UNIVERSITY OF CALIFORNIA, SAN DIEGO
UC SAN DIEGO LIBRARY
3 1822 04429 7497
OPTICAL SYSTEMS GROUP
TECHNICAL NOTE NO. 223
November 1990
Offsite
(Annex-Joi
rnals)
QC
974.5
T43
no. 223
Factors Influencing the Development
of a Short Term CFARC Prediction
Technique Based on WSI Imagery
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
SITY.
SAIN
Technical Sponsor, R. W. Endlich
HEL Support Branch
ut
LIG
• THE
RNIA
*1868
Hed
Prepared for
The U. S. Army Atmospheric Sciences Laboratory
White Sands Missile Range, NM 88002-5501
under contract No. N00014-87-C-0127
SCRIPPS
INSTITUTION
OF
OCEANOGRAPHY
MARINE PHYSICAL LAB San Diego, CA 92152-6400
UNIVERSITY OF CALIFORNIA, SAN DIEGO
1
1
11
III
11
II
11
LU
3 1822 04429 7497
TABLE OF CONTENTS
. . . . . . . . . . .....
1.0
2.0
3.0
4.0
Introduction ...........
............
...............
Background. . . . . . . . . . . . . . . . ......
Anticipated Sequence of Operations in the Predictive Algorithm ............
Important Considerations in Developing Short-Term CFARC Predictions ...
4.1 Variable Grab Rates...................................
4.2 Multisensor Arrays ..
4.3 Coordination with Other Cloud Information Sources ...
5.0
......
..........
.............
...
..
.....
Relevant Future Upgrades......
5.1 Use of Wider Angle Fisheye Lenses .
5.2 Active Ranging Devices ....
5.3 Image Acquisition at Night ..........
Concluding Remarks
va e cerere een AAN
6.0
Appendix A.
List of Figures .......
1.0 INTRODUCTION
TheLSTCIMager Display and Analysis Sub-System
(LIMDAS) located at the HELSTF facility at the White
Sands Missile Range currently provides near real-time
cloud cover imagery for the HEL Support Branch. The
system currently has the ability to looppreviousimagery
acquired at 5-minute intervals, providing a user with
some information concerning the motion of cloud
elements in the field of view of the WSI instrument. A
more reliable means to predict the short-term future
state of the cloud field is needed by those making the
decision to fire or not fire the laser. This note describes
the initial effort by the Optical Systems Group of the
Marine Physical Laboratory to develop prediction
techniques that could be incorporated into LIMDAS.
2.0 BACKGROUND
field near the target track, but also by the expected state
of the cloud field at some specified time in the future.
Two important questions to be answered by a predic-
tion scheme depend on the current state of the cloud
cover along the target track: 1) if the track is currently
cloud free, will it be cloud free at the time of firing, or
2) if clouds currently obscure the track, will it become
cloud free.
11
During periods of completely clear skies or complete
overcast, sky cover is often quite persistent, making a
short term prediction (less than 1 hour) almost trivial.
The sky cover climatology in Table 1 from Stallion Site
in the White Sands Missile Range indicate that, on
average, completely clear and overcast situations are
present over 40% of the time. On a month-by-month
basis, however, this total varies from a minimum of
17.2% during the summer rainy season, to a maximum
of 54.7% in February, a period dominated bymidlatitude
weather features, such as fronts, low pressure systems
and jet streams.
The presence of clouds can have adverse effects in
many applications, particularly when lasers are used to
direct enough energy at a target to damage or destroy
it. The presence of cloud along the track of a target is
thus an important parameter in the decision of whether
or not to fire such a laser weapon. Furthermore, since
it takes a finite amount of time and resources to prepare
certain lasers for firing, the decision-making process is
influenced not only by the current state of the cloud
At the other extreme are periods when sky cover
changes dramatically in very short periods of time.
Fig. 1 illustrates an example of such a situation as
observed in a set of daytime images acquired at Co-
lumbia, MO, in the spring of 1989. Several times on

Table 1
Monthly climatological sky cover amount frequencies (%) from
Stallion Site in the White Sands Missile Range, NM.
