clear sky. This point can be removed from the sample by pressing the center mouse
button, which also has the effect of changing the color of the point to red. After all the
desired deletions are completed, the user presses the rightmost mouse button, causing the
sample to be written to the LLL###c.OBB file then initiating the next image read. A sample
entry from 15:20 on 5/JUL/89 at Columbia is provided in Table 7-2.
7.2.2 Combining the Sample Files Using COMBNORM
After all the sample days have been processed through OBSBETA, they can be
combined into one data set using the COMBNORM program. Table 7-3 lists the times on
the 13 sample days used in developing the Columbia, MO normalized clear sky ratio
background tables. The LLL###c part of the .OBB filenames are given below the date
string. Since two samples per day are possible at a given solar zenith angle, information
from up to 26 sample images could be combined. The actual number of cases combined
varied from a maximum of 21 for 60° and 65°, to a minimum of 2 for 15º.
The COMBNORM program reads each sample distribution from a specified input file,
and adds the extracted values of normalized ratio and scattering angle to the appropriate
position in the sum arrays. The number of points included in the grid position's sum is
incremented if a non-zero normalized ratio was present. Since the distribution should be
symmetric about the solar azimuth, the points displaced from solar azimuth be the same
amount are also combined. For example, on the 35° zenith line in Table 7-2 both the 90
value from relative azimuth 165º and the 88 value from relative azimuth 195° would be
added to the composite sum at 165° relative azimuth. After all the input tables are
processed, the average.normalized ratio is computed at each grid location. The following
sample run illustrates how the Columbia samples were combined.
To Execute: Enter COMBNORM at the prompt, i.e.
D:\CMPDEC>COMBNORM
Screen Output:
Y
Y
Y
Enter the input file name : COL001B.OBB
Finished this file - get another? (Y or N):
Enter the input file name : COL003A, OBB
Finished this file - get another? (Y or N):
Enter the input file name : COLO06A.OBB
Finished this file - get another? (Y or N):
Enter the input file name: COL006G.OBB
Finished this file - get another? (Y or N):
Enter the input file name : COL010A.OBB
Finished this file - get another? (Y or N):
Enter the input file name: COL016D.OBB
Finished this file - get another? (Y or N):
Enter the input file name : COL018A.OBB
Finished this file - get another? (Y or N):
Enter the input file name: COL018G. OBB
Finished this file - get another? (Y or N):
8 fil file met anoi, coco
Y
Y
Y
Y
Y

.
0
.
1
.
1
.
.
.
.
.
.
.
.
.
.
.
. . .
......
A
.
.
.
9
.
.
.
0
.
.
.
.
.
.
.
.
...0
....
.
.
.
.
...
.
.
.
...
.
.
.
.
.
.
1
.
.
.
...
.
.
......
.. . ..
.
•
.
.
5
..
t
..
.
.
101
....
.......
.
.
#
.
.
.
OMSO
.
.
#
...
. .
.
...
.
.
.
.
....
..
...
0
..
.
......
.
.
S
.00
.
1
...
:
...
iii ...
.. .. ..
11
.
.
O
10.
.
...
DO
........
.
.
.
.
.
.
.
.
.
.
.
.
....
.
...
.
.
.
.
.
.
.
C000
..
.
.
.
.
.
..
.
.
010
.
.
.
.
..
..
.
.
..10
.
.
.
.
.
.
.
.
-
.
10..
.
.
100
V
DUR
..1
1
.
:
.
.
00
1
.
..
.1
.
10
C.RO
.
.
..
+
•
10
.
.
.
:
1
Pro0.50.
:
.
.
.
.
+10
.
.
.
)
00...10
#000 . . ... . ....
..10.2160010..
.. .... .. .. ...
. . . . . .. .
0 . 1 . . . . . . . . . . .
...... .......
.......... .
A
1
000
.
.
.
07
.
.
.
.
.
.
.
.
.. .....
9 . .
€0...
....
.
TOA
0
1
.
...
.
..
...
.
..
.
.....
.
.
i
.
.
.
.
.
.
+
..
..
.
!
.
60
.
..
...
.
.
...
+ 0
.
1.005
. 1 . ..
....
.....
.
.
1
.
.
..
0 .
.
$
...
.
..
+
.
..
.
•
.
•
W
11
.
•
10
0 0
.
.
1
.
.
O
.
O
20
.
...
....
..
.
..
.
.
..
. .
...
. . . .
.
.
.
..
.
.
.
.
10.
.
. . .
. ..
1
.
.
1
.
..
.
.
....
.
1
.
.
.
.
.....
0
.
.
.
.
.
.
.
.
.
..
.
.
......
.
O
...
.
.
.
....
t
.
.
.
....
.
.
.
0.06
..
.
1
...
.
.
.
..
.
....
.
.
.
.
O
.
.
..1.
.....
0
.
20
.
.
..
..10...
10.
.
.
.
D .
. . .
. • OOO
.
101
0 . ..
00000.
.. ..
.. ...
.
..
1.00
.. ...
......
.
.
.
.
...
.
D
.
.
.
16
•
DO
. CO
..
1
..
..
000
... .
.. .
.. .......
..
... ...
.
.
.
.
.
1
.
.
.
.
1.101.00
.. . . .
. . . .
.
....
.. .
......
.
.
.
.
. . . . .
00DD.
.
40
...
.
11
.
.
.
.
.
.
.
.
....
.
.
.
+
0
0 .
OTO +
.
.
.
..
O
..
..
0...
...
. ...
.
.
.
.
POD..
01.0.
.
0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
SNO
. .
. 0
.
.
......
.. ..
.
.
.
.
.
...
.
.
.
..
1.
S .
. *
100
.
002
.
0.00
.
.11
...
.
.
.
.
1
.
19t..
.
.
•
.
.
...
.
.
0
.
.
. ....
....
.
.
.....
.
.
.
.
.
.
...
.
.
.
.10
000.
.
.
.
0
.
..
.
0 . 0 .
...
. .. ..
. . . .
.
+
.
.
.
20
........
.
.
....
•
.
.
.
.
.
i
.
.
...
.
0
.
.
.
. ........
... .....
1 . 4 . .
. . . . .
.
.
.
+
.
-
.
.
.
.
.
.
.
.
1
1
.
.
0
.
.
0
.
.
9
.
U
.+
.
2
.
0
.
.
0
.
100
1.00
1
+
0
.
.
+
.
•
10000
.000 .
. .
n .
.
•..•°000.0 00 . 000
. . . 0 . .
..... ... 000.0 00.
0
00
..O
019.000.000X1000+
.... . . . ..
. .... .
... .. ...
.
. .. . .
. . . 1 .
01. ..... ... . . .. .. .
. . . . . . . .
0 . . . . .
.000000
100001
+
.
+
.
.
+
.
2000.
TO
.
1
+
0
.
.
0
0
.
DO
001
..
..
+
1
.
1
.
0
0001
.
.
+
.
.
+
.
.
.
.
....
.
0 .
910
...
.
...
.
0
0
.
1
0
0
0
0
0
.
0
.
LL
.
A
..
.
.
.
.
.
.
.
.
1
.
...
.
0
0
.
.
.
.
...
.
.
.
..
.
...
.
.
.
1
...
1
.
.
..
....
1
.
.
..
i
...
...
......
.
10
..
.
.
.
.
.
O
.
.
+
..
.
0
.
0.
0
.
1
1
.
T
CD-
.
_
_001
. .
•
.
.
.
1
1
Tonnt
COTT
.
.
11
. .
OCCO
.1. .
.
$
t
.
.
.
.
.
.
.
.
1
.
.
.
.
.
1
.
11
.
.
+
+
.
.
.
.
.
101.
-
1000
.
.
.
.
6
.0
.
