SEA SURFACE TEMPERATURE, THE RELATIONSHIP
BETWEEN ANALYSIS DETAIL AND DATA DENSITY
AND DISTRIBUTION - SOUTH CHINA SEA - NORTH-
EAST MONSOON

by

Harold Patrick Sexton, Jr,

United States
Naval Postgraduate Schoo

THESI

Sea Surface Temperature, the Relationship between

Analysis Detail and Data Density and Distribution

— South China Sea - Northeast Monsoon

by

Harold Patrick Sexton, Jr,

September 1970

Tki& document kcu> been apphovtd ^on. pub tic. k<l-
ItaAQ. and i>aJL<i', JLtb diAtnAhiitlori aj> unVjtuXz.d,

T136i:?:5

3AOT ■■•- .

IL POSTGRADUATE SCHOOtJ

DERBY, CALIF. 93940

Sea Surface Temperatures, the Relationship
Between Analysis Detail and Data Density

and Distribution
- South China Sea - Northeast Monsoon

by

Harold PatrickJJexton, Jr.

Lieutenant Commander, United States Navy

B.S., Boston College, 1959

Submitted in partial fulfillment of the
requirements for the degree of

MASTER OF SCIENCE IN OCEANOGRAPHY

from the

NAVAL POSTGRADUATE SCHOOL

September 1970

ABSTRACT

A daily subjective analysis of sea surface temperatures in the South China Sea
for three monthly periods during the Northeast (Winter) Monsoon was perfoiined. The
analysis model developed showed a three-part division of the South China Sea into a
cold eastern side, a warm central region and a cold western side. Bottom topography
and surface currents played a major role in delineating analysis patterns. The rela-
tionship between data density and distribution and analysis detail was found to be
affected adversely by continuity, when data density was small, and by the distribu-
tion of the observations, when the data were unevenly distributed. When the area
per observation was relatively small and the data were evenly distributed, the number
of observations and the number of analysis features which resulted from the analysis
had a good linear relationship.

TABLE OF CONTENTS

PAGE

I. INTRODUCTION - 9

A. STATEMENT OF THE PROBLEM _ 9

B. APPROACH TO THE PROBLEM 9

C. DATA SOURCE 10

D. DESCRIPTION OF THE STUDY AREA 12

1 . Bathymetry 12

2. Climate 14

II. BACKGROUND 15

A. SEA SURFACE TEMPERATURE ANALYSES 15

B. SEA SURFACE TEMPERATURE - SOUTH CHINA SEA 16
1. Mean Monthly Conditions During the Period of the Study 16

III. DATA TREATMENT AND PRESENTATION 18

A. ANALYSIS PHILOSOPHY 18

B. ANALYSIS PROCEDURES 18

C. ANALYSIS FEATURES AND SUB-REGIONS 20

D. - COMPARISON WITH OTHER AVAILABLE ANALYSES 23

IV. RESULTS 24

A. ANALYSIS PATTERNS - - 24

B. ANALYSIS DETAIL AND DATA DENSITY AND DISTRIBUTION 25

C. EFFECT OF CONTINUITY ON THE OBSERVATION TO FEATURE 45
RELATIONS

D. EFFECTS CAUSED BY ADDING AND REMOVING DATA 45

E. ANALYSES FROM OTHER SOURCES - 46

V. CONCLUSIONS 50

VI. RECOMMENDATIONS 51
APPENDIX A - SURFACE CURRENT CHARTS _ 52
APPENDIX B - SAMPLE SEA SURFACE TEMPERATURE ANALYSES 57
APPENDIX C - CALCULATION OF THE AVERAGE AREA PER OBSERVATION 64
BIBLIOGRAPHY 67
INITIAL DISTRIBUTION LIST 68
FORM DD-1473 69

LIST OF TABLES

Table Page

I Summary of Observations 1 1

II Percent Frequency of North or Northeast Winds 16
at Pratas Reef

III Monthly Average of the Daily Values for the Average 43
Area per Observation and for the Average Number of
Observations in Each Sub-region

LIST OF FIGURES

Figure Page
1 . Study Area 13

2. Sample SST Analysis 19

3. Sub-regions 21

4. Example of How Features Were Counted 22

5„ Monthly Average of the Number of Observations per 28

Feature vs Features - March-April

6. Monthly Average of the Number of Observations per 33
Feature vs Features - November

7. Monthly Average of the Number of Observations per 38
Feature vs Features - December

8. Monthly Average of the Area per Observation per Day 44
vs Monthly Average of the Number of Observations

per Day

9. Sample FWC Guam SST Analyses 48

I. INTRODUCTION

A. STATEMENT OF THE PROBLEM

The purpose of this study was to determine the effect of very dense sea surface
temperature (SST) data on daily analysis patterns. Subjective relationships between
data density, data distribution, and the amount of analysis detail have been sought.
Normally, only sparse SST observations are available for a particular area and
a description of the SST patterns on a day-to-day basis is therefore difficult to ob-
tain. However, due to the increased shipping activity - both civilian and military -
in the South China Sea in recent years, sufficient daily observations have been
obtained to make a description of daily SST patterns meaningful for that area.