Sky Cover Amount (tenth)
2 3 4 5
0
1
6
7
8
9
10
9.0
7.0
6.5
5.3
6.1
7.3
5.7
3.7
4.8
5.9
4.9
6.8
6.5
3.5
2.8
5.1
4.5
3.8
3.3
3.0
5.1
5.7
8.3
3.6
5.1
6.5
5.3
4.5
4.1
5.1
6.6
6.7
6.9
6.2
6.0
6.7
January
February
March
April
May
June
July
August
September
October
November
December
8.7
28.4
30.4
26.0
32.0
29.8
29.9
8.3
13.6
27.1
42.0
31.7
28.
0
9.6
10.6
13.2
10.1
12.2
13.8
10.5
9.1
1 1.7
7.9
8.2
9.5
7.4
7.2
5.7
8.6
8.4
5.5
9.3
9.1
10.6
11.5
8.6
6.7
5.3
5.6
8.2
10.3
11.2
7.2
5.9
6.4
4.9
4.2
5.1
7.2
6.9
4.3
4.5
9.7
7.5
4.6
5.2
4.4
8.7
6.3
6.4
3.8
4.5
5.5
19.0
24.3
20.3
11.8
10.7
6.1
8.9
6.3
8.9
8.7
15.3
18.6
5.1
8.3
7.2
7.0
9.0
6.7
4.9
5.6
4.8
3.5
3.7
5.3
5.2
4.4
3.5
3.7
2.9
7.3
6.5
6.9
8.0
Combined
27.3
10.4
7.4
7.1
6.2
4.8
4.4
5.4
6.4
7.4
13.2
this day, the sky cover went from nearly clear to nearly
overcast in a matter of minutes. These dynamic situ-
ations are often associated with strong tropospheric
winds that propagate the cloud elements rapidly across
the whole sky images, presenting quite a challenge to
any short term prediction technique.
T he current LIMDAS runs as a slave off of the WSI
field unit that grabs images at a fixed time interval of
1 minute. As specified by the user, LIMDAS currently
performs a grab every 5 minutes. The additional time
is allocated to the ratio and cloud decision processing,
and the looping display. Straightforward upgrades in
software would allow more frequent grabs, but the
master-slave relationship with the standard WSI im-
poses an upper limit on the grab rate at once perminute.
Grabs at faster rates would also be possible ifLIMDAS
were to drive its own WSI, and if the downstream ratio
processing were delayed until a sequence of grabs were
completed.
Step 2: Compute the Red/Blue Ratios
Ratio processing now takes roughly 30 seconds for
a full resolution image. Using an accelerator board, the
ratio computations take only 8 seconds.
Step 3: Perform the Cloud - No Cloud Decision
The following sections describe by means of ex-
ample, some of the important factors that must be
considered in developing a cloud element prediction
scheme, and the work completed thus far by our group.
We assume that a wide range of other cloud informa-
tion from radar and satellite imagery is also available to
the decision maker, and that such information may
eventually be incorporated into the short-term cloud-
free arc predictive technique.
3.0 ANTICIPATED SEQUENCE OF OPERA-
TIONS IN THE PREDICTIVE ALGORITHM
At lead times of more than a few minutes, the
predictive algorithm will most likely be based on the
extrapolation of the translation of cloud element edges
into the future. At shorter lead times, an alarm system
might also be employed to warn of the approach of
cloud elements into the target region. Since the ex-
trapolation forward in time is the more complicated
procedure, the discussion here will concentrate on that
technique. The sequence of operations is as follows:
Step 1 - Grab the current red and blue image pairs.
Step 2 - Compute the red/blue ratios.
Step 3 - Perform the cloud-no cloud decision on the
ratio image
Step 4 - Transform the fisheye view to a pseudo-
Cartesian coordinate
Step 5 - Detect cloud edges and elements on a single
image
Step 6 - Combine with recent previous images to detect
translations
Step 7 - Extrapolate current cloud elements into the
future
Step 8 - Validate the prediction made for the current
time
LIMDAS employs our preliminary cloud decision
algorithm that uses fixed ratio values as thresholds in
delineating thin and opaque clouds, and computes the
percent sky coverage. Work is currently underway to
develop a new algorithm that allows the background
blue sky ratio to vary as a function of scattering angle
from the sun and distance from the horizon. The newer
technique would require more computing time than the
current version.