.
.
.
t
.
•
.
.
.
.
01
.
.
.
.
.
.
- 11
.
.
.
100
Tr
.
.
.
.
:
.
.
.
.
.
.
.
.
.
.
.
+0.0001
.
.
.
.
.
.
.
+
.
.
11
.
.
.
.
.
.
.
.
.
.
_
_
$
6
0
.
.
.
.
1
0 .
016
1.
0
.
.
.
.
.
_
+
.
06
.
_
.
.
+
.
.
-1000
110
.
.
0
.
..OP
.
..
1
CO.
+
.
.
.
a
+
.
.
....
.
.
10
.
+
+
.
.
1
0
+
.
.
.
2
.
.0
+
•
1
+
+
.
.
.
.
.0.
.
-
.
1
11
.
+
+
OT+
. .
DI
.
a
.
TO
.
.
+
.
1
.
.
.
.
-
.
+
.
1
1
1
1
0
0
.
+
..
+
.
.
:
C1
.
+
.
1..
.
I
+
.
1
.
3
..
.
1
+
+
.
1
0
.
.
.
+
.
.
.
.
.
.
.
.
.
.
.
.
.
1 1
102
1
.
.
.
.
.
.
.
.
.
.
+
.
1
.
.
.
.
.
.
.
.
.
.
.
-
.
.
.
.
.
.
.
.
0
6
.
.
.
.
.
.
.
.
.
.
.
.
1
.
.
.
.
.
.
.
.
.
.
-
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
!
0.00
1
.
+
1
.
.
.
.
.
.
11
7
.
. . . . . . . .
. .
0001
. . . .
AUTO
. . . . . . . . . . .
. . . .
. .
1
0 .00 . . 0 .
OOO.
. ..
CON
.
4
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
-
.
0
.
+
.
.
10.
1
(
.
+
...
.
.
1
..
....
•1
.
.
-
.
..11
00
...
.
..
0
.
...
t
.
..
.
.
+
.
.
.
.
.
.
.
.
.
.
.
.
S
.
.
.......
.
....
0
DO!
.
.
.
.
.
..
.
..
.
.
.
..
.
.
.....
.. . .
.. .....
....
.
.
€
..
.
..
.
&...
...
.
.
.
.
.. ..
. .. .
C
.
_
.
.
.
..
.
1
....
.
.
.
.
.
.
SO
.
.
.
0.00000
.
.
11.
. .
.
..+00:00
.
006
to
11.
.
Fig. 7-2. OBSBETA sample points.
100
Table 7-2. Sample Entry into File COL018G.OBB
Normalized Ratio File COL018G.OBB
Lat = 38.817
Solar Declin =
Long = 92.217
22.740 Eq. of Time =
-.077
Site=COL Date= 5/ JUL/89 Time=15:20 Solzen=40.12 Coszen= .765 SolAzi= 100.65 Refval = 72.7
z/RA 0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 246 255 270 285 300 315 330 345
.0 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97 97
5.0 101 99 99 100 100 99 99 98 98 96 97 97 96 97 96 97 98 99 99 99 99 99 100 99
10.0 100 100 109 103 103 100 102 99 98 97 97 98 98 95 95 96 97 98 98 99 100 99 102 103
15.0 107 109 104 104 103 102 99 99 98 95 96 98 96 95 95 96 96 97 99 99 100 102 106 106
20.0 0 107 108 107 103 103 100 99 97 95 96 96 95 94 94 94 95 95 99 100 102 108 103 107
25.0 0 0 108 107 106 102 100 98 96 94 94 94 93 92 92 97 100 98 99 100 103 107 107 108
30.0 0 0 108 106 104 102 98 95 92 93 92 93 92 90 95 96 99 104 101 100 105 107 108 0
35.0 0 0 0 109 105 100 98 94 91 90 92 90 88 88 94 95 99 102 106 100 104 105 108 0
40.0 0 0 0 108 103 98 95 91 91 88 88 88 88 89 90 91 92 103 105 100 101 104 107 0
45.0 0 0 0 110 105 99 94 87 87 87 88 85 86 88 86 88 93 95 101 106 103 109 108 0
50.0 0 0 112 110 106 98 92 87 85 85 89 88 81 84 84 85 87 92 99 105 100 108 108 0
55.0 0 0 109 113 108 96 91 85 83 82 92 88 79 81 82 83 85 89 96 104 104 109 114 107
60.0 111 112 0 111 107 94 89 86 81 820 87 80 80 79 81 86 90 96 100 106 111 110 110
65.0 116 115 84 114 105 95 90 87 81 84 0 78 79 78 82 85 88 95 100 105 113 115 113
70.0 118 119 120 117 108 98 93 90 93 84 0 0 84 81 81 82 85 89 98 105 102 115 115 117
75.0 124 123 123 121 115 110 0 97 92 91 0 0 87 88 86 88 94 98 107 115 110 117 122 123
NNW
Z/RA
.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
45.0
50.0
55.0
60.0
65.0
70.0
75.0
0
40.1
35.1
30.
1
25.
1
20.1
15.1
10.1
5.1
.1
4.9
9.9
14.9
19.9
24.9
29.9
34.9
1 5
40.1
35.3
30.6
25.9
21.3
17.0
13.2
10.4
9.6
11.2
14.5
18.5
22.9
27.5
32.2
37.0
30
40.
1
35.9
31.8
28.0
24.6
21.7
19.7
18.8
19.2
20.7
23.2
26.4.
30.0
34.0
38.2
42.5
45
40.1
36.
7
33.7
31.1
29.0
27.7
27.1
27.4
28.5
30.4
32.8
35.7
39.0
42.6
96.4
50.4
60
40.1
3 7.8
36.0
34.7
34.0
34.0
34.6
35.8
37.
5
39.8
42.4
75
40.1
39.1
38.6
38.6
39.1
40.2
41.8
43.8
46.1
48.8
51.7
54.9
58.2
61.7
65.3
69.0
90
40.1
40.4
41.1
42.4
44.1
46.1
48.5
51.2
54.1
57.3
60.6
64.0
67.5
71.1
74.8
78.6
105
40.1
41.6
43.6
45.9
48.6
51.5
54.6
57.9
61.4
65.0
68.7
72.4
76.2
80.1
84.0
87.9
120
40.1
42.8
45.8
49.1
52.5
56.2
59.9
63.8
67.7
71.8
75.8
79.9
84.1
88.2
92.4
96.5
135
40.1
43.8
47.6
51.6
55.8
60.0
64.3
68.6
73.0
77.4
81.8
86.3
90.7
95.2
99.6
104.0
150
40.
1
44.5
49.0
53.5
58.
1
62.8
67.5
72.2
76.9
81.6
86.3
91.1
95.
