UNIVERSITY OF CALIFORNIA, SAN DIEGO
www
c
......
...
RAR
c
ccc
.
.
.
WWWWW
.....
............
........
. ....
......
.
...
WWW PURIFWWW.SARA
.
.
.
.
..
w
......
ww
.....
wwwwwwww
..........
...
.....
w
3 1822 04429 0799
ATMOSPHERIC OPTICAL SYSTEMS
TECHNICAL NOTE NO. 210
NOV 1988
Offsite
(Annex-Joi
mnals)
QC
974.5
. T43
no. 210
.
A MULTI-STATION SET OF WHOLE SKY IMAGERS
AND A PRELIMINARY ASSESSMENT OF THE
EMERGING DATABASE
FYN

M
R. W. Johnson
T. L. Koehler
J. E. Shields
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
1
Contract Monitor, Dr.J. W. Snow
Atmospheric Sciences Division
1
.
19
(FOR
• TH
Prepared for
Air Force 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
..
.:
w
ww
.ZIP
UC SAN DIEGO LIBRARY

UNIVERSITY OF CALIFORNIA, SAN DIEGO
.........
www
.
.
..
...
wowwwww
wWwW.MA..
RESEP
www....
S
..
W
..
..
.
www
.........
.
...
www.
WW225.0
..
...
3 1822 04429 0799
TECHNICAL NOTE NO. 210
A MULTI-STATION SET OF WHOLE SKY IMAGERS
AND A PRELIMINARY ASSESSMENT OF THE EMERGING DATABASE
This Technical Note contains the annotated vu-graph representations, and extended summary
which were presented to the Cloud Impacts on DOD Operations and Systems 1988 Workshop,
CIDOS-88, which was held at the Naval Surface Warfare Center, White Oak, Silver Spring,
Maryland on 18—20 October 1988.
w
ww.
A MULTI-STATION SET OF WHOLE SKY I MAGERS,
& PRELIMINARY ASSESSMENT OF THE EMERGING DATA BASE
R. W. Johnson, T. L. Koehler, J. E. Shields
University of California, San Diego, Marine Physical Laboratory
San Diego, California 92152-6400, USA
ABSTRACT
The Optical Systems Group of the Marine Physical Laboratory, in support of several DoD sponsored
programs has developed a visible-spectrum, computer controlled, imaging system for the automatic
assessment of the optical state of the atmosphere. A six-station network of these systems, oriented
specifically toward cloud cover assessments has been deployed throughout portions of the western United
States. Approximately 1.8 Gigabytes of whole sky image data is collected automatically each week from
each of these stations. Current algorithm development for the exploitation of this new data base is
slanted toward the determination of the spatial (1-1000 mi) and temporal (1 min - 12 hours) variability
of cloud free lines of sight and cloud free intervals. The operational status of the network is
reviewed, as is the status of the rapidly growing raw data base. Results from a first generation
software routine for automatic cloud/no cloud discriminations, and follow-on statistical analysis
procedures are demonstrated.
1.
INTRODUCTION
The development, testing and deployment of a family of compact, solid state imaging systems for
the automatic measurement of atmospheric optical and meteorological properties is an expanding activity
of the Marine Physical Laboratory of Scripps Institution of Oceanography.
The basic operating
characteristics of these video based devices have been outlined previously at the Fifth Tri-Service
Modeling Workshop held at the U. S. Naval Academy in June 1987. This report summarized our progress
on the deployment of a multi-station set of these Whole Sky Imagers, and an introduction to their
rapidly accumulating data base.
2. OPERATIONAL STATUS OF DATA COLLECTION
The as-built configuration of the hardware comprising the Whole Sky Imager system of each field
site is illustrated in Fig 1.
The more important features associated with the external sensor
installation are the camera and the optical filter assembly.
The camera is a solid state CID class
black and white camera which outputs standard RS-170 composite video.
It runs at full video rate and
in fixed gain configuration. The optical filter changer enables both blue and red spectral bands to be
selected for image acquisition, as well as neutral density range shifting. The major features within
the interior control console are built in to the AT class micro-computer.
They are the video frame
grabber which accommodates a 1024X1024 image memory, and the 8mm cartridge tape system which provides
2.2 Gigabytes of on-board archival memory. The field systems are deployed as illustrated in Fig 2.
The geographical distribution is currently biased toward the southwestern desert areas, however the
sites in Missouri and Florida will be operational by the end of calendar 1988 generating a better
balance between several different geographical and meteorlogical regimes.
Fig 1.
Image Acquisition & Analysis System
Hardware Block Diagram E/O System 5
Fig 2. WHOLE SKY IMAGER DATA SITES