Step 4: Transform the Fisheye Image into Pseudo-
Cartesian Coordinates
11
Detecting the motion of cloud elements is an inte-
gral part of the prediction scheme. Consider a simple
case in which a single cloud element is passing directly
overhead, travelling at some constant velocity without
changing its shape. In the WSI fisheye view, radial
distance from the center of the image is roughly pro-
portional to zenith angle. As the cloud enters the image
at high zenith angles, a fixed displacement in space
over a fixed time interval will result in a smaller zenith
angle change than would be produced by the same
displacement later when the cloud is directly overhead.
Therefore, in the fisheye view, the cloud will first
appear to move slowly, accelerating as it passes over-
head, then decelerating as it approaches the horizon.
Extrapolation of cloud element positions forward in
time is far simpler if the elements appear to translate in
a linear fashion.
It is impossible to map a single fisheye image into a
true three-dimensional Cartesian coordinate, because
Steps 1 - 3 are already available on LIMDAS.
Software for Step 4 has been completed, but is not yet
incorporated into LIMDAS. Some samples of identi-
fying specific cloud features on consecutive images
and detecting their motion (Steps 6 and 7) are presented
in the following sections. The prediction and valida-
tion steps await further progress on the previous steps.
Step 1: Grab the Red and Blue Image Pairs
1
move more slowly than low clouds moving at the same
speed. Figures 3 and 4 compare a sequence of 4 WSI
images before and after the transformation has been
performed. The transformed images only include
points within the 65° zenith angle, outlined by a blue
circle in the fisheye images (Fig. 3). Note how the the
long, thin nature of the cumulus near the occultor is
changed in the transformation process. The transfor-
mation also enlarges the narrow cloud band along the
bottom edge of the 65° region.
Step 5: Detecting Cloud Edges and Elements
the fisheye projection preserves only the zenith and
azimuth angle for a particular target, and provides no
range information. A pseudo-Cartesian coordinate
must then be developed that can be defined from the
two geometric parameters available from the fisheye
projection, and, under most circumstances, preserves
the linear translation of cloud motion. The simplest
method is to rescale distance from the center of the
image from the roughly linear dependence on zenith
angle in the fisheye projection, to a dependence that
varies linearly with the tangent of the zenith angle. As
illustrated in Fig. 2, the tangent of the zenith angle is
proportional to the horizontal distance of the target
from the imager divided by the elevation difference.
Assuming a cloud remains at a fixed elevation during
its motion, the distance travelled in the transformed
coordinate will the be proportional to its horizontal
displacement in Cartesian coordinates. The transfor-
mation equations used in preparing this report are
provided in Appendix A.
The transformation does have some noteworthy
limitations. Table 2 presents the incremental area
ratios for the fisheye and transformation images (see
Appendix A). A pixel of unit area in the fisheye image
would thus have an area of 0.4 in the transformed image
at the center of the image, and an area of 5.02 units at
65°. In practical terms, the transposed image would
use only 2 out of 5 pixels at the center, but would repeat
the same fisheye pixel 5 times at the edge of the
transformed image. If the transformation were extended
to the outer edge of the fisheye images (near 80°), the
distortion near the outer edge would be even more
pronounced.
Given a single cloud decision image, boundaries
between regions with and without cloud must be de-
termined. During this process, pixel groupings that
form individual cloud and clear elements will also be
defined. That is, the cloud image indicates whether
each picture element (pixel) is cloud or no cloud; the
next step would be to automatically determine which
pixels belong to the same cloud element. The detection
of cloud edges and elements for the examples in this
report were performed interactively by a human ob-
server. Objective methods for performing these rela-
tively straightforward tasks can either be developed or
possibly extracted from standard image processing
programs.
Step 6: Combine with Recent Previous Images to
Detect Translation
Table 2
Incremental area ratios (transformed/fisheye)
Zenith Angle
Ratio
0°
0.40
10° 20° 30° 400 500
0.43 0.50 0.63 0.91 1.50
600 650
3.09 5.02
The motion of cloud edges becomes quite apparent
when a sequence of WSI images are viewed in a
standard loop. In order to quantify the motion, the
more difficult task of identifying the same cloudy or
clear element on consecutive images becomes impor-
tant. To further complicate matters, several processes
are responsible for the observed motion, including: 1)
bulk movement of the clouds, 2) their growth and
decay, or 3) some combination of both. Considering
the broad spectrum of cloud field combinations, and
rates of motion, the development of a reliable, fully
automated cloud motion detection system is no trivial
matter. For this report we have selected a few cases that
illustrate certain characteristics of cloud motion that
must be considered in a general algorithm. Video loops
from these cases are available on an accompanying
video tape.