8
100.5
105.2
109.9
165
40.1
45.0
49.8
54.7
59.6
64.5
69.4
74.4
79.3
84.2
89.2
94.1
9 9.0
103.9
108.9
113.8
180
40.1
45.1
50.1
55.1
60.1
65.1
70.1
75.1
80.1
85.1
90.1
95.1
100.1
105.1
110.1
115.1
45.4
48.6
52.0
55.6
59.4
101
Table 7-3. Columbia, MO Ratio Images Processed by OBSBETA
in the Clear Sky Background Determination
Solar
Zenith
13/APR/89
COL010A
90
85
O
O
10/FEB/89
COL001B
13:00 23:40
13:40 23:10
14:10 22:40
14:30 22:10
15:10 21:40
15:40 21:00
16:20 20:20
17:20 19:30
O
O
23/FEB/89
COL003A
23:50
23:20
23:00
14:20 22:30
14:40 22:00
15:20 21:30
15:50 20:50
16:30 20:10
17:30 19:10
O
WWD Aurin aavvo
16/MAR/89
COLO06A
24:10
12:50 23:50
13:10 23:20
22:50
22:30
14:40 22:00
15:00 21:30
21:00
16:10 20:20
16:50 19:50
18:20
Nmmm
22/MAR/89
COL006G
12:20
12:40
13:00
13:30 23:00
14:00 22:40
14:20 22:10
14:50 21:40
15:20
15:50 20:40
16:30 20:00
17:20 19:10
O
12:10
12:30
13:00
13:20 23:00
13:50 22:30
14:20 22:00
14:40 21:40
21:10
15:40 20:40
16:10 20:10
16:50 19:30
17:40 18:30
Solar
Zenith
27/MAY/89
COL016D
29/JUN/89
COLO18A
5/ JUL/89
COL018G
9/AUG/89
COLO23G
18/SEP/89
COLO30E
90
85
80
75
70
65
60
50
12:13 23:53
12:43 23:33
13:03 23:03
13:33 22:33
14:03 22:13
14:23 21:43
14:53 21:23
15:13 20:53
15:43 20:33
16:13 20:03
16:43 19:33
17:13 18:53
ܐ ܚ ܢ ܚ ܙܙ
لا لا لا لا لا لا
23:40
23:20
22:50
22:20
14:10 22:00
14:40 21:30
15:00 21:00
15:40 20:30
16:10 19540
17:00 19:10
12:20 24:10
12:40 23:50
13:10 23:20
13:30 23:00
14:00 22:30
14:20 22:00
14:50 21:40
15:20 21:10
15:40
16:10
16:50
17:30
12:20
12:50 23:40
13:10 23:20
13:40 22:50
14:00 22:20
14:30 22:00
14:50 21:30
15:20 20:10
15:50 20:40
16:10 20:10
16:40 19:40
17:20 19:10
18:10
Ол ол
40
30
NN
20
15
16:40
17:10
18:10
30/NOV/89
COLO42A
Solar
Zenith
90
85
80
13:40
14:10
14:50
10/OCT/89
COL033F
23:30
23:10
13:10 22:40
13:40
14:10
14:40
15:10 20:40
15:40 20:10
16:30
18:00
30/NOV/89
COL0422
22:40
22:10
21:40
21:10
20:30
75
70
65
16:20 19:40
17:50
55
50
45
102
Y
Y
Enter the input file name: COLO23G.OBB
Finished this file - get another? (Y or N):
Enter the input file name: COL030E.OBB
Finished this file - get another? (Y or N):
Enter the input file name: COL033F.OBB
Finished this file - get another? (Y or N):
Enter the input file name: COLO42A. OBB
Finished this file - get another? (Y or N):
Enter the input file name: COL042Z.OBB
Finished this file - get another? (Y or N):
Y
Y
N
Stop - Program terminated.
D:\CMPDEC>
mann wr can be found in file murun
The resulting average distributions are written to a file named COMB.OUT. This file
should be renamed LLLCOMB.OUT to avoid future confusion. It is also helpful to assign
a new designator for each portable deployment. For example, the combined clear sky
tables for Madison, WI can be found in file MSNCOMB.OUT instead of
PORCOMB.OUT. The combined normalized ratio tables from COLCOMB.OUT at 90°
and 40° solar zenith are presented in Table 7-4. Note that some grid locations near the sun
position (0° relative azimuth at the zenith for the table) were always blocked by the occultor
mask, and thus have a 0 combined value. Other points near the occultor also have spurious
values. The next step is to use interpolation and/or extrapolation to fill in the missing
values, and to correct the spurious values. Table 7-5 shows the corrected 40° solar zenith
table from Table 7-4. Underlines indicate corrected values. The corrected files are stored
under both the LLLCOMB.REV and LLLSKY.DAT file names.
103
Table 7-4. Sample Output from COLCOMB.OUT
l
90
82.0
81.3
80.4
81.
6
80.7
80.3
Relative Ratio for Solar Zenith 90
Z/RA O 15 30 45 60 75
0 82.0 82.0 82.0 82.0 82.0 82.0
5 83.7 81.7 82.9 82.6 82.0 81.1
10 86.1 84.3 83.1 81.6 81.1 80.6
15 85.9 84.6 89.4 83.
6 81.1 82.1
20 86.7 86.1 86.4 89.6 83.3 81.9
88.7 88.3 87.1 84.0 83.9 81.7
89.6 88.6 87.6 85.4 82.7 81.4
35 90.9 90.1 89.0 86.3 82.7 81.6
40 95.1 92.4 90.7 86.3 83.4 82.0
45 101.2 97.7 93.6 89.3 85.6 80.7
50 109.2 106.9 97.1 91.4 86.3 82.3
55 118.4 114.8 103.4 94.1 89.0 84.3
60 128.8 123.2 110.1 97.9 90.9 85.3
65 137.0 135.0 120.4 104.4 95.4 88.9
70 0 156.0 134.0 115.0 103.9 96.4
75 0 180.3 158.1 136.3 118.7 108.9
80.7
78.
9
79.
3
79.7
80.7
80.
6
82.
3
85.
4
91.
4
102.7
7 data sets)
105 120 135
82.0 82.0 82.0
80.9 81.0 80.0
81.4 80.9 79.7
79.3 80.1 79.9
79.7 80.0 79.9
79.7 79.4 79.1
79.1 78.0 79.4
78.7 79.1 79.4
78.3 79.0 79.7
79.4 79.3 80.6
79.6 79.9 81.6
81.0 81.3 83.7
81.9 83.6 85.1
84.4 86.
9 90.0
90.9 92.7 96.0
101.6 106.0 111.4
150
82.0
80.3
79.4
79.9
78.9
78.7
79.4
77.7
79.7
80.9
82.7
84.4
88.1
90.7
98.6
113.4
165
82.0
80.6
79.9
80.3
79.9
80.0
79.9
78.4
80.7
82.0
83.4
84.9
88.4
93.0
100.4
117.0
180
82.0
82.0
80.3
79.4
79.9
80.4
79.7
78.7
81.1
82.4.
84.3
86.0
89.0
94.5
102.6
120.2
Corresponding BETA Values
Z/RA 0 15 30 45 60
O 89.6 89.6 89.6 89.6 89.6
5 84.6 84.8 85.3 86.1 87.1
10 79.6 80.0 81.0 82.6 84.7
15 74.6 75.2 76.7 79.1 82.2
69.6 70.4 72.4 75.7 79.8
25 64.6 65.6 68.2 72.3 77.5
59.6 60.8 64.0 68.9 75.2
35 54.6 56.0 59.9 65.7 73.0
40 49.6 51.3 55.8 62.7 71.0
44.6 46.6 51.9 59.7 69.0
50 39.6 41.9 48.1 56.9 67.2
34.6
37.4 44.5 54.4 65.6
60 29.6 32.9 41.1 52.0 64.2
65 24.6 28.6 38.1 50.0 62.9
70 19.6 24.5 35.3. 48.2 61.8
75 14.6 20.8 33.1 46.8 61.0
75
89.6
88.4
87.1
85.8
84.6
83.4
82.2
81.2
80.1
79.2
78.3
77.6
76.9
76.3
75.8
75.4
90
89.6
89.
6
89.7
89.
7
89.7
89.
7
89.7
89.7
89.
