UNIVERSITY OF CALIFORNIA, SAN DIEGO
UC SAN DIEGO LIBRARY
3 1822 04429 7471
OPTICAL SYSTEMS GROUP
TECHNICAL NOTE NO. 221
August 1990
Offsite
(Annex-Jo
rinals)
QC
974.5
..143
no. 221
THE WHOLE SKY IMAGER/LIDAR
INTERCOMPARISON EXPERIMENT
Janet E. Shields
Richard W. Johnson
Thomas 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 Air Force Geophysics Laboratory
CiTV
)
VNS
Contract Monitor, Dr. J. W. Snow
Atmospheric Sciences Division
ce
:
FORN
IGHT
*1868
NUVOLOUR
Prepared for
The Geophysics 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
mersin
w
ww............
www
UNIVERSITY OF CALIFORNIA, SAN DIEGO
3 1822 04429 7471
Sale
KOHA
TABLE OF CONTENTS
A
Introduction..
1.1 The Whole Sky Imager ......
1.2 Impetus for the WSI/Lidar Experiment ........
..
2.
Data Acquisition .....
2.1 Hardware Overview..
2.2 System Calibration.
2.3 Field Data Acquired.
0
1
3.
PIE
Preliminary Analysis and Processing....
3.1 Sample Field Imagery......
3.2 Data Quality .........
3.2.1 Tape QC....
3.2.2 Special Calibrations ......
3.2.3 Standard Camera Calibrations .
3.2.4 Final LAN Image Analysis....
3.3 Ratio Processing.........
ooouuuA AWNN
4.
Evaluation of WSI and LIDAR Data Intercomparison.......
4.1 Preliminary Evaluation.
4.2 Application of Data to Beta Algorithm Development.
4.3 Evaluation of Beta Cloud Data...
5.
Summary..
6.
References ......
..............
Appendix 1:
Appendix 2:
Appendix 3:
Tape QC Summary for WSI/Lidar Experiment.
Sample TAPRAT Diagnostic File: Tape POR001.................
Illustrative Charts: ....
Phase 1:
Preliminary Data Processing......
Phase II:
Analysis of WSI Data....
Red/Blue Ratio as a function of Solar Scattering Angle, Beta..........
Samples from Geometric Calibration Images....................
LIST OF FIGURES
Fig. #
Figure Title
2.1 Whole Sky Imager Camera Assembly....
2.2 Whole Sky Imager Hardware Block Diagram
2.3 Portable Whole Sky Imager at Madison........
VladISOI.............................................
3.1 WSI Imagery from 1 December 1989
Sample red images are shown at 10 minute intervals ....
3.2 WSI Imagery from 5 December 1989
Sample red images are shown at 10 minute intervals .......
3.3 WSI Imagery from 1 December 1989
Sample red images, reproduced to retain high brightness detail......
3.4 WSI Basic Image Processing Flow Chart...
3.5
Sample Cloud/No Cloud Decision Images from 1 December 1989
Image processed using fixed ratio threshold.......
3.6 Sample Cloud/No Cloud Decision Images from 5 December 1989
Image processed using fixed ratio threshold..........
1. INTRODUCTION
This document describes the use of the Whole Sky
Imager(WSI) in a joint experimental program involving
the WSI and two lidar systems at the University of
Wisconsin. The Whole Sky Imager has acquired an
extensive data base of ground based digital imagery
which is being used to provide cloud field assessment.
The joint WSI/idar experiment was designed to pro-
vide information pertaining to the optical depth of the
clouds, for use in our cloud algorithm development and
assessment.
magnitude of the calibrated red/blue radiance ratio.
The assessment identifies the sky dome (in the direc-
tion corresponding to that pixel location) as clear,
optically thin cloud or optically thick cloud.
At the present time, this distinction is based on a
ratio threshold; that is, the calibrated red/blue ratio is
computed from the raw imagery, and thresholded to
yield a cloud discrimination. The intent of this dis-
crimination is that transparent clouds should be iden-
tified as thin, and opaque as thick; in general most of a
cirrus cloud field will be identified as thin, and a
cumulus field will be identified as thick.
The WSI system was developed and installed by the
Marine Physical Laboratory at Scripps Institution of
Oceanography, University of California San Diego.
The lidar systems, the Volume Imaging Lidar (VIL)
and the High Spectral Resolution Lidar (HSRL), were
developed and operated by Edwin Eloranta and
Christian Grund, of the University of Wisconsin at
Madison.
1.1 The Whole Sky Imager
The Whole Sky Imager is an automated digital
imaging system designed for use in high spatial and
temporal resolution cloud assessment studies (ref.
Johnson, 1989). This ground-based electronic imag-
ing system monitors the upper hemisphere every minute.
It is a passive, i.e., non-emissive system, which acquires
calibrated multi-spectral images of the sky dome. The
system is an automated unit operating under micro-
computer control, gathering digital imagery suitable
for automated processing and analysis of cloud fields.
The resulting cloud discrimination is in general
quite good. The results compare well with standard
observers (ref. Shields, 1990), yet they are in general
much more consistent overtime than could be obtained
with standard observers. The further advantages of
high temporal frequency and spatial resolution as well
as data base size make this an outstanding and unique
data set, for several applications.
There is considerable interest however in more
specific discrimination of the thin clouds. In particular,
we would like to determine the approximate range of
optical depth values corresponding to the thin cloud
category and the thick cloud category. This was the
initial impetus forthe lidar tests. By taking simultaneous
data with the WSI and the HSRL, which yields optical
depth as well as other related parameters, we could
begin to test and improve our techniques, and finally
quantify the relation between ourcloud discriminations
and the optical depth.
These data are archived and subsequently converted
to cloud/no cloud decision images. Data are acquired
in 512 x 480 format, which yields 1/3 degree spatial
resolution. Four digital images are acquired every
minute, saved at full resolution every 10 minutes, and
saved at reduced resolution every minute, for a total of
1.2 gigabytes of data archived per week.
The WSI systems have been operating at several
locations during the last two years, acquiring data 12
hours a day. These units are collecting a cloud database
which can be utilized for a variety of applications
including extraction of predictive cloud free line of
sight (CFLOS) and cloud free arc (CFARC) statistics,
and cloud model evaluation and development.
1.2 Impetus for the WSI/Lidar Experiment
With our preliminary fixed threshold algorithm, we
do not expect a one for one relation between optical
depth and identified thin cloud at all points within the
sky dome. This is because there is a small directional
dependence inherent in the clear sky red/blue ratio,
caused by the directional properties of Mie scattering,
which causes some directional bias to the cloud dis-
crimination. Whereas the fixed threshold algorithm is
quite adequate for discrimination of two classes of
clouds, creating an accurate map of optical density is a
much more demanding task, which may require a
correction for the directional dependencies.
