MAN (alo
NASA CONTRACTOR
REPORT *%
NASA CR-474
<=
~
=
1
on
we
<—
[7a]
<=
=
| FEASIBILITY OF SEA SURFACE
TEMPERATURE DETERMINATION
USING SATELLITE INFRARED DATA
by James R. Greaves, Raymond Wexler, and Clinton J. Bowley
Prepared under Contract No. NASw-1157 dy
ALLIED RESEARCH ASSOCIATES, INC.
Concord, Mass.
for
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION » WASHINGTON, D.C. © MAY 1966
wun
eb [P00 ToEO”
LQ
IOHM/181N
NASA CR-474
THE FEASIBILITY OF SEA SURFACE TEMPERATURE
DETERMINATION USING SATELLITE INFRARED DATA
By James R. Greaves, Raymond Wexler,
and Clinton J. Bowley
Distribution of this report is provided in the interest of
information exchange. Responsibility for the contents
resides in the author or organization that prepared it.
Prepared under Contract No. NASw- 1157 by
ALLIED RESEARCH ASSOCIATES, INC.
Concord, Mass.
for
NATIONAL AERONAUTICS AND SPACE ADMINISTRATION
For sale by the Clearinghouse for Federal Scientific and Technical Information
Springfield, Virginia 22151 — Price $3.00
ABSTRACT
A first investigation, based on actual infrared data, has been made of the
feasibility of observing sea surface temperatures or temperature gradients from a
satellite. It has been found that the satellite-measured large-scale patterns are
generally persistent, with only limited day-to-day changes,and that the smaller
scale features are similar to those observed in intensive conventional measurement
programs over relatively restricted areas. These and other evidence suggest that:
1) The satellite data provide good measurements of the gradients of sea
surface temperatures.
2) At least one conventional sea surface temperature measurement is
required to serve as a benchmark if good absolute values of sea surface temperatures
are required. This is due to uncertainties in atmospheric attenuation of infrared
radiation and in sensor calibration and degradation for the TIROS radiometer.
3) In many cases, the satellite data appear to be detecting synoptic scale
changes which take place over a period of one or more days.
A critical problem is the detection and elimination of data points where
clouds prevent an uncontaminated view of the sea surface. It is found that, in day-
time, the TIROS Channel 5 data, converted to an albedo, are suitable for detecting
cloud contaminated points. No method for reliable cloud detection at night is
presently apparent.
It is found that other sources of cloud cover information, such as conventional
meteorological observations and satellite TV data, are insufficient for determining
all the significant clear areas for satellite observations of sea surface temperature.
This suggests that comprehensive studies or operational analyses of sea surface
temperatures will require the processing of relatively extensive quantities of
satellite radiation data, using these data themselves to determine the daytime cloud-
free areas. In this manner maximum use can be made of whatever cloudfree areas do
exist both on an individual day basis, and for the averaging of data for short periods of
more or less successive days. A single case (for a period of six days) is analyzed
using this approach, and strongly substantiates these conclusions. It is recommended
that a more extensive set of pilot studies, using the approach of the processing of
relatively extensive samples of data, be conducted. \
ital
FOREWORD
This report by the ARACON Geophysics Company, a division of Allied Research
Associates, Inc., Concord, Mass., presents the results of a first study of the
feasibility of observing sea surface temperatures, or temperature gradients, using
satellite data for an infrared window. These studies were performed for the
Meteorological Programs Division, Headquarters, National Aeronautics and
Space Administration under Contract No. NASW-1157.
Acknowledgement is due to Mr. Richard G. Terwilliger of the Meteorological
Programs Division for his continuing interest and assistance during the course of
this investigation; and to Messrs. William R. Bandeen and Robert Hite of the Physics
Branch, NASA Goddard Space Flight Center, for their cooperation in processing and
providing sample cases of the otherwise unavailable TIROS "real time''IR data.
The authors further wish to acknowledge the kind assistance and encouragement
offered by Messrs. Joseph Chase and Robert Alexander, and other scientists at the
Woods Hole Oceanographic Institute.
Mr. David Chang of ARACON Geophysics Company performed the atmospheric
attenuation calculations for most of the cases discussedherein. Dr. William K. Widger
also of ARACON provided useful suggestions throughout the study, and assisted in the
editing of this report.
atatal
Ap
it
'
YE Se
TABLE OF CONTENTS
ABSTRACT
FOREWORD
LIST OF FIGURES
1, INTRODUCTION
1,1 The Importance of Sea Surface Temperature Data
1.2 Use of Satellite Measurements
1.3 Brief Summary of Results of These Studies
2. DATA SOURCES
3. SELECTION OF DATA
3.1 Choice of TIROS VII
3.2 Case Selection Criteria
3.2.1 Clear Skies Criterion
3.2.2 Temperature Gradient Criterion
3.2.3 Conventional Data Criterion
4. DATA PROCESSING
4.1 General
4.2 Elimination of Cloud Contaminated Points
4.2.1 Daytime Cases
4.2.2 Nighttime Cases
4.3 Atmospheric Attenuation
4.4 Sensor Degradation
5. CASE STUDIES
5.1 North Atlantic, Labrador Current
5.2 Western Pacific Cases
17
21
TABLE OF CONTENTS (Cont. )
Page
5.2.1 Kuroshio Current Case 21
5.2.2 Yellow Sea Case 23
5.3 Consistency Tests 24
5.3.1 Sea of Okhotsk Case 24
5.3.2 Gulf of Alaska Case 28
5.3.3 Indian Ocean Case 31
5.3.4 Results of SST Consistency Tests as Compared to
Other Studies 34
5.4 Real-Time Studies 37
6. DISCUSSION OF THE PROPOSED EXTENSIVE DATA PROCESSING
APPROACH 4]
7. GENERAL SUMMARY AND CONCLUSIONS 44
REFERENCES 46
vi
Figure No.
LIST OF FIGURES
Comparison of Channel 5 and Channel 2 - 4 Difference
at a Cloud Boundary
Clear-Sky Temperature Corrections
North Atlantic, Passes 1126 and 1111
North Atlantic, Pass 1082
Kuroshio Current, Pass 779
Sea of Okhotsk, Pass 51
Sea of Okhotsk, Pass 65
Sea of Okhotsk, Pass 95
Gulf of Alaska, Passes 735 and 748
Indian Ocean, Pass 706
Indian Ocean, Night-time Pass 744
Sandy Hook Marine Laboratory Airborne IR Mappings
Real-time Data, Pass 107-D
Real-time Data, Pass 238-D
Six Pass Averaged SST Patterns
vii
Page
12
14
18
20
22
(42)
27
29
30
37
33
36
38
40
43
1. INTRODUCTION
1.1 The Importance of Sea Surface Temperature Data
The need for accurate measurement of the sea surface temperature(hereafter
referred to as SST),on a daily basis and for large areas of the world's oceans, is
becoming increasingly acute. Nearly all of man's sea-oriented activities (such as
fisheries, navigation, marine weather forecasting, and naval operations) rely to
some extent upon knowledge of the SST patterns and their variation with time.
Prediction of the spawning grounds of certain varieties of fish can be made using
accurate SST information, while good fishing areas frequently occur at the confluence
of oceanic currents which can in turn be detected by the packing of sea surface
isotherms. Determinations of the probability of icing or fog formation, require
1 é
» 25 onenre demonstrated direct correlations
SST data. Various investigators
between the SST and such meteorological phenomena as the growth and travel of
hurricanes ,and extratropical cyclonic development. The SST can be related to the
subsurface temperature structure, which in turn affects the propagation of sound,
a significant parameter in submarine warfare.
Despite the importance of SST data, this quantity is generally known only ina
gross climatological sense, and even then not adequately over all the oceans of the
world. Temperature distributions and their variations on the time and space scales
required for the needs outlined above are far less well determined.
1.2 Use of Satellite Measurements
As indicated above, the theory and techniques which can make practical use of
greatly improved SST information are at hand; what is lacking is an adequate amount
of data. A satellite equipped with IR sensors is a world wide observing tool which
may be used to considerably close this ''data gap". 2 This first study of the feasibility
of such an approach employed principally the Channel 2 (8-12 micron) radiometer
data from the TIROS VII meteorological satellite. A polar orbiting, sun-synchronous
satellite such as Nimbus, which could provide daily coverage of all the earth's oceans,
can drastically improve the capabilities demonstrated in this study. While the
Nimbus I HRIR (3. 7u) has serious limitations for this application (see Section 3. 1),
the Nimbus C MRIR should be an improvement over the TIROS data in many respects,
including coverage, sensitivity, calibration, and atmospheric attenuation.
It seems significant that when scientists at the Woods Hole Oceanographic
Institute were consulted regarding certain aspects of our investigations, they
unanimously expressed considerable enthusiasm that such a feasibility study was
taking place. They stressed the long standing lack of adequate data of the scale
and measurement frequency that only a satellite can practically provide.
1.3 Brief Summary of Results of These Studies
These studies have conclusively demonstrated that a satellite can measure
pertinent patterns and gradients of SST where cloud cover does not prevent the
radiometer from observing the ocean surface. When cases of repeated coverage
of a given area within a period of several days were analyzed, large scale
temperature patterns were found to have significant persistence. While smaller
scale features often changed somewhat from day to day, their patterns were
repeatedly similar to those observed in special programs which used conventional
measurements.