B. APPROACH TO THE PROBLEM

The approach used in this study can be outlined as follows: One would expect
that for a given sea surface feature.there is a minimum number of SST observations
necessary to specify it. Further, there should be a minimum number of SST obser-
vations which would bring out all the analysis features for a given area and any
observations beyond the minimum number would serve only to support those features.
It was apparent that, even though the amount of data for this study was considered
dense compared to data available on a daily basis for other oceanic areas, it was
not sufficient to establish the minimum number of observations which would show all
the features for this area. Therefore, it has been assumed that there is a certain
minimum number of observations which would show all features present in a given
area, but the minimum number was not reached in this study.

Accordingly, the minimum number of observations necessary to specify a single
analysis feature was the aspect upon which the emphasis was placed in this study.
A relationship was sought between the number of observations available and the
features which appeared in the analysis.

In another somewhat related study, Duing [13] has examined the structure of
SST analysis patterns in the Indian Ocean. However, the approach and emphasis
differ considerably from those of this study. Duing was concerned with the size of
computer contoured analysis features based on mean SST values in one degree squares,
and his results cannot be compared directly with the results from this subjective study.
C. DATA SOURCE

Five-day composite SST data for the periods 1 1 March - 10 April 1969 and
1 November through 31 December 1969 were obtained from Fleet Weather Central
Guam (FWC Guam). SST data had been plotted on five-day composite charts in a
color code which permitted extraction of daily SST values. The color code used by
FWC Guam for the five-day composites consisted of four colors. The first two days
of each composite period were indistinguishable from each other and were therefore
replotted together. Data for days 3, 4, and 5 could be separated by individual day
and were so plotted. In terms of number of analyses, twenty-four charts thus were
prepared from a month's data. This method of handling the data was not considered
biasing to the study since the stated purpose was to treat the observations in as small
a time frame as possible - daily if the data were available.

Table I is a breakdown of the number of observations which were analyzed. No
attempt was made to evaluate the accuracy of the observations reported nor to weight
them in accordance with the method by which they v/ere obtained (i.e0, injection

10

TABLE I
SUMMARY OF OBSERVATIONS

NUMBER OBSERVATIONS PER DAY OBSERVATIONS REJECTED

OF

PERIOD OBSERVATIONS

AVERAGE HIGH LOW NUMBER PERCENT

11 Mar-
10 Apr 2,833 91 152 35 7 0.25%

91

152

35

65

98

26

63

102

23

1-30 Nov 1,950 65 98 26 11 0.56%

1-31 Dec 1,946 63 102 23 29 1.49%

Totals 6,729 73 152 23 47 0.69%

11

thermometer, bathythermograph, XBT, etc.). However, data with obvious errors due
to coding and/or transmission were disregarded in the analyses. The percentage of
data rejected by this criterion was very small - less than one percent. (See Table I.).
D. DESCRIPTION OF THE STUDY AREA

The South China Sea embraces an area of 1.8 x 10 square nautical miles. This
largest of the marginal seas in the western Pacific is bordered on the west by Vietnam,
by the southern limit of the Gulf of Thailand and by the Malay Peninsula; on the east
by Formosa, the Philippines and Borneo; on the north by China; and on the south by
the rise between Sumatra and Borneo (about 3S). The specific part of the South China
Sea which was involved in the study is shown in Figure 1 . This chart is the area of
the South China Sea on operational synoptic weather charts of FWC Guam.
1 . Bathymetry

The major topographical feature of the South China Sc*a is a rhomboid-shaped
basin with depths in excess of 2000 fathoms. (See Figure 1.). Reef-studded shoal areas
occur within the basin, e.g. Macclesfield Bank and Scarborough Shoal. Along the
northwest side of the basin a shelf extends about 150 nautical miles offshore and in-
cludes the Formosa Strait and the Gulf of Tonkin. The islands of Hainan and Formosa
are situated on this shelf. The Gulf of Tonkin gradually deepens toward the center
reaching a maximum depth of less than 50 fathoms. The main entrance to the South
China Sea from the Pacific Ocean is by way of the deep Bashi Channel (north of the
Philippines), where most of the water exchange takes place. Relatively little water
is transported through the other connecting channels, on the south, east and west,
due to shallow sill depths.

12

13

2. Climate

The period covered in the study was the end of the Northeast Monsoon
(cold season) of one year and the first part of the Northeast Monsoon of the follow-
ing year. Typically, during the Northeast Monsoon, cold air moves from its Siberian
source equatorward as a high pressure system around which the winds circulate clock-
wise o Off Vietnam, this circulation is characterized by winds having a northeasterly
component o The surface air of the Northeast Monsoon flows southwestward gaining
heat from the warmer surface of the South China Sea. This situation creates a semi-
permanent cloud cover extending down to near 15 N. [5].

14

II. BACKGROUND

A. SEA SURFACE TEMPERATURE ANALYSES -

Sea surface temperature (SST) is one of the most easily measured oceanic proper-
ties and one of the few properties reported on a synoptic basis. SST is indicative of
other conditions and processes in the sea. The value of the knowledge of SST distri-
bution in synoptic oceanography has been well established, especially as applied to
navigation, naval operations, weather forecasting and fisheries problems.