We are currently developing a second generation
cloud algorithm, which predicts the small variations in
clear day ratio as a function of scattering angle from the
sun (beta), and look angle. The "beta-algorithm”,
based on the Hering FASCAT radiance model, will be
tested in a number of ways including evaluation of
In generating the cloud decision image, an automated
assessment is made at each pixel location, based on the

sun occultor may be seen slightly to its right. The
occultor shades the unit from direct sunlight, thus
providing a large measure of stray light control.
clear day ratio images. The lidar data should be an
invaluable aid in the testing, allowing us to evaluate
known thin clouds in addition to clear skies.
In particular, we would like to develop our algorithm
to the point that we can, from a specific lidar return(i.e.,
a measured value of optical depth at the zenith) accu-
rately determine the optical depth distribution over the
sky dome. Applying these potential results to the full
WSI data base, processed with the beta model, should
allow us to make our cloud discriminations correspond
to specific values of optical density. The extent to
which we can do this obviously depends both on the
algorithm development and on the subsequent WSI
lidar intercomparison.
The next section of this report documents the ac-
quisition of the WSI/lidar test data. This is followed by
an overview of the analysis and processing which has
occurred to date. The last section discusses the future
application of the data to the issues addressed above.
2. DATA ACQUISITION
This section gives an overview of the WSI system
hardware which was fielded, the calibrations acquired
for the unit, and the extent of the field data acquisition.
2.1 Hardware Overview
Fig. 2.1 Whole Sky Imager Camera Assembly
GE 2710
SOLID STATE
VIDEO
CAMERA
SONY PVM 1271 0
MONITOR
TMI COMPUTER
(IBMAT CLONE)
The WSI hardware is described in more detail in
Johnson, 1989, and Shields, 1990. The WSI is an
automated system designed for archival of cloud field
imagery at one-minute intervals. The sky is viewed
through a series of spectral and neutral density filters
using a fisheye lens to acquire most of the upper
hemisphere. The sensor is a fixed-gain solid state CID
(charge injection device) camera. The sensor is shown
in Fig. 2.1.
The WSI is controlled by an IBM AT-class micro-
computer. This fully automated system acquires four
digital images every minute, at 512 x 480 resolution.
Images are grabbed using a 1024 x 1024 imaging
board, and archived using a 2 gigabyte capacity tape
drive. The hardware components are shown sche-
matically in Fig. 2.2.
For the WSI/lidar experiment, the normal environ-
mental housing was redesigned to provide a fully
portable sensor unit, which included environmental
temperature control. This portable housing, shown in
Fig. 2.3, was designed so that the unit could be shipped
with the sensor in place, and the legs added and the unit
leveled on arrival. In this figure, the dome which
protects the lens may be seen at the top, and the black
AUTOMATIC
EQUATORIAL
SOLAR
OCCULTOR
ASSY.
VIDEO
IMAGE
PROCESSING
SUB - SYSTEM
(ITI FG 100)
ARCHIVAL
1/0
SUB - SYSTEM
(SEAGATE
65 Mbyte H.D.)
REMOTE
CONTROLLED
IRIS
ANALOG
ASSY.
EXABYTE EXB - 8200
2.2 Gbyte 8 mm CATRIDGE TAPE SYSTEM
REMOTE
CONTROLLED
OPTICAL
FILTER
ASSY.
ACCESSORY
CONTROL PANEL
STOWED KEYBOARD
EXTERIOR SENSOR
INSTALLATION
INTERIOR CONTROLLER
INSTALLATION
Fig. 2.2 Whole Sky Imager Hardware Block Diagram

.
...........
.................
Fig. 2.3 Portable Whole Sky Imager at Madison
2.2 System Calibration
The WSI camera systems are calibrated against
standards of radiant intensity traceable to N.B.S. using
standard radiometric procedures in association with
optical calibration facilities established at the Marine
Physical Laboratory. A three-meter bench, in con-
junction with a selection of standard lamps and cali-
bration targets, is used to characterize both the radio-
metric performance and geometric image characteristics
of the WSI. These calibrations, which normally require
portions of several days to complete, are outlined
below.
Electronic Calibration
The solid state sensor electronic calibration is
documented in Varah, 1989. Unlike the remainder of
the calibrations, which characterize the unit, this cali-
bration is an interactive procedure in which data are
acquired and the sensor is adjusted to optimize per-
formance. The characteristics which are evaluated and
optimized in this calibration are: noise characteristics,
array uniformity, dynamic range, sensitivity, set point
balancing, and overall performance of the interlaced
RS-170 video images.
Linearity Calibration
In the ideal CID sensor, the system output or signal
would be linearly related to the input radiance. Because
this ideal linear performance is not precisely obtained
in the sensors, it is necessary to measure and charac-
terize this relative system response. The linearity
calibration is a relative (not absolute) radiance cali-
bration which measures the relation between changes
in input radiance and changes in output signal. Non-
uniformity of the array may also be extracted from the
measurements obtained during this calibration.
Absolute Calibration at selected neutral density and
aperture settings
Three sets of these calibrations are normally ob-
tained; the primary calibration in neutral density 1, a
back-up neutral density 1 calibration at another aperture,
and the primary calibration in neutral density 2. Each
of these three sets of calibrations consists of 8-10
measurements taken in each of 4 spectral filters. These
are the measurements that give us the absolute radiance
corresponding to given signals in each spectral filter.
From these absolute measurements, one can also de-
termine the relative response of the system in the
different spectral bands, caused primarily by the
variations in spectral filter bandwidth and the spectral
response characteristics of the sensor chip.
Absolute Calibration vs Neutral Density
In this calibration, the density of each neutral den-
sity (ND) filter is measured in each of the spectral filter
bands. It is important for us to know both the actual
density of each ND filter, and the spectral variation in
that density.
Absolute Calibration vs Aperture
In this calibration the effects of variations in the
aperture settings are quantized. Although this does not
affect the calibrated ratio computation, it is required for
the generation of absolute radiance.
Edge Calibration
This calibration determines the physical location of
the image edges. The use of interference filters means
that there are slight differences in the image size and
placement for the various filters. These differences are
quantized in this calibration, so that the ratio compu-
tation can utilize the appropriately corrected pixel
location in each filter.
Flux Calibration
This calibration determines the necessary constants
for use by the flux control portion of the field program.
The flux control is designed to choose the optimal
neutral density/aperture combination for obtaining on-
scale data in the field.
Geometric Calibration
This calibration supplies the required imagery to
generate the equations relating pixel location in image
3
space to direction (azimuth and zenith angle) in object
space.
The above calibrations are the primary calibrations,
acquired both before and after the deployment. They
are processed and analyzed and then applied to the ratio
processing. Special calibration tests may also be
acquired as necessary.
For the Wisconsin test, the Pre deployment cali-
brations were acquired during November 7 - 11, 1989.
The Post deployment calibrations were acquired dur-
ing January 31 - February 9, 1990.
2.3 Field Data Acquired
During this period, the HSRL was operated during
the following intervals.
Date Operation Times
20 Nov 12:43 - 15:15
21 Nov 07:23 - 13:32
28 Nov 17:01 - 23:59
29 Nov 00:00 - 03:00
01 Dec 07:47 - 19:10
05 Dec 07:39 - 09:48,
12:07 - 18:04
06 Dec 07:05 - 07:48,
07:54 - 08:14,
08:17 - 14:07
Thus only tapes POR001-004 are directly applicable
for lidar intercomparison studies. The U.W. personnel
have indicated that both 1 and 5 December had par-
ticularly favorable cirrus cloud conditions.