The absolute temperatures were occasionally uniformly lower or higher from
day to day. These discrepencies could not always be completely accounted for in
terms of sensor degradation or changing air masses. An occasional conventional
temperature measurement is apparently required as a bench mark to properly
calibrate the IR measurements, at least until improved techniques for in-flight
sensor calibration, and improved knowledge of atmospheric attenuation, are
available.
One of the primary deterrents to observations of sea surface temperatures
or temperature gradients is the presence of intervening cloud cover. When clouds
partially fill the field of the sensor, that data point must be identified and discarded.
Our studies have indicated that this problem can usually be solved by establishing a
cloud-no cloud threshold value for the Channel 5 (0.5-0.7 micron, visible reflected
radiation) albedo listing, and discarding points for which the Channel 5 value exceeds
this threshold.
Significant difficulties were encountered in locating even single cases of open
mode IR data which simultaneously fulfilled the combined criteria of (1) mostly
clear skies, (2) distinct SST patterns or gradients, and (3) adequate conventional
ship coverage for comparison. The problems obviously multiplied when attempts
were made to find examples of recurrent coverage of a given area for purposes of
pattern persistence studies which could serve as one confirmation of the validity of
the observations, and particularly for observations of synoptic scale changes in
temperature patterns.
A review of these problems has led to the conclusion that the only feasible
approach to a comprehensive study of satellite observed SST appears to be the
computer processing of FMR tapes without specific prior knowledge as to whether
a given pass will contain useful data. In this approach, the Channel 5 albedo data
would be the primary source for determining adequately cloud free areas. While
this approach is inherently inefficient (since some of the processed data may include
not usefully cloud free areas) there would appear to be no alternative.
An initial trial run of this approach was made for an area off the western
coast of Australia, averaging clear sky data from six more or less consecutive
passes. The average map showed obvious similarity to the mean monthly charts
for this area. This case and others will be discussed in greater detail in the
following sections.
2. DATA SOURCES
The primary sources of data for all cases considered were the Final
Meteorological Radiation (FMR) magnetic tapes prepared by NASA. Listings of
the available data tapes and descriptions of the data contained therein have been
published. ° Extraction of the actual data from the FMR tapes is accomplished by
processing programs developed by NASA for use on the IBM 7094.
The processing can make the data available in any of several formats. The
listing program (MS-500) converts the binary FMR data to decimal form and
prints it in a convenient tabular format. This tabulation includes data point
location information, and values measured by each of the five channels of the TIROS
radiometers. Mapping programs are available which map the data to various
scales in either a mercator projection (MSC-2) or a polar stereographic projection
(MS-501).
3. SELECTION OF DATA
3.1 Choice of TIROS VII
All of the cases selected for investigation used the TIROS VII data. This
satellite was chosen because the degradation of the Channel 2 sensor (8-12 micron)
is well known. ¢ Moreover, because IR sensor degradation generally increases
with time, an attempt was made to the degree feasible to choose cases early in
the lifetime of TIROS VII (TIROS VII was launched 19 June 1963).
It was determined that the use of the Nimbus HRIR data during the current
study weuld not be feasible for the following reasons:
l. The availability of these data is extremely limited at the present time,
and until the digitization is completed, full documentation of the magnetic tape
format (to appear as Nimbus I High Resolution Radiation Catalog and Users' Manual,
Volume 2, ''Nimbus Meteorological Radiation Tapes - HRIR"') is not available.
2. There appears to be no reliable way to discern the presence or absence
of cloudiness, other than by abrupt changes in the recorded temperature, which
may not detect scattered to broken or low cloudiness. The daylight HRIR cannot be
expected to be applicable because of the reflected component. The problems of
detecting partial cloud cover at night from infrared data have no apparent solution,
as will be discussed in some detail in Section 4. 2. 2.
3. Those areas which appear clear in the Nimbus I HRIR photofacsimile film
strips are of minimum interest insofar as the detection of sea surface temperature
patterns are concerned. The probability of finding an area with distinctive patterns
which was known to be clear and has presently available digitized Nimbus I HRIR
data seemed too small to justify an exhaustive search.
It should be emphasized that the increased resolution capabilities of the
Nimbus HRIR data represent a potentially significant advantage over those of the
TIROS IR, and any future investigation of sea surface temperature determination
should reconsider the Nimbus data when they become available in digitized form.
However, the success of such a study will depend on solving the problem of detection
of small or low clouds. Because of the HRIR resolution, even very small isolated
clouds may be a greater source of error than they would when using the lower
resolution TIROS data. One possibility may be a joint use of the digitized data
and the analog visicorder record, as discussed in our proposal for this study, 2
but not attempted for the reasons discussed above.
3.2 Case Selection Criteria
As has been mentioned previously, much time and effort was spent in
attempting to locate cases which simultaneously satisfied the combined criteria of
(1) generally clear skies, (2) significant SST gradients, and (3) adequate conventional
ship coverage to provide data for comparison. Another significant restriction is
that, for all practical purposes, periods of closed mode IR operation must be
avoided due to the frequent errors in geographical location of the data. This, of
course, is simply a limitation of the measurement device itself.
3.2.1 Clear Skies Criterion
In our initial case selections, IR passes were chosen on the basis of generally
clear skies as determined from nearly simultaneous TV coverage by TIROS VII or
other meteorological satellites. Cases were selected by first examining the
nephanalyses contained in the Catalogs of Cloud Pinckommaplsr’. and then from an
examination of the actual satellite pictures for areas that the nephanalyses suggested
as promising. This approach, combined with efforts to concentrate on areas with
significant SST patterns, led to an obviously insufficient number of usable cases. In
particular, it excluded IR passes with no essentially concurrent TV data. Later in
the study, attempts were made to use conventional weather maps over the ocean as
a means of determining the cloudiness of an area. Since the conventional maps were
rarely drawn for the precise time in question, this meant interpolating cloud cover
between maps which were already based on sparse information. After reviewing
the amount of time required to arrive at some conclusion regarding the cloudiness
of an area from conventional data alone, and the undependable nature of the
conclusion, this approach was rejected.
The remaining alternative was to determine cloudiness from the IR data
themselves. This means, unfortunately, incurring the expense of FMR tape
processing before it can be decided whether or not the area scanned includes adequately
cloud free portions. The methods of determining cloudiness from the IR data will be
discussed in a later section.
This method was used as the sole method for determining cloud cover in
only one set of cases, that discussed in Section 6.
3.2.2 Temperature Gradient Criterion
The second criterion for case selection in this study, that of significant
SST gradients, proved to be generally incompatible with the first. Areas of
significant SST gradients (where observable synoptic scale activity seems most
likely) usually are areas of persistent or recurring meteorological fronts, and
so of obscuring cloud cover. (The comparative locations of the Gulf Stream and
of the mean position of the western North Atlantic frontal zone provide one illustration
of this problem.) There are several such ocean areas, but only daily observations
over a period of several days to a few weeks can be expected to show significant
SST changes.
3.2.3 Conventional Data Criterion
The third criterion of adequate conventional SST data could not in general be
met. (If it could, much of the need for satellite measurement of the world's SST
patterns would vanish.) Even in the North Atlantic cases, the number of conventional
ship reports proved inadequate. The only area of significant SST gradients where
there appear to be sufficient available ship SST measurements is in the vicinity of
the Gulf Stream off the east coast of the United States. But it is precisely here
that there is seldom readily available IR data.
Most of the data for this area, for the immediate west coast North Pacific
waters, and for much of the continental U.S. are the so-called ''real time'' TIROS IR
data and remain in undigitized form. These data are read out in ''real time''’ concurrent
with the playback of the IR data for the remainder of the preceding orbit, using a
frequency multiplexing technique. The data for the remainder of the orbit are read
out at 30times the record speed and frequency, and the automatic processing system
is designed to handle these data. Processing the ''real time'' data, whose sub-
carriers are thus one-thirtieth of those of the majority of the data, require an
accelerated rerecording, and considerably more time and manual attention than
does the conventional IR data. Moreover, timing errors of the order of seconds are
very likely to be introduced because of the manual determination of the ''End of Tape"!
time at the CDA stations. For these reasons, only three cases of "'real time'' data*
were included in our studies. We suggest that the reduction of a sizable sample of
* Made specially available, through the courtesy of Messrs. William R. Bandeen
and Robert Hite of GSFC.
the ''real time'' data on a routine basis would be of significant value to meteorological,
geological, and especially oceanographic research; and might well be given serious
consideration.
4. DATA PROCESSING
4.1 General
The data were generally extracted from the FMR tapes using the listing
program and were hand plotted onto blank mercator mapping paper. In so far as
feasible, these plotted data were corrected for data misplacement errors, following
guidelines suggested by Sherr. 9 In his report, Sherr discusses several classes of
misplacement errors together with methods for their recognition and correction. In
the mapping format, these errors cannot be detected, and hence may introduce some
pattern distortion.
Initial analyses were performed without reference to existing maps of mean
monthly SST patterns or local currents. Later the analyst consulted maps of mean
temperature patterns and currents for the area in question, and used them to guide
the analysis, particularly in areas where the plotted data were ambiguous and
permitted a choice. In no case was violation of the plotted data permitted. These
subsequent re-analyses of previously run cases resulted in a significant improvement
in the correlation between IR and conventional temperature patterns without violation
of the data.