The first case is from C-Station on 24 March 1989.
The period examined started completely clear. Several
individual bands of cirrus propagated from south to
north over the region, eventually producing a complete
overcast. The case was quite suitable for this prelimi-
nary investigation for many reasons. As noted in the

Two other features of the transformed images also
deserve attention. First, the vertical extent of the
clouds viewed may lead to confusion in the cloud
motion determination if the motion is assumed to be
horizontal in nature. Also, the apparent speed of cloud
element propagation in the transformed images is in-
versely proportioned to cloud height. Thus, high
clouds moving at some specified speed will appear to
coordinate transformation section, high clouds appear
to move more slowly than low clouds, enhancing our
ability to identify the same cloud element on full-
resolution images spaced 10 minutes apart. It is also
easier to identify along cloudedge thanmany individual
cloud elements. The case also exhibited several of
features relevant to this discussion: large feature (line)
motion, small feature (element) motion, and feature
evolution (line spreading).
period, and the propagation of the main cloud mass
from west to east is primarily due to the development
of new cells that amalgamate to form the larger cloud
mass. Only a slight driſt of the cells toward the north-
northeast is evident.
Steps 7 & 8: Extrapolate the Cloud Element Dis-
tribution into the Future, and Validate Previous Pre-
dictions
Fig. 5 shows a sequence of 4 images featuring what
appears to be two crossing cirrus lines. A discernable
edge has been highlighted, with the edge positions at
previous times being reproduced on later images. The
edge position from the image immediately preceding
this sequence (not shown) is included on the first
image. As expected, the translation of this line feature
is very uniform in the pseudo-Cartesian projection.
The average velocity of the line is at 45 pixels per 10
minutes toward the 3° azimuth. North (0° azimuth) is
toward the bottom of the images, east (90°) toward the
right, west (270°) toward the left and south (180°)
toward the top. In contrast, the motion of the crossing
point of the lines is much more rapid (85 pixels per 10
minutes toward the 62° azimuth), as shown in Fig. 6.
It is not unusual in meteorology to have large scale
features that move differently than the winds flowing
through them. For example, surface fronts often move
at a much slower rate than the winds, and squall lines
oſten propagate more slowly and in a much diſferent
direction than the individual cells within them. Such
is also the case in our example. The crossing point
motion is probably a good wind velocity tracer, while
the propagation of the cirrus band in general differs
considerably. A later sequence is presented in Fig 7.
Two individual cirrus bands have been outlined in
different colors. Note how the bands spread, showing
considerable overlap at the end of the sequence.
The predictive scheme used will be highly depen-
dent on the required lead time, the method used to
identify individual cloud elements, and the dominant
dynamic process for the current situation (such as line
vs. element motion). During its development, the
technique will evolve from requiring a high level of
user interaction at the beginning, to less and less
interaction in later versions. Growth and decay pro-
cesses will probably be modeled using some statistical
techniques. The accuracy of the statistical forecast can
be improved if growth and decay rates are continually
updated using current observations. These rates can
either be climatological in nature, or computed from an
earlier period from the same day or last couple of days.
4.0 IMPORTANT CONSIDERATIONS IN DE-
VELOPING SHORT-TERM CFARC PREDIC-
TIONS
II
Developing a predictive algorithm that can provide
useful cloud field predictions over the wide range of
meteorological conditions possible is not a trivial op-
eration. The motion of cloud elements within the WSI
field of view is the result of several different dynamic
processes. The discussion that follows outlines some
of the features that must be incorporated into a final
working technique.
4.1 Variable Grab Rates
LUI
The movement of cloud elements across the WSI
field of view is a function of both the wind motion in the
atmospheric layer in which clouds are present, and, as
shown earlier, the elevation of those clouds above the
instrument. The most difficult situation arises from
fast moving low clouds. The rate that images are
grabbed to best identify element motion falls within a
limited range. If grabs are made at too slow a rate,
identifying a particular cloud element is nearly im-
possible. For example, the choice of cases for this
report was limited to cases in which cloud propagation
was slow, so that features could still be identified in
consecutive 10 minute full resolution images. Grab
rates can also be too high. The individual elements
must move farenough to reliably estimate their motion.