"
...
SONY PVM 12710
MONITOR
GE 2710
SOLID STATE
VIDEO
CAMERA
IM COMPUTEA
(IBM AT CLONE)
NE
AUTOMATIC
EQUATORIAL
SOLAR
OCCULTOA
ASSY.
VIDEO
IMAGE
PAOCESSING
SUB-SYSTEM
(ITI FG 1001
AACHIVAL
110
SUB-SYSTEM
SEAGATE 65 M byte
HO.)
REMOTE
CONTROLLED
TRIS
ASSY,
EXABYTE EX8 - 8200
22 G byle 8mm CAATRIOGE TAPE SYSTEM
ANALOG
ACCESSORY
CONTAOL PANEL
CONTAES PANEL
REMOTE
CONTROLLED
OPTICAL
FILTER
ASSY.
STOWED KEYBOARD
STOWED KEYBOARO
• WSMAC
• OPERATIONAL O PENOING
• CLIC
• MAFB O COLUMBIA
KAFB
MALABAA
EXTERIOA SENSOR
INSTALLATON
INTERIOR CONTROLLER
INSTALLATION
• WSMRH
The WSI system acquired multispectral imagery at every minute during the twelve hour day bounded
by Local Apparent Noon 6 hours. During the first eight seconds of each minute a set of four images,
Aach
each similar to that illustrated in Fig 3 are grabbed, digitized and archived on tape. From each one
I
Strated
are
IC
From e
ne
minute set, only the radiances along the 33 rows and columns shown in the Fig 3 screen overlay are
archived resulting in 2880 screen images per day. At each ten minute interval, the entire 512X512
images are archived resulting in an additional 288 images. This composite 173 megabyte archive
US
represents the basic daily data set.
Post archival data processing is organized to follow the general procedure illustrated in Fig 4.
In this processing sequence, the intent is to impose strict quality and calibration controls upon the
raw data, while simultaneously reducing the number of images required for insertion into the optimized
cloud/no cloud decision processes. Preliminary test sequences run in fully automatic mode illustrate
reasonable processing times at the case study levels described herein, and should provide the necessary
insights for further optimization.

10:
M.
MM
,10'"
MAS MAGERI
CORRECTER MAGERI
COST AD
x
.
.
aaaaoo
CALIBRATION
FUNCTIONS
RAOLANCE
CONVERSIONS
CALO
BLUE
BLUE
MAGE
512 512.0