7
89.8
89.8
89.8
89.8
89.8
89.9
89.9
105
89.6
90.9
92.2
93.5
94.7
95.9
97.1
98.2
99.3
100.3
101.2
102.0
102.8
103.4
103.9
104.4
120
89.6
92.1
94.6
97.1
99.5
101.8
104.1
106.4
108.5
110.4
112.3
114.0
115.5
116.8
117.9
118.8
135
89.6
93.2
96.7
100.2
103.6
107.0
110.4
113.6
116.7
119.7
122.5
125.1
127.5
129.7
131.5
132.9
150
89.6
94.0
98.3
102.6
106.9
111.1
115.3
119.4
123.5
127.4
131.3
134.9
138.3
141.5
144.3
146.6
165 180
89.6 89.6
94.5 94.6
99.3 99.6
104.1 104.6
108.9 109.6
113.7 114.6
118.5 119.6
123.3 124.6
128.0 129.6
132.7 134.6
137.4 139.6
142.0 144.6
146.4 149.6
150.8 154.6
154.9 · 159.6
158.7 64.6
Table 7-4. Page 1
104
Table 7-4. (Cont.)
25
30
Relative Ratio for Solar Zenith 40 ( 12 data sets)
z/RA 0 15 30 45 60 75 90 105 120 135
o 97.4 97.4 97.4 97.4 97.4 97.4 97.4 97.4 97.4 97.4
5 102.0 101.2 100.6 100.3 98.
9 99.2 98.1 97.8 97.1 96.9
10 105.0 104.3 104.4 102.3 101.1 99.7 98.
5 98.0 97.3 96.5
15 108.3 107.3 105.8 104.9 102.8 101.5 99.6 98.4 97.0 96.7
20 112.8 110.8 108.2 105.5 102.8 101.6 99.5 97.6 97.1 96.7
.0 112.0 107.9 106.0 103.6 100.4 98.3 97.2 96.2 95.4
0 108.4 106.8 103.6 99.7 97.7 95.4 94.0 93.1
.0 115.2 105.7 102.8 99.3 96.3 93.4. 92.1 91.0
0 .0 115.2 105.3 101.8 98.3 94.8 91.8 89.9 88.6
.O 112.8 105.2 101.8 97.1 93.7 90.0 88.6 87.4
0 .0 111.7 105.2 101.0 97.0 91.9 89.2 87.
1 85.5
.0 116.8 110.3 106.2 100.8 95.8 92.3 89.4 86.2 85.0
60 116.5 116.4 111.9 106.5 100.9 96.7 92.7 88.8 86.6 84.9
65 118.1 116.1 113.7 109.0 102.6 98.0 93.5 90.3 87.6 85.8
70 120.1 118.4 115.5 110.7 104.4 99.9 96.
7 92.4 89.3 87.8
75 129.4 125.3 121.5 118.5 112.7 108.3 106.3 101.1 97.1 95.9
35
150
97.4
97.1
96.2
97.0
95.8
94.4
93.1
90.8
88.4
86.1
84.6
83.4
83.1
84.8
87.0
94.3
165
97.4
97.0
96.2
96.0
95.9
94.8
93.1
90.3
88.5
87.9
85.3
83.6
82.8
83.8
86.7
95.3
180
97.4
97.6
97.4
96.8
96.4
95.2
93.7
91.2
89.5
87.8
85.6
83.9
82.8
84.7
87.6
96.6
50
55
Corresponding BETA Values
z/RA on 15 30 45 60
0 40.2 40.2 40.2 40.2 40.2
5 35.2 35.4 35.
9 36.8 37.9
10 30.2 30.6 31.9 33.7 36.1
15 25.2 26.0 28.1 31.1 34.8
20 20.2 21.4 24.7 29.1 34.1
15.2 17.1 21.8 27.7 34.0
10.2 13.3 19.8 27.2 34.6
35 5.2 10.5 18.9 27.4 35.8
40 .3 9.6 19.2 28.5 37.6
45 4.8 11.2 20.7 30.4 39.8
9.8 14.4. 23.2. 32.8 42.4
55 14.8 18.4 26.4 35.7 45.4
19.8 22.8 30.0 39.0 48.6
65 24.8 27.4 34.0 42.
6 52.0
70 29.8 32.1 38.
1 46.4 55.6
75 34.8 36.9 42.5. 50.3 59.4
75
40.2
39.2
38.6
38.6
39.2
40.3
41.9
43.8
46.2
48.8
51.8
54.9
58.2
61.7
65.3
69.0
.90
40.2
40.5
41.2
42.5
44.1
46.
48.6
51.3
54.
2
57.
3
60.
64.0
67.5
71.2
74.9
78.6
105
40.2
41.7
43.7
46.0
48.6
51.6
54.7
58.0
61.5
65.0
68.7
72.5
76.3
80.1
84.0
87.9
126
40.2
42.9
45.9
49.1
52.6
56.2
60.0
63.9
67.8
71.8
75.9
80.0
84.1
88.3
92.4
96.5
135
40.2
43.9
47.7
51.7
55.8
60.0
64.3
68.7
73.0
77.4
81.9
86.3
90.8
95.2
99.7
104.1
150
40.2
44.6
49.1
53.6
58.2
62.9
67.5
72.2
76.9
81.7
86.4
91.1
95.9
100.6
105.3
110.0
165
40.2
45.1
49.9
54.8
59.7
64.6
69.5
74.5
79.4
84.3
89.2
94.2
99.1
104.0
108.9
113.9
180
40.2
45.2
50.2
55.2
60.2
65.2
70.2
75.2
80.2
85.2
90.2
95.2
100.2
105.2
110.2
115.2
Table 7-4. Page 2
105
Table 7-5. COLSKY (COLCOMB.REV) Sample Table
Relative Ratio for Solar Zenith 40 ( 12 data sets)
Z/RA 0 15 30 45 60 75 90 105 126 135
O 97.4 97.4 97.
4 97.4 97.4 97.4 97.4 97.4 97.4 97.4
5 102.0 101.2 100.6 100.3 98.9 99.2 98.
1 97.8 97.1 96.9
10 105.0 104.3 104.4 102.3 101.1 99.7 98.5 98.0 97.3
965
15 108.3 107.3 105.8 104.9 102.8 101.5 99.
6 98.4 97.0 96.7
20 112.8 110.8 108.2 105.5 102.8 101.6 99.
5 97.6 97.1 96.7
25 116.1 112.0 102.2 106.0 103.6 100.4 98.3 97.2 96.2 95.4
30 119,0 114.2 100 106.8 103.6 99.7 97.
7 95.4 94.0 93.1
35 123.] 120.6 115.2 105.7 102.8 99.3 96.
3 93.4 92.1 91.0
40 130.1 124.5 115.2 105.3 101.8 98.3 94.8 91.8 89.9 88.6
45 126.3 119.5 112.8 105.2 101.8 97.1 93.7 90.0 88.6 87.4
50 122,7 117.1 111,7 105.2 101.0 97.0 91.9 89.2 87.1 85.5
55 119.1 116.8 110.3 106.2 100.8 95.8 92.
3 89.4 86.2 85.0
60 116.5 116.4 111.9 106.5 100.9 96.7 92.
7 88.8 86.6 84.9
65 118.1 116.4 113.7 109.0 102.6 98.0 93.
5 90.3 87.6 85.8
70 120.1 118.4 115.5 110.7 104.4 99.9 96.7 92.4 89.3 87.8
75 129.4 125.3 121.5 118.5 112.7 108.3 106.3 101.1 97.1 95.9
150
97.4
97.1
96.2
97.0
95.8
94.4
93.1
90.8
88.4
86.1
89.6
83.4
83.1
84.8
87.0
94.3
165
97.4
97.0
96.2
96.0
95.9
94.8
93.1
90.3
88.5
87.9
85.3
83.6
82.8
83.8
86.7
95.3
180
97.4
97.6
97.4
96.8
96.4
95.2
93.7
91.2
89.5
87.8
85.6
83.9
82.8
84.7
87.6
96.6
106
7.3
Clear Sky Reference Value Specification
At this point in the processing, the opaque ratio thresholds have been defined and the
normalized clear sky background distributions have been computed. What remains is to
define how the 45° - 45° reference parameter changes for each day being processed. Fig. 7-
3 shows how this is accomplished.