The WSI unit was installed on November 14, 1989.
Installation of the portable unit proved to be straight-
forward. The unit was installed on the roof of the Van
Hise building at the University of Wisconsin, Madison,
by the Marine Physical Laboratory personnel, with the
very helpful support of U. W. personnel. The unit was
left in an automated operation mode; in this configu-
ration, it ran without supervision 12 hours a day,
requiring weekly change out of the archival tape. This
changeout and routine inspection were provided
courtesy of the U.W. personnel.
3. PRELIMINARY ANALYSIS AND PRO-
CESSING
Eleven data tapes were acquired, containing the
dates listed below.
Funding was available to build and field the WSI
unit at the lidar site. This has been accomplished
successfully. Although only marginal funds were
available for subsequent analysis and interpretation,
the experimental potential of the joint exercise was
recognized by both parties concerned. Therefore
preliminary analysis of the WSI data has proceeded,
and evaluation of the data intercomparison is anticipated
once the lidar data are available.
Tape Ident
POR001
POR002
POR003
POR004
POR005
POR006
POR007
POR008
POR009
POR010
POR011
Operation Dates
14 Nov - 20 Nov
28 Nov - 02 Dec
21 Nov - 27 Nov
04 Dec - 10 Dec
11 Dec - 17 Dec
18 Dec - 24 Dec
26 Dec - 01 Jan
02 Jan - 08 Jan
09 Jan - 15 Jan
17 Jan - 23 Jan
25 Jan - 26 Jan
Just as the lidar data, with its accurate determinations
of optical depth, should be quite useful in the analysis
of the WSI data base, there is interest in applying the
WSI data at U.W. That is, the WSI data base, with its
full sky representation and high temporal resolution,
may help U.W.evaluate certain aspects of the lidar data
collection.
With this in mind, there has already been some
transfer of WSI data. MPL has sent a 9-track tape to
U.W. (since they had not yet acquired Exabyte capa-
bility), with sample images from two days. The images
sent were raw data only, appropriate for qualitative
evaluation of the general cloud field during data ac-
quisition. Being unprocessed data, they were not
appropriate for direct comparison with optical density,
however they give the user a general feel for the
character of the sky and the degree of variability in the
scene. It should be emphasized however that these data
have not had calibrations applied, and should not be
On each day, data are normally acquired for the 12
hours surrounding local apparent noon. The 68 days on
which data were acquired represent approximately
215,000 images, or 11.6 gigabytes of data. At the end
of each 7 day data period, the system remains in
standby until the tape is changed. On a few occasions,
it was not possible for the U.W. personnel to change the
tape immediately. As a consequence, the following
data days were not acquired: 03 Dec, 25 Dec, 16 Jan,
and 24 Jan.
Both image sets illustrate that the cloud structure
can change quickly, and have highly variable cloud
thicknesses over the sky dome. Thus, it will be extremely
important to obtain careful time and space registration
in the data intercomparisons with the lidar. An accu-
rate geometric calibration will be essential for this
registration.
3.2 Data Quality
19
used in any quantitative analysis.
Similarly, U.W. personnel have indicated a willing-
ness to send lidardata to MPL, once they have completed
their calibration and quality analysis. Until this data is
transferred, it is not possible to proceed with the
evaluations. In the mean time, however, progress has
been made with the processing and quality analysis of
the WSI data. The processing which has occurred to
date (August 90) is discussed in this section.
3.1 Sample Field Imagery
Since U.W. personnel indicated that there were
particularly interesting cloud fields on 1 and 5 De-
cember, sample images were extracted from these
days. A series of 4 images acquired at 10 minute
intervals on 1 December is shown in Fig. 3.1. These
images were acquired at 1910 through 1940 Greenwich
(Z) time. Fig. 3.2 shows a similar series, acquired from
1720 through 1750 Z on 5 December. In each case, a
650 nm red image is shown.
The images in Fig. 3.1 show that cirrus clouds were
present and obvious in the imagery. During this 30
minute period, the cloud densities and amounts are
quite variable over the sky, and they appear to change
significantly with time. The first image shows cirrus
over much of the sky. In the next image shown, the
cirrus covers less of the sky, and also appears to be less
dense. The contrail of an aircraft proceeding to the
west in the image was visible in the original imagery.
In the following image, this contrail appears to have
been advected south (toward the top of the image), and
has expanded considerably. The final image shows
even less cirrus than in the initial images.
The WSI data go through a series of quality evalu-
ations as they are received and processed. This section
discusses the results of the various checks that have
been made.
3.2.1 Tape QC
The Tape QC process is an automated program that
checks each WSI tape as soon as it arrives in-house.
The QC procedure is documented in detail in Karr,
1989. In general, the QC program evaluates the tape
with respect to the following criteria:
a) Proper spectral filter sequence occurred.
b) Filters did not time-out in the field (i.e., a flag is
put in the header in the field if the spectral or neutral
density filter do not have time to go to the selected
position).
c) Occultor did not time-out.
d) Occultor reading was correct for given time and
longitude.
e) Flux control operation was reasonable; i.e., iris
may be open in and only if no 1
may be open if and only if no neutral density is used.
f) Flux control did not time-out; i.e., there was
adequate time for flux control.
adequate tim
g) Signals within the image are on-scale, i.e.,
brighter than 15 in the brighter two filters, darker than
240 in the darker two filters.
The images in Fig. 3.2 also show great variability,
both in cloud distribution and amount. The second
image in particular shows clearly the multiple layers,
with a much thinner layer appearing in the east above
the somewhat thicker layer to the south.
It should be noted that the above comments are
based on visual assessment of the red images. We find
that in these evaluations, the human is an excellent (but
unfortunately not automated or consistent) image
processor. Also note that much of the detail apparent
in the original imagery is lost in the conversion to
photographs for the report. Fig. 3.3 illustrates the same
images as Fig. 3.1, photographed at a different brightness
setting. In this rendition, the low brightness details are
lost, but the high brightness details, such as the contrail
near the sun at 1910, are retained.
h) Sensor dark response outside the image was
normal; i.e., dark signal is less than 15, and vertical and
horizontal ramping are less than 10.
i) All indicated days are present on tape.
j) Time record is continuous, with no missing
minutes. Both ten and one minute images
appear in the proper sequence.
k) WWV time was used, rather than BIOS.
YTYTY
1) Camera temperature readings stay below 40
Centigrade (below 30 Centigrade with the new
chiller unit).

em
aan
Time: 1910 Z
| Time: 1920 Z
te•12
Time: 1930 Z
Time: 1940 Z
Figure 3.1 WSI Imagery from 1 December 1989. Sample red images are shown
at 10 minute intervals.