The factors determining and modifying the surface temperature distribution and
subsurface thermal structure may be divided into five groups: (1) heat exchange,
(2) convective mixing, (3) advection, (4) mechanical mixing and (5) factors usually
of a local nature, such as runoff, tidal currents, and upwelling. [6] For this study
of the South China Sea, emphasis was placed on the last two of the factors listed,
since extensive meteorologically motivated studies of the other factors have been
made. [See especially 5, 6, and 9]. Numerous charts of long-term monthly average
SST's have been published which depend primarily on the period of the data collected
and the averaging techniques used. None of these "mean monthly" presentations takes
into account relatively sudden changes in SST which can be caused by any of the five
factors listed above. According to Wolff [6], the greatest variations in SSJ are caused
frequently by advection due to wind-driven currents. Table 2 shows the percent fre-
quency of north or northeast winds and the percentage greater than 25 knots at Pratas
Reef ( see Figure 1 ) for the months of November and December. [5]

15

TABLE II

PERCENT FREQUENCY OF NORTH OR NORTHEAST WINDS AT PRATAS REEF

NOV DEC
N/NE Winds (%) 86 80

%>25 knots 63 52

B. SEA SURFACE TEMPERATURE - SOUTH CHINA SEA

The values and variations of temperature in the waters of the South China Sea
are affected primarily at the surface by the monsoons and at depth by the bottom
topography. [9] The surface temperature decreases northward. It is lowest in
January and February, when the circulation induced by the winter monsoon (North-
east Monsoon) drives the cold waters southward; and highest in August, when the summer
monsoon (Southwest Monsoon) reverses the current pattern and insolation is at a maxi-
mum. As a result, surface temperatures are most variable in the shallow extreme

northern portions of the South China Sea. The mean annual range of surface tempera-
s' o
ture is about 32 F, and in exceptional years, the range can be as large at 40 F.

However, in the southernmost portions of the South China Sea, the annual variability

in SST rarely exceeds 3 F. [9].

1. Mean Monthly Conditions During the Period of the Study

The mean monthly sea surface temperature conditions for the South China

Sea have been taken from the FWC Guam Oceanographic Outlook for the months of

March, April, November and December. During March, the SST maximum varies from

o o

less than 70 F along the Vietnam coast and in the Gulf of -Tonkin to 83 F south of

ION. Isotherms are oriented generally SW to NE. During April, there is a period of

general warming as the Northeast Monsoon season comes to an end*. South of 15N,

the SST has the value of 82-84 F. In the Gulf of Tonkin and northern regions, the

16

o
SST varies between 75-77 F. During November, the prevailing Northeast Monsoon

causes the surface waters to cool due to decreased insolation caused by the semi-
permanent cloud cover extending down to 15N. [5] In addition, during the winter
monsoon, evaporation exceeds precipitation, heat is conducted to the colder atmos-
phere from the warm surface waters below, and wind mixing extends down through the

o
thermocline. SST varies between 79-83 F in the central regions compared with

values of 70-75 F along the China coast. During December, the Northeast Monsoon

°
continues to cool the South China Sea. The SST is between 75-80 F in the central

o
regions and 68-73 F along the China mainland. The December SST in the Gulf of

o
Tonkin is about 68-75 F. [1, 2, 3, 4]

17

III. DATA TREATMENT AND PRESENTATION

A. ANALYSIS PHILOSOPHY

In analyzing the sea surface temperature charts, the following principles were
observed:

1. All observations were considered valid, except the very small percentage
noted in Table I .

2. All valid observations were drawn to in the analysis.

3. Continuity was taken into account but it was not the determining factor.
In order not to filter out the daily changes, a moderate degree of change was per-
mitted on consecutive charts.

4. Analysis features were related to the bottom topography and surface currents.

5. Excessive "tonguing" of analysis features was avoided except in the Tonkin
Gulf sub-region where the dense data required, at times, very complicated and speci-
fic analysis patterns. (See Figure 2 for a sample SST Analysis).

B. ANALYSIS PROCEDURE

Seventy -two separate analyses were prepared for the study period using bottom
topography and surface current charts as underlays in the interpretation of the data.
Bottom topography and surface current charts were used as underlays: -

1. To permit isotherms to be drawn around islands and reefs - an important
factor considering the highly irregular topography of the South China Sea.

2. To allow relatively small irregular features to be drawn in shallow water
depths when the number of observations permitted.

3. To allow isolated features to be drawn when the surface current patterns
indicated a feature could have been broken off and isolated by a surface current eddy.

18

19

4. To Impart a reasonableness to the entire analysis; for, like Duing [13], it was
assumed that the sea surface temperature patterns were linked to the wind driven sur-
face circulation.

C. ANALYSIS FEATURES AND SUB-REGIONS

After the analyses were prepared, carefully checked and reviewed, a deter-
mination was made of the number of analysis features present with the corresponding
number of observations for each chart for each sub-region, (sub-regions are indicated
in Figure 3). Any discernible crest or trough and any closed isotherm was designated
an analysis feature. Figure 4 is an example of how the number of analysis features
were counted for three sub-regions on the actual analysis chart for 15 March 1969.

Answers to the following questions were sought:

1. Does data density and distribution determine the analysis detail?, and the
corollary

2. If one adds/removes data, does one get more/less analysis detail?

To answer these questions, for each of the three monthly periods of the study, the
number of observations in each sub-region which went into analyzing a particular
number of features was averaged and plotted against that number of features. The
total number and the variability of the observations (standard deviation) were also
indicated. In addition, the frequency of occurrence of a particular number of
features during the monthly period was plotted on the same graphs. (See Figures
5, 6, and 7).

The area (in square nautical miles) of each sub-region was determined from
standard U.S. Navy Hydrographic Office charts and the area per observation for
each day in each sub-region was then calculated. (See Appendix C). For each of

20

*"s

1/1

1/1

01

z

™*

o

E

o

«

LLl

3

(X.