When scientists at the Woods Hole Oceanographic Institute were consulted
regarding the resultant maps, it was noted that the IR temperature patterns were in
many ways similar to those seen in intensive conventional measurement programs
over relatively restricted areas.
The Woods Hole personnel also suggested that, in cases of particularly noisy
data or nearly isothermal areas, averaging over 1 squares or larger, and possibly
over repeated passes, would often be appropriate. Such averaging must be carried
out with some care,or small scale features and their changes in time will be quickly
lost. The validity of this suggestion will be demonstrated in several of the case
studies discussed later in this report.
4.2 Elimination of Cloud Contaminated Points
Whenever a cloud or portion thereof falls within the field of view (approximately
thirty nautical miles diameter at zero nadir angle) of the TIROS radiometer, it will
prevent accurate measurement of the surface temperature. In these cases, the
temperature recorded by the radiometer would be intermediate between that of the
cloud and that of the sea surface with the relative contributions of the cloud top
and surface temperatures depending on the percentage of cloud within the field of
view.
4.2.1 Daytime Cases
In daytime passes, both Channel 3 (0. 2-5 micron) and Channel 5 (0.5-0.7
micron) measure the comparatively intense reflection of solar radiation from the
cloud surfaces, and thus can serve as a discriminator between cloudy and clear
areas. For the TIROS VII data, Channel 5 exhibits somewhat less noisy characteristics
than Channel 3, and was used throughout the study.
Before the Channel 5 data can be used for this purpose, a clear-cloudy threshold
value must be determined. Channel 5 data as usually listed represents the irradiance
observed by the sensor in the 0. 5-0. 7 micron (visible) spectral region. This observed
value, however, depends on the solar zenith angle as well as the cloud reflectivity and
in this form can be used only by noting abrupt changes in its relative value. A more
significant parameter is obtained when the Channel 5 value is converted to the albedo.
The albedo is determined from the equation
I
N= Il cos 0
fe)
when @ is the zenith angle of the sun (which is included on the FMRT for each set of
data points), Iis the radiance measured by Channel 5, and te is the radiance that
the satellite would measure from a perfectly reflecting isotropic surface illuminated
by the sun at its zenith and with no intervening absorption or scattering. The value
of I, which is a constant, may easily be calculated from the solar constant and the
known properties of the sensor. In the albedo format, the effect of the solar zenith
angle is removed, and thus comparisons from one part of a pass to another can be
more easily made. Techniques for the calculation and use of this format were
developed midway through the study, and it was used for all subsequent cases.
To determine the required no cloud-cloud threshold, cases were chosen using
satellite photographs, in which a distinct cloud edge running nearly perpendicular to
the IR swath lines was observed. It was noted that, over clear ocean areas, the
measured albedo ranged from 0 to 10 %, while over overcast areas the albedo ranged
from 30 to 70 %, or even higher. Threshold values were chosen in the range of
from 10 to 20 % (the exact thresholds used in each case are specified in the subsequent
case study discussions). Trade-offs must be considered between the greater safety
10
of lower threshold values (which provide greater probability of eliminating all cloud
contaminated points), and the improved coverage when higher values are established.
These results are in agreement with the findings of an earlier study by Wexler.
Using the early orbits of TIROS III data, Wexler found an uncorrected average
Channel 5 albedo of 15 % for the range of concurrent Channel 2 temperatures from
242 to 268°K (clear conditions were assumed for temperatures greater than 268°K,
and an overcast was assumed for temperatures less than 242°K). Using the sensor
deterioration information presented in the TIROS III and TIROS VII Radiation Cat-
Bloeues 0’ oF this was found to be equivalent, during the early life of TIROS VII,
to a 24 % average albedo for approximately 50 % cloudiness. Thus, the 10-20 %
threshold values employed for individual data points should eliminate a very large
proportion of the seriously cloud contaminated data points.
4.2.2 Nighttime Cases
During nighttime passes, no meaningful measurements are available from
either Channel 3 or Channel 5. An investigation was made of the difference between
Channel 4 and Channel 2 measurements as a cloud detection mechanism. This
difference is relatively constant at a given temperature, but should decrease in the
presence of high water vapor amounts. Daylight cases were used to attempt to
correlate these differences with the presence or absence of clouds. Unfortunately,
but not totally unexpectedly, the measured differences showed no detectable changes
in the transition from clear to overcast conditions. This seemed due primarily to
the high noise level of the Channel 2 minus Channel 4 difference, relative to the
peak to peak range of this difference. The estimated short term relative accuracy
of these differences was + 3°K, while the entire range of the difference values
rarely exceeded g°K. The failure of this approach may, however, also be due to
the lack of sufficient or sufficiently abrupt humidity variations over a partially cloudy
area or at a cloud boundary.
Figure 4-1 shows data from a typical daytime swath across a cloud boundary.
Channel 5 shows an abrupt change, while the Channel 2 minus Channel 4 difference
reveals no obvious breaking point.
It was consequently concluded that the use of nightime IR measurements to
eliminate cloud contaminated points from SST data is impractical at the present time.
Unfortunately, this result approximately halves the number of usable cases over a
given area. There is no reason for believing the Nimbus MRIR data will be sig-
nificantly better in this regard.
10
AGL
12
Channel 2 minus Channel 4
TIROS VII
IR Orbit 1081
31 August 1963
( watts /m2)
18
16 | Cloudy
14
ine)
@
Channel 5 ( watts/m?)
ro)
Data Points
Fig. 4-1 Comparison of Channel 5 and Channel 2-4 Difference ata
Cloud Boundary
4.3 Atmospheric Attenuation
The radiation emanating from the earth's surface or clouds is attenuated by
atmospheric absorption; hence,the satellite sensor measures radiation with an
equivalent black body temperature, T less than the temperature, T, of the
E’
radiating surface. The difference, AT = T-T depends on the amount of absorbing
gases between the radiating surface and the oe of the atmosphere,and so, in general,
increases with increasing nadir angle of view of the satellite sensor.
The absorbing gases in the atmosphere which affect the measurement from a
satellite of terrestrial radiation are water vapor, carbon dioxide and ozone. Water
vapor has a strong vibration-rotation band centered at 6. 3, and rotation bands
beyond 20; CO, has strong bands centered at 4. 3u.and at 15y; O, has a strong but
relatively narrow band centered at 9.6. The region 8 to 13, to which Channel 2
of the TIROS satellites is sensitive, is termed a 'twindow!'! region, because of
relatively little atmospheric absorption. However, it is a "dirty"! window at best
because of absorption by Oz, and because of the residual absorption by H,0 due to
many small bands scattered throughout the region.
It is possible to calculate the amount of atmospheric absorption if the distri-
bution of absorbing gases in the atmosphere is known. The problem is complicated
because the absorption varies with pressure and temperature in a non-linear fashion.
The calculations utilize experimental measurements of absorption by atmospheric
gases. However, because environmental conditions in the experiments are different
from those in the atmosphere, theory must be applied. Good results have been ob-
tained by use of the Curtis approximations except in the case of ozone.
Computations of the outgoing radiation for 106 model atmospheres have been
made by Wark et al. 3 For clear conditions, water vapor was the primary variable.
A correction for an average ozone distribution was made; the error due to different
ozone distributions is likely to be relatively small. The results for clear sky
conditions are shown in Fig. 4-2 for nadir angles of 0° and 45°. In these diagrams,
the ordinates are the amount of precipitable water in the atmosphere. Horizontal
lines indicate "average" conditions (w= 2.25 cm) and "moist" conditions (w = 3.9 cm).
The nearly vertical lines are labeled T the equivalent black body radiation as
Ee’
measured by the Channel 2 sensor. Thus for a 0° nadir angle with Tee 290° the
correction to be added to obtain the temperature of the radiating surface is 6. BiG
for average humidity and 9°C for moist conditions; at 45° nadir angle the corrections
are 9.5 and 12°C.
13
¢l dl VI
$U01}99110D ainjOsadway AyS-4Da| 2-p BI4
(Ho) LV
8
2
/
9 G
(Ho) 1V
|
a — | ———
—
\)
%
A
==
i.
°Oz5
=
oSb = albuy 4IpoN
oO = a|buy 4!1PpON
( 48}0M aqo}Idisasd WO) M
14
The humidities should be determined from a nearby, essentially concurrent
radiosonde observation. In the absence of such humidity data, approximate corrections
can be estimated from synoptic-climatological considerations, or even from
climatological data.
In addition, it has been noted that the temperatures measured by Channel 2
and corrected for atmospheric absorption (and also degradation; see Section 4. 4)
are frequently lower than the actual surface temperature. Me The discrepancy has
been attributed to thin ice crystal clouds or layers of atmospheric particulates not
visible from the ground or in the TIROS TV pictures.
4.4 Sensor Degradation
It has been recognized that the TIROS sensors have been subject to degradation
which increases with time after launch. Corrections for this effect have been pro-
vided by NASA in the TIROS VII manuals. These corrections were based on changes
in the average signal output with increasing orbit number. For Channel 2, the
corrections are shown in Rigs 78 of the TIROS VII manual. ° For a sensor reading
of 300°K, the NASA correction is 2°K at orbit 780 and 4°K at orbit 1500.
An examination of sample printouts of outgoing radiation, as prepared from
Channel 2 data by the National Weather Satellite Center of the U.S. Weather Bureau
(now of ESSA), was made in order to determine the adequacy of the NASA corrections.