| AADIOMETAK
UNCAAITY
BUT AD
IONEVO
BIL
RADIOMETRK
SENSIINITY
?
MAGE
$12=5121.
*
MAGE PATIO
COMPUTATIONS
OPICAL
DISTORTIONS
100 TL
парфарини
AULAD MO
.
ime
:
NO.1
FELD OVEW
ONTON
Jogo
1
NO
CAO
X
$175176
SON
FILL UCTOR
--
ALGSTRAIO
ALIMENTS
8
AD.NO
MAGE
$12:$171.
-
@30
.
AU CONTRO
ITALSKUS
L
.
.
Fig 3. WSI Ten Minute Image
Fig 4.
WSI BASIC IMAGE PROCESSING FLOW CHART
3.
PRELIMINARY DATA ANALYSIS
To illustrate the types of products that will become available from the WSI system in the near
future, a limited case study from one site (C-station) was prepared.
Ten minute images from 19 days in
late July and early August (14-18 July, 20-26 July and 5-11 August, 1988) were processed using prototype
versions of our calibration and cloud-no cloud decision algorithms.
This yielded a total of 1319
images representing a wide range of sky cover conditions. Meteorological observations of total and
opaque cloud cover were also available at C-station on an hourly basis.
We decided to concentrate on thick opaque cloud cover in preparing this case study. A particular
point was designated as either clear or cloudy based solely on a single threshold in the blue to red
ratios.
This threshold was selected manually from a selected subsample of the cases. Most of the
cirrus cloud cover was not included in this classification. While several approximations were used in
preliminary processing of the blue to red ratios and the resulting cloud images, the quality of the
results was reasonable for illustrative purposes. Selected statistical evaluations of the resultant
cloud-no cloud images were then performed, a sample of which follows.
The modeling of the downtime duration for a multiple site GBL system due to the effects of clouds
is dependent upon many statistical relationships that define the temporal and spatial variability of
cloud cover, and the interrelationships between point, line and area probabilities. To illustrate some
of these relationships, a two line sample of data from the northern half of the sky was extracted from
the full images, with the east-west oriented lines centered on the north-south axis at zenith angles
of approximately 15° and 65º.
We now have the capability to compute many of the important statistical relationships mentioned
above for the two lines in the current sample.
The dependence of these parameters on the total sky
cover is a crucial factor.
For the C-station sample, three types of total sky cover information can be
employed:
the meteorological observations of total and opaque sky cover provided on an hourly basis,
and an estimate of total opaque sky cover from the WSI available from each ten minute cloud-no cloud
image.
In using the meteorological observations, the sky cover for the beginning of the hour is
applied to all six ten minute images taken during that hour. Obviously, the WSI sky cover will be more
sensitive to rapid changes in sky cover, but much of the CFLOS and CFARC modeling depends on site
parameterizations .based on the observed sky cover climatologies. Also, since the computed WSI sky
cover is fairly sensitive to the approximations made in the preliminary processing, observed opaque
cloud cover was used in these initial comparisons..
W.
WWW.Mw
Wwwww
.goud
i
Some of the parameters that were produced for the case study include CFLOS frequencies for the
center points on the two lines, the frequency distributions of line cloud coverage when the center
point is either cloudy or clear, correlation between the two center points, point to line correlations,
clear and cloudy recurrence frequencies for the two center points, and the time autocorrelations of the
center point cloudiness. All of these statistics can be broken into subsets based on the whole sky
cover
As an example of our computations, time autocorrelations for the center point cloudiness are shown
in Fig 5 for the complete sample 10 to 10 tenths sky cover), and for mostly clear (0 to 5 tenths) and
mostly cloudy (6 to 10 tenths ) subsamples. The autocorrelations were computed using a tetrachoric
correlation approximation. A binary contingency table is produced at each ten minute time lag, with
the table being updated for a given time pair only if the sky cover at the earlier time falls between
the specified thresholds. For example, if the cloud cover at 10 am LST were 7 tenths, the matches
between that time and each subsequent time would be included in the tables for the complete sample and
the mostly cloudy subsample, over the time lags up to 480 minutes (8 hours remaining to 6 pm), but not
in the mostly clear subset. Note, that for both zenith angles, the autocorrelation for the mostly
cloudy subset drops off more rapidly with time than does the mostly clear subset, but both curves cross
at a lag of approximately 180 minutes (3 hours), with the mostly clear cases : exhibiting smaller
minimum correlations than the mostly cloudy cases.
These sample autocorrelations illustrate the type of statistical evaluation procedures that have
been developed by our group. Considering the limited 19 day sample and the preliminary calibration and
cloud decision algorithms applied, great care should be exercised in interpreting these figures.
15 DEGREES
65 DEGREES

1.00
1.00

0.75
0.75
10
AUTOCORRELATION
AUTOCORRELATION
.........
-
1.-2.3
see
0.00
....
....
.....
....
..
ht.
0.25
0.25
400
100
o
400
600
500
100
500
600
200 300
TIME LAG (MIN)
200 300
TIME LAG (MIN)
Fig 5. Time autocorrelations for the 15° and 65° arc centerpoints