0

c

CO

cr

3

M

t/>

_c

rS

0

oi

if

L.

u.

0

>-*

21

22

the three monthly periods of the study, the area per observation per day was averaged
and plotted against the average number of observations for that sub-region. (See
Table III and Figure 8).

D. COMPARISON WITH OTHER ANALYSES AVAILABLE

Two sources of analyses were available for comparison with the analyses of this
study, the mean monthly analyses based on a climatological approach to the data
[12], and daily operational analyses of FWC Guam for the period 11 March - 10
April 1969. FWC Guam's analyses were not influential in the formation of those
of this study because they were not available to the author until after the present
analyses were completed. A relation between analysis features and data density and
distribution was sought. Also, the method by which FWC Guar., utilized bottom
topography and surface current information in making the analyses was examined.
Additionally, FWC Guam layer depth and below layer gradient synoptic charts for
the same period were studied to see if a relationship existed between them and the
SST analyses - both those prepared by Guam and those of the present study.

23

IV. RESULTS

A. ANALYSIS PATTERNS

The seventy -two analyzed sea surface temperature charts prepared showed one
overall common characteristic, a three-part division of the South China Sea into a
cold western side, a warm central portion, and a cold eastern side. (See Appendix B
for sample analyses). The three parts were oriented generally SW to NE. The topo-
graphical arrangement of the South China Sea actually lends itself to the three-part
characteristic indicated in the analyses. The cold saline surface water of the Pacific
Ocean comes into the South China Sea via the Bashi Channel [10], is picked up by the
monsoon-induced surface currents, swept counterclockwise around the top of the South
China Sea, and moves down the coast of Vietnam picking up the cold less saline water
moving out of the Tonkin Gulf as it goes. The maximum surface current velocities are
experienced along the coast of Vietnam. (See Appendix A.).

The cold eastern side of the South China Sea is characteristic of a deep water
mass with relatively low surface current velocities which maintains a more or less
permanent equilibrium adjusting by convective processes for the warmer surface water
advected slowly from the south.

The warm waters advected from the south are confined to a warm tongue extending
up into the west central region of the South China Sea. Relatively shallow waters,
low surface current velocities, and numerous islands and shoal areas help to keep the
surface temperatures in this central areo warm.

The three-part characteristic indicated that one major wave length was present
with various smaller wave lengths superimposed on it. The smaller wave lengths were

24

affected by the character of the monsoon (since the monsoon determines cloud cover,
wind direction, and the amount of runoff), but more importantly, the smaller wave
lengths appeared to be caused by the irregularities of the bottom topography and
surface current eddies, especially off the coast of Vietnam.

B. ANALYSIS DETAIL AND DATA DENSITY AND DISTRIBUTION

Figures 5, 6 and 7 represent the relationship by sub-region of the number of
observations to the number of features analyzed. The relationships were considered
either good or poor using the following criteria:

good - the number of features showed a linear increase with increasing number
of observations; standard deviations were included in determining the linear character
of the plot; in some cases, one or , at most, two data points were omitted; in all cases
where data points were omitted, the data point had a frequency of occurrer ce of one.

poor - no consistent pattern of the data points was discernible.
A good linear relationship existed between the number of observations present and the
number of features analyzed when the area-to-observation ratio was relatively small
and the data were more or less evenly distributed throughout the sub-region.

Sub-region 'a' indicated a good relationship for each of the three monthly peri-
ods of the study. The data in Sub-region 'a' were relatively dense and evenly distri-
buted. Sub-region 'a' contains the main shipping lanes in the South China Sea- both

1

There should be a minimum number of SST observations which would
bring out all the analysis features for a given area. On reaching this minimum number
of observations, the relationship between observations and features will cease to be
linear. The number of features will remain constant while, theoretically, the number
of observations will increase without bound. This minimum number of observations was
not reached in this study and, therefore, only the linear part of the relationship is
considered.

25

north-south and east-west. It is unencumbered by any large islands or shoal areas.
Three to six analysis features occurred most frequently.

Sub-region T also indicated a good relationship. In two of the three monthly
periods studied, Sub-region T had more observations than Sub-region 'a'. Sub-
region T also indicated a good distribution of data was present although the area
north of 20N was not always covered very extensively. Four to seven analysis
features occurred most frequently.

For the three monthly periods of the study, Sub-region W indicated the best
relationship between the number of observations and features analyzed. The standard
deviations of the data points were the smallest of any other sub-regions and from one
to four analysis features occurred most frequently. Sub-region W was the smallest
sub-region in total area but nevertheless the observations were dense and distributed
evenly - probably due to the large amount of coastal ship traffic moving in the sub-
region.

Slopes of the lines drawn through the data points of the figures for Sub-regions

2

'a', T and W were between 0.4 and 1 .7. (The average slope was 0o9).

Sub-regions 'c' and N indicated a poor relationship between the number of
observations and features analyzed for each of the three monthly periods of the study.
These two sub-regions had the smallest data density not only because the main shipping
lane of the South China Sea passes only through the very eastern side of Sub-region N

2

" The line drawn through the data points from which the slope was determined

included subjective considerations of the standard deviation of the data points, the
total number of observations making up each data point, and the frequency of occur-
rence of the feature during the monthly period. At times, one or two data points had
to be disregarded. The disregarded data points were, in all cases, tho.'; which had a
frequency of occurrence of one.