The second case, from 9 July 1989, had overall
qualities similar to the March case, with a practically
cloud free sky changing to complete overcast in less
than an hour. Since cumulus is the predominant cloud
type, identifying particular cloud elements on the 10
minute images in Fig. 8 is difficult. To aid in element
identification, 1 minute images were examined. Fig. 9
shows a time series of 1 minute intervals over a 30
minute period for this case. (The time interval between
images is 2 minutes.) It is clear from this figure and
from the corresponding loop on the video tape, that
translation of elements is not the dominant process in
the cloud element evolution for this case. In fact,
several cumulus cells are seen to develop during the
1
TIS
The optimum grab rate for any two situations may thus
be quite different. Here user input will probably be
helpful. In terms of system design, there will be a limit
on how fast grabs can be made in the current WSI,
designed to provide multispectral images for ratio
processing. Ifmore rapid grab rates become necessary,
consideration would have to be given to performing
motion detection on raw radiance imagery, nominally
available at full video rates (30 frames per second).
4.2 Multisensor Arrays
increase in camera chip resolution to 1024 or 2048
pixels, compared to the current 512 pixel arrays, the
new lenses could significantly enhance our resolution
of cloud elements near the horizon. The image trans-
formation to pseudo-Cartesian coordinates could then
be performed to greater zenith angles, permitting an
earlier detection of cloud elements.
5.2
Active Ranging Devices
If the range to a particular cloud element could be
determined by an active ranging device, such as a lidar,
image transformations could be made to a true Carte-
sian coordinate. Map projections of the clouds could
then be produced and imagery from multiple sensors
could be better coordinated.
5.3
Image Acquisition at Night
Placement of the target track relative to the cloud
layer wind can have a significant impact on the reliability
of forecasts. In a single imager system, forecasts made
for tracks on the downwind side of the image will be far
more reliable than a track near the upwind edge of the
image, because reasonable estimates of cloud element
translation can be made before they impinge on the
target track. Placing additional imagers upwind of the
prevailing winds for the region may provide the in-
formation needed to derive better predictions for the
upwind target tracks. Additional climatological in-
formation regarding the prevailing cloud layer winds,
and frequency of particular cloud elevations would
affect location selection for additional sites.
4.3 Coordination with Other Cloud Information
Sources
Work is currently underway to develop a night
system. The best alternative employs a slow scan
camera. Thus, the lower limit on the grab interval
would be greater than that available during the day,
decreasing the forecasting ability for fast moving low
clouds at night. However, the benefits of acquiring any
data at night far outweigh the limitations.
111111
The decision makers already have satellite cloud
imagery at their disposal. The satellite images suffer
from poor time resolution. However, the satellite data
can provide a “heads up" warning in many situations.
Forexample, in the frequent cases with eithercompletely
clear skies or complete overcast, the satellite images
may detect large scale regions of clearing or cloud
approaching the site long before the clouds appear on
the WSI imagery. This information could then alert the
decision makerto pay more attention to the WSI, which
would fine tune the propagation rates.
5.0 RELEVANT FUTURE UPGRADES
6.0 CONCLUDING REMARKS
The results presented here are reasonably encour-
aging. We have demonstrated that the translation of
cloud elements can be reliably tracked on consecutive
images after projecting the WSI fisheye view into a
pseudo-Cartesian coordinate. While development of
short-term prediction capability not trivial, it does
appear tractable if sufficient long-term support is
available, and if a clear set of requirements are speci-
fied. Future upgrades to both the hardware and the
software of the system could significantly enhance the
forecasting capability. The development of a highly
automated version of the predictive algorithm is not a
simple task that can be solved by finding the right off-
the-shelf software. The feasibility of producing a
fully-automated predictive scheme, able to handle all
possible scenarios without any human interaction, is
still open to question.
Several of the following changes in our image
acquisition ability are currently under consideration.
Some of these could enhance our forecasting ability.
5.1 Use of Wider Angle Fisheye Lenses
The fisheye lenses currently used in our WSI sys-
tems have a nominal 160° field of view and are no
longer available. Possible replacements include 180°
and 220° prime lenses, with better focusing charac-
teristics than the current fisheye. Coupled with an
There are limits to a single WSI being able to
provide reliable predictions, particularly for longer
lead times. If longer and longer lead times are needed,
data from other sources, such as satellite images, be-
comes more important. Eventually, developing pro-
cedures for incorporating non-WSI information into
the decision-making process will have to be addressed.