The analysis was restricted to clear areas of the tropical Pacific between 5°S and
15°s, where the ocean temperature was anticipated to remain reasonably constant
over the period of interest (June to September 1963). Any change could thus be
attributed to degradation. Although there were some discrepancies, it was concluded
that, by and large, the NASA corrections were approximately valid.
A periodic degradation of Channel 2 was found by NASA to be related to the
orbit-sun phase geometry (Fig. 75 in the TIROS VII manual). The NASA charts
showed that, over a 76 day period, errors could be as much as 3° greater or less
than those shown in the average degradation curve. A correction for the effect would
require determination of the magnitude of the error for individual days during the
76 day period. Corrections to this degree of precision have not as yet been published
(the scale in the previously mentioned Fig. 75 is far to gross for this application).
It is understood that GSFC is continuing its investigation of the periodic degradation,
and that more precise corrections may be available in the future. Asa result, no
quantitative correction for periodic degradation was attempted in our analyses. The
15
average error due to this effect is estimated to be less than 2°C, which is insufficient
to account for the several discrepancies between conventionally observed sea surface
temperature and the equivalent Channel 2 temperature corrected for average degra-
dation and atmospheric absorptions, as will become apparent in later sections of this
report.
16
5. CASE STUDIES
5.1 North Atlantic, Labrador Current
Early in this study, four cases were selected for areas of the North Atlantic.
The areas chosen exhibit high climatological SST gradients during summer months,
and it was expected that adequate conventional data from ship reports would be
available. The selected cases all occurred during late August or early September
of 1963. Nearly coincident TV pictures and nephanalyses were used to assure some
regions of clear skies within the selected areas.
Of the four cases selected, one was found to be unusable since it included only
closed-mode radiometer data. While it was not a criteria for case selection, there
were areas of overlap between each of the remaining three cases.
The analysis of the best of these cases is shown in Figure 5-1. These data
are taken from Pass 1126 of TIROS VII, on 3 September 1963 between 1234Z and
1238Z. The isotherms shown in this and all subsequent analyses are uncorrected
for atmospheric attenuation or sensor degradation. Rather, attenuation and degra-
dation corrections were determined for individual points or areas as required for
comparisons with conventional data. The isotherms are drawn at 2-1/2°K intervals.
The cloud edge shown in Figure 5-1 was determined by using a 10 % albedo threshold
and the Channel 5 data. The dashed extensions to the solid isotherms, both into the
cloudy area and beyond the edge of the data, are merely extrapolated best estimates.
The 267. 5°K isotherm shown below the edge of the Pass 1126 data was taken from an
earlier pass.
Fortunately, the weather ship Bravo was located within the clear area of the
analysis, and provided upper air data for use in atmospheric attenuation corrections.
These corrections were calculated as discussed in Section 4. 3 above. They indicated
that, due to atmospheric attenuation, the apparent radiative temperature of the sea
surface would be 6°K less than the actual sea surface temperature. From the sensor
degradation estimates in the TIROS VII Radiation Data Cxmllon’, it was found that
the measured temperatures would be approximately 2°K less than actual for these
temperatures and time period. The average temperature of the waters off Labrador
in this season is 40°F, or 277°K. Figure 5-1 suggests that 265°K is a good
(uncorrected) representative temperature for the area involved. When increased
by the 8°K combined correction for attenuation and degradation, an average corrected
sea surface temperature of 273°K is obtained, or 4°K less the anticipated value. As
discussed at the end of Section 4.3, such a difference between conventionally observed
surface temperatures and the corrected Channel 2 data is not unusual.
17
(€96] ‘sequiajdas Z) 1/11 PUD (E96) ‘sequiajdes ¢) 92]| SessDq ‘OIjuD]JW YYION |-G B14
Age MoSS MoO09 MoS9
No ay Esa NoGG
11/11 $S0qg WOl,
(7 ~H
4 SSS _ ——
See +s; 5s0
lett IS ee eerie
| OAveE
ays 43a4]08M
l $292 OL
6292)
=
i=}
ins)
S
a
L a\le ral ESA : 3°
Si S292 =
L ESN a |
no
Sod 4 iS
iL, a
Aa \
Ser a
S9¢\- =e oa
NoO9 si ; ; NoO9
$292 092 GlS2 S62
MoOG MoS MoO09 MoGS9
18
Conventional surface ship data concurrent with the selected cases were re-
quested from the National Weather Records Center (NWRC), but unfortunately not
a single ship had filed reports for the desired areas during the appropriate time
period.
There are several interesting features in the temperature pattern presented
in Figure 5-l. There is a strong indication of the cold Labrador Current, which
moves southward off the eastern coast of Labrador. To the east of the current
there is a nearly isothermal region, probably due to mixing with a warm counter -
current which moves northward along the western coast of Greenland. ne There is
some indication of this warm current in the extreme eastern edge of the analysis,
but the significance of the pattern there is speculative.
A pass on the previous day was found to be largely cloud covered (based on
a Channel 5 10 % albedo threshold), but a small clear area of the earlier pass did
overlap a restricted area in the southwest corner of Pass 1126. This was Pass 1111
of TIROS VII, on 2 September 1963 at 12132. The limited SST data from Pass 1111
seemed to fit well with those from Pass 1126 and are included in Figure 5-1.
Figure 5-2 shows the analysis ofa still earlier pass over this same area.
These data are taken from Pass 1082 of TIROS VII, on 31 August 1963 at approximately
1309%Z. The useful SST data were somewhat more restricted by cloudiness in this
case, and seemed in general more noisy. When compared with Figure 5-1, Figure 5-2
shows that the detailed structure is considerably changed, but the gross pattern
remains the same. Whether or not actual synoptic changes are shown by the difference
in detailed structure could not be determined because of the apparent noise in the
IR data and the lack of conventional data. There is still an indication of the Labrador
current off the east coast of Labrador, with a generally isothermal area east of the
current.
The uncorrected Channel 2 values shown in Figure 5-2 average about 2. 5°K
higher than those for Pass 1126. Because of a decrease in the moisture content of
the upper atmosphere, the attenuation correction in this case is only 5°K (aire less
than for Pass 1126) while the correction for sensor degradation remained at DOK
After these corrections, the approximate average temperature of the field of view in
Pass 1082 is 274.5°K, or 1.5°K higher than that seen in Passes 1111 and 1126.
There are indications that such an average temperature change over an area of this
size is quite Tose o nubile ls but, without conventional data, the reality of this change
could not be substantiated.
19
(E961 ‘LSNONV IE) Z8Ol Ssdd ‘UO YON Z-G “B14
MoOS MoGG MoO09 NoGS9
NoGS | NoGG
2
afpg 2100
= SS aes
| JOPDIGOT
NoO9 SEER a fe ph ileee ee Oe eee ! ! Nees cafes aN fee eS NoO9
MoOG MoGG Mo09 MoS9
20
5.2 Western Pacific Cases
Due to the difficulty in finding sufficient cloud-free cases in the high temperature
gradient areas of the North Atlantic (due especially to the lack of reduced "real time"!
data as discussed in Section 3.2.3), it was decided to investigate the relatively
cloud-free areas of the Western and mid-Pacific. It was hoped that parts of the
Kuroshio or North Pacific Currents could be observed. Three cases were selected
for the Western Pacific, and one for the mid-Pacific. Of these four, the mid-Pacific
case and one of the Western Pacific cases could not be properly analyzed due to
problems associated with near closed-mode conditions. The remaining two cases
will be discussed separately:
5.2.1 Kuroshio Current Case
The data for this case were run in both the listing and the mapping format.
The mapping was done at a scale of 1:2,500,000, but the data proved to be much
too noisy for adequate analysis on this seale (adjacent data points at an approximate
20 n. mi separation differed by as much as 20°K). Presumably much of the apparent
noise may be attributed to errors in geographical locations due to the near closed-
mode conditions. It was decided to hand average the uncorrected data over 1°
Squares, and to consider abrupt changes in the Channel 5 values as outlining the
cloudy areas. (Listings in the albedo format were not available at the time these
analyses were prepared.) Since the area was partly cloudy, the nephanalyses were
insufficient for precise determination of clear areas.
The analysis is shown in Figure 5-3. The data are from Pass 779 of TIROS VII,
on 11 August 1963 at approximately 02382. In addition to the problem of the Channel 2
noise, there was occasionally conflicting information as to the cloudiness of an area
where individual swaths of the Channel 5 data overlapped each other. As a precaution,
all areas indicated as cloudy on any swath were eliminated from the analysis. The
resulting cloud pattern, as shown in Figure 5-3 is remarkably similar to that of
the nephanalysis from a concurrent TV pass; the nephanalysis showed a series of
narrow cloud bands extending westward across a mostly clear area from a larger
cloud system.
The gross outlines of the warm Kuroshio current can be seen to the southeast
of Japan, although detailed structure was lost in the averaging process. The usual
north-south temperature gradient in this region for the month of August is from
21
140°E l45°E SOse
e Normalized Ship Reports
Fig.5-3 Kuroshio Current, Pass 779 (I! August, 1963)
22
294 to 301°K ve or a difference of about 7°K, which is in agreement with the tem-
perature gradient as analyzed from the IR data.