26

and the northwestern corner of Sub-region 'c', but also because the military shipping
in the South China Sea did not concentrate in these areas and further reduced a
potential source of data. Sub-region 'c' has extensive shoal areas and Sub-region
N contains Pratas Reef. Slopes of the lines drawn through the figures for Sub-regions
'c1 and N have a maximum value of 0. 1 . (Most figures had no slope).

The sub-regions which indicated a poor relationship between the number of
observations and features analyzed plotted above the 11,5000 sq. n.m. line on
Figure 8 where the average number of observations per day was between four and
nine. This indicates that in areas of about the same size (for example, Sub-regions
a, b, c, d and N) greater than nine observations per day were needed to specify the
analysis features for that area. (Sub-region N which has an area of about 74,000 sq.
n.m. appears close enough in total area to include in the comparison above.).

Sub-region 'b' for November (Figure 6b) was the only poor sub-region which
plotted below the 11,500 sq. n.m. cut-off line on Figure 8. A possible explanation
for this is, although the data were relatively dense, the distribution was uneven.
It is Sub-region 'b' that contains the Paracel Island and Macclesfield Bank which
cause shipping traffic to be diverted around them and hence SST observations to be
distributed unevenly in the sub-region. (Sub-region 'b' for December was in the
poor category and plotted with the cluster of other poor sub-regions on Figure 8.
Sub-region 'b' for March-April showeda relatively good relationship but had over
a hundred more observations than in November which were evidently distributed
sufficiently for an observation-to-feature relationship to result).

27

5a

40

A
V

30

Sub-regioti a (Mar-Apr)

N
O

O
B

P

E
R

20

10

E
A

«*>«* J-

X

/

\

1 2 3 4 5 6

FEATURES

LEGEND for Figures 5,6,7

Standard Deviation

Number of Observations
Making up ( ) the Average

Frequency of Occurrence of
Featurep''Duri>i£ the Month

5b

40

A
V
E

N
O

O
B

30

a 20

P
E
R

A

Sub-region b (Mar-Apr)

/
/
/
/ -*•

a 1 ©

1 2 3 4 5 6 7 8

FEATURES

Figure 5: Monthly Average of the Number of Observations

per Fe ^ure vs Features

(March-April)

28

5c

40

Sub-region c (Mar-Apr)

A
V
E

N
O

O
B
S

P
E
R

F
E

A

30

20

10

24) /

/ 'A >

12 3 4 5

FEATURES

5d 40f Sub-region d (Mar-Apr)

A
V
E 30

N
O

O
|20

P
E
R

F 10
E

A

@ (S>

^.,

12 3 4 5 6

FEATURES

Figure 5: Monthly Average of the Number of Observations

per Feature vs Features

(March-April)

29

Sub-region T (Mar-Apr)

60

70

A
V
E
R
A
G
E

N
O.

50
O
B
S
E
R

V 40
A
T
I

O
N
S 30

P
E
R

A
T
U
R
E 10

— */

• \

V

123456789
FEATURES

Figure 5: Monthly Average of the Number of Observations

per Feature vs Features

(March-April)

30

5f 40

A
V
E

N
O

O
B

30

20

10

Sub-region N (Mar-Apr)

59 40 Sub-region W (Mar-Apr) 5h 40 Sub-region S (Mar- Apr)

A

V
E

N
0

0
B
S

P
E
R

F
E
A

30

20

10

>*»* — t

A
V
E

N

0

0
B
S

P
E
R

F
E
A

12 3 4 5

FEATURES

6

30

20

10

©-'

(W f\®

©

12 3 4 5 6

FEATURES

Figure_5: Monthly Average of the Number of Observations

per Feature vs Features

(March -April)

31

5i

40

Sub-region Ficticious Maximum
(Mar -Apr )

A
V
E

8

O
B

30

20

P
E
R

10

©

1,2 34 567 89

FEATURES

5j

40

30

8

i

20

P

E
R

F
E

10

Sub-region Ficticious Minimum
(Mar-Apr)

©

\«. ^

12 3 4 5 6
FEATURES

Figure 5: Monthly Avetage of the Number of Observations

per Feature vs Features

(March -April)

32

6a

40

30

A

V
E

N
0

0 20
B

S

/
F

E
A

10

Sub-region a (November)

©

N-*^

2 3 4 5 6

FEATURES

7 8 9 10 11

6b

40 Sub-region b (November)

A

30

V

E

N

0

20

0

B

S
/

10

F

E

A

©

©

1234 56789 10 11

FEATURES

Figure 6: Monthly Average of the Number of Observations

per Feature vs Features

(November)

33

6c 40 Sub-region c (November)

A
V

E 30

N
0

0 20

B
S
/

F

E
A

10

A fe

,<®

4 ^

' T \y'\

\. ,

12 3 4 5

FEATURES

6d 40 Sub-region d (November)

A
V
E 30

N
0

0 20

B
S
/

F 10

E

A

12 3 4 5 6 7 8

FEATURES

Figure 6: Monthly Average of the Number of Observations

per Feature vs Features

(November)

34

6e

40 Sub-region T (November)

A

V

E 30

N
0

0 20
B

S

/

A

no;,

1 2 3 4 5 6

FEATURES

61 A

V
E

N
0

0
B
S
/
- F
E
A

20

10

Sub-region N (November)

/ 9- &\

1 2 3 4 5 6

FEATURES

7 8 9

Figure 6: Monthly Average of the Number of Observations

per Feature vs Features

(November)

35

69 . 30. Sub..

reeion * (Wo

vember)

FEArUREs

egio» s (Nov

ember)

FEATURES

F,'9ure6-^MyAverogeof

(November)

36

6i i Sub-region Ficticious Maximum

( November)

A
V
E

N
0

0

B

S

/

F

E 10

A

30

20

©

4 5 6 7 8 9

FEATURES

10 11

6j A 30

V
E

N
0

20

Sub-region- Ficticious Minimum
(November)

3 4 5 6

FEATURES

Figure 6: Monthly Average of the Number of Observations

per Feature vs Features

(November)

37

7a 40 Sub-region a (December)

A
V
E

N
0

30

0
B20

s

/

F
?10

/ -- -

*^.