Appendix A
Transformation from Fisheye to Pseudo-Cartesian Coordinates
The coordinate transformations used for this note
are based on the following expressions. Angles are
measured in degrees.
Fisheye:
R= [1.25(X, – 255)| +(Y, –240)?
= 0(3.032–.002590)
The only real difference between the two coordinate
systems is the R vs O relationship.
The procedure for constructing the pseudo-Carte-
sian image from the fisheye image begins with iden-
tifying a pixel (Xp, Yp) on the transformed image. A
corresponding (0, 0) pair is then determined. From
these the corresponding fisheye pixel (Xf, Yf) is
computed, and the value for the fisheye pixel is entered
into the transformed array at (Xp, Yp). Near the outer
edge of the image, a single value from (Xf, Yf) will be
used at several (Xp, Yp) locations.
A parameter called the incremental area ratio quali-
fies the ratio of the area mapped into the transformed
array from the fisheye array.
TU
0 = 585 (1-V1-.001128R,)
© = tan"" [1.25(x,-255)/(Y, –240)]
or inversely,
(AR)RE
208/ 285 / sin 0 )
cose)
(173.72 -8.50250) (173.72–17.050)
tan 65
X, = 255+0.8 R, sino
Y, = 240+ R, cosº
where 0 is expressed in radians. Results from this
formulation are given in Table 2 of the text.
Pseudo-Cartesian:
R2 = \[1.25(x,-255)* + (Y, –240)*
= 235(tan® / tan 65°)
R, tan 650 7
-1
0 = tan-
P
235
0 = tan-1
X, = 255+0.8 R, sino
Y, = 240+ R, coso
List of Figures
Fig. #
Figure Description
A sample time series of sky cover based on 1-minute WSI images from Columbia, MO. A
comparison with hourly meteorological observations is included.
A graphic depicting the relationships between zenith angle, cloud height and horizontal distance
from the imager used in developing the pseudo-Cartesian coordinate.
A time series of 4 WSI red/blue ratio images in the original fisheye view. The time interval
between images is 10 minutes. The blue circle is drawn at the 65º zenith angle, the outer edge
of the transformed imagery.
Same as Fig. 3, except it displays images transformed into the pseudo-Cartesian coordinate.
An example of cirrus band motion. An identifiable linear feature is outlined in blue. Previous
positions of the feature are also drawn, with the corresponding minutes appearing on the left edge
of the image.
The same set of images used in Fig. 6, only the crossing point in the X pattern is tracked using
green X's.
An example of cirrus band spreading from the same day as Figs. 6 and 7. The outer edges of the
leading band are colored blue, and the edges of the following band are colored orange.
8
Full-resolution 10-minute images from a summertime cumulus case.
WILT
9
1-minute subset imagery from 1908 Z to 1938 Z spaced 2 minutes apart from the same period
Fig. 3

**
*
**
XXX
*
ST
*
**
14
..
!
st
*
112
72get:
.
.
XXXN
**
**
20 April 1989
MPC
COTON
1.99
1V
NEX
KYY.
*
*
*
*
*
**
*
TULUI
TY
NO
22
***
2
PvXXXXXX
.
ITI0211
*
**
*
XXKDA
XXX
WIN
*
LX
*KET
XXX
*
**
*
TRIXX
**
2
***
*
RA
2020
2040
R
hia?
!!
IT
XXII TVEIRA:CAXXIS
W
**EUX
Zit
ta
3
Us
XXXN
RA
***
8. X
YE*:
.
P
*
BOX
EN
194.23
TYYNXXXXKXL
41.
IT.
F
***
*
Petit
ast
23 24
XXXX
**
****
sit
XXX
xxxxxx
*
**
****
yu
BAR
RUS
*
**
31*
*
XXX
2
XXXXXX
SARRER
2
2212
X
..
.
1
PELLET
34
CH
.
SE
***
***
*
XXX
X1 222.222
*
24.ee
*
**
*
EL
2
SUS
YS
*
*
******
foto
Writ
Xi
tit
2010
2030
14
126
WSMR
C Station
Fisheye view
0!
Fig. 4

*****
***
*
*
.