to
11
tt
.12
sce
Time: 1720 Z
Time: 1730 Z
TENE +12
fe
*12
Wall
Konto
DE
Time: 1740 z
Time: 1750 Z
Figure 3.2 WSI Imagery from 5 December 1989. Sample red images are shown
at 10 minute intervals.

tt = =12
www
Time: 1910 z
Time: 1920 Z
爹爹舞臺旁囊囊
​WWW
BARD
N
2000
Dm
Time: 1930 Z
Time: 1940 Z
Figure 3.3 WSI Imagery from 1 December 1989. Sample red images, reproduced to
retain high brightness detail.
m) LAN (local apparent noon) images appear nor-
mal visually
n) Histogram on LAN image has a reasonable
signal distribution.
o) On LAN image, iris, occultor, and filter read-
ings are reasonable and properly recorded on
the header.
compared with histograms for sample field images
from before and after the abnormality occurred. The
histograms compared very closely, indicating that the
double termination was probably the source of the
problem.
The TapeQC program gives several pages of diag-
nostic information per tape. The information is
evaluated, and a summary of tape conditions put to-
gether. The summary for the tapes from the Madison
deployment is given in Appendix 1.
As a result, post deployment calibrations were ob-
tained both with and without the double termination.
The termination was found to affect both the linearity
(i.e., the relative response curve) and the absolute
calibration constants. Since it was possible to determine
the source of the problem and take appropriate cali-
brations, the processed data should be fairly reasonable.
There should be more offscale bright data than normal,
however, due to the reduced sensitivity range of the
camera in the truncated mode.
3.2.3 Standard Camera Calibrations
As mentioned earlier, the significant tapes, for the
lidarintercomparison, are POR001-POR004. The data
were good in most respects up until the fourth tape. On
this tape, the signal was truncated around a signal of
176, rather than the normal 255. We later determined
that this problem started on 2 December, and that cloud
identification from the truncated imagery is possible.
This data truncation is discussed in more detail in
Section 3.2.2.
The next step in the data evaluation and preparation
for processing is the reduction and evaluation of the
radiometric calibration data. Presentation of the full
data set and its reduction is beyond the scope of this
report, however an overview of the results is appropriate.
For convenience, will use the terms “Pre” and “Post"
for Pre-deployment calibrations and Post-deployment
calibrations in the following section.
In addition, there were occasional images or inter-
vals that were indicated as missing on some of the days,
particularly on the first tape. Through judicious use of
more sophisticated tape read procedures, the data were
retrieved. The data were successfully processed dur-
ing the later data processing stages, so there is appar-
ently little actual data loss.
3.2.2
Special Calibrations
As noted above, the WSI recorded signal became
truncated starting 2 December. On return from Madison,
it was found that one of the switches on the back of the
monitor had apparently been bumped. The 75 ohm
termination switch forincoming video on Data Line A
is normally switched off, since the video is terminated
on the FG100 board. When the switch is on, the data
are doubly terminated, and the signal is compressed.
On return, the switch was halfway between the on and
off settings, and the termination would change modes
with a slight bump of the monitor. The computer rack
was physically moved at the end of the day on 1
December, so it is reasonable to guess that the bumping
may have caused the termination to become enabled.
In order to verify whether this Line A termination
switch was the source of the abnormally truncated data,
images of the sky were acquired at MPL both with and
without double termination under a variety of lighting
conditions. Histograms for these images were then
Both the Pre and Post calibrations appear to be
reasonably self consistent. The full dark value, which
should be close to 0, was 0.2 (on a 0 to 255 scale) on the
Pre calibration, and 0.5 on the Post calibration. Full
bright values, which need not be at 255 but should be
near 255, were 245.7, and 246.2 on Pre and Post
calibrations respectively. Thus the camera output was
well matched to the image board A/D input. The noise
was reasonably low, with typical signal standard de-
viations of 2 to 3 counts, for a 20 x 20 pixel block. The
system was slightly noisier at the high end, with STD's
near 4. System response was quite stable during both
linearities and absolutes. During the linearities the full
measurement cycle is repeated, and the repeat values
differed from the initial values by 1 or less in most
cases. Similarly, duplicate measurements during ab-
solute calibrations generally differed by 1 or less.
During absolute calibrations, measurements are
taken redundantly at a variety of lamp positions. Each
of the redundant measurements, though at different
signal levels and lamp settings, should yield the same
calibration constants. Variance may be caused by stray
light, instabilities in the camera relative response curve,
and measurement noise. For the portable WSI cali-
brations, the redundant measurements were very con-
sistent. For example, in the absolute calibration at
The Date File
ND=1, Aper=160, which is the primary absolute cali-
bration, 10 measurements were taken in each spectral
filter. For spectral filters 1,2,3, and 4, the STDs for the
10 measurements were 0.6%, 0.8%, 0.7%, and 1.4%,
all of which indicate very good stability and consistency.
For processing field data to yield ratio images, two
very significant numbers computed from the calibra-
tions are the SPR value and the NDR value. Both of
these numbers have to do with the relative spectral
response of the system, and both are used in the ratio
computation. For the SPR value, there are two redundant
measurements. The two measurements differed by
between 1% and 2% on both Pre and Post calibrations,
and the Post differed from the Pre by 2%. With the
NDR value, there is only one measurement; the Post
differed from the Pre by less than 2%. Thus the Pre and
Post calibrations yield very consistent estimates of
these parameters.
This file lists the starting date and number of days on
each tape, for verification by the ratio program. This
file also lists a date correction, if required. That is,
sometimes the computer in the field will either skip or
jump a date, due to improper handling of the WWV vs
BIOS correction near midnight. The analyst must find
the occasional occurrences of these abnormalities by
looking for missing or duplicate dates, checking the
tape date on known field site visit dates, and by com-
parison of the cloud cover at LAN with the standard
weather reports. Any input date offsets are corrected in
the ratio processing.
The Time File
There were significant differences between the Pre
and Post aperture calibrations. In general, the aperture
calibration is approximate, because the camera iris is
not a precision device. This is why the 4 images used
to create a composite ratio are always acquired at one
aperture setting; in this way, the effects of the aperture
inconsistencies are ratioed out in the red/blue ratio.
Thus, this difference between the Pre and Post cali-
brations is not ideal in terms of absolute radiance, but
it has no impact on the computed ratio images.
There are still some aspects of the calibration pro-
cess that are under development. For example, the
spectral response curves for the filters are being mea-
sured, and the spectral emittance curves of the lamps
are being updated. As a result, the current calibration
procedure includes the verification and possible ad-
justment of the computed calibration constants, using
the field data. For this data set, adjustments of less that
5% to the computed values were required.
In summary, the Pre and Post calibrations show no
unexpected instrumental problems. The results are
very consistent within each set, and the Pre and the Post
results compare very well. The Pre calibration was
chosen for generation of the calibrations forinput to the
ratio processing, since most of the field data were
acquired closer in time to the Pre calibration set.
3.2.4 Final LAN Image Analysis
Once the calibrations are prepared, it is necessary to
create four more files for use by the ratio program;
these four files in general involve further evaluation of
the data quality. The four files are discussed below.