-_ OPTICAL SYSTEMS GROUP
uu
FIGURE 1
The Marine Physical Laboratory is a 100 person multi-disciplined
research organization within the University of California system. It is
located in San Diego, Calif. and is administratively within the jurisdiction
of Scripps Institution of Oceanography. The Optical Systems Group is
one of several inter-related research teams devoted to a variety of optical
and acoustical studies of the marine environment.
MPL
MARINE PHYSICAL LABORATORY
FIGURE 2
The Optical Systems Group of the Marine Physical Laboratory,
in support of several DoD sponsored programs, is developing a visible
spectrum, computer controlled imaging system for the automatic as-
sessment of the optical state of the atmosphere.
The following annotated illustrations describe the current ver-
sion of the whole sky imager (WSI) and its emerging database.

WHOLE SKY IMAGING NETWORK
DISCUSSION OUTLINE
1. OPERATIONAL STATUS OF NETWORK
2. RAW DATABASE STATUS
3. FIRST GENERATION CLD / NO CLD SOFTWARE
4. INITIAL STATISTICAL APPLICATIONS DEMONSTRATION

Image Acquisition & Analysis System
Hardware Block Diagram E/O System 5
FIGURE 3
SONY PVM 12710
MONITOR
The major E/O System 5 sub-assemblies are physically inter-
related as illustrated in this block diagram. The as-built configurations of
the exterior sensor housing is illustrated in the figures which follow.
GE 2710
SOLID STATE
VIDEO
CAMERA
TMI COMPUTER
(IBM/AT CLONE)
ARCHIVAL
AUTOMATIC
EQUATORIAL
SOLAR
OCCULTOR
ASSY.
This block diagram illustrates the system used for automatic
cloud cover archival and analysis. It is somewhat more powerful than its
sister system which is used for automatic visibility determinations.
110
VIDEO
IMAGE
PROCESSING
SUB-SYSTEM
(ITI FG 100)
SUB-SYSTEM
(SEAGATE 65 M bytel
H.D.)
REMOTE
CONTROLLED
IRIS
ASSY.
EXABYTE EXB - 8200
2.2 G byle 8mm CARTRIDGE TAPE SYSTEM
ANALOG
ACCESSORY
CONTROL PANEL
REMOTE
CONTROLLED
OPTICAL
FILTER
ASSY.
The key hardware items are the 2710 camera which provides an
RS-170 composite video output, the FG-100 Image Processing Board
which provides the digitized 512 x 512 x 8 imagery, and the EXB-8200
cartridge tape sub system which provides 2.2 Gigabytes of data storage
capacity.
STOWED KEYBOARD
EXTERIOR SENSOR
INSTALLATION
INTERIOR CONTROLLER
INSTALLATION
FIGURE 4
The as-built exterior sensor assembly is illustrated in the adja-
cent photograph. This relatively weatherproof housing contains the
fisheye lens adapter, its related relay optics and the necessary peripheral
items outlined in the preceeding block diagram.
The device runs in fully automatic, unattended mode under the
control of its associated computer system to provide multi-spectral
imagery customized for automatic cloud discrimination studies.

WHOLE SKY IMAGER DATA SITES
ME
WA
ND
MT
NH
MN
WI
NY
OR
SD
ID
FIGURE 5
MU
RU
WY
PA
+NJ
CT
TA
NE
OH
X
IN
WU
NV
VA
UT
СО
KY
НО
KS
NC
CA
TN
As of October, 1988 there are five Whole Sky Imager Systems
installed and operational as illustrated in the adjacent figure. The
geographical distribution is currently biased toward the southwestern
desert areas, however the sites in Missouri and Florida will be operational
by the end of calendar 1988 generating a better balance between several
different geographical and meterological regimes.
SC
AZ
OK
NM
AR
MS
AL
GA
TX
LA
FL
In addition to the fixed base sites illustrated herein, a portable
package is being developed to enable more convenient quick response
case study applications.
• OPERATIONAL
O PENDING
• WSMR/C
(KAFB
O MALABAR
• CLNWC
O COLUMBIA
WSMR/H
.MAFB
FIGURE 6
The exterior sensor installation at the White Sands Missile
Range Weather Observation Station (WSMR/C) is shown in the adjacent
photograph.
Note the WSI sensor is mounted in an auxilliary support pedestal
which provides both heating and cooling to the housing as required to
cope with the high summer temperatures and the colder winters associ-
ated with the various test sites.