TY
X*XXXS
*
*
RE:
HE
2x
X2
XXXXXX
AYUHET
XXXX
20 April 1989
MPC
Taxi
* T*
*
OLL.
*
**
.*
2
X
XK XX1
***
*
*
XXXX
1707
TEKER
XXX
K
27S
ER
UDHIB
LU
*!
A
-25.
Y+*-
+
*
***
*
*
**
XI
*
**
2020
2010
TXT
TER
:
20
itit
*
KOHA
H
HER
SITE
.
*
***
*
2
TEX
.*
*
FXXU
.
S
Cros
2 XXX
*
*
**
LRT
28
1111
2.X
XXX
indi
111
ATTI
X
7
.3
SC
XXX
om
**
*
Oy
LET
212312
*
XXX
*
***
*
****
2
.1
OUT
T
Het
*
***
YX
***
****
*
SO
WS
**
*
x
2010
2030
.
K
1940
It
SX
XXX
H
Air
Pseudo Cartesian
C-Station
WSMR
.
SSSSSSS
***
HUTZ
Fig. 5

1659
1649
own
.
.
.
,
**
*
*.:.:
...
.
...
.
.
.
.
.
.
.
.
....
!
!!
!
!
!
!
!
.
!..
.
..
.
.
.
.
..
..
..
*
.
.
*
.*
...
*
T.
.
.
*
**
*
.
::
....:'
...........
*
.....**
...
VO**
-
X
m
.
***
..
O
ost
.
.
***
SO
*-..
.
.
.
.
*
.
11
T
**
*
**::.
.
+
**
:: Fr
.
.
.
**
**
W
XXXXX
**
**
*
..
.
XXXXXXX
#
klik
WA
OKIE
421
*
3
SAXO
**
***
-
1
*
2+
ta
...
*...
.
.
.
..
***
c*
.
.
.
st
Sex
EN
XX
ATA
*
****
***
**
**
1639
6291
.
!
!
.
..
+
.
i
i
.
.
.
.
.
.
.
.
.
.
.
.
.
i
.
..
...
.
.
**
*
*
.
***..
*****
*
.
..
.
.
.
.
:
..
:
.
+
+
,
*
**
**
.
.
..
.
.
*
..*...
**
1.
04
*..
.
*
:.:.
**
*
1
.
*
.
:
WWW
**
•
".
.
!
.
.
***
*
*
***
*
2
**
*
*
!!
.**
.
*
.
*
:
*.
.
*
e
*
:
.
.
*
.
**
*******
.
.
..
*
*
.
X
22
*
*
*
**
*
*
**
*
*
.
*
*
*
.
.
*******
*
*
A
2
.
XX
*
.
**
*
CS
3
1
::
.......
...
.
.
..
.
.
.
.
.
:::
*
M
**
*
*
..
!
.
***
***.
**
*
*
*
*
:
.
.:,:*******
*
.
1
.
XXX
*
.
...
.
-
.
.
.
.
.
.
.
.
.
.
***
.
**
*
.
***
********
.
.
.
..
MPC
11
..
.
Pseudo cartesian
C Station
WSMR
-
-
-
-
24 March 1989
Fig. 6

.
ASR
XXX
.*
XXXXXX
TV+
XX
**
4.
++
4:
+
+
24 March 1989
MPC
*
*
*
*
*
*
*****
******
**
nis
*
**
*
.
.
.
-
.
-,
*
*
*
*
*
.
.
.
.
*.*
.
2
..
*
**
*
*****
*
.
Ia
ER
.
***
*
::
1639
1659
*.. .**.
*...*...*
**
**
*
**
.
.
.
.
.
.
**
.
:
***
.
.
..
.
...
.*.
*.***.....
...
***
ti
.
200
.
.
.
+
.
:
..
.
.
*
*
****
*
*
*
**********.*:*:
***
.
.
.
.
.
.
.
.
..
***15 . 1
**** ..
*
.
.
..
.
. .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
li
.
:
.
*
*
..
. Ai
Kitong
I
X
*.
.
2
W
*********.:.:.:.::*39
39.
33
*
*
**
**
.
GLA
..
:::
**
**
**
:
*
b
....
:
...
*
.
++........
*
*
.
...
.
.
.
.
.
.
A
.
.
T-
:
..
*
*
*
. *.***:
***::: ::.**
.:.:.
0
.
....
.