Upper air data from the Japanese station at Tateno indicated a hot and moist
troposphere, with a slight subsidence inversion at 820 mb. Taking 275°K as the
average temperature of the field, and considering the 20° minimum IR sensor
nadir angle of the pass, an approximate atmospheric attenuation correction of 9°K
was determined. This, plusa 2°K sensor degradation correction, adjusts the
average observed temperature to 286°K, still short of the 300°K indicated in mean
monthly charts for this region.
Conventional ship measurements of SST were extracted from the National
Meteorological Center (NMC) charts over a five day period centered on the date
of this pass (11 August). These conventional measurements were generally in
line with the climatological values, leaving the approximately 14° discrepancy in
the IR data unexplained. The ship data were converted to degrees Kelvin, and then
reduced by 25°K to simplify the comparisons with the IR data. (The conversion of
the ship data to a Channel 2 base, rather than vice versa, was used to reduce the
amount of data to be modified, since only a very limited amount of conventional
data was available.) The modified ship data are plotted in Figure 5-3. Considering
the lack of dependability of typical ship reports of SST, and the fact that the ship
data cover a five day period, the two types of data correlate quite well. It seems
reasonable to infer that, given a suitable conventional observation to serve as a
benchmark, absolute temperatures as well as temperature gradients may be deduced
from the TIROS IR measurements.
5.2.2 Yellow Sea Case
The FMR tapes for this case were processed in the same manner as those for
the Kuroshio Current case discussed in the previous section. Again the noise level
was such that averaging in is squares was necessary. When relative changes in the
Channel 5 data were used to eliminate cloud points, only a narrow clear strip re-
mained, running southwestward from the southern tip of Korea to about 25° North.
The data taken from Pass 780 of TIROS VII, at 0417Z on 11 August 1963,
revealed an unpatterned, essentially isothermal field with an average uncorrected
temperature of 279°K. Climatological aadyas reveal a temperature gradient of
less than 2.5°K from the tip of Korea to 25° North.
23
Sparce upper air data from Mosulpo in Korea indicated similar, but slightly
drier conditions than reported at Tateno that same day (11 August); the Tateno data
were discussed in the previous section. Due to the effect of higher nadir angles,
however, the appropriate attenuation correction in this case is 10°K. The total
error is thus approximately 12°kK, raising the corrected average to 291°K, still
9° less than the anticipated value for this region. No ship reports were available
to substantiate the finding of isothermal conditions.
5.3 Consistency Tests
As can be seen from the previous discussions, the problem of finding con-
ventional data to substantiate the IR observations and analyses is a formidable one.
Accordingly, it was decided to apply a series of consistency tests to the IR data.
If the IR data from different days can be shown to be consistent over areas where
it is expected that day to day temperature changes will be small, then a greater
level of credability may be assigned to the day to day small and large scale SST
pattern changes observed in the absence of conventional measurements. Three
areas were chosen for this part of the study: (1) the Sea of Okhotsk, which is
partially sealed off from the North Pacific by the Kamchatka Peninsula and the Kuril
Islands; (2) the Gulf of Alaska; and (3) the relatively isothermal waters of the Indian
Ocean off the northwest coast of Australia.
5.3.1 Sea of Okhotsk Cases
The Sea of Okhotsk is an area of slow SST change. Nephanalyses for 22, 23
and 25 June 1963 indicated largely clear conditions over the area. On all three days
the TIROS radiometer was in the alternating-open mode when passing over the Sea
of Okhotsk.
The analysis for the first day is shown in Figure 5-4. These data were taken
from Pass 51 of TIROS VII, on 22 June 1963. As was to be expected for this remote
area of the world, no ship reports were available. Little is known of the local cur-
rents, nor even of the climatic temperature trends.
In Figure 5-4, a band of relatively warm water can be seen around the northern
coast of Sakhalin, and particularly around the northern entrance to the Tatar Strait,
which is between Sakhalin and the Russian mainland. This pattern at first suggested
a possible warm current, passing northward from the Sea of Japan. Closer inspection
24
60°N
fp
Nid VYLVHIWVY &
&
- w S
6 eG
a 8
a Bes
i ne)
ny wo
G) (op)
Ss es
a ee
VE
5
| ol
CaN}
| —
een
S 5
3 S 8
ee
sy
[o)
tS
AS
(e)
ue
(o}
oO
wo
wn
& GO aes
8 8 0
om
Le
[] rez
: “(Im *—B000
=
8
25
of this and the following days, however, reveals an indentation toward the Strait
entrance of the local isotherms, a pattern indicative of a southbound cool current.
From these maps alone, no conclusive decision as to current direction could be
reached.
Extending northwestward from the southern tip of Kamchatka is a suggestion
of a cool current. The cold Oyashio Current is known to flow southwestward along
the eastern coast of Kamchatka. He The cool current suggested by the IR analysis
may be branching from the main Oyashio Current. These suggested currents, and
a southbound return current off the east coast of Sakhalin, are indicated in Figure 5-4.
It should be noted that both this return current, and a southward current through
the Tatar Strait are indicated in standard current maps.
In the northeastern corner of the Sea of Okhotsk is a large area of relatively
warm water which does not appear in the analyses of later passes.
Figure 5-5 shows the analysis for this same area one day later, using data from
Pass 65 of TIROS VII. Useful coverage here is considerably less than that of the
previous day, while the minimum nadir angle is a high 35a lOn higher than in Pass 5l.
Again Sakhalin is surrounded by a band of relatively warm water, but the
warmer water at the northern entrance to the Tatar Strait is no longer evident. The
presumed cool branch of the Oyashio Current appears to extend further north, while
the warm waters in the northeastern sector of the Sea of Okhotsk can no longer be
identified. It is interesting to note that the small scale cold and warm spots, seen in
the area to the east of the warm water surrounding Sakhalin, can still be identified.
The generally lower temperatures throughout the field of view may be partially
attributed to the increased nadir angle. Another possible cause for these lower
temperatures is the periodic fluctuation in Channel 2 degradation with the orbit-sun
phase geometry. During this period there was a rapid downward excursion in sensor
sensitivity (see Fig. 75, Ref. 6) which would have the effect of decreased Channel 2
temperature values. Upper air data from the Alexandrovsk station on Sakhalin
indicated no change in air mass between the two days.
The complete disappearance of the northeastern warm waters can not be
accounted for. Considering the high level of consistency of the remainder of the
map, this disappearance may, at least in part, be real. Other investigators
have found, however, that the maximum 24 hour temperature change that one can
: : : fo)
expect over any extensive region is about 2 K.
26
There was no useful IR pass over the area on the 24th of June, but, on the
25th, TIROS Pass 95 again encountered clear skies over the Sea of Okhotsk. By
this time, minimum nadir angles were near or above AB”. As a result, the validity
of the analysis of this day is questionable, although, as may be seen in Figure 5-6,
the fundamental large scale features remained intact. The warm waters in the
northeastern part of the Sea of Okhotsk, seen in Pass 51, have not reappeared. Upper
air data, again from Alexandrovsk, indicated a net increase in moisture on this day
of such an extent that, under similar geometric conditions, the IR recorded temperature
would be expected to be from 1/2 to 1 degree lower.
5.3.2 Gulf of Alaska Case
The second area selected in the series of consistency tests was the Gulf of
Alaska, just southeast of the Alaskan Peninsula. This area appears on mean
monthly maps as a large, nearly isothermal region with an average August tem-
perature of about 284°K. Unfortunately there was only a restricted area of cloud
free (as determined by relative changes in the Channel 5 data) overlap between the
two passes available, and that in an isothermal region devoid of pattern. The data
as analyzed in Figure 5-7 were largely taken from Pass 735 of TIROS VII on 8 August
1963 at 0324%. The dashed line in the southeast corner of the figure represents
the northern extent of the overlapping data from Pass 748 on the following day at
002424. Pass 748 also showed this region to be isothermal, and at the same absolute
temperature as indicated by Pass 735.
Upper air data for both days from the Alaskan King Salmon station indicated
a well mixed lower layer with near saturation from the surface to about 800 mb.
Because of the relatively cold air temperatures, however, this did not represent
a significant amount of precipitable water, and the correction term is only 5°K.
Combined with a 2° correction for sensor degradation, the average corrected IR
temperature becomes about 280°K. Again, the corrected temperatures are some-
what less than the true values.
In spite of its limited applicability as a consistency test, this case is considered
worthwhile because of the suggestion of a warm current moving westward at about
58°N (see Fig. 5-7), whichis not analyzed in mean monthly charts. At about 42°N,
the North Pacific Current moves due east across the Pacific Ocean toward the
American mainland. ye (It is originally formed to the east of the Asian continent
28
MoSv/
NoSS
(E961 ‘ssnbny 6) Sb2 pud (S96 ‘ISNBNY g) Gey SessDd ‘DYSDIV jo yjn5 yYZ-c -bi4
MoO0S/ MoSS/
MeO0S/ MoSS/
Me09/
Me09/
No GS
NoO9
by a confluence of the southward flowing Oyashio Current and the northward flowing
Kuroshio Current, both of which have been discussed in connection with previous
cases.) Upon reaching the American mainland, it divides into (1) the southward
flowing California Current, and (2) the Alaska Current which curves counter-clockwise
past the Gulf of Alaska and then follows the Alaskan Peninsula. Itis suggested
that the elongated warm pattern which, in Figure 5-7, extends across the area from
east to west may represent the northern-most branch of the Alaska Current at the
longitude where it again recurves southward.