When the computer is unable to access WWV time,
the time is based on BIOS, which can drift several
minutes over a period of weeks. This drift is determined
by comparison of the actual sun spot in the image with
the predicted position based on clock time. This yields
a good measure ofclock drift, which is listed in the time
file, and then used to correct the image time in the ratio
program.
The Occultor File
Since the field personnel at some sites are not
always able to change the occultor arm on the normal
dates, it is necessary to determine when they were
actually changed. These dates are contained in this file,
so that the appropriate occultor mask may be drawn in
the ratio program. Also, the occultor is sometimes
misaligned by a few degrees; that is, the readout may
be offset from the physical position ifthe occultoris not
aligned properly. The amount of offset is determined
from the LAN image, and input using this file.
The Version File
This file lists which version of the hardware was in
the field on any given date, and therefore which version
of the calibrations to use. In addition, missing data
days are indicated in this file. Finally, this file has a
quality indicator, which indicates the following visu-
ally determined faults, if present:
1. No occultor present.
y
2. Stray light present, due to the wrong occultor
arm, a large time offset in the clock, or other
factors.
3. Split images (can be caused by faulty image
board).

4. Bad input look up table on image board.
5. Range truncation at the top of the range.
6. Range truncation at the bottom of the range.
7. Obstruction to vision on dome, due to condensa-
tion or other obstruction.
For the Portable data set at Madison, the data were
quite good in the above respects. There were few
missing dates, and no date slips. Time corrections were
not required, since the system ran on WWV. The
proper occultor arm was used, and the alignment was
quite close (a 3 degree correction was required). There
were no data quality abnormalities indicated from the
above list. Range truncation did occur starting 2
December, as mentioned above, however it was pos-
sible to calibrate the truncated system, so the ratios
should be essentially normal.
3.3 Ratio Processing
Once the myriad of quality control and calibration
evaluations are completed as described in Section 3.2,
the next step is the processing of the raw data to yield
ratio data, using Program TAPRAT. In anticipation of
receivinglidardata at some point in time, this processing
has been completed.
The ratio processing is part of the full processing
illustrated conceptually in Fig. 3.4. The TAPRAT
program starts with the raw data on the left of Fig. 3.4;
the right-hand-most green circle, the composite ratio
image, is the final product (we now generate red/blue
ratio, rather than the blue/red shown in the illustration).
The ratios have not been processed to the cloud
stage, since the ratios, rather than the cloud decision
images, are the appropriate parameter for lidar com-
parison. However, a preliminary look at the ratio data
may be obtained by using the fixed threshold cloud
algorithm to color the ratio, as shown in Figs. 3.5 and
3.6. As an illustration, the cloud cases illustrated in
Figs. 3.1 and 3.2 have been processed using the fixed
threshold cloud algorithm, to yield preliminary cloud
images shown in Figs. 3.5 and 3.6, respectively. In
these images, grey to white represents areas identified
as relatively thick cloud, and yellow represents areas
identified as thin cloud
One may see in these figures that many of the thin
clouds are properly identified, however the thin clouds
which are downsun are generally missed. Although the
distinction between the downsun clouds and immedi-
ately adjacent sky may be readily seen in ratio images
(not shown here), the fixed threshold algorithm does
not always make this distinction.
Like the Tape QC program, the TAPRAT program
checks a number of diagnostics, and creates diagnostic
output files. The diagnostic checks include verifica-
tion of time, occultor, and filter consistencies. A
sample of the summary output file for the first tape,
......
.
.
.......
...
BASIC IMAGERY
CORRECTED IMAGERY
COMPOSITE RATIO
DELIVERABLE DATABASE
CALIBRATION
FUNCTIONS
RADIANCE
CONVERSIONS
CALIB
BLUE
IMAGE
CONVERSIONS
E BLUE
IMAGE
512 X 512 x B
CLOUD
NO-CLOUD
DECISION
ALGORITHMS
RADIOMETRIC
LINEARITY
BLUEIRED
RATIOS
RADIOMETRIC
SENSITIVITY
CALIB.
UP-SUN
RED
IMAGE
512 X 512 x 8
DERIVED
PRODUCTS
IMAGE
T-
DOWN-SUN
OPTICAL
DISTORTIONS
IMAGE
RATIO
COMPUTATIONS
NEAR HORIZON
SPATIAL
DISTRIBUTIONS
-
COMPOSITE
BLUEIRED
RATIO
MAGE
OPTIMUM
CLOUD
NO-CLOUD
IMAGE
FIELD OF VIEW
DEFINITION
TWILIGHT
-
TEMPORAL
DISTRIBUTIONS
BLUE+N.D.
IMAGE
512x512 x 8
CALIB.
(BLUE)
IMAGE
DAWN
SENSOR CHIP
UNIFORMITY
STATISTICAL
PARAMETERS
PIXEL
SELECTIONS
FOR
OPTIMUM IZED
COMPOSITE
AS REQD
(BLUEY(RED)
RATIOS
REGISTRATION
ADJUSTMENTS
RED + N.D.
IMAGE
512 X 512 x 8
CALIB.
(RED)
IMAGE
FLUX CONTROL
THRESHOLDS
L
L-
Fig. 3.4 Whole Sky Imager Basic Image Processing Flow Chart
11

Time: 1910 Z
Time: 1920 Z
12
Time: 1930 Z
Time: 1940 Z
Figure 3.5 Sample Cloud/No Cloud Decision Images from 1 December 1989.
Image processed using fixed ratio threshold.

Time: 1720Z
Time: 1730 Z
Time: 1740 Z
Time: 1750 Z
Figure 3.6
Sample Cloud/No Cloud Decision Images from 5 December 1989.
Image processed using fixed ratio threshold.
13
POR001, is included in Appendix 2. The more detailed
diagnostic file is too large to include here.
At this point, all of the bulk processing of the data
has been completed. The next section outlines the
general direction of the further analysis that we intend
to pursue.
4. EVALUATION OF WSI AND LIDAR DATA
INTERCOMPARISON
As noted above, the previous sections detail what
has been accomplished to date. This section gives
further detail on analysis approaches that seem rea-
sonable at this point in time. Some of this analysis is
pending transfer of sample lidar data; other parts of the
analysis can proceed without access to lidar data.
4.1 Preliminary Evaluation
4.2 Application of Data to Beta Algorithm Devel-
opment
As noted earlier, the fixed threshold algorithm works
quite well for a two-class cloud discrimination. The
precise mapping of optical density is a much more
demanding task, however, which should require
tightening of both the calibrations and the cloud al-
gorithm. The development of a directionally-dependent
beta algorithm is an important aspect of the WSI/HSRL
intercomparison. The application of the algorithm
should then allow a more specific determination of the
directional impacts on the two-class cloud determina-
tion.
The development of the beta cloud algorithm is
currently proceeding. Sample calculations have been
made with the Hering FASCAT model. These results
are being parameterized, to enable a simple and
computationally fast correction to the ratio. This
correction currently takes the form of a correction for
solar scattering angle and a correction for look angle,
i.e. for zenith angle.