The first order of business is to load the 10-min ratio tape onto the read ExaByte, and
run the GETREF routine, which computes reference value estimates for each image on the
tape. The analyst then uses the SETREF routine to characterize the clear sky reference
parameter behavior.
7.3.1 Estimate the Reference Parameter for Each Image Using GETREF
11
Program GETREF is designed to read a 10-min ratio image from ExaByte, then divide
that image by the clear sky normalized ratio distribution interpolated to the observed solar
zenith angle from the LLLSKY.DAT tables, and to each individual pixel location from the
adjacent table entries. Ideally, in clear skies, the resulting quotient would be uniform
across the entire image. However, the normalized background values are only a first order
approximation, and the ratio values can be corrupted by clouds, stray light and other ratio
image features that create variations in the quotient. GETREF is designed to examine the
lower part of the quotient distribution where the most reliable reference parameter estimates
can be found. A seven count window is passed across the lower portion of the quotient
histogram distribution, and the window with the greatest percent contribution to the image
from zeniths of 0° to 70° is identified. The estimated reference value is assigned the center
value of the window. Due to the computationally intensive nature of this program, it must
be run on a computer equipped with the PL-1250 accelerator card. An example of a
GETREF run for Columbia Tape 035 is presented below. GETREF accesses the
LLLSKY.DAT file for the normalized background, and the CMPDEC.INP file for the site
azimuth correction. The program can be started after the tape is loaded into the read
ExaByte (drive 0).
To Execute: Enter GETREF at the prompt, i.e.
D:\CMPDEC>GETREF
Screen Output:
Welcome to GETREF - Version 1.2
This program computes the ratio, normalized sky ratio
distribution along with the 45/45 values.
(Ten Minute Full Resolution Images Only)
Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography:
University of California, San Diego
1991 - All Rights Reserved
107

LLLSKY.DAT
10-min
Ratio
Tapes
Compute the Boundary
Layer Reference Value
Estimate for Each
Ratio Image
(GETREF )
CMPDEC.INP
Reference Value File
(LLL###T.R45)
VERSION.LOG
Define Representative
Clear Sky Reference
Value Distributions Daily
(SETREF]
Reference
File
Archive
Tapes
DATES.LST
Daily Clear Sky
Reference Distributions
(LLL###.REF)
Fig. 7-3. Haze loading parameter definition flow diagram.
108
Site #
Location
Site #
Location
o
-
1
MPL
C-Station
HELSTF
Kirtland
China Lake
Malmstrom
Malabar
Columbia
Composite 1
Portable 1
4
n A
China
1
10
|
-> Please enter the site number: 7
** PL Setup Completed **
-> Enter the field tape sequence number (eg, 001):035
Opening Reference File COL0351.R45
Azimuth correction for calibration CAL7001C = -1.0 deg
Reading sky background from file COLSKY.DAT
Sky background read complete
1
1 19/OCT/89 15:50
56.8 159.6
4.0 162.6
4.7 154.0
6.0
1
66 66 25/OCT/89 23:20 91.4 135.9
*** GETREF processing finished ***
5.3
.0
.0 145.0
18.3
0
Stop - Program terminated.
D:\CMPDEC>
A description of the output line format is given below. Img Cnt is the image counter within
GETREF and Rat Seq is the ratio sequence number assigned in TAPRATPL. Points 1 and
2 refer to the two 45°-45° points with Avg being the average of the 5x5 sample and Dev
being the standard deviation. Window refers to the 7 count window used in identifying the
the reference value (Center) and % indicates the percentage of the image area included in the
window. Finally, ND is the neutral density indicator.
GETREF Output Line Explanation
Img Rat
Time Sol Pt. 1 Pt. 2
Window
Cnt Seg Date (GMT) Zen Avg Dev Avg Dev Center % ND
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1 1 19/OCT/89 15:50 56.8 159.6 4.0 162.6 4.7 154.0 6.0 1
These output lines are also copied into the GETREF output file called LLL###cT.R45.
(Normally the c is not present.) A sample from COLO35T.R45 is presented below.
109
Reference File COL035T.R45
Azimuth correction for calibration CAL7001C =
-1.0 deg
in WNA
1 19/OCT/89 15:50
2 19/OCT/89 16:00
3 19/OCT/89 16:10
4 19/OCT/89 16:20
5 19/OCT/89 16:30
56.8 159.6
55.7 165.9
54.6 169.3
53.6 170.4
52.7 173.6
Сллол
4.0 162.6
4.8 159.8
4.3 158.0
8.2 174.2
9.5 158.4
4.7 154.0
4.1 157.0
4.2 158.0
9.8 159.0
7.1 151.0
6.0
6.3
7.5
6.5
3.9
1
1
1
1
1
4
45 19/OCT/89 23:10 88.0 121.8
46 19/OCT/89 23:20 89.9 98.5
1 20/OCT/89 12:20 91.5 110.2
2 20/OCT/ 89 12:30 89.6 103.5
3.3 121.6
6.4 92.4
7.5 .0
2.4 103.3
4.0 115.0
5.6 99.0
.0 115.0
2.6 108.0
2.9 0
7.50
23.0 0
44.7 0
65 65 25/OCT/89 23:10 89.5 112.5
66 66 25/OCT/89 23:20 91.4 135.9
3.5 111.1
5.3 .0
4.0 121.0
0 145.0
31.5
18.3
0
0
7.3.2 Select Representative Clear Sky Reference Values
The next step in the procedure is to select a set of reference values that best defines the
clear sky variation for each day. This is done in a program called SETREF. It is an
interactive routine that plots the GETREF distributions on a color display, and allows the
user to select representative points using the mouse, then stores these representative points
in a output file. Tech. Note 230 provides detailed instructions on how to use the numerous
options in SETREF. A brief description is provided here. To initiate SETREF, not only
are the LLLSKY.DAT and CMPDEC.INP files required, but the VERSION.LOG and
appropriate DATEs.LST files must also be present on the run directory.
To Execute: Enter SETREF at the prompt, i.e.
D:\CMPDEC>SETREF
Screen Output:
Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
SETREF
Version 1.2
19 Jul 91
Sets up reference values used in CMPDECTP
eference V
SU
n CMPDEC
Switch over to RGB if you wish.
At this point, switch to the RGB screen and answer the prompts on the lower part of the
screen. The user then tells the program the site and tape ID information. Plots of the
reference value variation as a function of time can be displayed, and the user can select up
110
to 24 points that describe the clear sky reference value variation for input into the cloud
decision routine. SETREF adjusts the reference values in ND settings 1, 2 and 3 to appear
as if they came from ND 0. This adjustment is made using ratios between the ND 1, 2 and
3 opaque thresholds and the ND O opaque threshold. The percent of image covered by the
GETREF window is also plotted, to aid in the selection process. Higher percentage values
imply greater confidence in the associated reference value. Options allow the user to
substitute values from clear days on completely cloudy days, to superimpose already
selected reference values onto the original plots, to access another tape's files, and to save
current versions of the selected points into a file named LLL###c.REF. The cloud decision
routine will access this file when tape ### is being processed, and will linearly interpolate
the selected reference values to the time of the image being processed. A sample from the
reference file COL035.REF is reproduced below.