FIGURE 7
A similar installation at the White Sands Missile Range (WSMR/
H) using an auxilliary trailer van rather than a building rooftop is shown
in this photo.
KITA
At all WSI field sites, the interior and exterior installations are
essentially identical except for the support structure provided by the host
agency.
WSI DATA TAPE STATUS - 30 SEPT '88
TOTAL DAYS IN FIELD
DAYS QC COMPLETED
FIGURE 8
BACKLOG
NO DATA
DATA DAYS
The accumulation of the WSI imagery began in February 1988
and has continued as illustrated in the accompanying bar chart. The
installation dates for each site are noted along the horizontal axis.
69
"
D
"
106
100
The “F” and “P” designators within the "QC completed " bars
indicate “Full” or “Partial” data days, where a Full day contains 12 hours
of imagery centered around Local Apparent Noon.
79
F63
E
32
22
16
16
2
C-STA
30 MAR
HELSTE
29 MAR
KIRTLAND CHINA LAKE MALMSTROM
17 MAY
23 JUN
29 AUG
WSI DATA SITES
COLUMBIA
ОСТ
MALABAR
NOV

WSI DATA ACQUISITION & ARCHIVAL SCHEDULE
TYPICAL HOURLY DATA SET -
FIGURE 9
20
40
50
30
MINUTES
The WSI system acquires imagery in accordance with the
schedule illustrated on the adjacent time line. A basic set of four images
is captured during the first eight seconds of each minute. These four
images, two red and two blue, are digitized by an 8 bit A/D converter and
subsequently stored at either 512 x 512 resolution or in a low resolution
subset of 33 rows and 33 columns.
ARCHIVAL KEYS: Assume 12 hr day (LAN + 6 hrs)
B Once each 10 min : 2 red + 2 blue Images @ 512 X 512 :: 288 images/day @ 262k ea = 75.5 MB/day
© Once each 1 min : 2 red + 2 blue Images @ 512 x 66 :: 2880 images/day @ 33.8k ea = 97.3 MB/day
DAILY TOTAL: 3168 images/day 172.8 MB/day
(264 images/hr)
(14.4 MB/hr)
NEO
MOCCA
FIGURE 10
An example of the digitized resolutions appearing in the raw
imagery is shown in the adjacent photograph. This image displays both
the full 512 x 512 resolution initially captured, and as an overlay, the 33
rows and columns that comprise the one minute subsets.
As noted in the Archival Schedule, full resolution 512 x 512
imagery is archived every ten minutes, and low resolution, 33 x 33
imagery is archived every minute