2 3 4 5 6 7 8

FEATURES

9

7b

30

N

0 20

0
B
S
/ 10

.F
E
A

Sub-region b (December)

Figure 7: Monthly Average of the Number of Observations

per Feature vs Features

(December)

38

30, c (December)

FEATURES

7d A 30. «?mv

V / region d r

FUTURES

F'9ure 7; ^

°nf/l/X AverQge of fhe N ,
v^ecemter;

39

7e Sub-region T (December)

40

A
V
E

N
0

30

0

^20

S

/
F
E
A 10

#1

©

,-'x

' •*• H* '

1 2 3 4 5 6 7 8

FEATURES

9 10

71 A

V 30

E

Sub-region N (December)

20

0

B

S

/

F 10

E

A

Figure 7: Monthly Average of the Number of Observations

per Feature vs Features

(December)

40

79

A
V
E

N
0

Sub-

region w (December)

0

20

B

S

/

F

E 10

A

/ \

LA

1 2 3 4

FEATURES

Sub-region s (December)

^e7:MonfMyA f mberof

per Feature vs Features
(December)

ions

41

7i 40

A

30

V

E

N

0

20

0

B

S

/
F

10

E '

A

Sub-region Ficticious Maximum
(December)

T

--

1 23456789

FEATURES

7J

A
V
E

30

N
0

20

0

B

S

/

10

F

E

A

Sub-region Ficticious Minimum
(December)

I T © ©

2 3 4 5 6 7

FEATURES

8

Figure 7: Monthly Average of the Number of Observations

per Feature ys Features

(December)

42

TABLE III

Monthly Average of the Daily Values for the Average Area per Observation
and for the Average Number of Observations in Each Sub-region; as well as
in Two Ficticious Sub-regions as Defined. (Data plotted in Figure 8).

SUB-REGION

MARCH-APRIL NOVEMBER DECEMBER

Ave. Area/ Ave.# Ave. Area/ Ave.# Ave. Area/ Ave.#
Obs. Obs. Obs. Obs. Obs_. Obs.

7472 16 6969 18 9353 15

9976 15 10616 11 16979 9

16186 9 25642 7 23397 8

8592 14 7677 17 9243 13

39 7011 16 4967 19

16854 7 16548 8 17848 6

w 3167 10 9723 3 8413 4

s H201 8 24043 4 16318 6

5731 20 5746 19 7701 17

7 26719 6 27815 6

a
b
c
d
T 1853

* Fie. Max.

** Fie. Min. 18900

* Composite region, consisting of the average of the largest of the
number of observations on each chart for the four central sub-regions
(a,b,c,d) .

** Composite region, consisting of the average of the smallest of the
number of observations on each chart for the four central sub-regions
(a,b,c,d).

43

10000CH

Figure 8: Monthly Average of the Area per Observation per
Day vs Monthly Average of the Number of Obser-
vations per Day. (See Table III)

MAR--APR t

NOVEMBER X

DECEMBER O

50000

10000

5000

s 0

t

11,500

0W

t

0 o

t

x, x

X t

tw

Note: Subscripted symbols refer to specific sub-regions

Non -subscripted symbols refer to the central sub-regions
a, b, c, d.

1000,

AVE NO OBS PER DAY

"42

36

44

C. EFFECT OF CONTINUITY ON THE OBSERVATION TO FEATURE RELATIONS

As stated previously, continuity was not the determining factor in the analyses,
but it was considered. When there were few data and when these did not conflict
with patterns determined previously, continuity was used in shaping analysis features.
As a result, comparing data density to analysis features is necessarily biased by the
continuity. As an extreme example, on one chart six features could appear in a
sub-region based on twenty-two observations; the next consecutive chart based upon
continuity could have six features in that sub-region, although there were only five
observations.

In order to see just what effect continuity had on the relationship between anal-
ysis features and data density and distribution, and how extensively continuity biased
the results of this study, selected sub-regions were re-analyzed removing all influ-
ence of continuity, therefore using analyses which depended entirely on the observa-
tions actually present. After these analyses were made, the average number of
observations per featui o was plotted against the number of features which appeared.
The relationship between the number of observations present and the number of
features analyzed was good for those areas which showed a good relationship using
the original data which included continuity; and poor where the relationship was
poor previously. Therefore, the relationship which exists as a result of the sub-
jective method used in the present study appears to be valid.