The next step in the algorithm development will
probably be a comparison with measured ratios from
clear day cases, perhaps with variable aerosol haze
loads. Once the results are reasonably satisfactory, the
next step is evaluation of thin clouds. In terms of ratio,
thin clouds appear as a small perturbation on the clear
sky ratio. Even without the lidar data, cirrus data such
as found in this data set can be extremely valuable.
Contrail cases such as found in Fig. 3.1 can give good
insight into the nature of the perturbation to the ratio
field caused by a thin cloud. Studies of the azimuthal
dependence ofthe cloud covercan yield furtherinsights.
Even though the fixed threshold cloud algorithm
tends to underestimate the down-sun thin clouds, it is
able to detect most of the thin cloud, and it detects thick
clouds quite well. Therefore, it may be useful to make
a first rough evaluation using the HSRL, to determine
the approximate value of optical depth for cases which
are identified as clear, thin cloud, and thick cloud.
Fortunately, the WSI and HSRL were co-located. It
will still be necessary to be very careful with directional
and time coordination. As noted earlier, this will
require careful interpretation of the geometric cali-
bration.
It will also be useful to determine the relative
sensitivity ranges of the two instruments; for example,
can the thickest clouds detected by the WSI be accu-
rately measured by the HSRL, and can the thinnest
clouds measured by the HSRL be detected by the WSI.
The VIL lidar unit may also provide useful com-
parisons. This unit was located several miles from the
WSI and HSRL, so the comparison will perhaps be
somewhat limited. We are not certain of the scanning
time required by the VIL; this will be important to
know. Even though the VIL is a relative device, the fact
that it provides a "slice" through the upperhemisphere,
rather than a single point, makes it well worth evalu-
ating the possibilities.
Another potentially useful data point is the sun point
in the WSI image. The sun occultor utilizes a 4 log
14108
neutral density filter, allowing the apparent solar ra-
diance to be detected. For those cases with this radi-
ance on-scale, this yields an additional indication of
optical depth.
At this point the lidar data would then be an addi-
tional ground truth. By looking at cases with similar
HSRL results, which are at different times of day and
therefore different sun angles, it may be possible to
further test the efficacy of the algorithm. The VIL data,
if it turns out to be applicable, would be especially
useful. That is if the scan time and site separation do
not cause too much of a problem, a measurement of the
relative optical density over a slice of the image would
be most useful in checking for directional bias in the
algorithm.
4.3 Evaluation of the Beta Cloud Data
Once the beta algorithm development is satisfactorily
completed, it would be appropriate to determine, using
the lidar data, the range of optical depths which are
identified as clear, thin cloud, and thick cloud. It may
1A
i
14
be possible to determine the thinnest cloud the WSI can
detect. If the ratio vs density relation is tight enough,
this would enable the WSI image to be converted, in a
sense, to an optical density map. It might be reasonable
to chose thresholds based on optical density, with the
selected optical density depending on the application.
Since this November 89 through January 90 deploy-
ment, the WSI data have undergone extensive quality
analysis. The data quality appears to be quite good.
Sample data have been sent to U.W. In addition, the
processing of the WSI data has been completed.
1
Once lidardata are received, preliminary evaluation
of the data comparison can proceed. In the mean time,
it is our feeling that the most productive comparison
should proceed in conjunction with the development of
a second generation beta cloud algorithm. This algo-
rithm development is proceeding independently. As
further progress is made on the algorithm develop-
ment, lidar data comparisons will become increasingly
important.
6. REFERENCES
How much of this will be possible is difficult to
estimate at this point. The WSI was designed to sort the
clouds into only two categories. In spite of some
directional bias, this two-category sorting in general
yields results that are quite reasonable and compare
well with the standard obseryer. Through the lidar
intercomparison, it should at least be possible to relate
these two categories to optical depth. Whether finer
discriminations can be made is something we are eager
to know.
In order to apply these results to the full WSI data
base, some further work with the calibrations will also
be advised. Specifically, calibration of the spectral
response of the filters and lamp must be completed and
applied to the calibrations. Further analysis of the
calibration accuracy, in the form of system inter-
comparisons, may also be required.
5. SUMMARY
Johnson, R. W., W.S. Hering, and J.E.Shields, (1989).
Automated Visibility and Cloud Cover Measure-
ments with a Solid-State Imaging System, Univer-
sity of California, San Diego, Scripps Institution of
Oceanography, Marine Physical Laboratory, SIO
89-7, GL-TR-89-0061.
Karr, M. E. and J. E. Shields, (1989). Whole Sky Im-
ager Management of Raw Database, University of
California, San Diego, Scripps Institution of
Oceanography, Marine Physical Laboratory,
Technical Note No. 211
A network of Whole Sky Imagers has been operating
over the past two years at a number of sites. The
extensive data base which has resulted is being suc-
cessfully processed to yield assessments of thin and
opaque cloud cover. Although these results compare
well with standard observers, it would be very useful to
quantitatively relate the results to optical depth values.
Shields, J. E., T. L. Koehler, M. E. Karr, and R. W.
Johnson(1990). Automated Cloud Cover and Vis-
ibility Systems for Real Time Applications, Uni-
versity of California, San Diego, Scripps Institution
of Oceanography, Marine Physical Laboratory,
Technical Note No. 217
In order to provide data toward this end, a portable
WSI unit was built, and was fielded for just over two
months at the lidar facility at U.W. Simultaneous data
were acquired by the two instruments during times
when cirrus clouds were prevalent. The instrument
fielding and data acquisition were successfully com-
pleted.
Varah, J. R. (1989). Whole Sky Imager Solid State
Sensor Electronic Calibration, University of Cali-
fornia, San Diego, Scripps Institution of Oceanog-
raphy, Marine Physical Laboratory, Technical Note
No. 212
APPENDIX 1
Tape QC
Summary for WSI/Lidar Experiment
Portable system - Madison, Wisc.
immary
WSI
ir
POR001 - (14 - 20 Nov), Rec'd 29 Nov.. Nov. 16 & 20 had halo of speckles
(over saturation). Nov. 14 started at 20:10 and had tape errors at:
20:50, 21:10, 22:10, 22:20 & 23:30. Nov. 15 had tape errors at 11:40,
13:10 & 13:20. Nov. 17 had tape errors at 12:30 & 23:00. Nov. 18 had
tape errors at 14:30, 15:30, 15:40, 16:20, 16:30, 16:46, 18:10, 18:50 &
19:10. Nov. 19 had tape errors at 19:50. Nov. 20 had tape errors at:
11:40, 13:10, 15:10, 15:20, 15:30, 15:50, 21:20 & 23:39. LAN's and
histograms looked o.k.. System used WWV time. Highest LAN temperature
was 12 degrees and the lowest was 3 degrees.
POR002 - (28 Nov. - 2 Dec.), Rec'd 11 Dec. Tapeqc was rerun on another
machine and two more days were extracted. Nov. 28 starts at 21:00 and
has no LAN. Dec. 2 had a data gap between 19:40 and 23:41 and ended at
23:47. LAN images, histogram and diagnostic files look ok. System used
WWV time. Lowest LAN temperature was 2 degrees.