JAN
FEB
AY
SUNE
JULY
AUG
SEPT
OCT
NOV
WAME
W
.
WEEK OF
C-STA
INSTALLATION
MAINTENANCE
DATA QUALITY
DOWN
HELSTF
INSTALLATION
MAINTENANCE
DATA QUALITY
DOWN
FIGURE 11
KIRTLAND
INSTALLATION
MAINTENANCE
DATA QUALITY
DOWN
CHINA LAKE
INSTALLATION
MAINTENANCE
DATA QUALITY
The adjacent chart illustrates the overall performance of each
fieldinstallation. Computerized quality control procedures examine each
data tape as it is returned to the Laboratory, and assess the data's
completeness, stability and suitability for subsequent automatic process-
ing.
DOWN
MALMSTROM
INSTALLATION
MAINTENANCE
DATA QUALITY
AUG
DOWN
COLUMBIA
INSTALLATION
MAINTENANCE
DATA QUALITY
Hardware down times are also indicated.
DOWN
MALABAR
INSTALLATION
MAINTENANCE
DATA QUALITY
DOWN
TOTAL
• GOOD DATA
SPECIAL HANDLING REQUIRED
= SYSTEM DOWN
• INSTALLATION MAINTENANCE
BASIC IMAGERY
CORRECTED IMAGERY
COMPOSITE RATIO
DELIVERABLE DATABASE
CALIBRATION
FUNCTIONS
RADIANCE
CONVERSIONS
BLUE
IMAGE
512x512x8
FIGURE 12
CLOUD/NO-CLOUD
DECISION
ALGORITHMS
CALIB.
BLUE
IMAGE
RADIOMETRIC
LINEARITY
BLUE/RED
RATIOS
RADIOMETRIC
SENSITIVITY
UP-SUN
RED
IMAGE
512x512 x 8
CALIB.
RED
IMAGE
DERIVED
PRODUCTS
IMAGE RATIO
COMPUTATIONS
DOWN-SUN
Post archival data processing is organized to follow the general
procedures illustrated in the adjacent chart. The intent is to impose strict
quality and calibration controls upon the raw data, while simultaneously
reducing the number of images required for insertion into the optimized
cloud/ no cloud decision processes.
OPTICAL
DISTORTIONS
NEAR HORIZON
SPATIAL
DISTRIBUTIONS
COMPOSITE
BLUE/RED RATIO
IMAGE
OPTIMUM
CLOUD/NO-CLOUD
IMAGE
FIELD OF VIEW
DEFINITION
TWILIGHT
TEMPORAL
DISTRIBUTIONS
DAWN
SENSOR CHIP
UNIFORMITY
STATISTICAL
PARAMETERS
AS REQ'D
PIXEL SELECTIONS
FOR
OPTIMUMIZED
COMPOSITE
It should be noted that the functions indicated within the dashed
areas of the chart represent a necessary but not exclusive set of technical
operations.
(BLUE)/(RED)
RATIOS
REGISTRATION
ADJUSTMENTS
RED - ND
IMAGE
512 X 512 x 8
CALIB.
(RED)
IMAGE
FLUX CONTROL
THRESHOLDS
L-
WSI BASIC IMAGE PROCESSING FLOW CHART

38EUXS.
st:
TUINE: IKI: 2
TOT 1 Bium IN
NOK DUN/10
TIHK. 16:wii
IRIXON
CI-TIS
NIM
FIGURE 13
The adjacent photograph illustrates a red image acquired by the
WSI during its normal operational cycle. It is one of the four "basic
imagery" images archived every ten minutes.
This high contrast, whole sky image can be usefully employed as
the bogey "sky truth” image against which the automatic determination
of cloud/no cloud can be judged.
FIGURE 14
This false colored image represents the results of a computer-
ized, bit by bit discrimination of whole sky cloudiness. It was derived in
automatic mode by exercising the sequence illustrated in the preceeding
image processing flow chart.

IAT 12UKL THINN
TIHK-1M:11:
IRIS-MH
ULTI-MG
FIGURE 15
This photo illustrates a red image acquired under different cloud
conditions than the preceeding example. The accurate discrimination of
cirrus clouds is one of increasing interest, and also difficulty as the cloud
thinness approaches the sub-visual regime.
FIGURE 16
This false colored image illustrates the automatic system's
discrimination of thin cloud conditions. The ability of these early
automatic decision algorithms to function accurately in this more difficult
discrimination task encourages one to press forward even further with
this technique.
NO
ADVANTAGES OF THE WSI DATABASE
FIGURE 17
• HIGH TEMPORAL FREQUENCY
• GOOD SPATIAL RESOLUTION
The WSI image data provide many advantages over previous
data sets in determining the spatial and temporal variability of cloud cover
parameters such as CFLOS. Some of the items listed improve cloud
detection, while others will enhance the statistical descriptions of cloud
cover variability.
· CALIBRATED DIGITAL RADIANCES
• BOTH RED AND BLUE IMAGES
• MULTIPLE SITES AND SEASONS
!
!
!
!
!
.
W
WW
*
.
IMPORTANT STATISTICAL RELATIONSHIPS
.
*
........
• Frequency of CFLOS reflecting zenith angle and sky cover dependence
FIGURE 18
• Change in autocorrelation coefficient as a function of time
lag (relaxation time)
• Site to site correlation as a function of separation distance
(relaxation distance)
Models of down time duration for multiple site configurations
require parameterizations of many important statistical relationships,
including those listed here. The WSI data will provide the ground truth
needed to validate these models.