SETREF Output Sample
Filename: COLO35.REE
Blue Sky Ratio Background Reference Values
Number of days = 7
ܝܕ ܢ ܚ
808
DAY #1 : 19/OCT/89
Number of Reference Points = 9
Ref. Point Time (min) Ref. Value
746 109.5
82.5
890
72.5
994
69.0
1068
66.5
1170
69.5
1280
73.0
1334
85.0
1400
116.5
ܩ ܟ
ܗ ܢ
ܣ ܩ
DAY #2 : 20/OCT/89
Number of Reference points = 9
Ref. Point Time (min) Ref.Value
746
109.5
1344
1382
97.5
114.5
DAY #7 : 25/OCT/89
Number of Reference points = 7
Ref. Point Time (min) Ref. Value
748
116.0
794
103.5
890
72.5
878
97.0
990
87.5
1268
88.5
1344
97.5
1382 114.5
vous A WWNH
This completes the preparation of the input files for the composite cloud decision
program.
111
8.0
Composite Cloud Decision Image Construction
Every step thus far has focussed on preparing input files for some subsequent
procedure. This section describes the routine that produces the final data product, the
composite cloud decision images. Fig. 8-1 shows the input and output ingredients of the
algorithm that performs the composite cloud decision on the 10-min ratio images,
CPMDЕСТР.
All the necessary parameters have been defined as described in the previous section,
except the thin cloud acceptance level. If the clear sky ratio background for each image was
accurately defined, and the hemispheric dome protecting the lens on the WSI was always
shaded from direct sunlight and kept clean, then one could classify a pixel as thin cloud if
the observed ratio exceeded the clear sky ratio by some small percentage. The current
normalized clear sky background estimates are good only to about 10%. For example, The
lowest normalized ratios in Table 7-2 in the downsun direction (values in the high 70's at
65° zenith and 180° to 210° relative azimuth) are 4-6 counts lower than the average vales in
Table 7-5, because 5/JUL/89 was hazier than most of the other days in the clear sky
sample. Also, the occultor design kept the lens shaded, but not the entire protective dome.
When the dome got dirty or scratched, direct sunlight was scattered into the lens from
unshaded portions of the dome. The problem was further amplified when the sun was not
centered in the occultor for any number of reasons, including incorrect time, not changing
the occultor arm on the right date, misalignment of the arm, and a large azimuth correction.
The ratios in scattered light areas were greater than the surrounding clear sky ratios,
bahaving in a similar manner to thin clouds.
To avoid overdetermining thin cloud regions due to the above mentioned problems, we
forced the observed ratio to exceed the clear sky ratio estimate by 20% before the pixel is
classified as thin cloud. This is equivalent to setting the thin cloud acceptance level to 1.20.
Preliminary runs of CMPDECTP with representative samples can help to further refine this
parameter. All routine processing to this date has used the 1.20 factor.
112

10-min
Ratio
Tape
CMPDEC.INP
LLLSKY.DAT
: Perform Comparisons
Between the Observed and
Clear Sky Background Ratios
and Construct Composite
Cloud Decision Images
(CMPDECTP)
LLL###.REF
CCv •
Archive
Composite Cloud
Cover Information
( LLL###Teccv )
10-min
Cloud
Decision
Tape
Tapes
Fig. 8-1. Composite cloud decision algorithm input
and output specifications
113
8.1
Description of CMPDECTP
CMPDECTP is more complex that its fixed threshold counterpart, CLDDECM. The
steps performed in the composite algorithm are described below. Steps 5 to 7 use the PL
processor.
Step 1: Find the next 10-min ratio image on the input tape.
This image in transferred to the FG-100 board and important header information extracted.
Step 2: Compute the solar position for this image.
Step 3:
Interpolate the normalized ratio tables to the observed solar zenith angle.
The LLLSKY.DAT tables are available in 5° solar zenith angle increments. A linear
interpolation of the tables to the image solar zenith angle is performed.
Step 4:
Interpolate between the reference values in time to the image time.
The LLL###c.REF file contains selected reference values for each date on the tape.
Interpolation in time is used to estimate the reference value at the time of the image. A
reverse ND adjustment to that performed in SETREF is applied to the reference value if the
image used a no-zero ND setting.
Step 5:
Compute the clear sky background ratio at each active pixel.
Bilinear interpolation is used to interpolate the normalized ratios from the OBSBETA grid
to the zenith-azimuth position of the pixel. The resulting normalized value is multiplied by
the ND-adjusted reference value, yielding the estimated clear sky ratio value.
Step 6: Construct the cloud decision image.
Each pixel is assigned a cloud decision value based on comparisons between its observed
ratio (RATIO), the clear sky ratio estimate (CLEAR), the appropriate opaque threshold
(OPAQUE), and the thin cloud acceptance level (ACCEPT). The cloud decision byte
values (DECIS) are assigned in the manner illustrated in Table 8-1.
Step 7: Determine % of image covered by each category and sky coverages using an
image histogram.
Step 8: Update the header information and LLL###cT.CCV file, and transfer the image to
the output ExaByte.
A dummy DOS header is written to tape before each image. EOF's are written at the end of
each day.
114
CMPDECTP has several hotkey options, similar to those in TAPRATPL. The
CMPDECTP image tape format and header information are described in Tech. Memo
AV91-031t.
Table 8-1. Composite Decision Categories
Category
Description
Decision
Criteria
Decision
Values
Mathematical
Expressions
No Data
RATIO = 0
Off Scale Bright
RATIO = 240
201
Indeterminate
CLEAR >= OPAQUE
202
Opaque Cloud
RATIO >= OPAQUE
140-200
100-139
DECIS = 100(RATIO/OPAQUE) + 40
DECIS = 40 (RATIO/CLEAR) /ACCEPT + 60
Thin Cloud
RATIO/CLEAR >= ACCEPT
Clear
RATIO/CLEAR < ACCEPT
1-99
DECIS = 100 (RATIO/OPAQUE)
115
8.2
Running CMPDECTP
After all the CMPDECTP input file information is prepared, the 10-min ratio tape
being processed to cloud decision is loaded into the read ExaByte (drive (), a blank tape is
loaded into the write ExaByte (drive 1), and program CMPDECTP executed. The
following example is for Tape 035 from Columbia MO.
To Execute: Enter CMPDECTP at the prompt, i.e.
D:\CMPDEC>CMPDECTP
Screen Output:
Welcome to CMPDECTP - Version 1.4
This program makes the cloud/no cloud decision
after correcting for scattering and zenith angles.
The corrected images are written to ExaByte
(Ten Minute Full Resolution Images Only)
Marine Physical Laboratory - Optical Systems Group
Scripps Institute of Oceanography
University of California, San Diego
1991 - All Rights Reserved
Site #
*
Location
1
Site #
Location
6
uns WNHO i
MPL
C-Station
HELSTF
Kirtland
China Lake
Malmstrom
Ovo I
Malabar
Columbia
Composite 1
Portable 1
1
11
-> Please enter the site number: 7
-> Enter the field tape sequence number (eg, 001):035
** PL Setup Completed **
Note: You have several options which may be changed during a run.
Options may be changed by entering the appropriate
keystroke,
without pressing the ENTER key. All of these options take
extra time and are intended only for occasional use during
a run to visually check the decision images.
The Key enter options are:
S to change Show Color flag
D to change cover Display flag
I to change Image save flag
X exit after current image
? to print current values of flags
Any other key will cause the available options to be printed
116
-> Press ENTER to continue
Opening Cloud Cover File COL035T.CCV
Azimuth correction for calibration CAL7001C.RAT = -1.0 deg
Opaque Thresholds: ND (0) = 145. ND (1) = 135.
ND (2) = 153. ND (3) = 153.
Thin Cloud Acceptance Level = 1.20
Ratio reference tables successfully extracted from COLO35.REE
Writing initial EoF to target tape ***
Image Cover
Sky Cover
cld Rat
Off (%) Off Tenths
Seq Seg Date Time Drk Clr Thn Opq Brt Ind Opq tot
Reading sky background from file COLSKY.DAT
Sky background read complete'
1 1 19/0CT/89 15:50 10 0 0 89 0 0 10 10
46 images processed for this date.