POR003 - (21 - 27 Nov.), Rec'd 11 Dec. Nov. 21 had 19 tape errors
throughout the day. Nov. 22 had 4 tape errors throughout the day. Nov.
23 was missing the 11:40 ten minute image. Nov. 25 only had 1 tape
error. Otherwise, LAN images, histogram and diagnostic files look o.k.
System used WWV time. Highest LAN temperature was 13 degrees and lowest
was 3 degrees.
POR004 - (4 - 10 Dec.), Rec'd 2 Jan. The signal is cutting off at around
176. Otherwise, LAN images, histogram and diagnostic files look 0.K.
System used WWV time. Dec. 4 started at 18:30 and has no LAN. Highest
LAN temperature was 12 degrees and lowest was 2 degrees.
POR005 - (11 - 17 Dec.), Rec'd 2 Jan. Signal compression occurs at 178
now at this station. Dec. 11 started at 19:30 and had no LAN image.
Dec. 12 had spectral errors at 11:55, 12:13, 12:48, 14:41, 14:53, 20:48,
21:53, 22:43 and the ten minute image at 12:50 is missing. Dec. 13 had
spectral errors at 14:27, 16:03, 18:04 & 19:03. Dec. 14 had spectral
errors at 16:33, 16:55, 17:07, 17:17, 17:22, 17:26, 17:29, 17:40, 18:10,
18:51, 19:05, 19:42, 22:09, 22:19, 22:37 & 22:48. Dec. 15 had a tape
error at 16:10. Dec. 16 had spectral errors at 11:50, 12:03, 17:44,
17:50, 18:06, 18:12, 18:28, 18:49, 22:26, 22:42, 23:05, 23:10 & 23:24.
Dec. 17 had spectral errors at 14:59, 15:07 and a halo of speckles on
LAN image. System used WWV time. Lowest LAN temperature was -01
degrees.
POR006 - (18 - 24 Dec.), Rec'd 8 Jan. Signal compression occurs at 178.
Dec. 18 starts at 19:00 (no LAN image) and has spectral errors at 19:04,
19:11, 20:57 & 21:11. Dec. 19 has spectral errors at 14:36, 14:45,
15:30, 15:34, 16:16, 18:00 & 18:20. Dec. 20 has spectral errors at
20:54 & 21:01. On Dec. 21 data is unrecoverable from 12:00 to 19:40 -
looks like FG100 board was not working properly, then the problem cleared
up. Otherwise, the rest of the LAN images are ok. System used WWV time.
Lowest LAN temperature was -01 degrees.
16
APPENDIX í
POR007 - (26 Dec. - 1 Jan.), Rec'd 9 Jan. Signal compression occurs at
164. Dec. 26 starts at 22:00 and has no LAN image. Dec. 29 had a tape
error at 16:02 and two image signal errors at 14:00; QD1 = 5.56 and
QD2 = 10.86. Dec. 31 had a tape error at 16:50. Dec. 30 & 31 LAN
images had snow on the dome. Otherwise, LAN images and diagnostic files
look ok. System used WWV time. Lowest LAN temperature was 3 degrees.
POR008 - 12 - 8 Jan.), Rec'd 23 Jan. Signal compression at 173 now.
Jan. 3 had a halo of speckles. Jan. 4 had snow over 2/3 of the dome.
Jan. 3 had spectral errors at 21:52 & 22:11. Jan. 4 had two image signal
errors at 14:00; QD1 = 7.86 & QD2 = 5.11. Jan. 5 had a spectral error
at 21.44. Jan. 7 had spectral errors at: 16:13, 16:53, 19:00, 19:17,
20:06, 20:17, 20:58, 21:01, 21:18, 21:27 & 21:37. Jan. 8 had spectral
errors at 22:13, 22:15 & 22:22. System used WWV time. Lowest LAN
temperature was 5 degrees.
POR009 - (9 - 15 Jan.), Rec'd 23 Jan. Signal compression at 174. Jan.
9 started at 20:10 and has no LAN image. The 9th has seventeen
occurances of tape errors starting at 20:40 and ending at 23:00. It also
had a spectral error at 22:07. On Jan. 10 at 13:10 Occultor movement of
-140 degrees was flagged. Jan. 11 had one image signal error at 22:10.
Diagnostics for Jan. 12, 13 & 14 look ok. System used WWV time. Lowest
LAN temperature was 4 degrees.
POR010 - (17 - 23 Jan.), Rec'd 31 Jan. Signal compression to 174.
Occultor looks slightly off-center. All of the days except Jan. 23 are
missing the midnight ten minute images. Jan. 17 starts at 18:20 and
misses LAN by 10 minutes. It has two image signal errors at 22:20;
QD1 = 8.39 and QD2 = 12.53. Jan. 23 has one image signal error at 22:10;
QD1 = 10.97. System used WWV time. Lowest LAN temp. was 6 degrees.
POR011 - 125 - 26 Jan.), Rec'd 31 Jan. Jan. 25 starts at 20:30 and has
no LAN. Jan. 26 ended at 14:19 when stopped by Dick and Gene to prepare
system for removal. Diagnostic files looked ok.
34
17
APPENDIX 2
Sample TAPRAT Diagnostic File: Tape POR001
Starting Mo/Da/Yr on tape is 11 14 89
Field Sequence Number = 001 Confirmed
* Results of the Calibration File Selection *
Data
Qual.
1
Day Seg.
Day Day-MO-Yr
684 14/NOV/89
685 15/NOV/89
686 16/NOV/89
687 17/NOV/89
688 18/NOV/ 89
689 19/NOV/89
7 690 20/NOV/89
1
vous as WN
Quality Hard. Soft. Calibration
Indicators Vers. Vers. File Selected
0000000000 1A 2.90 CAL9V01A.RAT
0000000000 1A 2.90 CAL9V01A.RAT
0000000000
2.90 CAL9V01A . RAT
0000000000
2.90 CAL9V01A.RAT
0000000000
2.90 CAL9V01A.RAT
0000000000
2.90 CAL9V01A.RAT
0000000000
2.90 CAL9V01A.RAT
1
Initial header found: Dale = 14/NOV/89
Initial date confirmed.
FIELD tape format confirmed: FLDTYP = 3
Initial image type = 1
Log file 19001.LOG
opened successfully
New calibration file CALIVO1A.RAT opened successfully.
Calibration file date is :02/ JUL/90
Basic Time Information for 14/NOV/89 :
Estimated Start Time : 1143 GMT
Computed Sunrise = 1245 GM'T
Computed Sunset = 2239 GMT
Time correction = 0 min
Occultor Arm 1 selected: Offset - 3
Tape read error in GETI MG - Line 8 Return to MAIN
Summary for 14/NOV/89
Number of 10 min images processed to ratio =
14
FREQUENCY
PROBLEM
Tine skips ahead :
Time skips back :
Occultor out of range :
Occultor disagrees with time :
Neutral density inconsistencies :
Spectral filter inconsistencies :
Quadrant inconsistencies :
....