5.3.3 Indian Ocean Case
The third series selected in this set of tests was located in that portion of
the Indian Ocean just off the northwest coast of Australia. This area was one of
generally clear skies throughout the period 6 August through 8 August 1963, as
evidenced by both nephanalyses from TIROS TV passes, and conventional weather
maps. This was also one of the few occasions in this investigation when it seemed
relatively safe to assume clear skies during a nighttime pass. Two passes were
selected for analysis; one from TIROS VII, Pass 706, on the 6th of August 1963
at 0348Z or 1148 local time; and the other 2-1/2 days later from TIROS VII,
Pass 744, on 8 August at 16552 (or 0055 on 9 August local time).
The first of these analyses is shown in Figure 5-8. As indicated previously,
the area is largely isothermal, with only a slight increase in temperature toward
the north (equatorward). The cloud mass shown in the western portion was deter-
mined using a 10 % albedo threshold applied to the Channel 5 listings. The satellite
was in alternating-open mode in both cases, so the data as originally plotted were
somewhat noisy, particularly in view of the small over-all gradients. Much of this
problem disappeared when the data were re-plotted as averages over ih squares.
Figure 5-9 shows the corresponding night-time analysis from Pass 744. With
the exception of the relatively cool waters along the Australian coast, the pattern
looks quite different, although the absolute temperatures are not drastically changed.
The area of the warmest waters in Figure 5-9 was largely cloud covered at the time
of Pass 706 (Fig. 5-8). The cooler temperatures indicated in the northwest corner
of the Pass 744 data seem out of place, and may well be the result of some cloudi-
ness. Of course Channels 3 and 5 provide no information as to.the cloudiness of an
area in night-time passes. The periodic variation of Channel 2 degradation with the
orbit-sun phase geometry is relatively steady during this period, and hence cannot
account for the generally warmer temperatures found in the second case.
31
FoSel
SoG/
FoS2/
082 > L
(E961 "HSNBNY 9) QOL SSDd ‘uda90 UDIpU|
Je02/ JeGI/
—
RX
08Z <1< G22 |
G28E<1< G82
g-g bl4
F002! JeS//
EAS
SG!
FoOl!
Se0¢
(E961 *
O82
082 <1< S282
isnbny 8) pp SSDg awl}
7002!
F002
JUBIN
UD990 UDIPU|
6-S"
614
oO!
S0G/
Jo0//
33
No upper-air data were available for this area, and of course there were no
ship reports. Assuming a large atmospheric attenuation correction of 8°K, anda
sensor degradation correction of DR, the approximate average temperature of the
Pass 706 analysis, 282°K, becomes 292°K. The average August temperature for
these waters is 298°K. ie Part of the discrepancy may be due to the relatively large
minimum nadir angle of 30° found in both cases, and part to the presence of invisible
ice crystal clouds or atmospheric particulates.
5.3.4 Results of SST Consistency As Compared to Other Studies
As has been the case in all phases of this feasibility investigation, lack of
conventional shipboard SST measurements was a major stumbling block. In this
particular series of tests, we have taken repeated looks at a given area at intervals
of a day or two, and then checked the resultant IR analyses for consistency. Inherent
in the procedure is the assumption that there is consistency to be found. Without
simultaneous intensive shipboard measurements, the validity of this assumption
is always somewhat subject to question.
Numerous investigators have studied both the short and long term fluctuations
in SST, and their possible causes. Wolff et eit indicate that 48 hour changes in
temperature, averaged over relatively large areas, may reach £ 4°K in areas of
sharp temperature gradients. On the average, however, these changes are of the
order of a OL SR, Te ag suggested that these changes in SST are caused predominantly
by advection patterns which are large in scale, corresponding in area and time scale
to atmospheric disturbances at the surface. These findings are substantiated by
Chase”, who found that warming generally occurred in southwest winds prior to
the passage of a cold front, while cooling occurred in northerly winds after the
frontal passage.
During the discussions which led to the 'Recommendations of the Panel on
Sea Surface Meaapemameem: ” of the Conference on the Feasibility of Conducting
Oceanographic Explorations from Aircraft, Manned Orbital, and Lunar Laboratories,
several of the oceanographers present stated that, over large areas of the oceans,
day-to-day changes in SST, and also year-to-year changes for the same calendar
month, are considered to be so small that they place greater credence in a climato-
logical value than they do in direct observations from any single ship. It was these
opinions that led to the stringent requirements for accuracy of satellite-observed SST,
to fractions of a degree C, which are stated in the Recommendations. yy Our meteor -
ological experience, however, leads us to presently view these opinions with some
34
doubt, feeling that they may result more from lack of adequate data and the conse-
quent necessity of using primarily climatological techniques, rather than any real
knowledge, for many areas, of the typical day-to-day and year-to-year variability.
Many meteorologists can remember when conditions in both the tropics and the
stratosphere were considered to be highly persistent, with little or no departure
from climatology to be expected; adequate subsequent observations have since
rudely dispelled these illustions. The extent of coverage and the frequency of
observations of SST,which only satellites can make practically possible, probably
represents our best and perhaps our only chance for determining the real degree
of variability of SST.
Some work has been done using airborne IR measurements of SST. oe Errors
at present are of the order of whole degrees. Ina study undertaken by the Sandy
Hook Marine Laboratory of the U.S. Fish and Wildlife Service, repeated IR measure-
ments of the SST patterns of the middle Atlantic continental shelf were made from
heights of 200 to 500 feet at approximate two week intervals. Figure 5-10 shows a
successive pair a of these measurements for the month of May, 1964. The data
are plotted in degrees Farenheit at 1° intervals. It should be noted that here also
large scale patterns persist, while the smaller scale patterns have changes in the
two week interval. The absolute temperature in some areas of the maps have also
changed by as muchas 10°F. Of course, this is an area near the western edge of
the Gulf Stream where significant SST changes are not unlikely.
Our tests indicate that where there is no general change in the sea surface
temperature as averaged over an area of reasonable size, the observed IR patterns
within this area also remain relatively constant; this result seems reasonable in
terms of scale considerations. This is particularly true of large scale current
features such as were seen in the Sea of Okhotsk, and in the Western Atlantic
in Section 5.1. Only in the coastal waters of Australia was there a complete change
in pattern, and here there were other influences or possible problems such as a
generally isothermal sea, a long Dail [2 day interval, and the use of a nighttime
case in which the presence of scattered or low cloudiness will always remain a
possibility. The persistence of the small cold and warm spots off the eastern coast
of Sakhalin in the Sea of Okhotsk indicated that even small scale features may at
times change only insignificantly. The only really negative result of these tests was
the disappearance of the large warm area in the northeastern portion of the Sea of
Okhotsk. The high level of consistency in the other portions of this same case
suggests that there may have been either an actual change in SST, or, more likely,
undetected change in cloudiness or atmospheric absorption. In view of the general
35
{ aeriaL SHELF SURVEY
| SURFACE ISOTHERMS - °F.
from Infra-Red Radiation Thermometer A SS fi Gti ect aes
Sandy Hook Marine Laboratory
| U.S. Bureau of Sport Fisheries
and Wildlife
in cooperation with
The U.S. Coast Guard and
The Committee for Scientific
Exploration of the Atlantic Shelf.
Depth Contours
Flight Path
eooccce Drop Points for Drifters
“, WEATHER: Legs A-E, Wind-NNE-ESE
35 kts. Air Temp. 51.8°-53.6° Fj Sky clear to
fog, Calm Sea.
eqs F-K, Wind-NE-N kts. Air Temp 536°F
Sky PC. Extensive Fog, Calm Seo.
Legs L-M, Wind-S.W. 5-lOkts, Air Temp
53.6°-68.0°F, Sky Clear-P.C.- Slight Haze,
Large Ground Swell
{AERIAL SHELF SURVEY feces : \
re erg re ae > it at == seniatasiieeeerescal WN PorRPoise
‘SURFACE ISOTHERMS~-°F: 2), : hd WHALE
from Infra-Red Radiation Thermometer X “to + , M& shank
Or FISH
Sandy Hook Marine Laboratory <3 ScHoors
U.S. Bureau of Sport Fisheries
and Wildlife
in cooperation with
The U.S. Coast Guard and
The Committee for Scientific
Exploration of the Atlantic Shelf.
Depth Contours
Flight Path
eccoce Drop Points for Drifters
WEATHER: Legs A-E, Sea Calm, Sky Clear
to P.C., Wind N.Skts Air Temp 67.1°-750°F.
Logs F-K, Sea Calm, Sky Clear. Wind S.W.
| 1Okts., Air Temp 75°F.
=) Legs L-M, Sea Calm, Sky Clear, Wind S.W.
5-7 kts, Air Temp. 60.8 °F.
Fig. 5-10 Sandy Hook Marine Laboratory Airborne IR Mappings
pattern consistency revealed in these tests, it is felt that even a single pass over a
given area can usually provide an accurate indication of at least the large scale SST
gradient patterns. If a conventional measurement is available for a benchmark,
reasonably accurate absolute values of SST can be deduced.