• Point, Line and Area interrelationships
• Recurrence vs. Persistence probabilities
• Combined multi-site probabilities
Bil
.
.
.
,
..
....
....
..
.
.
..
...
... ...
.
WSI IMAGE OBSTRUCTION LIMITS
FIGURE 19
NORTHERN LIMIT
OF OCCULTOR
This figure illustrates the orientation of the WSI images. The
curved line traces the path followed by a corner on the northernmost edge
of the occultor during a summer day.
ZENITH
W
150
H
ity
Cloud cover information was extracted during late June and
early August at C-station along the two line segments shown. The center
points of these lines are at the zenith angles indicated. Subset lines of
increasing length centered on the north axis were also examined. Tic
marks indicate the end points of the second, fourth and sixth subsets.
65°Hetty
N
SAMPLED SEGMENTS AT 15 & 65 DEGREE ZENITH ANGLES
O TO 10 TENTHS CLOUD COVER
1.00
FIGURE 20
0.75
-
Time autocorrelations of cloud cover at the center points of the
line samples at 15°(solid) and 65°(dashed) for all the ten minute images
in the 19 day case study period. These were computed using a
tetrachoric approximation described in:
-
-
wanna
-
-
-
-
-
AUTOCORRELATION
--
Panofsky, H. A. and G. W. Brier, 1968: Some Applications of Statistics
to Meteorology. Penn State Univ., University, PA, 224 pp.


0.00
.
-0.25
0
100
200
400
500
300
TIME LAG (MIN)
600
15 DEGREES
1.00
..
.
.
..
.
2
........
C
.
.
FIGURE 21
......................
0.75
6 - 10
AUTOCORRELATION
Time autocorrelations of cloud cover at the 150 sample point
categorized by total sky cover. The 0-10 line (solid) corresponds to the
total sample and is identical to the 150 line in Fig. 20. The 0-5 line
(dashed) is for the subset of cases with clear or scattered total opaque
sky cover at the earlier of the paired times. Broken and overcast cases
(6-10) are represented by the dashed-dot line. Note the crossover of the
latter two lines at roughly 3 hours (180 min.)
I
0.00
.
-0.254
0
100
400
500
600
200 300
TIME LAG (MIN)
..
.
O TO 5 TENTHS CLOUD COVER
.
..
FIGURE 22
650
AUTOCORRELATION
Similar to Fig. 20, only for the clear and scattered initial sky cover
cases. The rapid decrease at both zenith angles out to 4 hours and the
return to higher values after seven hours are driven by the strong diurnal
cloudiness cycle at C-station in July and August.


150
..
.
............
...
...
..
www.
0
100
400
200 300
TIME LAG (MIN)
500
600
6 TO 10 TENTHS CLOUD COVER
1.00
0.75
FIGURE 23
one
030
AUTOCORRELATION
Similar to Figs. 20 and 22, only for the broken and overcast initial
sky cover cases. Note the less rapid decrease in the autocorrelations at
650 zenith than at 150 during the earlier lags, and a larger difference
between the two lines at the longer lags than in the previous diagram.
SO
.
0.00
-0.25
100
400
500
600
200 300
TIME LAG (MIN)
..
.
.
............
...
.....
.
.
65 DEGREES
1.00
.
.......
.
0-10
FIGURE 24
6 - 10
AUTOCORRELATION
Same as Fig. 21, except for the 650 zenith angle sample. Note
the lines for the mostly clear and mostly cloudy samples cross at roughly
3 hours, just as they did for the 150 sample.


..
-0.25
0
100
200 300
TIME LAG (MIN)
400
500
600