0
1
2
CLOUD COVER SUMMARY
Cloud Cover (Tenths)
2 3 4 5 6
0 0 0 0 1
0 1 0 1 0
1
0
0
Total
Opaque
O
7
0
1
8
0
1
9
1
4
10
43
36
Image Cover
Sky Cover
cid Rat
Off (%) Off
Tenths
Seq Seg Date Time Drk Clr Thn Opq Brt Ind Opq Tot
i 1 20/OCT/89 12:20 6 80 4 2 0 8 0 1
O
Image Cover
Sky Cover
cld Rat
Off (%) Off Tenths
Seq Seg Date Time Drk Clr Thn Opq Brt Ind Opq Tot
1 1 25/OCT/89 12:30 5 56 18 12 0 9 1
*** CMPDECTP processing finished ***
Writing final EOFs to output tape.
66 images processed for this date.
0
17
30
Total
Opaque
CLOUD COVER SUMMARY
Cloud Cover (Tenths)
2 3 4 5 6
9 8 8 6 4
16 3 0 0 0
1
14
17
7
0
0
oooo
9
0
0
10
0
0
117
The 6 image cover and 2 sky cover column headers are reasonably self-explanatory.
The off scale dark and indeterminate category percentages are not included in the tenths of
sky cover computations. Off scale bright is presently included in the opaque category
during the sky cover computations. Excerpts from the cover file (LLL###cT.CCV) for our
example are listed below.
Cover File COL035T.CCV
Azimuth correction for calibration CAL7001C = -1.0 deg
Opaque Thresholds: ND (0) = 145. ND (1) = 135.
ND (2) = 153. ND (3) = 153.
Thin Cloud Acceptance Level = 1.20
Ratio reference tables successfully extracted from COL035.REF
Image Cover
Sky Cover
cid Rat
Off (%) Off
Tenths
Seq Seg Date Time Drk Clr Thn Opq Brt Ind Opg Tot
1 1 19/OCT/89 15:50 10 0 0 89 0 0 10 10
2 2 19/OCT/89 16:00 10 0 0 90 0 0 10 10
45
46
45 19/OCT/89 23:10
46 19/OCT/89 23:20
6
33
31 57
64 0
1
0
0
0
4
2
0
0
6
0
46 images processed for this date.
0
1
2
Total
Opaque
CLOUD COVER SUMMARY
Cloud Cover (Tenths)
2 3 4 5 6
0 0 0 0 1
0 1 0 1 0
1
0
0
O ON
O OD
7
0
1
8
0
1
9 10
1 43
4 36
WO
و هم
4
(%)
Image Cover
Sky Cover
Cid Rat
Off
Off
Tenths
Seq Seg Date Time Drk clr Thn Opq Brt Ind Opq Tot
1 1 20/0CT/89 12:20 6 80 4 2 0 8 0 1
2 2 20/0CT/89 12:30 5 86 0 4 0 6 0 0
65 65 25/OCT/89 23:10 4 82 0
66 66 25/OCT/89 23:20 4 33 32
*** CMPDECTP processing finished ***
6
20
0
0
8
11
1
2
1
6
66 images processed for this date.
0
17
30
Total
Opaque
CLOUD COVER SUMMARY
cloud Cover (Tenths)
2 3 4 5 6
9 8 8 6 4
16 3 0 0 0
1
14
17
7
0
0
8
0
0
9
0
0
10
0
0
118
9.0
Decision Tape Stacking Procedures
As mentioned in the previous section, the cloud decision tapes coniprise the WSI data
set of greatest interest to most outside users. It was important to protect against loss of
these tapes, and at the same time, reduce the number of tapes being handled. The ratio and
decision processing steps compress the volume of image information for a given tape from
4 images per sample for both the 10-min and 1-min resolutions on the original raw radiance
tapes, to 2 images (ratio and corresponding dark red) for either the 10-min or 1-min
resolutions on the ratio tapes, to finally 1 cloud decision image at either resolution. Thus,
the cloud decision image tape would contain about one-eighth the amount of data on the
original field tape. To make the data handling less cumbersome, we decided to stack 1
month of cloud decision data onto a single tape.
The stacking procedure in outlined briefly in Fig. 9-1. Two computer algorithms are
used: EXCOPY to copy the data from a cloud decision master tape, and EXQC that
evaluates the stacked 1-min decision copies and produces a summary of the information on
the stacked tape. Only brief descriptions of how these algorithms work is provided here.
More detailed information concerning the output from these routines is provided in Tech.
Note 230.
119

10-minute
Processing
Copy Diagnostic
Output Files
(LLL###.CPY)
1-minute
Processing
ee
10-min
Cloud
Decision
Tapes
Stack Cloud
Decision Tape Copies
(EXCOPY]
1-min
Cloud
Decision
Tapes
-
-
-
-
Stacked
10-min
Cloud Dec.
Tape
Stacked
1-min
Cloud Dec.
Tape
-
-
-
-
-
-
Check the Stacked
1-min Cloud
Decision Tapes
( EXQC)
-
-
-
-
-
Tape Summary
File (*.SUM)
-
-
Fig. 9-1. Decision tape stacking procedures.
120
9.1
The Tape Copying Program - EXCOPY
EXCOPY is general program that copies the data available from a tape loaded on the
read ExaByte (drive 0) directly onto a tape loaded on the write ExaByte (drive 1).
EXCOPY is used to back up both the original FIELD tapes (as noted in Section 2.5) and
the cloud decision tapes. The following example shows how the EXCOPY program is
executed.
To Execute: Enter EXCOPY at the prompt, i.e.
D:\COPY>EXCOPY
Screen Output:
EXCOPY
version 2.
5
03 Jun 90
Enter a filename, usually the source tape ID,
up to 8 characters long: COLO04
Filename entered: COLO04.CPY
-- Be certain read (source) tape is in drive 0 and is -WRITE PROTE-
CTED-.
--
and load -WRITE ENABLED - (target) tape in drive 1.
NOTE: In this version of excopy, filemarks ARE
copied from the source tape. Regardless of source
tape contents a filemark is always written at the
beginning and at the end of data on the target tape.
4/4/90:
ALL READ MEDIA ERRORS ARE IGNORED FOR TESTING.
Hit any key to start copying...
The format of the subsequent screen output and the information written to the *.CPY file is
described in Tech. Note 230.
In creating a stacked copy, multiple runs of EXCOPY are made, one for each of the
original tapes being copied. For example, cloud decision tapes COL004, COL005,
COL006 and COL007 contain cloud decision images from 2/MAR/89 through 29/MAR/89.
These tapes were combined to create the March stacked tape from Columbia. Four
consecutive runs of EXCOPY were made, with one of the 4 original tapes being loaded
into the read ExaByte before each run. The write ExaByte was not touched between runs,
allowing the copy from a subsequent run to be written on tape immediately following the
previous run's output. The final stacked copy would then be labelled COL004-COL007
cloud decisions, either 1-min or 10-min. In the case of the 1-min preliminary TASC data
release, the stacked tape was also copied with the copy being provided to TASC.
121
9.2 Stacked Decision Tape Quality Evaluation
Another routine called EXQC was developed to provide a final quality control check
on the 1-min fixed threshold cloud decision tapes sent to TASC. This program performed
two functions. First, by reading the entire tape, it could identify problems on the tape,
particularly media errors. If a significant problem was encountered, the copy could be
rerun and checked again. Second, EXQC evaluated the header information of all the
images on the tape, and summaries of the information from each day on the tape were
produced in the *.SUM files. EXQC can be run from a tape on either the read or write
ExaBytes, and is started in the following manner.
To Execute: Enter EXQC at the prompt, i.e.