D. EFFECTS CAUSED BY ADDING AND REMOVING DATA

To answer the question of whether adding or removing data will result in more
or less analysis definition, one needs to look at the local conditions that exist. In
an area such as the deep basin of the eastern side of the South China Sea, adding or

45

removing portions of the data had little effect since the water mass is relatively
stable and sea surface temperatures vary little from day-to-day, only gradually
becoming colder or warmer as the seasons change. However, off the coast of Viet-
nam and in the Tonkin Gulf, the water masses are very much disrupted by the pre-
sence of shallow water depths, runoff - especially in the Mekong and Red River
Delta regions, surface current eddies, and small islands and shoals in the local area
which cause perturbations in the SST isotherms. In these areas, adding or removing
data very much affects the analysis detail.

E. ANALYSIS FROM OTHER SOURCES

Neither the oceanographic atlas [12] nor the operational analyses of FWC Guam
showed the three-part division of the South China Sea indicated by this study. The
atlas displayed isotherms which showed a cold region close to the coast of Vietnam
but then indicated diagonal lines across the South China Sea with an orientation
generally SW to NE. The FWC Guam analyses for the period available for compari-
son also showed a cold region along the Vietnam coast. However, as the isotherms
moved away from the Vietnam coast, they crossed the South China Sea in an undula-
ting wave pattern without regard for the land masses and shoal areas which intervened,
(See Figures 9a and 9b).

The sea surface temperature analyses of FWC Guam appeared to be principally
based on continuity. There seemed to be no correlation between the number of analy-
sis features and the data density or distribution (indicated by dots in Figures 9 a and
9 b). For example, in the Tonkin Gulf region, Guam's analyses consistently showed
one or two analysis features; whereas l! analyses of the present study showed an

average of six features in the same region. (Only once in the period of comparison
did the present analyses show as few as two features there).

46

In addition, Guam's analyses were characterized by smooth wave patterns which
at times seemed to ignore sizable quantities of data. There did not appear to be any
relationship between Guam's analyses of mixed layer dapth or below layer gradient
and either the Guam SST analyses or the SST analyses of the present study.

47

o
o

i_
3
CO

t)

CTi

<JD

CT5

H

I/O

W

q:

£

CL

3

<

VS

00

o

s

HI

U_

<

O

43

<c

cry

h-

—

to

(/)

or

3

<

U

OS

2

LU
<

49

V. CONCLUSIONS

The subjective sea surface temperature analysis model for the South China Sea
during the Northeast Monsoon season developed in this study showed a three-part
division of the South China Sea into a cold western side, a warm central region, and
a cold eastern side. Bottom topography and surface currents played a major role in
delineating the analysis patterns.

There are two important factors which must be considered when deriving a rela-
tionship between the number of observations and the analysis features resulting from
them. These factors are continuity, when the data density is small, and data distribu-
tion, when the observations are unevenly distributed.

When the area per observation was relatively small and the data were evenly
distributed, the number of observations and the number of features which resulted
from the analysis had a good linear relationship. As the area per observation increased
and the data became unevenly distributed, the number of observations and the number
of features had little or no relationship. In the latter situation, continuity had to
be used to determine the analysis features and the extent to which continuity had
to be used was directly related to the kind of relationship between the number of
observations and the number of features which resulted. -

50

VI. RECOMMENDATIONS

Sea surface temperature analyses should be Investigated for the Southwest
(Summer) Monsoon season in order to complete a sea surface temperature analysis
model for the South China Sea for both monsoon seasons.

The number of sea surface temperature observations available as a result of the
increased shipping activity in the South China Sea during the last five years should
provide an excellent data base to work out more specific relationships between the
number of observations and the analysis features.

Analysts should use underlays of bottom topography and surface currents while
analyzing sea surface temperatures in the South China Sea.

51

APPENDIX A

52

53

54

55

en

H

2

UJ

oc

cr

D

o

UJ

s

E

u.

01

o:

u

^

0)

«n

Q

til

t-

<

o

56

APPENDIX B

57

58

59

60

61

n

to

>

<0

*0

u

c

bl

<

Q

M

H-

I/O

l/>

ill

1-

<

a

62

63

APPENDIX C

64

CALCULATION OF THE AVERAGE AREA PER OBSERVATION

The average area per observation for each monthly period was calculated by the
following method:

1. The area per observation for each sub-region for each chart was obtained by

taking the area of the sub-region and dividing by the number of observations in that

sub-region for that chart. (There were 24 charts for each monthly period.)

Area
X = . =Xi'X2'X3' X24

i No. of Observations

2. The daily values for the area per observation were averaged for the monthly

period.

24 Area

Lj l£_ X. 24 1=1 (No. of Observations).

i=l

X = 2* 9_,

Example : X. = Area/No. of Observations

X = Average Area/No. of Observations

Area = 100 sq. n. m.

No. of Obs. = 5, 10, 20, 10, 5

100 ^r~

* = ^ti; (.2+.1+.05+.1+.2) = (20) (.65)
5 i— I

X = 13 sq. n. m.

The average area per observation (for the five-day period of the example) is

13 sq. n. m.

The average number of observations (for the five-day period of the example) is 10,
The average area per observation was not calculated by taking the total area and

dividing by the monthly average of the observations. Referring back-to the example,

65

this method would result in a value of 10 sq. n. m. for the average area per observa-
tion for the five-day period.

The reason that the average area per observation was calculated by the method
described above was a by-day average of the area per observation was more represen-
tative of what the analyst faces in the spatial distribution of data to be analyzed.

66

BIBLIOGRAPHY

1. Fleet Weather Central Guam, "South China Sea", Oceanographic Outlook,

p. A-l, March 1969.

2. , "South China Sea", Oceanographic Outlook,
p. A-l, April 1969.