5.4 Real-Time Studies
As discussed in Section 3.2.3, the greatest density of conventional ship SST
reports occurs off the eastern coast of the United States. It is here also that the
most investigated of the ocean's major currents, the Gulf Stream, provides an
excellent opportunity to observe synoptic scale temperature changes, especially
along its edges. Unfortunately, there is very little regularly processed TIROS IR
data for this region; most of the IR data for this area are the so-called "'real-time"
data, which currently require a special, long and laborious conversion to usable
form.
For these reasons, only three "real-time"! cases for this area were selected
and ordered from the Computations Group at NASA's Goddard Space Flight Center.
Of these three cases, timing errors caused the rejection of one of them, and were
prevalent ina second. The remaining case proved to be satisfactory, although
noiser than the usual IR data. The analysis of the first of these cases is presented
in Figure 5-11. These data are from Direct Pass 107 of TIROS VII, on 26 June 1963.
This is the better of the two usable "'real-time'! cases, and clearly shows evidence
of both the southbound Labrador Current and the Gulf Stream. These patterns show
considerable similarity to the most recent mean monthly charts available for this
area. ee The dashed isotherms in Figure 5-11 represent best estimates, used to
bridge areas of poor data.
Upper air data from Washington, D.C. for this date revealed a low relative
humidity throughout the atmosphere, except very near the surface. Recorded air
temperatures were relatively high. Atmospheric attenuation calculations indicated
that an approximate 5°K temperature correction should be added to the recorded
Channel 2 values. This pass occurred rather early in the lifetime of TIROS VII, and
no sensor degradation correction is required. With these corrections, the recorded
temperatures are about 12°K lower than those indicated for this region in mean
monthly charts.
37
75°W
4O°N
JIN
Cloud
38
282
e
© 283
Fig. 5-11
7O0°W
SS AEM
NO Ne 280
Z 282.5 9 295
N 2825
Fy N
Bs ap mare
282 Y),
265
a pers 285 e/
2S ia
e 28/ /;
hy
AN
aria 0)
‘ 9
5 Br «
289 288 Ve
r) V &
yi Yi
, 7
90 y Za
Oe ya YL
a 7
le 289 w7
ae 287
i
75°W 7O°W
Real-time Data, Pass 107-D (26 June, 1963)
4O°N
J5°N
The second usable pass (238-Direct) occurred 9 days later, on 5 July 1963.
The data here had considerably more noise, but the basic SST patterns are still
visible. (See Fig. 5-12).
Temperatures for this pass were 2.5 to 5°K lower than in Pass 107-D. Upper
air data, again from the Washington station, indicated a slightly drier atmosphere
for this day, although the calculated correction was still approximately 5°K. As in
the previous case, there was no significant correction for sensor degradation in-
dicated by the usual degradation graphs. ° It should be noted, however, that the
periodic orbit-sun phase fluctuation does suggest a rapid degradation during this
period. This may account for part of the over-all temperature drop. The minimum
nadir angle was 10" higher than in the previous case, but was stilla relatively low
20°. Thus, there seemed to be no adequate explanations for the general 2.5 to 5°K
temperature drop across the field of view. As this is the time of year when one
would expect a gradual warming of the sea surface, it is unlikely that this temperature
drop is real.
Conventional surface ship data were extracted from NMC charts for five day
periods centered on the dates of each of the selected "real-time! cases. These
data indicated no significant change between the two cases, and agreed reasonably
well with mean monthly charts. They were converted to °K and uniformly reduced by
12°K for easier comparison with the IR data. The '!\reduced'! data are plotted in both
Figure 5-1l and 5-12. In Figure 5-11 it is seen that there is good correlation between
the adjusted ship data and the recorded IR data. Good correlation can also be seen in
Figure 5-12, if the ship data are reduced another 2.5 to 5°K to compensate for the
over-all drop in the IR data.
The first of these two cases is a particularly good demonstration of how a
dependable surface temperature measurement may be used as a benchmark to
calibrate the recorded IR temperatures. Using this procedure false shifts in the
absolute values of a temperature field can be avoided. It may be that only an
occasional benchmark (occasional in both space and time) will be required to properly
calibrate recurring IR coverage.
39
4O°N
3S5°N
40
75°W
7O°W
>
ee 270
eg i
@
\ 275 272.5
‘ 282
\
283 | I
e775 ) |e 28 oe 277.5
286
@
S
XN
a3 282
@
280 Cs
BER pay 0289
283
@ 0286
284 y fi
0286 vy,
282.5 WA
288
@ nave —
Grom Ti panne age
75°W 7O°W
Fig. 5-l2 Real-time Data, Pass 238-D (5 July, 1963)
40°N
J5°N
6. DISCUSSION OF THE PROPOSED EXTENSIVE
DATA PROCESSING APPROACH
The problems encountered in attempting to find suitable cases which simul-
taneously satisfied the three criteria of clear skies, large SST gradients, and
adequate conventional data included the following: (1) frequent absence of usable
IR data due to high nadir angle or closed mode conditions; (2) frequent lack of
TIROS TV picture data for use in determining whether the area of interest was
adequately clear; (3) inadequacy or lack of conventional surface meteorological data
over the oceans for determining clear areas; (4) the impossibility of determining
the clear areas from the IR data themselves, (using the threshold albedo techniques)
until the expenses required for FMRT processing have already been incurred; and
(5) the tendency for areas of greatest synoptic interest and change, suchas the Gulf
Stream, to have frequent and persistent fronts and cloud cover.
Using the currently available (TIROS) IR data, little can be done with regard
to Point (1) above; it will be totally alleviated as regards closed mode, and con-
siderably as regards nadir angle, when Nimbus MRIR data becomes available.
Points (2) and (3),however, can presently be avoided by facing upto Point (4) and
accepting the costs and inefficiencies inherent in FMRT runs without specific prior
knowledge as to the presence or absence of adequate clear areas. (Of course, a
general examination of any available data, prior to an FMRT run, is desirable to
rule out obviously hopeless cases, or to select the more promising of alternative
cases.) Rejection of cloud contaminated data on the basis of a Channel 5 albedo
threshold value seems feasible, as demonstrated above, and resonable modifications
of the existing data reduction programs should permit automatic rejection of these
data. Point (5) is really the only condition inherently imposed by nature, and like
Point (1) must be lived with. The long and tedious chore of trying to find areas of
clear sky over such difficult regions of the oceans, however, can be relegated toa
computer when repeated passes over a given area for some specific time period
are to be automatically processed. In all probability, the savings in human search,
comparison, and decision time will more than compensate for the increased computer
costs.
For the initial pilot investigation using these procedures, areas with a reasonable
probability of synoptic change should be chosen, and cases would then be run for
those periods which, by reference to other readily available meteorological sources
(seasonal tendencies, conventional weather charts, TIROS TV data, etc), present
Al
the greatest likelihood of having at least minimally adequate periods of clear skies.
Analyses for individual days and running averages of the data over several more
or less consecutive days should be investigated. These averaged maps should
provide a good basis for tracing the gradual changes of large scale SST patterns,
while comparisons of the individual days should reveal capabilities as regards
smaller scale and more rapid changes and developments.
As a first pilot study to investigate the general feasibility of such an approach,
a six day case was run, using both the listing and the mapping formats, for the area
off the western coast of Australia. The time span covered by the six days processed
was from 29 July to 8 August 1963; and the TIROS VII data used included Passes
589, 647, 662, 691, 706, and 735. For the first few passes, an albedo threshold
value of 10% was used with the Channel 5 listings to outline clear areas. This
resulted in areas which were too restricted for practical use. By locating abrupt
changes in the Channel 2 temperature values, it was determined that a 20 % albedo
threshold could be safely used in this case, revealing a much larger apparent clear
sky area. All six passes were then processed using the 20 % albedo limit. Of the
six passes, only one appeared completely overcast. The outlined clear areas of
the several maps were then hand averaged producing the final average analysis
in Figure 6-1.
Mean monthly SST maps for this region show isotherms running east-west
across the Indian Ocean, and then dipping southward as they reach Australia. From
the southern to the northern tip of western Australia the mean monthly SST rises
approximately TK , in agreement with the results in Figure 6-1. No upper air
data or ship reports were available for this area, but, even with reasonable atten-
uation and degradation corrections, the IR observed SST's would again be colder
than those anticipated for this region. The individual maps which were used to
make up the final analysis in Figure 6-1 were also analyzed, but revealed no
obvious continuous or significant change in SST patterns. The individual analyses
showed more complex SST patterns than are seen in the averaged mapping. Apparently
one advantage of the averaging process is its ability to ferret out the more significant
patterns from the background of extraneous noise of both the natural and the sensor
produced varieties. The positive results of this trial case have seemed to reaffirm
both the utility of the TIROS IR observations as a source of valid SST data, and our
feeling that, for extensive studies or operational uses of satellite IR data for SST
determinations, the types of extensive processing methods proposed above will have
_to be employed.
42
120°E
WINE
105°E
Austra/ia
eae 275 <1 <2775
[| e7e5<r<e75
RSS§ 270. <7 < 2725
Fig. 6-1 Six Pass Averaged SST Patterns (29 July-8 August, 1963)
43
7. GENERAL SUMMARY AND CONCLUSIONS
These studies have clearly demonstrated that the satellite IR data provide
good measurements of the gradients of sea surface temperatures. This has been
confirmed, in part, by consistency of the IR data for areas of known SST persistence,
which has given some indication of the validity of the observed SST patterns on single
passes. These consistency tests have revealed that larger scale patterns and
gradients are easily recognizable and change little from day to day, implying that
these patterns are real, and can be accepted without requiring confirmation from
conventional measurements (although such conventional substantiating data would
still be desirable as we build up further experience regarding these matters). The
tests have also revealed that, where there is persistence of small scale patterns,
they also can be seen on repeated passes. Whether they will always be seen, and
whether observed changes in these small scale patterns are real, remains unanswered.