D:COPY\>EXQC
Screen output:
EXQC
Version 1.0
2 May 90
Enter the drive # (0 - 1) to be read from: 0
Ente
com:
EXQC creates a detailed .TQC logfile and a .SUM summary.
Enter a filename for the output log and summary file,
usually the source tape ID, up to 8 characters long:
COL0389
Filename entered: COL0389.SUM
Hit any key to start reading...
Table 9-1 taken from Tech. Memo AV90-72t shows the summary information produced by
EXQC for the stacked March 1-min decision tape from Columbia, MO.
At the time of this note, a version of EXQC that would work on the full resolution, 10-
min data sets had not yet been developed.
122
One-Minute Cloud Decision Tape Summary
TAPE: COL004-007 (Copy) produced May 1990
Summary generated 18 May 1990
Oooo
START END SOURCE FIELD #CLOUD SEQ# TIME PARTIAL DATAQUAL
DATE TIME TIME TAPEID IMAGES IMAGES GAPS GAPS IMAGES VALUES COMMENTS
2/MAR/89 12:36 00:06 004 691 691 0
0 1 0000000
3/MAR/89 12:34 00:07 004 694 694
0 1 0000000
4/MAR/89 12:33 00:08 004 696 696
1 0000000
5/MAR/89 12:31 00:09 004 699 699 0 0
1 0000000
6/MAR/89 12:30 00:10 004 701 701 0 0
1 0000000
7/MAR/89 12:29 00:11 004 703 703 0 0
1 0000000
8/MAR/89 12:27 00:10 004 704 704
1. 0000000
9/MAR/89 12:25 00:08 005
704
704
1 0000000
10/MAR/89 12:24 00:07 005 704 704
0 1 0000000
11/MAR/89 12:22 00:15 005 714
714
1 0000000
12/MAR/89 12:21 00:16 005 716 716
1 0000000
13/MAR/89 12:20 00:17 005 718 718
1 0000000
14/MAR/89 12:20 00:18 005 719 719
0 1 0000000
15/MAR/89 12:20 00:19 005 720 720
1 0000000
16/MAR/89 12:20 00:19 006 719 719
1 0000000
17/MAR/89 12:20 00:19 006 719 719
1 0000000
18/MAR/89 12:20 00:19 006 720 720 0 0
1 0000000
19/MAR/89 12:20 00:19 006
720
1 0000000
20/MAR/89 12:20 00:19 006 720 720 0 0
1 0000000
21/MAR/89 12:20 00:19 006
720
720 0 0
1 0000000
22/MAR/89 12:20 17:33 006 314 314
1 0000000 . 1
22/MAR/89 17:34 00:19 006 710 406 0 0
1 0000000 1
23/MAR/89 12:20 00:19 007 720 720
1 0000000
24/MAR/89 12:20 00:19 007 720 720 0 0
1 0000000
25/MAR/89 12:20 00:19 007 720 720 0 0
1 0000000
26/MAR/89 12:20 00:19 007 720
720
1 0000000
27/MAR/89 12:20 00:19 007 720 720
1 0000000
28/MAR/89 12:20 00:19 007 720 720
0 1 0000000
29/MAR/89 12:20 00:19 007 720 720 0 0 0 1 0000000
oooooooooooooooooooooooooo
ooooooooooooo hoo ooooooooooo
720
Unless otherwise indicated, time gaps and date gaps were either missing from
the original field tapes, or were inaccessible during the ratio processing.
Footnotes:
1 - the image sequence numbers backtrack by 10 between 17:33 and 17:34
on 22/MAR/89. This happens occasionally when multiple runs are
needed to process a single field tape to ratio. There are actually
720 images available on this day, as indicated by the sum of the
number of cloud images.
Table 1
Sample stacked l-min cloud decision tape summary
for March 1989 at Columbia, MO
123
ACKNOWLEDGEMENTS
The work presented in this note was sponsored by the Geophysics Directorate of the
Phillips Laboratory of the Air Force Systems Command under Contract No. F19628-88-K-
0005. I wish to thank Mr. Donald Grantham for his continued support and encouragement
of our efforts. The development of the computer algorithms described in this report was a
team effort. Thanks to those involved, including Janet Shields, Gene Zawadzki, Monette
Karr, John Malo, Peter Pak and John Fox. Monette and Janet developed the initial quality
control procedure, and Janet organized the calibration check in and data reduction
procedures. Melissa Ciandro played a crucial role not only in the development and
implementation of the processing procedures, but also wrote many of the algorithm run
descriptions used in this note. Special thanks are due to Carole Robb and Melissa for their
publications support. My final thanks go to Richard Johnson for his supervision of the
entire project and his steadfast encouragement during the development and documentation
of these procedures.
124
REFERENCES
The bulk of the references in this note are to Technical Memos circulated within the
Optical Systems Group of the Marine Physical Laboratory, and to Technical Notes that
have a wider, yet limited circulation. References in the text to these nonstandard sources
are made using the OSG identifiers. Titles and authors of the Notes and Memos referred to
in this note are listed below, preceded by the only standard reference to a CIDOS-91
presentation.
Standard Reference
Koehler, T. L., R. W. Johnson and J. E. Shields, 1991: Status of the whole sky imager
database. Cloud Impacts on DoD Operations and Systems, 1991 Conference, El
Segundo, CA, Phillips Laboratory, Air Force Systems Command, in press.
Optical Systems Group - Technical Notes
Tech. Note 226
"WSI Data Base Summary (Status as of 31 Dec 1990)" by R. W. Johnson, J. E.
Shields and T. L. Koehler, Feb 1991.
Tech. Note 228
"WSI Data Base Catalog" by M. L. Ciandro, Apr 1991.
Tech. Note 230
"WSI Processing Manual" by M. L. Ciandro, Oct 1991.
Optical Systems Group - Technical Memos
OS
AV88-0430
"READTPQC -- Tape Read Program" by J. A. Malo, 22 Aug 1988.
AV88-046t
"Tape Check-in Procedures" by M. E. Karr, 24 Aug 1988.
AV88-065t
"Cloud Program Data Handling Overview" by J. E. Shields, 16 Nov 1988.
AV89-028t
"Field Unit Hardware Summary, January & February 89" by J. E. Shields,
8 Mar 1989.
1
AV89-0567
"Software Documentation: Linearity Processing" by J. E. Shields, 2 Jun 1989.
AV89-0577
"Software Documentation: Filter Program" by J. E. Shields, 2 Jun 1989.
125
AV89-0587
"Software Documentation: Absolute Processing" by J. E. Shields, 5 Jun 1989.
AV89-0741
"Oriel Spectral Filter Curves" by J. E. Shields, 17 Jul 1989.
AV89-082
"Format of the One Minute Cloud Image Tapes" by J. E. Shields, 1 Aug 1989.
AV90-026t
"Calibration Log Update" by J. E. Shields, 5 Mar 1990,
AV90-042t
"New Ratio & One-Minute Cloud Decision Image Header Format" by
T. L. Koehler, 4 Apr 1990.
AV90-049
"Derivation of Calibration Constants" by J. E. Shields, 18 Apr 1990.
AV90-057t
"Format of the Ten Minute Cloud Decision Image Tapes" by T. L. Koehler,
3 May 1990.
AV90-070t
"Field Unit Summary, March and April 1990" by J. E. Shields, 16 May 1990.
AV90-072t
"Preliminary Data Release: One Minute Cloud/No Cloud Tapes, Supplemental
May 90 Release" by M. L. Ciandro, 30 May 1990.
AV90-123t
"Format of Ratio Tapes: One and Ten-Minute" by J. E. Shields, 13 Sep 1990.
AV91-031t
"CMPDEC Image and Header Format", by T. L. Koehler, 18 Jun 1991.
126