.
APPENDIX 2
Basic Time Information for 15/NOV/89 ;
Estimated start Time = 1143 GMT
Computed Sunrise = 1247 GMT
Computed Sunset = 2238 GMT
Time Correction = 0 min
Occultor Arm 1 selected: Offset = 3
Sumnary for 15/NOV/89
Number of 10 min images processed to ratio =
59
FREQUENCY
PROBLEM
Time skips ahead :
Time skips back :
Occultor out of range :
Occultor disagrees with time :
Neutral density inconsistencies :
Spectral filter inconsistencies :
Quadrant inconsistencies :
Basic Time Information for 16/NOV/89 :
Estimated Start Time = 1143 GMT
Computed Sunrise = 1248 GMT
Computed Sunset = 2237 GMT
Time Correction = 0 min
Occultor Arm 1 selected: Offset = 3
Summary for 16/NOV/89
Number of 10 min images processed to ratio :
59
FREQUENCY
PROBLEM
Time skips ahead :
Time skips back :
Occultor out of range :
Occultor disagrees with time :
Neutral density inconsistencies :
Spectral filter inconsistencies :
Quadrant inconsistencies :
Basic Time Information for 17/NOV/89 :
Estimated Start Time = 1143 GMT
Computed Sunrise = 1249 GMT
Computed Sunset = 2236 GMT
Time Correction = 0 min
Occultor Arm 1 selected: Offset - 3
APPENDIX 2
Summary for 17/NOV/89
Number of 10 min images processed to ratio =
59
FREQUENCY
PROBLEM
Time skips ahead :
Time skips back :
Occultor out of range :
Occultor disagrees with time :
Neutral density inconsistencies :
Spectral filter inconsistencies :
Quadrant inconsistencies :
Basic Time Information for 18/NOV/89 :
Estimated start Time = 1143 GMT
Computed Sunrise = 1250 GMT
Computed Sunset = 2235 GMT
Time Correction = 0 min
Occultor Arm 1 selected: Offset = 3
Summary for 18/NOV/89
Number of 10 min images processed to ratio =
59
FREQUENCY
www
PROBLEM
Time skips ahead :
Time skips back :
Occultor out of range :
Occultor disagrees with time :
Neutral density inconsistencies :
Spectral filter inconsistencies :
Quadrant inconsistencies :
Basic Time Information for 19/NOV/89 :
Estimated Start Time = 1144 GMT
Computed Sunrise = 1252 GMT
Computed Sunset = 2235 GMT
Time Correction = 0 min
Occultor Arm 1 selected: Offset - 3
20
APPENDIX 2
Summary for 19/NOV/89
Number of 10 min images processed to ratio =
58
FREQUENCY
PROBLEM
Time skips ahead :
Time skips back :
Occultor out of range :
Occultor disagrees with time :
Neutral density inconsistencies :
Spectral filter inconsistencies :
Quadrant inconsistencies :
Basic Time Information for 20/NOV/89 :
Estimated Start Time = 1144 GMT
Computed Sunrise = 1253 GMT
Computed Sunset = 2234 GMT
Time Correction : 0 min
Occultor Arm 1 selected: Offset = 3
Tape read error encountered - processing stopped.
Summary for 20/NOV/89
Number of 10 min images processed to ratio :
58
FREQUENCY
PROBLEM
Time skips ahead :
Time skips back :
Occultor out of range :
Occultor disagrees witli time :
Neutral density inconsistencies :
Spectral filter inconsistencies :
Quadrant inconsistencies :
21
APPENDIX 3
Illustrative Charts
This appendix contains charts illustrating some of the concepts discussed in
this note. These charts were prepared for a meeting concerning the WSI/Lidar
experiment.
The first chart illustrates the preliminary WSI data processing, as discussed in
Section 3. The second chart illustrates the analysis of WSI data using the joint
experimental data, as discussed in Section 4. The plot illustrates the variance in
background sky ratio, as discussed in Section 4.2. The illustration shows samples
from a geometric calibration image. Both the original image with moderate lens
misalignment, and the image with a corrected geometric calibration are shown.
22

PHASE I: WSI DATA PROCESSING
FOR WSI / HSRL EXPERIMENT
09900.00
jensia
Process
Field Data
Tape QC
Run
Assessment
Verify
Time, Occultor
Accuracy
ocultor
ol
sepewj 10
Ratio
Images
SO
999
00
POPCESSSSSSSSS
ерпрәЯ
Data
CS
Pre-deployment
calibrations
@jenjeng
josues
aquewioped
CONCEN444
100
9999999049..400000
Prepare
Calibration
Files
99999999
Reduce
Data
Evaluate
Post-deployment
calibrations
Josues
APPENDIX 3
Performance
10.000.
000
23

ejea ISM JO SISÁLLUY leseyd
juewuədxa TUSH/ISM Buisn.
090
1. Preliminary Evaluation
Geometric
Calibrations
Preliminary
Ratio Images
Relative
Variance.
WSI vs HSRL
7USH
ejed
24
II. Ratio vs Optical Depth Evaluation
28
ty
S32
CoMATYI
Enhanced
Beta Algorithm
2020
Geometric
Calibrations
39
SASSAR
D
SEN
NA
Preliminary
Ratio Images
Beta - corrected
Ratio vs
Optical Depth
2222
28
AO
APPENDIX 3
26
33033
Enhanced
Calibration
Studies
HSRL
Data

FASCAT RESULTS
Zenith Angle = 45 Deg.
NI
- Opaque Cloud
- V = 12 km, Solar Zen. = 66 deg.
- V = 12 km, Solar Zen. = 37 deg.
0 V = 100 km, Solar Zen. = 66 deg.
mm V = 100 km, Solar Zen. = 37 deg.
RED / BLUE RATIO
.
.
.
.
.
.
www
.
NI
.
HI
.
NANT
w
WE
indemningen
AN
SINH
w
www
MUNAUNGANA
ww
S
S
ww.
SU
www
ch
IN
w
DIDELIFE
www
www
.
.
ben 10
MPE
090
0.0
20.0
0°08
0 001
1 20.0
40.0
60.0
SOLAR SCATTERING ANGLE B (Degrees)
APPENDIX 3
25

pərbənog
leuibuo
Vio
WA
MA
DAHA
FACES
Lahore
26
SEO
iniz
S
ESERLE
A
MERICA
Casa
ANG
NE
TE
RESTRELAS
R
ESCEUTA
TELL
MEDERE
ANTALASEVA TASCA
DETSERT
SERIALITETAS
ca
BASINUTEN
RADIONIRA
T
IE
ARRIBA
LINH
REPERERERE
REN
TE
TATTACI
RAREASCA
DALAR
BE
RSS
ANA
JENS
FHHHHHHH4+44**
1899
5
SER
st
Salut
ONLINE
WINE
S882
SE2
1601
TOMOS
TIL
TOTA
2
MARTIN
THE
.
OM
NieR
SZOWWI NOLIVhaltvo bul3w035 Wo: 537dWVS