The answers can probably come only from finding adequate samples of satellite SST
data with concurrent conventional observations, or from the processing and analysis
of statistically significant quantities of satellite observed SST data.
Subsequent investigation aimed at determining the extent of the validity of IR
observed small scale patterns will be significantly helped by concurrent ship reports,
and it may be that such reports will be indispensable. The Gulf Stream provides an
ideal area to observe both mesoscale and synoptic scale changes, and has the greatest
density of conventional ship measurements. If significant amounts of "real-time"
TIROS IR data were to become available, this would be an obvious area to conduct
further studies along these lines; this by itself constitutes a strong justification for
routine processing of a full year of such "real-time"! data. Otherwise, such invest-
igations will be hampered until the Nimbus C MRIR data become available.
The tendency of the satellite IR temperatures, even when corrected for
atmospheric attenuation and sensor degradation, to be significantly cooler than
conventionally measured temperatures, indicates the need for an occasional bench-
mark to calibrate the data. It also re-emphasizes the already existing need for further
investigation of these discrepancies, which have been repeatedly noted in various
analyses of the TIROS IR data. Even if one of the previously proposed explanations
can be shown to be the answer, there will still remain the need for improved methods
for estimating the proper corrections.
The determination of clear sky areas from the IR data themselves uses the
Channel 5 sensor to distinguish cloudy from clear areas, and has the advantage
of reducing or eliminating dependence on satellite TV or conventional meteorological
44
data. Our studies indicate that this can be done for daytime cases for overcast
or broken conditions. Areas of small scattered cloud, however, may erroneously
decrease the Channel 2 temperature values without being detected by Channel 5.
Nighttime detection of cloud points using Channel 2 and 4 differences is not presently
possible, and no alternative useful approach to the nighttime cloud detection problem
is yet apparent. The use of sharp changes in the Channel 2 values is insufficient,
since it may not reveal areas of low or scattered clouds.
The pilot study off the western coast of Australia indicated that methods of
extensive data processing can usefully be applied to SST studies. The final map
in this study was produced by manually outlining clear areas and then hand averaging
point by point. With reasonable modifications of existing computer programs, this
tedious process can be automated, and individual or averaged clear sky maps
produced as a direct computer output. The human with his experience and knowledge
of local currents and mean temperature patterns is still required for the optimum
final analysis. We believe that, for comprehensive studies or operational uses of
SST, such extensive data processing techniques will be required. The application
of these techniques to satellite IR data can significantly increase our knowledge
of sea surface temperatures and their patterns, gradients, and variations over the
several scales of time and space at which they have significant scientific and prac-
tical applications.
45
10.
Wik.
12.
13
14,
46
REFERENCES
Palmer, E., 1958: 'On the Formation and Structure of Tropical Hurricanes,"
Geophysica, 3, pp. 26-38.
Fisher, E. L., 1958: "Hurricanes and the Sea Surface Temperature Field,"
YOUR, OH METEOR, 5 UB(S)), Wis S2H-333.,
Petterssen, S., Bradbury, D. L., and Pedersen, K., 1962: 'The Norwegian
Cyclone Models in Relation to Heat and Cold Sources," Geofys. Publikosjoner,
24, pp. 185-222.
Bradbury, D. L., 1962: "A Note on the Use of Sea Surface Temperatures in
Synoptic Weather Analyses,'! J. Applied Met., 1 (3), pp. 421-425.
Glaser, A. H.; Merritt, B.S.) Wexler, R., and Widger, Jr., W.-K, 19653 ihe
Applicability of TIROS and Nimbus Data to Investigation of the Feasibility of
Sea Surface Temperature Measurements from Satellites, '' Oceanography from
Space (G. C. Ewing, Editor), Woods Hole Oceanographic Institution, Ref. No.
65-10, pp. 219-227.
Staff Members, Aeronomy and Meteorology Division, 1964: TIROS VII Radiation
Data Catalog and Users! Manual-Volume I, NASA, Goddard Space Flight Center.
Bandeen, W.R., Hanel, R.A. and Strange, I., 1964: A Radiation Climatology in
the Visible and Infrared from the TIROS Meteorological Satellites, NASA Technical
ote D- , Goddar pace Flight Center.
National Weather Satellite Center, 1963: Catalogue of TIROS VI and VII Cloud
Photography, USWB.
Sherr, P.E., 1965: Techniques for Improving Geographical Locations of TIROS
Radiation Data Listed from FMRT, Reprinted from Final Report, Contract
Cwb-10812Z, ARACON Geophysics Company, Concord, Massachusetts.
Wexler, R., 1964: "Infrared and Visual Radiation Measurements from TIROS III,"
Applied Optics, 3, pp. 215-219.
Staff Members, Aeronomy and Meteorology Division, 1962: TIROS III Radiation
Data Catalog, NASA, Goddard Space Flight Center.
Curtis, A.R. and Goody, R.M., 1954: "Spectral Line Shape and its Effect on
Atmospheric Transmissions," Quart. J.R. Met. Soc., 80, p. 58.
Wark, D. Q., Yamamoto, G., and Lienesch, J.H., 1962: "Methods of Estimating
Infrared Flux and Surface Temperatures from Meteorological Satellites," J. of
the Atmos. Sciences, 19, pp. 369-384.
-Zdunkowski, W., Henderson, D., and Hales, J. V., 1964: The Influence of
Haze on Infrared Radiation Measurements Detected by Space Vehicles, Final
Report, Contract No. Cwb-10648, Intermountain Weather, Inc., Salt Lake
City, Utah.
IB),
16.
Io
18.
19.
20.
elke
(are
References (Cont. )
U. S. Navy Hydrographic Office. 1944: World Atlas of Sea Surface Temperatures,
H.O. Pub. No. 225, Washington, D.C.
Sverdrup, H. U., Johnson, M.W., and Fleming, R.H., 1942: The Oceans-Their
Physics, Chemistry, and General Biology, Prentice-Hall, Inc., New York.
Wolff, P.M., Laevastu, T,, and Russell, J., 1965: Short-Term Fluctuations
of the Sea Surface Temperature, Their Magnitudes, Causes and Effects, Tech.
Note No. 6, Fleet Numerical Weather Facility, Monterey, California.
Chase, J., 1959: '"Wind-Induced Changes in the Water Column Along the East
Coast of the United States," J. Geophys. Res., 64, pp. 1013-1022.
Tully, John P. (Chairman), 1965: "Recommendations of the Panel on Sea Surface
Temperature, '!' Oceanography from Space (G. C. Ewing, Editor), Woods Hole
Oceanographic Institution, Ref. No. 65-10, pp. 117-118.
Oshiver, A. H. and Berberian, G.A., 1965: "Sensing Sea-Surface Temperatures
by Airborne IR,' Geo-Marine Technology, March 1965, pp. 22-26.
Aerial Shelf Survey, 1964: Surface Isotherms - °F from Infra-red Radiation
Thermometer, Sandy Hook Marine Laboratory and U.S. Bureau of Sport
Fisheries and Wildlife, G.P.O. Nos. 947-719 and 947-852, Maps for May 4,
5, and 7 and May 18, 19, and 21.
Schroeder, E.C., 1965: Average Surface Temperature Maps, Woods Hole
Oceanographic Institute, Woods Hole, Massachusetts, (to be published).
NASA-Langley, 1966 CR-474 47
v 5 i 4 ae bho a oth
" ‘ 6 7 ye a svat vayapene Sila pedis by
= ) if p b i 5
: ‘ ¥i i : ‘
ie, : | ? 1
Daten ay ¢ ny '
st fl ’
ji ; ey i : Aa
\ . Abr e 7
tb oe He |
Toe iy tie Ay i Diet Hlth as ‘
i ont hits Y vv :
1g
; 4
: Sy : ' \ pe re y
) hal i
HAIKU a ila WeeKapis t } SBA
: ‘ ; 4 SCR 9 Gs .
s . t ry } Wy t fi ‘
hy ‘ :
"
i eee | { F | i
‘oes 4 ne Cyl i
; Ral Sony bee y ees ‘
i -~ ~ a. ti vr
h ay j in ; Caen OAR
{ ere oa ‘ it ee oie Ch
€ y j
4 Pa
tA f “
a VT mi
" sf J }
{ : , hea aA, fe
- » a)
y las ‘ | re : $ r} \ a ‘ ik
= i i j ; ‘ eal! 2 AN RR ar
HP fb td py YR ‘ ie Pee :
4 A ne
i j ;
Z i Oy { q
, \ / _
t es “
\ C ( ' \ chs j ‘.
r a ‘ " i
i wi ; i }
j I
} .
}
. y s / | ,
ss ; f , ~
uu i { ; 4
]
iy ya vA ? \ hive Sek he ~ f
BIS Ry ce SIT n { Sian ) i i }
; ~
| *, if 39 (i 1
iv |
j
i ,
Gayl
i
é
Ware dl }
Ni tarh
i peanec eae
and ae Masia aa
‘f
aeons D.C 2054