3. , "South China Sea", Oceanographic Outlook,
p. B-l, November 1969.

4. , "South China Sea, Oceanographic Outlook,
p. B-l, December 1969.

5. Navy Weather Research Facility Report NWRF 12-0669-144, The Diagnosis and

Prediction of SEAS I A Northeast Monsoon Weather, p. 1-66, June 1969.

6. , Report NWRF 36-0667-126, Analysis and Forecast-
ing of Sea-Surface Temperature (SST), by P. M. Wolff, L. P. Carstensen,
and T. Laevastu, p. 1-48, June 1967.

7. U.S. Naval Hydrographic Office, Currents in the South China, Java, Celebes

and Sulu Sea, (H. O. Pub. No. 236), 1945 (Reissued 1965).

8. Marine Geography of Indochinese Waters, (H.O.
Pub. No. 754), 1951.

9. U.S. Naval Oceanographic Office Informal Manuscript 67-5, Temperature,

Salinity, and Density of the World's Seas: South China Sea and Adjacent
Gulfs, by P. E. LaViolette and T. R. Frontenac, p. 1-134, February 1967.

10. Informal Manuscript 67-10, Oceanographic Cruise
Summary South China Sea, by S. G. Tooma, Jr. and H. Iredale, III,
p. 1-20, February 1967.

11. Informal Report 68-67, Oceanographic Cruise Summary
Cam Ran h Bay -N ha Trang-Poulo Condore Group, by D. E. Kenney, p. 1-24,
September 1968.

12. Special Publication SP-99, Monthly Charts of Mean,

Minimum, and Maximum Sea Surface Temperature of the Indian Ocean,
by P. E. LaViolette and C. Mason, p. 1-48, 1967;

13. University of Miami Scientific Report, The Structure of Sea Surface Tempera-

tures in Monsoonal Areas, by Walter D'uing, p. 1-25, April 1970.

67

INITIAL DISTRIBUTION LIST

No. Copies

1. Library, Code 0212 2
Naval Postgraduate School

Monterey, California 93940

2. Dr. Glenn H. Jung 3
Department of Oceanography (Code 58Jg)

Naval Postgraduate School
Monterey, California 93940

3. LCDR Harold P. Sexton, Jr. 2
USS HEPBURN DE 1055

FPOSan Francisco, California 96601

4. Defense Documentation Center 2
Cameron Station

Alexandria, Virginia 22314

5. Department of Oceanography, Code 58 3
Naval Postgraduate School

Monterey, California 93940

6. Commanding Officer 1
U.S. Fleet Weather Central

COMNAVMARIANAS, Box 12

F. P.O.San Francisco, California 96630

68

Unclassified

Security Classification

DOCUMENT CONTROL DATA -R&D

[Security classification of title, body at abstract nnii indexing annotation must be entered when the overall report Is classified)

I ORIGINATING ACTIVITY ( Corpora te author)

Naval Postgraduate School
Monterey, California 93940

2a. REPORT SECURITY CLASSIFICATION

Unclassified

2b. GROUP

3 REPOR T TITLE

Sea Surface Temperature, the Relationship between Analysis Detail and Data Density
and Distribution - South China Sea - Northeast Monsoon.

4. DESCRIP

TivE NOTES (Type of report and. inclusive dates)

Master's Thesis; September 1970

9. AUTHORISI (First name, middle initial, last name)

Harold P. Sexton, Jr.

6 . FTE PORT DATE

September 1970

7«. TOTAL NO. OF PAGES

68

7b. NO. OF REFS

13

««. CONTRACT OR GRANT NO.

ft. PROJEC T NO

»a. ORIGINATOR'S REPORT NUMBERIS)

9ft. OTHER REPORT NOIS) (Any older num6»r« that may be assigned
this report)

10 DISTRIBUTION STATEMENT

This document has been approved for public release and sale; its distribution is
unlimited.

II. SUPPLEMENTARY NOTES

12. SPONSORING MILITARY ACTIVITY

Naval Postgraduate School
Monterey, California 93940

13. ABSTRACT

A daily subjective analysis of sea surface temperatures in the South China Sea
for three monthly periods during the Northeast (Winter) Monsoon was performed.
The analysis model developed showed a three-part division of the South China Sea
into a cold eastern side, a warm central region and a cold western side. Bottom
topography and surface currents played a major role in delineating analysis patterns.
The relationship between data density and distribution and analysis detail was found
to be affected adversely by continuity, when data density was small, ancTby the
distribution of the observations, when the data were unevenly distributed. When
the area per observation was relatively small and the data were evenly distributed,
the number of observations and the number of analysis features which resulted from
the analysis had a good linear relationship.

DD ,fr:..1473

S/N 0101 -607-681 1

(PAGE 1 )

69

Unclassified

Security Classification

A-31408

Security Classification

Unclassified

KEY WO R DS

Sea Surface Temperature
South China Sea

)D ,F.r..1473 (back.

/N OIOt-807-6821

70

Unclassified

Security Classification

A- 3 t 409

SHELF BINDER

^= Syrocuie, N. Y.
S3 Stockton, Calif.

Thes i s
S4193

Sexton 121928

Sea surface tempera-
ture, the relationship
between analysis detail
and data density and
distribution - South
China Sea - Northeast
Monsoon .

thesS4193

Se.?..surface temPerature, the relationshi

3 2768 001 94509 0

-,_ DUDLEY KNOX LIBRARY