*
..+
*
*
::
:.**
:.:.:.:.:.*
::::......
.
.
*
.
..
..
.
.
.
.
.
.
.
i
*
*
+
.
.
.
*
.
*
D
.
.
.
.
.
*
*
*
*
*
*
.
.
*
*
.
SO
.
.
.
1649
..
::::.
**
**
***
1629
+
:...
:*:
**
.
**
***********
*
*
**
*
***
.
*
.
******
.
***
*
..
..
.
.
.
...
.
.
.
.
**
.::.
- + +
+
+
+
+
+
+
+
+
+
+ +
+
+
--
-+
--
-
+ +
+
+
-
1
*
2
SE
*
*
*
..
.
..!
.
.
.
..
'
Pseudo cartesi31
C Station
WSMR
.
.
..****
.
.....
.
..
..
.
.
.
.
...
.
..
..
.
.
..
*
...
*..
..
.
.
.
.
.
..
.
.
.
+
+
+
+
+
+
+
+
+
+
+
+
+
.
.
.
.....
.
:,:***...
******
.
...+1
.
:
*
*
*
***
**
*
......
TOT
**
.
*
...
..
.X
.
. .
..
..
.
.
.
*
.......
.
XX
...
.
'
.
1...
.
22
Fig. 7

GELT
1729
***1. fibi
*
***
***
-1.
***
WYRWWX
*
612 LET
914
2
.
.
+2I.
.
vy
.
*
*
07
*
NANNY
2
WI
X
*
X
*
*
*
DE
2x
*
21
ext
t
.
*
2
TAKEN :14:1
XXXL
2
XX
101
**
*
$•
T
LE
**
**
*
*
PERS
HO
tet
Tom
..
*
*
*
TE
*
17.
*
*
*
VE
S
Hi!
KE
WINNO
1-18
*
Montil
XIX
2
XX
2
. W
t
.
2.21
GILL
1709
*RENTE. **********
*
!.
*
.
..
.
.
.
.
YA14
YXATUN
*
W
***
WAT
**
**
*
2
LEXE
3
XXX
****
SXXXX
*
ii.
*
:
*
***
CY
2
*
*
*
OS
+
**
ST
OS
.
AR
X
*
.
C
**
*
3
TITEL
.
3
22
**
a
ty
MTF
:
.
HTE
ST
27
ASE
**
A
*
M
37
.
RS
.
22
XXXX
2
HEM
ori
*
*
tis
RX
ALLINN
.
2
431 +14+
XXXX
TEMUAN
*
•15
+++++
MERITUD
***
..
.
HOAX
*****
***132*2
*
TES
***.XX
XL.
.
MPT
swivu
:
TUTVU
Pseudo Cartesian
C Station
WSMR
+
++
+++
24 March 1989
Fig. 8

X
NUX
XX
Tixit
22%2KO
VYNYX
K
2
NAARREXXX
T
***.
**
9 July 1989
*
*
**
*
***
MP
YA
* *
**
*****
C2EK2:4**
XXXX%.
XSUS SAX
3
*
*
*
*
7%
2
XXXX
TEST
RAS
XXX
SA
X
*
**
NE
2
.
SES
*
Zt
T.
2
22.
S
225
*
1948242
2
*
*
tit
Sex
+
X2
ES
*014
1920
1910
13:
ZX7
.
ANA
TTYY
TRIXIR
*
**
SXX
XXI
***
.
.
US!
*
**
:::
XXN
:::
CAN
*
XXXXXXXXXXXXX
..12
PER
2014
S
20
VYYS
X.
y
1.
14
**
*
"
RHITE
XIXLI
*
STA
21
*
22
*
***
2
XX.
*
t.
*
*
**
1
27
XXXXXXX
1.
XX
*
ex
XXX
*
*
'
1910
*
XX
*114
1930
.
..
*
*
**
*
*
F
*
XXY
XX.
TE
22
2
Pseudo cartesign
C Station
WSMR
XX
/
.
T
9
IT.
SUS
27
::
42.**
*
***
*
**
STATUT
IX
net
ut at
2S
SRETA
1919
1913
...!
: 119
1....t.
ア​..
本
​たれ
​Fig. 9

..
さ
​.
.
サイト
​11
.
.
.
….
イ
​This,
"
: 11
.
1
1998
14-1
T-
"-
すで
​7
レ
​・